{"id":"gen-1781867892-2ccEOFXhDFXMoPvOOspN","choices":[{"finish_reason":"stop","index":0,"logprobs":null,"message":{"content":"Below is a **physics-based Python script** that predicts the temperature at a point in steel inside a ladle using the **1D transient heat equation** through the bath depth.\n\nIt solves:\n\n\\[\n\\frac{\\partial T}{\\partial t} = \\alpha \\frac{\\partial^2 T}{\\partial z^2}\n\\]\n\nwith convective boundary conditions at the top and bottom:\n\n\\[\n-k\\frac{\\partial T}{\\partial z}\\bigg|_{z=0} = h_{top}(T - T_{top,env}), \\quad\n-k\\frac{\\partial T}{\\partial z}\\bigg|_{z=H} = h_{bottom}(T - T_{bottom,env})\n\\]\n\n> **Note:** This is a simplified model. It assumes constant steel properties and no solidification. \n> `h_top` and `h_bottom` are **effective** heat-transfer coefficients and can represent convection, radiation, and refractory losses.\n\n```python\n# pip install numpy matplotlib\n\nimport numpy as np\nimport matplotlib.pyplot as plt\n\nC2K = 273.15 # Celsius to Kelvin conversion\n\n\ndef predict_steel_temperature(\n z_query,\n t_end,\n H=1.5,\n T_init=1600.0,\n T_top_env=25.0,\n T_bottom_env=200.0,\n h_top=120.0,\n h_bottom=25.0,\n rho=7000.0,\n cp=750.0,\n k=30.0,\n dz=0.01,\n dt=None\n):\n \"\"\"\n Predict temperature in steel inside a ladle using a 1D transient heat-conduction model.\n\n Parameters\n ----------\n z_query : float\n Depth from the steel surface [m]. z=0 is top surface, z=H is bottom.\n t_end : float\n Prediction time [s].\n H : float\n Steel bath depth [m].\n T_init : float\n Initial steel temperature [°C].\n T_top_env : float\n Effective surrounding temperature above the free surface [°C].\n T_bottom_env : float\n Effective surrounding temperature / interface temperature at the bottom [°C].\n h_top : float\n Effective heat transfer coefficient at the top [W/m^2-K].\n h_bottom : float\n Effective heat transfer coefficient at the bottom [W/m^2-K].\n rho : float\n Density of steel [kg/m^3].\n cp : float\n Specific heat of steel [J/kg-K].\n k : float\n Thermal conductivity of steel [W/m-K].\n dz : float\n Spatial step [m].\n dt : float or None\n Time step [s]. If None, an explicit-stability-safe value is chosen.\n\n Returns\n -------\n times : np.ndarray\n Time history [s].\n T_query_hist : np.ndarray\n Temperature at z_query over time [°C].\n z : np.ndarray\n Depth grid [m].\n T_final : np.ndarray\n Final temperature profile [°C].\n \"\"\"\n\n # ---- Basic validation\n if H <= 0 or t_end < 0 or dz <= 0:\n raise ValueError(\"H, t_end, and dz must be positive.\")\n if not (0.0 <= z_query <= H):\n raise ValueError(\"z_query must be between 0 and H.\")\n if rho <= 0 or cp <= 0 or k <= 0:\n raise ValueError(\"rho, cp, and k must be positive.\")\n if h_top < 0 or h_bottom < 0:\n raise ValueError(\"h_top and h_bottom must be non-negative.\")\n\n # ---- Thermal diffusivity\n alpha = k / (rho * cp) # [m^2/s]\n\n # ---- Grid\n n_nodes = max(3, int(np.ceil(H / dz)) + 1)\n dz = H / (n_nodes - 1)\n z = np.linspace(0.0, H, n_nodes)\n\n # ---- Explicit stability limit (conservative)\n Bi_max = max(h_top, h_bottom) * dz / k\n Fo_limit = 1.0 / (2.0 * (1.0 + Bi_max)) # conservative boundary-safe limit\n\n if dt is None:\n dt = 0.45 * dz**2 / (alpha * (1.0 + Bi_max))\n else:\n if dt <= 0:\n raise ValueError(\"dt must be positive.\")\n\n Fo = alpha * dt / dz**2\n if Fo > Fo_limit:\n raise ValueError(\n f\"Chosen dt is too large for explicit stability.\\n\"\n f\"Need alpha*dt/dz**2 <= {Fo_limit:.4f}, got {Fo:.4f}.\\n\"\n f\"Try reducing dt or increasing dz.\"\n )\n\n # ---- Convert temperatures to Kelvin internally\n T = np.full(n_nodes, T_init + C2K, dtype=float)\n T_top_env_K = T_top_env + C2K\n T_bottom_env_K = T_bottom_env + C2K\n\n # Boundary Biot numbers\n Bi_top = h_top * dz / k\n Bi_bottom = h_bottom * dz / k\n\n # ---- Store history at the query point\n times = [0.0]\n T_query_hist = [np.interp(z_query, z, T) - C2K]\n\n # ---- Time loop\n t = 0.0\n while t < t_end - 1e-12:\n dt_step = min(dt, t_end - t)\n Fo_step = alpha * dt_step / dz**2\n\n T_new = T.copy()\n\n # Interior nodes\n T_new[1:-1] = T[1:-1] + Fo_step * (T[2:] - 2.0 * T[1:-1] + T[:-2])\n\n # Top boundary (z = 0)\n T_new[0] = T[0] + 2.0 * Fo_step * (\n T[1] - T[0] - Bi_top * (T[0] - T_top_env_K)\n )\n\n # Bottom boundary (z = H)\n T_new[-1] = T[-1] + 2.0 * Fo_step * (\n T[-2] - T[-1] - Bi_bottom * (T[-1] - T_bottom_env_K)\n )\n\n T = T_new\n t += dt_step\n\n times.append(t)\n T_query_hist.append(np.interp(z_query, z, T) - C2K)\n\n return np.array(times), np.array(T_query_hist), z, T - C2K\n\n\nif __name__ == \"__main__\":\n # Example inputs (replace with your plant data)\n z_query = 0.75 # m below the surface\n t_end = 1800.0 # s (30 minutes)\n\n times, Tq, z, Tfinal = predict_steel_temperature(\n z_query=z_query,\n t_end=t_end,\n H=1.5,\n T_init=1600.0,\n T_top_env=25.0,\n T_bottom_env=200.0,\n h_top=150.0,\n h_bottom=20.0,\n rho=7000.0,\n cp=750.0,\n k=30.0,\n dz=0.01,\n )\n\n print(f\"Predicted temperature at z = {z_query:.2f} m after {t_end/60:.1f} min: {Tq[-1]:.2f} °C\")\n\n # Plot temperature at the chosen point over time\n plt.figure(figsize=(7, 4))\n plt.plot(times / 60.0, Tq, linewidth=2)\n plt.xlabel(\"Time (min)\")\n plt.ylabel(\"Temperature at query point (°C)\")\n plt.title(\"Steel temperature at selected depth\")\n plt.grid(True)\n plt.tight_layout()\n\n # Plot final depth profile\n plt.figure(figsize=(5, 6))\n plt.plot(Tfinal, z, linewidth=2)\n plt.gca().invert_yaxis()\n plt.xlabel(\"Temperature (°C)\")\n plt.ylabel(\"Depth from surface (m)\")\n plt.title(\"Final temperature profile\")\n plt.grid(True)\n plt.tight_layout()\n\n plt.show()\n```\n\nIf you want, I can also generate:\n1. a **2D axisymmetric ladle model** for a true point `(r, z)`, or \n2. a version that **fits `h_top` / `h_bottom` from measured thermocouple data**.","refusal":null,"role":"assistant","annotations":null,"audio":null,"function_call":null,"tool_calls":null,"reasoning":"**Generating a Python script**\n\nI need to create a Python script for predicting temperature at a certain point of steel in a ladle. I think I could use a heat transfer model, possibly looking at 1D radial diffusion in a cylindrical ladle. Although this sounds complex, I might need to ask the user for more specific details, like the depth from the surface. So, I'll provide an example script with some assumptions and input parameters. I should also clarify that it's a simplified model.**Interpreting temperature in steel**\n\nI’m considering how to model the temperature at a point inside the steel mass in a ladle. Given that the steel is a hot liquid or solid, I think heat conduction and convection are key factors here. Since the ladle is axisymmetric, the simplest approach could be a radial heat transfer model, predicting temperature changes using a 1D transient conduction equation. If the steel is well-mixed, a lumped thermal model might work too. I should show the relevant equations, possibly using explicit finite difference with the thermal diffusivity of steel.**Creating a temperature prediction script**\n\nTo predict temperature at a specific point, I think using the heat equation is the simplest approach, represented as ∂T/∂t = α ∇²T in cylindrical coordinates. However, a full 2D PDE is complex. To simplify, I could reduce it to a 1D model, calculating T(x,t) from the wall to the center. Since the user hasn’t specified the geometry, I’ll create a parameterized 1D radial model. Using an analytical solution for transient cooling of an infinite slab or a finite difference solver might work best for the script.**Developing a temperature prediction script**\n\nI'm considering how to create a script to predict the temperature at a specific point in a ladle of steel. The user likely wants Python code, including mathematical models and comments. Inputs could involve initial steel temperature, ambient temperature, heat transfer coefficients, thermal conductivity, density, specific heat, coordinates, and time. I’d think about using finite difference methods to compute temperature distribution in cylindrical coordinates while considering the geometry of the ladle for accuracy. This script might need to be quite complex but should balance detail with usability.**Creating a heat equation script**\n\nI'm thinking about writing a Python script that solves the transient heat equation on an axisymmetric grid using finite differences. The user should be able to query the temperature at any point (r, z) in the steel at a specific time. I need to decide whether to use SciPy or just numpy with explicit time stepping. The explicit method may need a small time step for stability, but I could consider using an implicit method to enhance reliability even if it complicates the code. A self-contained script with comments might be the best approach!**Defining model assumptions**\n\nI'm considering the model assumptions for the heat equation in a steel bath. Here’s what I’m thinking: the properties of steel, such as density (\\(\\rho\\)), specific heat (\\(c_p\\)), and thermal conductivity (\\(k\\)), should be constant. I'll be using axisymmetric cylindrical coordinates for temperature \\(T(r,z,t)\\). The heat equation can be expressed using \\(\\alpha\\), which is \\(k/(\\rho c_p)\\). I also need to establish boundary conditions for convective cooling and symmetry at \\(r=0\\) to query the temperature at a specific point \\((r_{query}, z_{query})\\) after \\(t_{end}\\).**Evaluating temperature modeling**\n\nI’m considering a model for steel's thermal behavior, noting that if the steel is liquid and well-mixed, a simpler lumped model might be suitable. But I could also propose a \"well-mixed\" model as an alternative if needed. I should mention that if we’re deep within the steel, molten convection will likely dominate and the temperature gradients are minimal. I think including plant data could improve prediction accuracy. Next, I’ll work on designing a robust axisymmetric solver while addressing specific code correctness issues.**Developing thermal boundary conditions**\n\nI’m working on a standard cylindrical Laplacian discretization for interior points (r>0). The discretized version involves temperature at neighboring nodes. Then, at the boundaries with convection, I’ll handle the heat flux at the top surface (z=0). \n\nI need to ensure my sign conventions are accurate, especially when measuring the z-coordinate down from the top surface. The mathematical derivation shows the relationship between heat transfer and temperature gradients. Ultimately, I'll express the temperature change in terms of the heat transfer coefficient and the temperature differential.**Evaluating temperature modeling options**\n\nI’m thinking about deriving a discrete update but considering that using a finite volume update for boundary nodes with convective flux might be easier. There’s also the option of applying boundary ghost node formulas. Alternatively, I could use an explicit finite volume approach, which might improve accuracy. Simplifying to a 1D radial or vertical model could make things easier too, especially if depth from the top surface is the main focus in a well-mixed ladle of molten steel.**Deciding on heat modeling**\n\nIn a ladle, temperature gradients are typically vertical due to cooling from the top slag, the walls, and the bottom. A 1D vertical heat equation with convective boundaries could predict temperature at any depth, which seems like a good compromise. The user didn't specify if they need radial variations, so I think providing a simple 1D vertical model is sufficient for estimating temperatures. I might offer a Python script, clarifying that it's a simplified model while suggesting I can extend it into a more complex 2D model if needed.**Implementing a temperature model**\n\nThe user might expect to predict temperature at a specific point using a mathematical model. Without process data, the script should be parametric, perhaps using a lumped first-order cooling equation. This would give us temperature over time, but it relates to the whole steel mass, not just a single point. However, if the steel bath is well-mixed, the temperature could be nearly uniform. I might clarify that the temperature at a specific point would approximate the bulk temperature in a well-stirred ladle.**Considering heat conduction modeling**\n\nI'm thinking about using an analytical solution for 1D transient heat conduction with a convective boundary on one side. But I realize I might compute a series solution at a point inside the domain. The user requested a script, so numerical methods are okay. Maybe I can use a simple 1D finite difference explicit solver, which is likely less complex than a 2D model. I need to ensure mathematical correctness, especially for boundary handling. Let's focus on a vertical 1D slab model.**Exploring heat boundary conditions**\n\nI'm considering that the bottom could be adiabatic if the refractory is highly insulating, but that might not be entirely accurate. I could also use different heat transfer coefficients for the top and bottom. For a temperature at a certain depth, I can interpolate based on the heat equation. With boundary conditions that involve convective heat transfer, it's good to model the top with an effective heat transfer coefficient that includes both convection and radiation. However, including radiation could complicate things unnecessarily, so I need to reconsider that.**Considering lumped model limitations**\n\nI could also explore using a lumped model with a cooling coefficient UA that accounts for the top, sidewalls, and bottom surfaces. However, I'm realizing that point temperature would represent the bulk temperature rather than being point-specific, which isn't ideal for accurate modeling. I really want to make sure the approach captures the temperature variations properly, so maybe I need to rethink the use of this lumped model.**Considering script generation**\n\nIt looks like I should create a Python script related to temperature in a steel ladle. The user suggests they’re looking for a physics-based approach rather than machine learning, likely involving a partial differential equation (PDE) solution.\n\nI’ll consider writing a general script for 2D axisymmetric transient conduction with convective boundaries using explicit finite differences. It might be around 100-150 lines, focusing on managing boundaries robustly and implementing vectorized NumPy updates. I need to derive specific components now!**Exploring modeling approaches**\n\nI'm considering using finite difference methods in cylindrical coordinates, including boundary nodes. Alternatively, a more elegant option might be a finite volume method on a staggered grid. Each control volume would need an energy balance.\n\nFor the interior cells, conduction fluxes would go to neighbors, and for boundary cells, we'd apply convective flux at the exposed face. An axisymmetric 2D model depends on radial distance, which raises the question of how to set coordinates effectively. A radial 1D axisymmetric approach might be more suitable here, especially since natural convection plays a significant role in molten steel.**Refining temperature prediction**\n\nTo predict temperature at a specific point within the steel, using radial distance from the center could work well. A 1D radial model in a cylindrical domain (0 ≤ r ≤ R) offers symmetry at the center and accounts for convective cooling at the wall to predict T(r,t). The equation I’d use is ∂T/∂t = α (1/r ∂/∂r (r ∂T/∂r)). This approach simplifies things compared to a 2D model and aligns well with the ladle context. However, if the point has depth too, I lose that dimension.**Considering heat prediction models**\n\nI'm thinking about how steel ladles are typically stirred. Steel is well-mixed, so the temperature should be uniform. Most temperature losses come from the slag and air at the top, as well as the lining walls. A simple 1D radial/vertical model might work. If the user only wants a point temperature, we can identify any location. I need to clarify: \"This script predicts temperature at a point using a transient heat-conduction model,\" and consider providing both a bulk model and a 1D depth profile.**Developing a Python script**\n\nI’m thinking about providing a Python script. I could say, \"If you want, I can adapt it to your ladle dimensions, refractory thickness, or measured process data.\" Since the user just wants a script, I want to keep it concise yet useful. It might be helpful to include math equations in comments, such as the heat equation in 1D and explicit finite difference. I need to ensure the boundary formulas for the 1D slab with convective boundary conditions are derived correctly.**Using the ghost node approach**\n\nI'm considering the ghost node method for boundary conditions. At the z=0 boundary, the derivative can be approximated as \\((T_1 - T_{-1})/(2dz)\\), which equals \\(dT/dz\\) at the boundary. This equals \\(h/k (T_0 - T_\\infty)\\) because z is positive downward, and the top boundary derivative is positive when \\(T_0 > T_\\infty\\). I need to ensure I derive this correctly and clarify any assumptions so it makes sense in this context.**Evaluating modeling options**\n\nI'm thinking about providing an optional 2D axisymmetric script using explicit finite difference with convective boundaries. But, there's a risk of bugs in that. The 1D slab model has simple, reliable code, and the user didn’t specifically ask for radial variation. I could say this gives the temperature at depth z. Perhaps I should also offer a 2D axisymmetric model and mention simplifying if depth is all that's needed. I want to ensure the code isn’t too error-prone, so let’s see if this is manageable.**Deriving coordinates for modeling**\n\nI'm aiming to derive the 2D axisymmetric finite difference model. For this, I'll set the coordinates where r ranges from 0 to R, and z goes from 0 to H, with z measured downward from the top surface. This will help me structure the model accurately and ensure that I’m on the right track with the parameters. Let’s keep this organized as I move forward with the calculations!**Evaluating temperature derivatives**\n\nI’m thinking about how to approximate the second derivative of temperature at the boundary using a ghost node. I can use the formula (T_ghost - 2T_i + T_{i-1})/dr² for ∂²T/∂r² and apply central difference for ∂T/∂r as (T_ghost - T_{i-1})/(2dr). Since the ghost formula is based on the derivative at the boundary, it seems like a solid approach. I’ll then apply the same cylindrical formula for the boundary node with r=R, which should work out well.**Considering boundary conditions**\n\nI'm breaking down the radial operator with ghost nodes for my calculations. I see that d²T/dr² and dT/dr should be evaluated, and I'm considering how to handle corners along with edges. Using ghost nodes helps to account for edges and corners effectively. While I could potentially use loops to compute ghost nodes, corners present a unique challenge since they involve both directions. I’m contemplating a simpler 1D vertical slab approach since it's less risky, even though the user didn't request a full 2D model. Perhaps I could offer a 1D axisymmetric solution as a helpful alternative!**Formulating a temperature model**\n\nI’m considering creating a 1D model for temperature versus depth in the ladle, noting that if the ladle is well-stirred, the temperature should be uniform. So, for a query point, depth from the top could be sufficient. I think it’d be best to provide a Python script using the transient heat equation to predict temperature at any depth in the ladle over time. Simplicity is key, so I’ll focus on mentioning required units and parameters instead of complicating it with multiple layers. If I include matplotlib, that could enhance it too!**Analyzing input parameters**\n\nI’m working with some inputs for my temperature model: density (rho) is 7000 kg/m³, specific heat (cp) is 750 J/kg-K, and thermal conductivity (k) is 30 W/m-K. I'm considering a steel depth (H) of 1.5 meters and an initial temperature (T_initial) of 1873 K (which is around 1600°C). For the temperature at the top (T_air), I need to find an effective boundary temperature, since the top is exposed to slag and ambient conditions, not simply using the ambient temperature. This all seems quite reasonable!**Ensuring numerical stability**\n\nI need to make sure the numerical scheme is stable. When working with the explicit 1D heat equation, stability is achieved if the Fourier number (Fo) is less than or equal to 0.5. I wonder if the same stability condition applies to convective boundaries as well. It’s essential to keep track of these conditions to avoid issues in my calculations! Let's keep this in mind as I proceed.**Exploring time step options**\n\nI'm considering the time step calculation with the formula `dt <= 0.45 dz² / α`. I might set `dt = 0.4 * dz**2 / alpha` for safety. If the user wants longer times, we might have many steps, but that's okay. Crank-Nicolson could be useful for stability, but I wonder if it makes the script more complex. I'll think about whether I can implement an implicit solver using SciPy while balancing simplicity and performance.**Developing explicit boundary formulas**\n\nI'm considering implementing the script using `scipy.linalg.solve_banded` or sparse matrices, but an explicit method seems easier for a self-contained script. If I address stability concerns, I can allow automatic `dt` calculations, so the user can select smaller `dz`. Explicit updates for interior nodes involve the formula: `T[i] += Fo * (T[i+1] - 2*T[i] + T[i-1])`. For the boundaries, I'll carefully derive the updates for the top and bottom, ensuring to check the signs and units. This approach could work well with convective cooling scenarios.**Clarifying boundary conditions**\n\nI'm thinking that if the interior temperature is higher than the boundary, then the interior will warm up, which seems correct. However, I need to confirm if this formula is derived from the linearized boundary conditions. It’s crucial to ensure that I'm applying the right theory here, as my conclusion depends on it being based on valid assumptions. I might need to revisit the derivation to clarify how everything fits together.**Considering temperature units**\n\nI'm focused on the ghost node central difference and think I might need to use query point interpolation with `np.interp(z_query, z_grid, T)` since the z grid is monotonic downward. If the query depth is outside the range, I should raise an error. Celsius seems more intuitive, but for physics models involving radiation, Kelvin is necessary. Since my PDE handles only differences, I could choose Celsius for output, but Kelvin might be safer if radiation is added later.**Deciding on temperature units**\n\nI think the user wants the temperature in Celsius, so I can implement conversion functions that state inputs in Celsius while using Kelvin for internal computations if radiation is involved. To keep it simple, I’ll stick with Celsius for the calculations, noting that it works since only differences are involved. If we later add radiation, I’ll mention converting to Kelvin. I might add an optional `use_radiation=False` parameter to simplify things. I'll explore including radiation while managing complexity!**Calculating net heat flux**\n\nTo compute the net heat flux directly, I’ll focus on the top boundary node (i=0) using finite differences with a ghost node. The convective flux equation will be \\( q = h_{conv} (T_0 - T_{air}) + εσ(...) \\). Since heat flux is nonlinear, I could approximate it with \\( h_{eff} = h_{conv} + h_{rad} \\) and use \\( h_{rad} = εσ(T_{s_K} + T_{sur_K})(T_{s_K}^2 + T_{sur_K}^2) \\), referencing the current boundary temperature. I can then use the same boundary update formula with \\( Bi = \\frac{h_{eff} \\cdot dz}{k} \\) and set \\( T_{\\infty} = T_{air} \\).**Considering radiation and temperature**\n\nI think about how radiation relates to surrounding temperatures, wondering if `T_inf` should equal `T_sur`. If `T_sur` equals `T_air` or maybe a different wall temperature, that could be fine. This linearization seems approximate. For my script, I don’t need it to be exact; using `h_eff` might just complicate things. Keeping the code tight by excluding radiation could streamline things, although using eigenfunction expansion for complicated solutions is tempting, it complicates implementation.**Exploring 1D slab solutions**\n\nI'm looking at the solution for a 1D slab with half-thickness L and symmetry at the center. The equation involves a sum with \\(A_n\\) and \\(\\lambda_n\\), but our geometry has two convective surfaces, which complicates things. Implementing root-finding is trickier than using explicit finite differences. Since the user likely doesn't need an exact analytical solution, FDM seems fine. Also, should I note that if molten steel is well-mixed, conduction might not capture actual gradients? It’s okay if the user just needs a basic model.**Considering molten steel modeling**\n\nI’m thinking that if the steel is fully molten and stirred, a lumped-capacity model might be better. However, since the user is asking for a specific point, I’ll mention that my script assumes a vertical temperature profile. If the ladle is well-stirred, the temperature should be nearly uniform. It’s probably wise to provide a script along with a comment on how to adjust for a specific depth, `z_query` in meters. Also, if the point isn’t right beneath the surface, we’d need to consider a 2D approach.**Refining heat transfer script**\n\nI'm considering refining my approach to the heat transfer script, possibly simplifying it with function parameters. For my defaults, I'm thinking about values like density, heat capacity, thermal conductivity, and bath depth. \n\nI might set the bottom temperature to around 300°C for effective ambient but want to ensure it doesn't cool too aggressively. Maybe I'll use placeholders and suggest a top ambient value of 1000°C since the top would be near steel temperature. It's all about clarity and ensuring users can adjust the parameters as needed!**Designing temperature functions**\n\nI want to ensure that my script runs until `t_end` and outputs the temperature at the specified depth at that time. It could also be useful to output a temperature history since users might want to predict the temperature at certain points. I'll provide functions for both calculating temperature over time and for sampling at a specific query time. \n\nI might also consider plotting the temperature profile, so I need to think carefully about naming these functions!**Considering temperature profiling**\n\nI’m thinking about this code that simulates temperature over time. To optimize memory, I could store temperature profiles at intervals, but if `t_end` is large and `dt` is small, it could take up too much memory. So, maybe it makes sense to just return the final temperature at the queried point instead. It might also be useful to allow an option to save profiles at specific intervals with something like `save_every`. That could balance between detail and memory use!**Refining temperature prediction**\n\nI’m considering how to predict temperature at a specific point over time. The output can be the temperature at that point, `T_query`, and I could offer an option to plot it as well. It might be beneficial to adapt to time-varying boundary conditions, like changing surface temperatures, but using constants for `T_top` and `T_bottom` could simplify things. Perhaps the user would appreciate a script that predicts temperature at a depth for any given future time based on current conditions. I'll think about adding a function called `temperature_at_depth(t_end, z_query, ...)`.**Considering modeling choices**\n\nI’m thinking about validating `z_query` within the range [0, H], maybe using `np.clip` or throwing an error. Using `np.interp` at the end could work too. Should I incorporate the steel phase change? Well, it seems the ladle steel stays molten in relevant timescales, so maybe not. I could consider a more advanced script with a 1D radial coordinate and convective wall, especially since the user didn’t specify solidification. I'll mention a simple model and offer to extend it if needed.**Drafting a preface**\n\nI think it's a good idea to include a short preface to clarify the model I'm using. I could start with: \"Model: 1D heat equation in the steel bath.\" Then I can present the equation: \n\\[\n\\frac{\\partial T}{\\partial t} = \\alpha \\frac{\\partial^2 T}{\\partial z^2}\n\\] \nFinally, I’ll mention the convective cooling at the top and bottom of the system before sharing the code. Clarity is key here!**Exploring numerical methods**\n\nI'm considering using `scipy.integrate.solve_ivp` instead of doing manual time-stepping for my lumped model. This method allows me to discretize space using finite differences and integrate ordinary differential equations (ODEs) elegantly. I could represent spatial grid nodes as a state vector and compute dT/dt while incorporating boundary conditions. Using `solve_ivp` with methods like 'BDF' or 'RK45' would avoid stability concerns. Overall, it seems like a more elegant and less error-prone approach!**Considering solver options**\n\nI'm thinking about the availability of `solve_ivp` from SciPy; it might not be in all environments, but it's likely found commonly. If the user just wants a script, using NumPy could be more universally applicable. However, the stiff solver like `BDF` in `solve_ivp` would manage larger time steps and stable boundaries well—definitely a good balance! Ultimately, I might consider providing an explicit NumPy solver and offering a note about the SciPy version for those who might find it useful.**Considering heat conduction model**\n\nI'm thinking about the accuracy and physical realism of this 1D conduction model for molten steel in a ladle. Sure, it’s a mathematically based script, but I realize the user didn’t ask for high-fidelity detail. I should clearly state my assumptions to prevent any misrepresentation. I can mention that this model predicts temperature at a specified depth, but if the melt is well-mixed, a lumped model may work better. I'm imagining possibly using an energy balance for estimating temperature too.**Refining simulation code**\n\nI’m thinking about the temperature profile model \\(T(z,t) = T_{bulk}(t) + (T_s(t)-T_{bulk}(t))\\frac{z}{H}\\), but it feels like using a simple linear approach might be too simplistic. I see that using PDEs would improve accuracy. Maybe incorporating measured temperatures as boundary conditions could help, like using thermocouple data for calibration. \n\nFor the final code, I could add features such as `steel_properties`, simulation functions, query points, and plotting options. Including functions for `T_top` and `T_bottom` could be useful too, allowing for flexibility based on user preferences.**Considering boundary conditions**\n\nI'm thinking about the side boundaries and how they’re not included in the 1D model. I could mention that the top and bottom boundaries are approximated using effective heat transfer coefficients. For insulation via refractory, setting `h_bottom` to a small value might work well.\n\nMaybe I can use `np.clip` to prevent temperatures from going negative, though I’m not sure that’s essential. I also need to clarify whether the \"point temperature\" should be measured from the top surface or the bottom. I think using `z_query` from the top surface could be more intuitive!**Evaluating modeling options**\n\nI'm considering how to query depth in a bath, where if the point is 0.4 m below a 1.2 m surface, `z_query` would be set to that value. There’s a potential issue of it lying between grid nodes, but I think `np.interp` can handle that. Including `plot_profile=True` would be beneficial, though not strictly necessary. It could be good to mention installing `numpy` and `matplotlib`, and that `scipy` isn’t required. Let's also think about using `solve_ivp` for variable time steps to simplify the code.**Considering model options**\n\nI'm thinking about using a Gaussian process or regression model based on temperature sensor data, but I remember the user's request is specifically for a mathematics-based approach, which rules out ML. Since we're working with the heat equation and PDEs, I might mention that the most general model is a 2D axisymmetric solver, but for the script, we can stick to 1D. I should also check if the boundary formulas for an explicit scheme are correct and ensure the code handles this properly.**Defining temperature boundaries**\n\nI'm considering the scenario where `h_top` is large and `T_top` is lower, which means the boundary cools strongly. It might be helpful to include `T_init` as a uniform initial temperature. If the user knows the current temperature field, we could replace this with an array. Also, I need to choose the orientation for `z`. I'll define `z=0` at the top surface and `z=H` at the bottom, making the depth measurement intuitive as it goes downward from the top.**Considering code structure**\n\nI’m thinking about how to craft this code. I want `predict_point_temperature(...)` to return the temperature at `z_query` after `t_end`. Then, I could create `simulate_profile(...)`, but maybe just one function is enough with detailed comments, like the heat equation and boundary conditions. \n\nUsing `plot=True` sounds good to visualize temperature history, though it might not be necessary. Additionally, I'm considering how to model temperature losses in metallurgy with refractory layers, but maybe that’s more detail than the user needs.**Noting thermal behavior**\n\nI’m considering a potential issue: If the bath is molten and well-mixed, the internal conduction PDE might show little variation if both k and alpha are large. However, if the h values are small, small gradients may occur. I could query temperature at a point, noting that molten steel’s strong convection typically results in minor temperature differences. \n\nI need to calculate thermal diffusivity as alpha = k/(rho*cp), around 5.7e-6 m²/s. For H = 1 m, diffusion time would be approximately 50,000 seconds, meaning time steps of around 0.1 seconds for dz = 1 cm. This results in about 18,000 steps for 30 minutes, which isn’t too expensive.**Considering time step limits**\n\nI'm thinking about whether I should include a `min_time_step` to prevent having too many steps in the calculations. I could compute dt based on `dz`, ensuring it doesn't go below one second. However, it’s crucial to remember that stability relies on having an upper bound rather than just a lower limit. I want to make sure the time steps remain manageable while still adhering to the necessary stability requirements for the model.**Assessing temperature modeling**\n\nThe user's mention of a \"certain point\" is significant. Since the prompt is a bit ambiguous, I think I should clarify limitations in my answer. I could phrase it by saying, \"The script below estimates temperature at a point specified by depth from the steel surface; we can extend it for a fully 3D point.\" I'm considering whether to provide a 2D axisymmetric model instead, as it might yield more detail without being overwhelming. Let's explore if we can implement a 2D explicit solver with convective boundaries!**Evaluating grid efficiency**\n\nI'm considering how to approach this node-by-node with if statements, and it doesn't seem too bad. However, there might be speed issues with loops in Python for larger grids. For moderate sizes, though, it should work. Vectorization is an option with array slices for different regions, but boundary corner nodes pose a challenge. I could create helper functions like `radial_operator` and `axial_operator`, but the explicit 2D approach could be slow. However, a 1D method is simpler and faster. Maybe 2D is overcomplicated here!**Considering script details**\n\nI think D is fine, and I don't want to over-engineer it. I can mention that if the point isn't along the depth axis, the user should provide the coordinates, and I'll generate a 2D axisymmetric version. I need to decide whether to include `matplotlib` to plot the temperature versus depth profile for validation. Including the plot makes sense since the user asked for a script. However, I wonder if I should keep the code lean.**Exploring boundary conditions and outputs**\n\nI see a potential issue with constant boundary temperatures `T_top` and `T_bottom`: if `T_bottom` equals the initial temperature, we won’t have any cooling at the bottom. If `dz` doesn’t divide `H` exactly, I’ll need to adjust by calculating `n_nodes` and updating `dz` accordingly. If `dz` becomes larger than `H`, that would cause an error — I'll make sure to validate that. I also need to decide between using `np.arange` or `np.linspace` for creating nodes. Maybe storing temperature history could be useful, but I’m concerned about memory usage with full `T_hist` at each step.**Considering history storage options**\n\nI’m thinking that I could potentially only store the query point history instead of the full temperature history. This might help manage memory usage while still providing the essential data that the user might need later. It’s a simpler approach that still captures relevant details without overwhelming the system. I’ll consider how this fits into the overall structure of the simulation and whether it meets the user's needs adequately.**Considering stability in calculations**\n\nI’m thinking about how to handle a situation where `dt` is not provided. I want to compute a stable value using `dt_stable = 0.4 * dz**2 / alpha`. If `dt` is specified, we'll check that `Fo <= 0.5`, otherwise, a warning is raised. To ensure robustness, if `t_end` isn’t a multiple of `dt`, I should adjust the final step size to `min(dt, t_end - t)`. This way, I can manage any potential issues with the final step.**Evaluating modeling choices**\n\nI'm reflecting on the use of `numpy` and `numba` for modeling, but I see potential issues with explicit finite differences at boundary nodes, particularly with convective boundary conditions derived from a 1D slab assumption. Since I’m focusing on depth, the top and bottom boundaries are fine, but I need to note that sidewall losses are indirectly included if a larger effective `h_top` is chosen. However, in a 1D model, sidewall losses aren't modeled separately, which could be a limitation.**Setting expectations for modeling**\n\nI think it might be helpful to mention in the preface that if someone wants the temperature at a specific 3D point, they should provide details like the ladle radius, bath depth, and whether the steel is stirred. This way, I can offer a 2D axisymmetric solver. It’s a good approach for setting the right expectations upfront, ensuring users know what information they need to provide for accurate results.**Considering model alternatives**\n\nI'm thinking about an alternative approach and might note that if the steel is strongly agitated, using a lumped model might be better, but only if the user wants a simpler option. It feels like mentioning it could distract from the main script, so I’ll add a sentence at the end instead. I'm also checking if explicit formulas could lead to unphysical results with large Biot numbers, but it seems stable with a small time step. I wonder about the trade-offs in solver types.**Evaluating dependencies**\n\nI want to avoid adding dependencies beyond standard scientific libraries, but I might consider noting that if the user has SciPy installed, there's a more efficient version I could provide. However, since they just asked for a script, minimizing dependencies isn't their main concern. Using `scipy.integrate.solve_ivp` seems natural in a mathematical model context, as the explicit Euler method isn't as robust. It'll be interesting to compare the code quality between these options.**Evaluating numerical methods**\n\nI'm thinking about how to determine temperature at a certain point over time, using interpolation on the solution across different time evaluations. The BDF solver can deal with stiffness well, and writing a complete script with solve_ivp isn't too complicated. I wonder if using BDF with many points in `t_eval` could be heavy, but it should be fine for small nodes. I could define `jac_sparsity` for performance enhancements, but it might not be necessary. Ultimately, I’m considering providing a primary SciPy-based approach with an explicit fallback.**Considering script options**\n\nI might provide a pure NumPy version to keep dependencies minimal, while mentioning that I can rewrite it using SciPy's solve_ivp if needed. However, since the user is asking for a script rather than a discussion, I think simpler is better. Maybe I can use `scipy.sparse.diags` to create a matrix for the discretized PDE, but that might complicate things unnecessarily. Instead, I should start with a concise introduction: “Here is a physics-based 1D transient heat-conduction script,” followed by the code and a brief note on assumptions regarding temperature depths and inputs.**Considering temperature units**\n\nI’m thinking about using Celsius for convection to ambient, but if radiation is involved later, it’ll be necessary to switch to Kelvin. Maybe it's best to note that. Just to be safe, I could convert all Celsius inputs to Kelvin at the beginning, run the model, and then convert the outputs back to Celsius. Since the model only uses differences, the results will remain consistent. This way, I'm keeping things clear and accurate!**Considering script improvements**\n\nI’m working on a code that includes arrays `z` and `T`. I think I can improve it by adding a plot of `Tq_hist` versus time, potentially using `matplotlib`. This might make the answer longer, but it could be really helpful. Since the user asked for a script, that's probably fine. I’ll also mention that `h_top` and `h_bottom` need calibration with plant temperature data—this avoids unrealistic predictions. Plus, we could improve accuracy with a “lumped” coefficient for cooling.**Exploring steel ladle modeling**\n\nI'm considering that the user might ask about the temperature of a specific point of steel in a ladle, where a thermocouple is inserted at a known radius/height. A 1D depth model won't capture the radial variation, so I’m thinking about incorporating sidewall losses using an effective `h_side`, even while keeping it 1D. \n\nI could also model the total heat loss with a well-mixed bulk model, or maybe consider a linear temperature profile. However, 1D conduction with top and bottom boundaries feels better, along with adjusting `h_top` and `h_bottom` if needed.**Clarifying script capabilities**\n\nI think it’s important to note that “this script models temperature versus depth and not full 3D point coordinates.” If the user needs a 3D model, I could definitely extend it. This clarification helps address any limitations in the current version. Additionally, I'll make sure the code remains readable and only utilizes `numpy` for clarity and efficiency. Keeping things straightforward is essential for a good user experience!**Considering solver options**\n\nI’m thinking about using `solve_ivp` with BDF, which would eliminate the need to select a time step. Still, the explicit code might be easier for users to inspect, and the mathematical basis remains solid. In the final answer, I could note: \"If you want, I can convert this to an implicit solver or a 2D axisymmetric ladle model.\" It's worth mentioning that \"Steel properties are approximate and temperature-dependent.\" I should also check for any potential bugs in the `Bi_bottom` formula, especially considering the bottom environment’s temperature.**Generating Python Script**\n\nThe user is asking for a mathematics-based Python script, likely without extensive explanations. I'll include a brief introduction mentioning the equations first, outlining that this script applies a 1D transient heat-conduction model. The specific equation is \\(\\partial T/\\partial t = \\alpha \\partial^2 T/\\partial z^2\\), focusing on temperature at a specific depth. \n\nI think using `solve_ivp` and `scipy` could enhance the approach, offering a better mathematical basis while keeping the script clear and adaptable. An explicit solver with comments seems like the best choice!**Finalizing Script Details**\n\nI’ll keep the current approach, ensuring we set `np.interp(z_query, z, T)` with `z` in ascending order and temperatures in Kelvin. I also need to add error handling for conditions like `if H <= 0 or t_end < 0 or dz <= 0: raise ValueError`.\n\nFor cases where `t_end = 0`, it should just return the initial temperature since the while loop skips. It’s important to ensure there are at least three nodes, especially if `dz` is small relative to `H`. I'll include a note about calibrating `h_top` and `h_bottom`, explaining these coefficients account for more than just convection, since they also reflect radiation and refractory losses in a ladle context.**Refining h Values**\n\nI believe using effective `h` values can accurately reflect the conditions we're simulating. I should include a comment in the code to clarify this, so anyone reading it understands the significance of these values. Capturing these nuances will enhance the code's clarity and usefulness. It’s great to think about how to communicate these technical aspects well, ensuring that the script is not only functional but also easy to interpret for future users!**Defining environment parameters**\n\nI’m thinking that the environment isn’t well-defined, so I’ll keep the height parameters as `h_top=150` and `h_bottom=25`. I should set `H` based on the bath depth, maybe around `1.8 m` or possibly `1.5`. I’m considering `t_end` of `30*60`. For the final temperature at depth, I think it might stay around 1550. I should also note the model ignores buoyancy-driven convection, which makes it an approximation, but still useful for a first-order estimate.**Enhancing code documentation**\n\nI think we should tell the user, “If you know the exact point coordinates, I can create a 2D axisymmetric version.” This could be helpful if they need more details. Now, let’s consider adding documentation for the code, specifically a function docstring with units and behavior, like this: \n\n```python\ndef predict_temperature_at_depth(...):\n \"\"\"\n Solve the 1D transient heat equation in a steel bath. Returns:\n times (s), temperature_at_query (°C), depth_grid (m), final_temperature_profile (°C)\n \"\"\"\n```\n\nI’m deciding whether to include the temperature history or just the final temperature. Since they want a prediction at a point, I'll return both. I'll also consider using `np.asarray` for the return values and maybe add functionality for script execution as a standalone.**Finalizing the script**\n\nI’m thinking this makes it a complete script! I should include a `matplotlib` import because it handles plotting. Even though the user can run the function with just NumPy, it will need `matplotlib` for visualizations. I wonder if it’s smart to include the plotting code only in the main block. Maybe I could even comment it out if `matplotlib` isn't installed, but honestly, it seems pretty common for users to have it.**Analyzing stability conditions**\n\nI'm thinking about `Fo` stability and boundary conditions. If `h` is large, I might need a smaller `dt`. When `Fo` is less than or equal to 0.5, things should be fine, but boundary conditions might impose limitations. To be on the safe side, I should likely choose `dt` based on `min(0.4 dz² / alpha, 0.4 * rho * cp * dz / h_max)`. Then there's the stability constraint `2Fo Bi <= 1`, which I'm gearing up to derive. So, I should ensure coefficients remain nonnegative. Good catch!**Considering coefficient behavior**\n\nI'm realizing that in an explicit scheme, negative coefficients don't automatically mean instability. Instead, they can lead to oscillations, which is an interesting nuance. I need to keep this in mind when analyzing my results and any effects on the overall stability of the system. It's a delicate balance, and I want to ensure I'm accurately accounting for these scenarios to avoid unexpected behavior in my computations. That's something to watch carefully!**Calculating boundary stability limits**\n\nI’m trying to derive a boundary stability limit from the equation. If the Biot number (Bi) gets large due to high heat transfer, such as with h=100 W/m²K, the condition leads to a Fo value of about 0.484. If h increases to 1000, it drops to a limit of 0.375. It’s sensible to choose a smaller dt to ensure stability. I'll calculate a robust time step limit, defaulting to a conservative approach for safety. Let's go ahead and compute it!**Verifying Biot number effects**\n\nI'm verifying with Bi=0 to confirm that Fo should be less than or equal to 0.5, which checks out. With positive Bi values, we tighten the requirements for stability. I will compute the maximum Biot number using the formula and set a stable dt. Choosing 0.45 instead of 0.5 might be smarter for safety. I need to decide whether to automatically reduce dt or raise an error if the user specifies a larger dt, since silently changing it could be unexpected. A warning or ValueError might be the best way to go.**Calibrating values**\n\nI need to replace `h_top` and `h_bottom` with the calibrated values from my plant data. It seems like just doing that is sufficient for what I'm trying to accomplish. I wonder if there’s anything else I need to consider for accuracy, but for now, focusing on these values should help streamline the process. Let’s make sure I have the correct data ready to input and keep things straightforward!**Validating parameters and temperature**\n\nI’m considering whether I should validate if `k <= 0`, which isn’t necessary but would be nice for ensuring positive physical parameters. The user might be looking for information at a certain depth rather than surface depth. I could phrase it in the code docstring about the `z_query` as the distance from the steel surface. \n\nAdditionally, it might be useful to mention \"bulk temperature\" versus \"point temperature.\" Since the bath could be mixed, perhaps the point temperature would be close to average. \n\nLastly, I think I should highlight that this is a first-principles model, not a fitted regression, and include a brief note on the Fourier number, `Fo = α dt / dz²`.**Improving code accuracy**\n\nI see that including some mathematical explanations in code comments or prose could really enhance understanding. It’s a good idea to use something like `while t < t_end - 1e-12` to help avoid any floating-point issues, which can be tricky. This way, I’m ensuring that the calculations will run smoothly without unintended errors from precision problems. Just thinking through this helps me feel more confident in the accuracy of the code!**Considering code documentation**\n\nI'm thinking about whether to include instructions for `pip install numpy matplotlib` at the top. It might not be necessary, but maybe I should mention it if it's missing. Code comments could be enough for guidance. Now, I'm also considering using `np.full(n_nodes, T_init + C2K, dtype=float)` for initialization, and that seems like a solid approach. It's all about providing clarity and making things easier for anyone using the code!**Considering temperature limits**\n\nI’m thinking about whether to include `np.maximum` to keep temperatures above 0 K, but it might not be necessary for short times and certain physical parameters. However, if the user inputs odd values, I could update it to `T = np.maximum(T_new, 0.0)`, though that might distort physical results if the model is unstable. I also wonder if using `np.interp` could cause issues if `z_query` is out of range. I'll validate that. It’s important to note that this uses effective heat-transfer coefficients, combining convection and radiation.**Evaluating temperature considerations**\n\nI'm thinking about using effective `h_top` to account for heat transfer, which is good. Maybe `T_bottom_env` should represent the refractory interface temperature instead of ambient temperature. It might be valuable to note at the end that if a true point inside the ladle is needed, I can generate an axisymmetric version. I also consider using `scipy.integrate.solve_ivp` for continuous point temperature but we decided to stick with pure NumPy. Meanwhile, I need to ensure I handle `h_bottom=0` with the convective boundary formula, then apply the Neumann zero-flux condition for bottom node updates.**Considering boundary conditions**\n\nI could mention in a comment to \"set `h_bottom=0` to model an insulated refractory bottom,\" which seems useful. But should I include a more realistic bottom boundary by setting `T_bottom_env` with a low `h_bottom` instead of zero? That would be smart. Now, I’m also thinking about how to format the answer. Since the code is lengthy, using a single code block might be the most effective way to present it neatly.**Considering thermal modeling**\n\nI’m thinking about the code and whether to include options for a 2D or radial model, and even calibration against sensor data. I wonder if I should mention the expected output, though it may not be necessary. For clarity, I could replace \"mathematics-based\" with \"physics-based\" to avoid confusion with machine learning. I’ll explore whether the user is looking for a script using measured variables to predict temperature at a point, noting that coefficients often relate more to conduction than convection.**Refining the model**\n\nI'm considering how to handle the sidewall in the model. If we lump it into effective top and bottom losses, that might work, but we could also mention explicitly using a 2D axisymmetric model if needed. I wonder if we should add physical constants like the Stefan-Boltzmann for radiation, but that might not be necessary. Looking into making the code more mathematically rigorous with `solve_ivp` could help, as it avoids some stability issues. Comparing script lengths, using `solve_ivp` might yield a shorter code than using pure NumPy.**Exploring implementation options**\n\nI'm considering whether to implement both methods: a `rhs` function along with using `solve_ivp` from SciPy, but that might complicate things a bit. Instead, I think I could focus on crafting a script that uses `solve_ivp` since it's concise and mathematically sound. I need to evaluate if having a SciPy dependency is okay, but most scientific Python setups include SciPy, and it's generally easy to install if it's not already there.**Considering a Python script**\n\nI'm thinking about the final point of historical data access, and how we can use `sol.y[:, -1]`. It seems good that the `solve_ivp` function has adaptive time steps, making `t_eval` less impactful. I wonder if the boundary updates would be stable with BDF? It feels like a better option. But if SciPy isn't available, I should suggest `pip install scipy`. Since the prompt didn't limit dependencies, using SciPy may align better with the mathematical approach. Still, I should consider the user's likely setup. Would a simpler pure numpy solution be more user-friendly for those without SciPy?**Deciding on solver approach**\n\nI'm weighing which approach would satisfy the user best. Pure NumPy seems like a solid choice since it’s universally runnable, and explicit time stepping isn’t an issue when applying to data. I might go with a default pure NumPy solution since it's easier to grasp and modify. I could also mention an implicit version if they need stability, but it doesn’t seem necessary. I'll finalize it as a 1D explicit finite-difference solver using NumPy and Matplotlib, mentioning assumptions and a future 2D upgrade. I'll also check if the `h_top` and `h_bottom` units and values make sense.**Structuring the final answer**\n\nI’m considering setting `h_top = 150.0` and `h_bottom = 20.0` for sample usage, which feels good. Should I also include `steel_k = 30.0`, `steel_cp = 750.0`, and `steel_rho = 7000.0`? That seems alright. It might be useful to note that \"steel properties vary with temperature, but these values are common approximations.\" I think I'll structure the answer in four parts: a short introduction, an equation block, the full script, and a brief note on adapting it to actual ladle data. This could be concise yet informative!**Crafting code carefully**\n\nI’m planning to write about 100 lines of code, and I want to make sure it’s clear and understandable. I think it’s a good idea to include comments within the code to explain formulas and parameter names. That way, anyone reading it can follow along easily. I’m feeling motivated to ensure everything is well-explained and documented, so let’s focus on that as I start crafting the code!**Reviewing the Python code**\n\nI’m looking at a polished version of a Python function for predicting the temperature of molten steel in a ladle using a 1D transient heat-conduction model. It inputs various parameters such as depth, environment temperatures, and material properties. Validation checks ensure positive values for depth and time, while the thermal diffusivity is calculated using material properties. There’s also grid setup involving nodes along the bath depth, with a placeholder for stability limits. Let’s keep refining this!**Evaluating calculations**\n\nI should be cautious about using `Bi_max` computed with `dz` after rounding. If `t_end` is zero, the loop should skip and return the initial temperature, which is good. I see a potential issue: if `h_top` or `h_bottom` are zero, that gives `Fo_limit = 0.5`. That's fine, and I also realize `np.interp(z_query, z, T)` works even if `z_query = H`, which is great. Maybe I can include a note about the units for `rho`, `cp`, and `k` in the code comments.**Refining model parameters**\n\nI’m considering setting `cp=800.0` and `k=35.0`, but the user can adjust these values. Maybe I should suggest some typical approximate values and note that they can replace them with temperature-dependent properties if they have them. I think I should mention that the heating/cooling coefficients can be calibrated using measured ladle data. It’s also important to clarify that `h_top` may include radiation, while `h_bottom` represents overall refractory resistance. Let's keep things concise in the final response!**Structuring the code response**\n\nI’m thinking about using a short introduction, the code, and a one-line note at the end. I need to ensure that the code includes `numpy` and `matplotlib` imports. It’s important to mention that this model assumes the steel bath is homogeneous in the horizontal direction. Maybe I should include `z_query` as the depth from the top surface, with a comment like `# z_query measured downward from the free surface`. While I could provide a variant using `scipy` for robustness, it might not be necessary since the initial script meets the prompt requirements.**Enhancing the code explanation**\n\nI think it could be helpful to include a note stating: \"Set `h_top`, `h_bottom`, and `T_bottom_env` to values from your process; these are the main calibration knobs.\" This would be a practical guide for users. I should mention that after the code. I could also add a note that if users are measuring a point with coordinates (r,z), I can convert that to a 2D axisymmetric solver. I pondered adding an optional function for computing `bulk` temperature, but it might not be necessary. There's a potential issue relating to `Bi_max`, which uses `dz` and `k`.**Optimizing code efficiency**\n\nI'm considering optimizing some code where `Fo_limit` could drop below 0.5 if both parameters are large and `dz` is small. The code automatically selects a stable `dt`, which is great! I'm thinking about adding some transparency by printing `dt`, but I might hold off since the script shouldn't print by default. Also, it might be efficient to compute `Bi_top` and `Bi_bottom` outside the loop, since these constants don't change. That's definitely a good move for efficiency!**Refining temperature calculations**\n\nI'm adjusting the calculations for `Bi_top` and `Bi_bottom` to be defined in terms of heat transfer coefficients and depth. While working through the loop, I'll ensure `Fo` is calculated for validation. However, there’s a potential issue: if `dt` is based on `Bi_max`, it could vary due to the top and bottom heat coefficients. It might make sense to interpolate temperature at each depth step, though storing the times might not be necessary since history is useful. I'll consider returning temperature values in Celsius for clarity!**Noting limitations**\n\nI’ve made a note that the model doesn’t account for solidification or reaction heat. This seems important since those factors could influence the overall results significantly. I wonder how this might affect the conclusions drawn from the model. It’s a good reminder to consider all elements that might not be included in the analysis. I should keep this in mind as I proceed, just to ensure that I’m thorough in my evaluations.**Evaluating cooling parameters**\n\nI'm trying to avoid over-explaining. I need to consider the constant properties of the system. However, I see a potential issue if `T_bottom_env` is set at 25°C with `h_bottom>0`, as this could lead to excessive cooling of the bottom boundary. \n\nI’m thinking about calibrating `T_bottom_env` based on the actual refractory temperature or by adjusting `h_bottom=0`. It might be realistic to set `h_bottom=0` for cases with good insulation, yet it’s important to provide some flexibility. I wonder if `h_bottom=20` could work better since it balances reality with user adaptability.**Configuring environment settings**\n\nSince I'm working with a script, I think setting `h_bottom=10` and `h_top=150` could work well. The example can use those values along with `T_bottom_env=100`, but I realize the user will likely adjust them. \n\nI might consider creating separate variables for environment temperatures, like `T_slag_env` or `T_air_env`, though it doesn't seem necessary right now. I wonder if using `np.interp` on the final profile would oversmooth the query temperature, but maybe it’s workable. As for incorporating `pandas`, it doesn’t feel required, but tracking `point_history` at each time step sounds valuable for user flexibility.**Deciding on storage and libraries**\n\nI’m thinking about whether to reduce overhead by not storing `times` if the user only wants the final results, but I realize that having a history could be valuable. \n\nNext, I'm considering whether to use Python’s built-in `math` library instead of `numpy`, but I think `numpy` is still a good choice. It offers capabilities that could really enhance performance, so I should stick with it. This balance between performance and memory is interesting to navigate!**Considering modeling accuracy**\n\nI think the small time step might still incur some numerical diffusion, but it's acceptable for a simple model. It could be useful to mention alternative methods like Crank–Nicolson or BDF for more accurate predictions. A note could suggest an implicit solver to address performance concerns. Also, I should clarify that this isn't a CFD model. \n\nNow, for predicting temperature in a ladle, a 1D depth model seems like a reasonable first step, but we should be clear that this approach doesn't account for full 3D.**Finalizing temperature model**\n\nThe script predicts temperature at depth `z_query` from the steel surface. I plan to show the relevant equations and use a code block for clarity. At the end, I’ll suggest adjusting `h_top` and `h_bottom` to fit the user's ladle. If they need a full 2D axisymmetric model or something calibrated from data, I can generate that too. \n\nI’ll ensure the code only uses `matplotlib` in the example section. If the user doesn’t have it, they can remove the plotting lines. Including a comment about installing dependencies might be helpful, like: `# pip install numpy matplotlib`, but I want to keep the final answer concise.**Considering numerical methods**\n\nBefore finalizing, I'm wondering about using `solve_ivp` because it feels more mathematically grounded than explicit methods. But hold on—explicit methods are mathematically sound too. The user didn't specifically request high-order accuracy, so it's fine to stick with them.\n\nI’m also thinking whether the code could be simplified with `numpy.linalg`, though that might not be really necessary. It might be a good idea to check that `Fo_limit` isn't too small and verify that `dt` isn't tiny either. Lastly, I want to double-check the update formulas to ensure the top boundary sign aligns with `T_top_env_K`.**Considering temperature modeling**\n\nI’m thinking about how to explain that, for a vigorously stirred molten bath, the temperature is often uniform, which is a good first-order estimate. If I add this detail, my answer might get a bit longer, but that's okay. I could also mention that if the user needs temperature at an off-center point, they can provide the coordinates for a 2D axisymmetric model. I want to think about using `T_bulk(t)` from energy balance and adding a small vertical gradient. I should mention \"effective heat-transfer coefficients\" for calibration, too. Overall, the PDE is solid, but using actual convective values might not be enough, and calibration will be necessary.**Considering temperature adjustments**\n\nI’m pondering whether to set `T_bottom_env` lower than `T_init` but not at ambient temperature. The bottom refractory surface can get really hot, possibly around 1000°C. It might make sense to refer to the refractory interface temperature instead of the outside temperature. I can still use any `T_bottom_env` in the script. I think I should mention \"effective surrounding temperature or interface temperature.\" I’m also considering a simple linear interpolation of `h_top` with time for slag removal or pour time, though that might not be necessary.**Considering function adjustments**\n\nI'm thinking about how I might name the function — possibly `steel_ladle_temperature_point`, but that's not strictly necessary. I’m debating whether to return just `temperature_at_point` instead of an array, maybe adding a parameter like `return_profile=True`. To improve clarity, I could shorten the example output. Additionally, using a finite difference model in radial coordinates could be useful, but it seems easier to stick with vertical since I lack info. I’ll also add an initial comment about the governing equation and boundary conditions.**Exploring code comments and performance**\n\nI’m considering presenting the discrete equation in comments, like `T_i^{n+1} = T_i^n + Fo(...)`, but the code comments may already be clear enough. I want to ensure `T_top_env_K` and `T_bottom_env_K` remain unchanged over time. If the user wants temperatures in Celsius only, they can skip the K conversion since the code handles it. I’m also pondering memory and performance if `t_end` is long and `dt` is small. Even with 45,000 steps for an hour, this doesn't seem huge. I could include a note about adjusting `dz` for faster runs or higher spatial resolution, maybe even a comment about using `np.interp` for exact grid points.**Enhancing code functionality**\n\nI’m realizing that if the user wants a point not on the depth axis, we need to take note. I’m also thinking about improving the code by making `T_env` values time-dependent functions instead of constants. For instance, I could create functions like `def T_top_env_func(t): return ...` and `def h_top_func(t): return ...`. However, it may not be strictly necessary since, in a ladle, the top conditions do change as the process progresses, which is an important consideration.**Summarizing modeling approach**\n\nI’m considering how to frame the environment terms in relation to time, though it’s not essential. I think a \"mathematical model\" section could work well, including parts like PDE and BCs, and mentioning variables like `Fo` and `Bi`. I could even enhance the code with discrete formulas for interior and boundary updates through comments. There might be a question about setting `T_bottom_env` to 25, given insulation issues, but the user's calibration will ultimately take precedence.**Structuring the script**\n\nI'm thinking the answer could focus on a \"script only\" format, but adding a brief intro and notes might enhance clarity. I want to use SI units for variable names and provide outputs in Celsius. I don't think adding `from typing import Tuple` is necessary. For the script's organization, I’ll include sections like parameters, validation, grid, time step, simulation, and example usage. This should improve readability. A 1D model with convection should suffice for a mid-point analysis in a ladle, but I’ll encourage using a 2D model if near the wall or bottom.**Considering code approach**\n\nI’m exploring whether a `solve_ivp` version would be shorter and allow for a future 2D extension, but I think the explicit code is already straightforward. It would be good to include a note like, \"This is a first-order approximation,\" to set expectations. A potential issue arises if both `T_top_env` and `T_bottom_env` are lower than `T_init`, as that would cool the entire profile, while having one higher could allow for local heating.**Considering function parameters**\n\nI’m mulling over whether the equation `Fo_limit = 1/(2*(1+Bi_max))` is too strict, especially if `Bi_max` takes its maximum from `h_top` or `h_bottom`, but what if one boundary isn’t that large? It feels safe, though. Maybe printing the chosen dt could help with debugging, even if it’s not essential. Still, I think the function should remain silent overall. So, I’ll keep that in mind for now.**Considering effective coefficients**\n\nI'm thinking about using a comment regarding effective coefficients for radiation or refractory losses near the h parameters. It seems beneficial to clarify that for `T_bottom_env`, we could say \"use 0 if adiabatic bottom.\" That makes sense! For an example, I’m considering using `h_bottom` values of 10.0 or 0.0 to illustrate the point. Just want to make sure these comments help others understand what's going on!**Finalizing model notes**\n\nI’m thinking it would be helpful to provide a minimal additional note below the code. To model a point at coordinates (r,z), I should recommend using a 2D axisymmetric heat equation instead of the 1D depth model. It's good to incorporate `h_top` and `h_bottom` as time-dependent, if desired, but I don't think it's necessary. I need to clarify that the model is a physics-based approximation, without claiming it’s \"mathematically exact.\" I'll note that the steel has constant properties, and there’s no phase change. If solidification is needed, adding latent heat might be relevant. Lastly, I should clarify the model only considers conduction within the steel, as internal convection is absorbed into coefficients. For molten steel, the temperature is nearly uniform if well-mixed.**Preparing model expectations**\n\nI’m considering that our model may better approximate a solid bath or effective diffusion in a mixed melt. It's a simplification, so I want to be honest about that. I should say something like, \"This is a simplified thermal model. For molten steel with strong stirring, actual mixing can exceed conduction; in that case, consider using a lumped model or fitting effective coefficients.\" I'll ensure we provide the script as requested. It would help to label the model at the top as a \"Simplified physics-based model\" to set proper expectations. Let's finalize the response with moderate detail.**Evaluating default parameters**\n\nI'm considering whether the default `dt` should be `None` or a numeric value. Setting it to `None` allows for auto-selection of stable values, which is useful since the user can override it if needed. If the user changes `dz`, I might think about updating `dt` automatically. I’m noticing some potential issues with the while loop; if `t_end` is very small, it could affect the `dt_step`. Preallocating arrays like `times` and `T_query_hist` might improve performance, but it isn't strictly necessary. Let's reflect on the final code style!**Ensuring proper formatting**\n\nSince I'm working on the final answer, it's crucial to check that the indentation and formatting are spot on. I’m thinking of adding a comment to clarify the meaning, something like: `# z=0 -> steel surface, z=H -> bottom of bath`. That should help anyone reading the code understand the context better. It's all about making it clear and easy to follow! So yes, I’ll include that comment for sure.**Evaluating temperature boundaries**\n\nI'm considering that surface boundary can be significantly influenced by radiation, where h_top acts as an effective value. This brings up a potential issue: in a ladle, the top boundary temperature might not be the ambient 25°C due to the slag cover, possibly making T_top_env closer to 1200°C. Since h_top is effective and environment temperature is a reference, the user can set it accordingly. It’s good to mention in the introduction that effective coefficients and boundary temperatures should be calibrated.**Considering script accuracy**\n\nI'm wondering if the user might want a math-based script to predict the temperature at a specific point of molten steel in a ladle. However, I have some concerns since, in this situation, there's no conduction possible due to the movement of the molten steel. This might mean that our script won't be physically accurate enough for their needs. It's important to ensure that any predictions we make align well with the actual physical behaviors in this scenario.**Considering code adjustments**\n\nI'm thinking of including the return of `temperature_point_history` in Celsius and `final_profile`. There's also the name `z_query` vs. `point_depth` that I should evaluate. Perhaps I could add a snippet on calibrating `h_top` and `h_bottom`, suggesting fitting them to thermocouple trajectories using least squares, but that might not be necessary. I want to ensure that the final answer remains direct, focusing on the code since the user requested a script. A note about constant properties will be helpful too.**Improving function stability handling**\n\nI want to ensure the function correctly handles `dt` if a user supplies a value. I should compute `Fo = alpha * dt / dz**2` and then establish `Fo_limit`. If `dt` exceeds that limit, raising an error is appropriate, as users might be frustrated if they want to ignore stability but get an issue instead. If `dt` is `None`, I could compute a stable `dt` and allow the user to specify `steps_per_second`, but that might be unnecessary for the current implementation.**Evaluating code stability**\n\nI’m considering a potential issue: some users might copy-paste code without `matplotlib`, but they can comment out plots if they don’t need them. `numpy` and `matplotlib` are common imports, so that's fine. Maybe I should think about including a `try/except ImportError`? No, I need to ensure the finite difference update at boundaries uses `Bi_top` and `Bi_bottom` correctly. I’ll compute those outside the loop since they depend on `dz`. If `h_top` or `h_bottom` isn't supported, that could be a problem. Additionally, I’ll make sure the error message guides users on `dt`. That seems sufficient!**Considering analytical solutions**\n\nI’m wondering if there’s a more sophisticated analytical solution I could mention briefly. Maybe I could say, \"If you prefer an analytical series solution instead of numerical finite differences, I can provide that as well,\" but that’s not strictly necessary. I'm also thinking about the final length of my answer in relation to OV 6. The code plus introduction and notes should be moderately sized. Oh, and there’s probably no need to discuss the difference between \"steel\" and \"molten steel.\"**Evaluating temperature predictions**\n\nI see that `predict_temperature_at_depth(...)` returns `Tq` in Celsius, which is good. Instead of using `T_point_hist[-1]`, I might want to print `Tq[-1]`. Plotting `Tfinal` against `z` and inverting the y-axis could work well too. \n\nI'm considering if the user wants to predict the temperature at a specific point in steel within a ladle. Our model can help with that, although a point in the steel bath might show little temperature difference. I think adding a `plots=True` flag for optional graphs could be helpful.**Reviewing temperature model code**\n\nI’m considering the option of using a cell center grid instead of boundary nodes, although it doesn’t seem necessary. The final answer can stay as is. I should finalize the comment wording at the top of the code, like: “# Physics-based 1D model for temperature in a steel ladle,” and include details on the boundaries. \n\nUsing comments about the heat-transfer coefficients is a good idea. I’ll also check for any confusing variable names; for instance, `T_top_env` and `T_bottom_env` might be better as `T_surface_env` and `T_base_env`. 🤔 Also, computing `Bi_max` with `max(h_top, h_bottom)` makes sense!**Considering validation of coefficients**\n\nI'm thinking it might be important to validate the positive values of `h` because having negative coefficients would be unphysical and make no sense. If both coefficients are negative, that could indicate an invalid scenario. \n\nSince our current approach doesn’t include validation, I see the merit in adding a check for positive values only. This way, we can ensure that the coefficients make sense and align with physical principles. It's just a little tweak that could enhance the model's reliability!**Considering plotting figures**\n\nI'm thinking it might be a good idea to include a call to `plt.show()` after plotting each figure. That way, I can ensure that the figures are displayed properly. This will help in visualizing the data effectively, right? So, I will make sure to add that at the end to wrap things up nicely! I want to make sure I get everything polished. Let's finalize the code before running it.**Considering script adjustments**\n\nIn a non-GUI environment, plotting may not work, and that's okay. I can suggest commenting it out if needed. The initial lines of code should include `import numpy as np` and `import matplotlib.pyplot as plt`. There’s no need for `from __future__ import annotations`, but it might be helpful to annotate return types for Python 3.9+. I can phrase it as: \"This script predicts temperature at a depth `z_query` from the surface. To enhance realism, consider calibrating effective coefficients with plant data.\"**Clarifying parameters and notes**\n\nThis is looking concise and useful! I should mention that if the user's point is at a radial coordinate, they can let me know. I’ll consider adding a brief list of parameters outside the code for clarity: `H` for bath depth, `z_query` for depth from the surface, `h_top` and `h_bottom` for heat transfer coefficients, and `T_top_env` and `T_bottom_env` for surrounding temperatures. I want to keep the code self-explanatory but might add a compact note about time history resolution since it could be coarse due to larger `dt`.**Reflecting on boundary updates**\n\nI’m considering whether there might be a subtle issue with the boundary update formulas that use `Bi_top` and `Bi_bottom` defined without a half-step for `dz`. That's standard for boundary nodes, so it seems fine. However, maybe the update should use `Bi = h * dz / k` and consider `2Fo`. There's also a possibility of enhancing accuracy by using a one-sided finite difference of order 2. But honestly, the current method seems good enough for now.**Reviewing numerical code improvements**\n\nI'm aiming for higher accuracy with the Crank–Nicolson method, which sounds promising! I realize there's a potential issue with using `Fo_limit` based on `Bi_max`, as stability also depends on the largest `Fo` interior. Since I'm using the same dt, that’s fine. I think being more concise in my final answer would help too. Adding a comment about constant properties is a good idea. I should also clarify unit usage and temperatures in Celsius while indicating that they're converted to Kelvin internally. That improves clarity!**Enhancing code clarity**\n\nI think noting that this can represent a refractory interface temperature or an effective sink temperature is helpful! To make the code clearer, I could add small functions like `c_to_k` and `k_to_c`. They would enhance readability, even if it adds more lines. Alternatively, I could use a constant `C2K = 273.15`, which is neat! I also want to check if there's an issue with printing after `predict_temperature_at_depth`. Additionally, including \"mathematics based\" in comments might not be necessary, but it could clarify things. I should remember that this is a simplified model, especially for molten steel!**Clarifying boundary conditions**\n\nI want to consider the order of boundary conditions and equations in my explanation. The mathematical model could be presented as: \n\\[\n\\frac{\\partial T}{\\partial t} = \\alpha \\frac{\\partial^2T}{\\partial z^2}\n\\] \nwith convective boundary conditions. I think it’s crucial to mention that `alpha` represents thermal diffusivity. It makes sense to include that in the code comments instead of the prose for clarity. Keeping it concise while ensuring that the details are understood is important. Overall, I’m feeling good about this structure!**Refining update formula**\n\nI'm considering the explicit update formula at the boundary, which relies on the ghost node temperature formula. I think there might be a factor of 2 in the boundary convective term, and it's good we derived that. \n\nAs I'm preparing the final answer, I wonder if I should add the note for users to install the necessary libraries. A line like `# pip install numpy matplotlib` would be helpful without cluttering things up. \n\nI want to keep it concise, mentioning the main calibration knobs and the option for a 2D axisymmetric version too.**Clarifying script expectations**\n\nI'm considering whether the user just wants the script without much explanation, but I believe a brief introduction and note would be fine. I should emphasize that the model is 1D vertical rather than full 3D. For completeness, I could mention, \"For a full 3D point, you need radial and axial coordinates (r,z) and a 2D axisymmetric PDE.\" \n\nI think including a note about the temperature uniformity when the bath is well-stirred is helpful. I also need to clarify the term `T_top_env`, since the top surface is likely a slag cover. Using \"effective boundary temperature\" instead could be clearer, I think.**Optimizing the code**\n\nI'm considering whether to include `num_steps = int(np.ceil(t_end/dt)` to preallocate arrays, but it seems unnecessary. I noticed a potential issue with using `np.interp` since `T` is in Kelvin. Subtracting Celsius to Kelvin gives Celsius, which is important to clarify. It might be better if `T_init` is in °C. \n\nAlso, there's a possible improvement I could make: I could vectorize the update of the interior nodes with `T_new[1:-1] = T[1:-1] + Fo_step*(T[2:] - 2*T[1:-1] + T[:-2])`. That sounds good!**Considering temperature prediction**\n\nI’m thinking about how the user might want to predict the temperature at a point in a steel ladle using actual sensor data from a previous time series. The function I have takes `t_end` and model parameters without needing data input right now, which is okay since it’s about generating a script. I’m considering potential issues, like how rounding can affect grid spacing. Using `ceil` instead of `round` might be safer for ensuring accuracy in different situations.**Tweaking calculations for nodes**\n\nI’m considering how to set up my calculations for grid spacing using H and dz. If I set H to 1.535 and dz to 0.01, I might end up with 154 for both `ceil` and `round`. However, using `round` could lead to an unintended 153. To ensure I don’t end up with a coarser grid, switching to `ceil` seems smarter. I’ll set `n_nodes` with `ceil` and ensure the actual dz is calculated afterward. This adjustment should help maintain accuracy in my calculations!**Refining time steps**\n\nI'm considering setting the default `dt` based on the actual `dz`, which seems like a good idea. Another improvement could be raising an error if `dt` is provided and too large. I worry that `Bi_top` and `Bi_bottom` might not be symmetric if `h_top` and `h_bottom` differ. \n\nI think the time step stability limit formula should use `Bi_max = max(h_top, h_bottom) * dz / k` with the actual `dz`. It feels polished now! Maybe we should define `t_points = np.linspace(0, t_end, 200)` for equally spaced outputs.**Optimizing data storage**\n\nI’m considering whether we should store history at specific output intervals instead of every step to save memory. Since the function only records one value per step and one profile at the end, managing 22,500 data points seems fine. If `dt` is small and `t_end` is long, we might hit 45,000 points, but that’s okay. Maybe we can limit history by choosing every `n_output` steps, but the user might prefer the full curve. Since the model isn’t huge, storing everything should be fine. I also want to clarify that `h_top` and `h_bottom` are “effective” values, noting that the top surface radiates, which could be important!**Considering boundary conditions**\n\nI'm thinking about using both convection and radiation at the top boundary, which feels more realistic. Should I include optional radiation with an effective coefficient based on the current surface temperature? It adds complexity, but it might improve the model's realism without becoming unmanageable. It’s worth exploring to see if that increase in detail enhances the accuracy of our results. I like the idea of balancing realism with complexity carefully!**Considering radiation and code complexity**\n\nThe user didn't mention radiation, but maybe it would be nice to include it. The simpler code works well as is, and adding parameters for radiation could complicate things. Still, we could introduce a flag for it, but that might make the code harder to follow for the user. It's likely better to keep it simple and make a note about extending the top boundary for radiation if needed. Then I’ll check for any potential issues with the calculation parameters.**Evaluating interpolation code**\n\nI’m considering whether to return `final_point_temperature` as a separate value since that could be convenient for the user. But, it feels less necessary because they can use `T_query_hist[-1]`. Keeping the return signature simple is probably best. I wonder if we should mention that `z_query` can be chosen as `H/2` for center depth, but maybe that’s overkill. I should note in comments about the trade-off with `dz` for detail and stability. Also, the repeated `np.interp` calls seem fine since it's simple linear interpolation.**Considering code optimization**\n\nI’m contemplating whether to compute `idx = int(z_query / dz)` for manual linear interpolation to possibly speed things up. While that might work, I think `np.interp` keeps the code clearer, which is great. If optimizing becomes necessary, we could precompute weights for fixed query depths, but that might not be needed. I should mention commenting out the plot section for headless environments, though it's not crucial. Finally, I believe including a markdown code fence with Python syntax highlighting is a good idea. I'll also start with a brief introduction before presenting the code.**Generating 2D axisymmetric models**\n\nIf the user is looking for a 2D axisymmetric model for a specific point (r, z), or even if they need a version that fits parameters based on plant measurements, I'm ready to generate that. I think it’s great that I can provide different model types to meet their needs. It shows flexibility in the approach, so whatever they prefer, I’ll make it work! Let’s get started on that if they want it.**Reviewing temperature equations**\n\nI’m looking at the equations involving temperature differences and boundary conditions, and they seem standard, which is great! I want to mention the assumption about constant properties and no phase change. One potential issue is that using `z_query` from the surface could confuse users if they're thinking of measurements from the bottom. I’ll clarify this in code comments and documentations, especially around conversions to Kelvin. Also, maybe I should consider renaming `h_bottom` to `h_base`, but I think everything looks good for now!**Clarifying code notes**\n\nI'm thinking it would be useful to mention \"Set `h_bottom=0` to make the bottom adiabatic\" in the notes. For a well-stirred bath, the model will likely remain consistent across depth, which gives a good expectation for results. I want to make sure the final answer doesn’t have any contradictory statements. \n\nThe code name could be `predict_temperature_in_ladle`, but `predict_temperature_at_depth` might be more accurate. I'll add an example print statement for clarity and consider using depth from the surface, noting if it's at the centerline too.","reasoning_details":[{"type":"reasoning.summary","summary":"**Generating a Python script**\n\nI need to create a Python script for predicting temperature at a certain point of steel in a ladle. I think I could use a heat transfer model, possibly looking at 1D radial diffusion in a cylindrical ladle. Although this sounds complex, I might need to ask the user for more specific details, like the depth from the surface. So, I'll provide an example script with some assumptions and input parameters. I should also clarify that it's a simplified model.**Interpreting temperature in steel**\n\nI’m considering how to model the temperature at a point inside the steel mass in a ladle. Given that the steel is a hot liquid or solid, I think heat conduction and convection are key factors here. Since the ladle is axisymmetric, the simplest approach could be a radial heat transfer model, predicting temperature changes using a 1D transient conduction equation. If the steel is well-mixed, a lumped thermal model might work too. I should show the relevant equations, possibly using explicit finite difference with the thermal diffusivity of steel.**Creating a temperature prediction script**\n\nTo predict temperature at a specific point, I think using the heat equation is the simplest approach, represented as ∂T/∂t = α ∇²T in cylindrical coordinates. However, a full 2D PDE is complex. To simplify, I could reduce it to a 1D model, calculating T(x,t) from the wall to the center. Since the user hasn’t specified the geometry, I’ll create a parameterized 1D radial model. Using an analytical solution for transient cooling of an infinite slab or a finite difference solver might work best for the script.**Developing a temperature prediction script**\n\nI'm considering how to create a script to predict the temperature at a specific point in a ladle of steel. The user likely wants Python code, including mathematical models and comments. Inputs could involve initial steel temperature, ambient temperature, heat transfer coefficients, thermal conductivity, density, specific heat, coordinates, and time. I’d think about using finite difference methods to compute temperature distribution in cylindrical coordinates while considering the geometry of the ladle for accuracy. This script might need to be quite complex but should balance detail with usability.**Creating a heat equation script**\n\nI'm thinking about writing a Python script that solves the transient heat equation on an axisymmetric grid using finite differences. The user should be able to query the temperature at any point (r, z) in the steel at a specific time. I need to decide whether to use SciPy or just numpy with explicit time stepping. The explicit method may need a small time step for stability, but I could consider using an implicit method to enhance reliability even if it complicates the code. A self-contained script with comments might be the best approach!**Defining model assumptions**\n\nI'm considering the model assumptions for the heat equation in a steel bath. Here’s what I’m thinking: the properties of steel, such as density (\\(\\rho\\)), specific heat (\\(c_p\\)), and thermal conductivity (\\(k\\)), should be constant. I'll be using axisymmetric cylindrical coordinates for temperature \\(T(r,z,t)\\). The heat equation can be expressed using \\(\\alpha\\), which is \\(k/(\\rho c_p)\\). I also need to establish boundary conditions for convective cooling and symmetry at \\(r=0\\) to query the temperature at a specific point \\((r_{query}, z_{query})\\) after \\(t_{end}\\).**Evaluating temperature modeling**\n\nI’m considering a model for steel's thermal behavior, noting that if the steel is liquid and well-mixed, a simpler lumped model might be suitable. But I could also propose a \"well-mixed\" model as an alternative if needed. I should mention that if we’re deep within the steel, molten convection will likely dominate and the temperature gradients are minimal. I think including plant data could improve prediction accuracy. Next, I’ll work on designing a robust axisymmetric solver while addressing specific code correctness issues.**Developing thermal boundary conditions**\n\nI’m working on a standard cylindrical Laplacian discretization for interior points (r>0). The discretized version involves temperature at neighboring nodes. Then, at the boundaries with convection, I’ll handle the heat flux at the top surface (z=0). \n\nI need to ensure my sign conventions are accurate, especially when measuring the z-coordinate down from the top surface. The mathematical derivation shows the relationship between heat transfer and temperature gradients. Ultimately, I'll express the temperature change in terms of the heat transfer coefficient and the temperature differential.**Evaluating temperature modeling options**\n\nI’m thinking about deriving a discrete update but considering that using a finite volume update for boundary nodes with convective flux might be easier. There’s also the option of applying boundary ghost node formulas. Alternatively, I could use an explicit finite volume approach, which might improve accuracy. Simplifying to a 1D radial or vertical model could make things easier too, especially if depth from the top surface is the main focus in a well-mixed ladle of molten steel.**Deciding on heat modeling**\n\nIn a ladle, temperature gradients are typically vertical due to cooling from the top slag, the walls, and the bottom. A 1D vertical heat equation with convective boundaries could predict temperature at any depth, which seems like a good compromise. The user didn't specify if they need radial variations, so I think providing a simple 1D vertical model is sufficient for estimating temperatures. I might offer a Python script, clarifying that it's a simplified model while suggesting I can extend it into a more complex 2D model if needed.**Implementing a temperature model**\n\nThe user might expect to predict temperature at a specific point using a mathematical model. Without process data, the script should be parametric, perhaps using a lumped first-order cooling equation. This would give us temperature over time, but it relates to the whole steel mass, not just a single point. However, if the steel bath is well-mixed, the temperature could be nearly uniform. I might clarify that the temperature at a specific point would approximate the bulk temperature in a well-stirred ladle.**Considering heat conduction modeling**\n\nI'm thinking about using an analytical solution for 1D transient heat conduction with a convective boundary on one side. But I realize I might compute a series solution at a point inside the domain. The user requested a script, so numerical methods are okay. Maybe I can use a simple 1D finite difference explicit solver, which is likely less complex than a 2D model. I need to ensure mathematical correctness, especially for boundary handling. Let's focus on a vertical 1D slab model.**Exploring heat boundary conditions**\n\nI'm considering that the bottom could be adiabatic if the refractory is highly insulating, but that might not be entirely accurate. I could also use different heat transfer coefficients for the top and bottom. For a temperature at a certain depth, I can interpolate based on the heat equation. With boundary conditions that involve convective heat transfer, it's good to model the top with an effective heat transfer coefficient that includes both convection and radiation. However, including radiation could complicate things unnecessarily, so I need to reconsider that.**Considering lumped model limitations**\n\nI could also explore using a lumped model with a cooling coefficient UA that accounts for the top, sidewalls, and bottom surfaces. However, I'm realizing that point temperature would represent the bulk temperature rather than being point-specific, which isn't ideal for accurate modeling. I really want to make sure the approach captures the temperature variations properly, so maybe I need to rethink the use of this lumped model.**Considering script generation**\n\nIt looks like I should create a Python script related to temperature in a steel ladle. The user suggests they’re looking for a physics-based approach rather than machine learning, likely involving a partial differential equation (PDE) solution.\n\nI’ll consider writing a general script for 2D axisymmetric transient conduction with convective boundaries using explicit finite differences. It might be around 100-150 lines, focusing on managing boundaries robustly and implementing vectorized NumPy updates. I need to derive specific components now!**Exploring modeling approaches**\n\nI'm considering using finite difference methods in cylindrical coordinates, including boundary nodes. Alternatively, a more elegant option might be a finite volume method on a staggered grid. Each control volume would need an energy balance.\n\nFor the interior cells, conduction fluxes would go to neighbors, and for boundary cells, we'd apply convective flux at the exposed face. An axisymmetric 2D model depends on radial distance, which raises the question of how to set coordinates effectively. A radial 1D axisymmetric approach might be more suitable here, especially since natural convection plays a significant role in molten steel.**Refining temperature prediction**\n\nTo predict temperature at a specific point within the steel, using radial distance from the center could work well. A 1D radial model in a cylindrical domain (0 ≤ r ≤ R) offers symmetry at the center and accounts for convective cooling at the wall to predict T(r,t). The equation I’d use is ∂T/∂t = α (1/r ∂/∂r (r ∂T/∂r)). This approach simplifies things compared to a 2D model and aligns well with the ladle context. However, if the point has depth too, I lose that dimension.**Considering heat prediction models**\n\nI'm thinking about how steel ladles are typically stirred. Steel is well-mixed, so the temperature should be uniform. Most temperature losses come from the slag and air at the top, as well as the lining walls. A simple 1D radial/vertical model might work. If the user only wants a point temperature, we can identify any location. I need to clarify: \"This script predicts temperature at a point using a transient heat-conduction model,\" and consider providing both a bulk model and a 1D depth profile.**Developing a Python script**\n\nI’m thinking about providing a Python script. I could say, \"If you want, I can adapt it to your ladle dimensions, refractory thickness, or measured process data.\" Since the user just wants a script, I want to keep it concise yet useful. It might be helpful to include math equations in comments, such as the heat equation in 1D and explicit finite difference. I need to ensure the boundary formulas for the 1D slab with convective boundary conditions are derived correctly.**Using the ghost node approach**\n\nI'm considering the ghost node method for boundary conditions. At the z=0 boundary, the derivative can be approximated as \\((T_1 - T_{-1})/(2dz)\\), which equals \\(dT/dz\\) at the boundary. This equals \\(h/k (T_0 - T_\\infty)\\) because z is positive downward, and the top boundary derivative is positive when \\(T_0 > T_\\infty\\). I need to ensure I derive this correctly and clarify any assumptions so it makes sense in this context.**Evaluating modeling options**\n\nI'm thinking about providing an optional 2D axisymmetric script using explicit finite difference with convective boundaries. But, there's a risk of bugs in that. The 1D slab model has simple, reliable code, and the user didn’t specifically ask for radial variation. I could say this gives the temperature at depth z. Perhaps I should also offer a 2D axisymmetric model and mention simplifying if depth is all that's needed. I want to ensure the code isn’t too error-prone, so let’s see if this is manageable.**Deriving coordinates for modeling**\n\nI'm aiming to derive the 2D axisymmetric finite difference model. For this, I'll set the coordinates where r ranges from 0 to R, and z goes from 0 to H, with z measured downward from the top surface. This will help me structure the model accurately and ensure that I’m on the right track with the parameters. Let’s keep this organized as I move forward with the calculations!**Evaluating temperature derivatives**\n\nI’m thinking about how to approximate the second derivative of temperature at the boundary using a ghost node. I can use the formula (T_ghost - 2T_i + T_{i-1})/dr² for ∂²T/∂r² and apply central difference for ∂T/∂r as (T_ghost - T_{i-1})/(2dr). Since the ghost formula is based on the derivative at the boundary, it seems like a solid approach. I’ll then apply the same cylindrical formula for the boundary node with r=R, which should work out well.**Considering boundary conditions**\n\nI'm breaking down the radial operator with ghost nodes for my calculations. I see that d²T/dr² and dT/dr should be evaluated, and I'm considering how to handle corners along with edges. Using ghost nodes helps to account for edges and corners effectively. While I could potentially use loops to compute ghost nodes, corners present a unique challenge since they involve both directions. I’m contemplating a simpler 1D vertical slab approach since it's less risky, even though the user didn't request a full 2D model. Perhaps I could offer a 1D axisymmetric solution as a helpful alternative!**Formulating a temperature model**\n\nI’m considering creating a 1D model for temperature versus depth in the ladle, noting that if the ladle is well-stirred, the temperature should be uniform. So, for a query point, depth from the top could be sufficient. I think it’d be best to provide a Python script using the transient heat equation to predict temperature at any depth in the ladle over time. Simplicity is key, so I’ll focus on mentioning required units and parameters instead of complicating it with multiple layers. If I include matplotlib, that could enhance it too!**Analyzing input parameters**\n\nI’m working with some inputs for my temperature model: density (rho) is 7000 kg/m³, specific heat (cp) is 750 J/kg-K, and thermal conductivity (k) is 30 W/m-K. I'm considering a steel depth (H) of 1.5 meters and an initial temperature (T_initial) of 1873 K (which is around 1600°C). For the temperature at the top (T_air), I need to find an effective boundary temperature, since the top is exposed to slag and ambient conditions, not simply using the ambient temperature. This all seems quite reasonable!**Ensuring numerical stability**\n\nI need to make sure the numerical scheme is stable. When working with the explicit 1D heat equation, stability is achieved if the Fourier number (Fo) is less than or equal to 0.5. I wonder if the same stability condition applies to convective boundaries as well. It’s essential to keep track of these conditions to avoid issues in my calculations! Let's keep this in mind as I proceed.**Exploring time step options**\n\nI'm considering the time step calculation with the formula `dt <= 0.45 dz² / α`. I might set `dt = 0.4 * dz**2 / alpha` for safety. If the user wants longer times, we might have many steps, but that's okay. Crank-Nicolson could be useful for stability, but I wonder if it makes the script more complex. I'll think about whether I can implement an implicit solver using SciPy while balancing simplicity and performance.**Developing explicit boundary formulas**\n\nI'm considering implementing the script using `scipy.linalg.solve_banded` or sparse matrices, but an explicit method seems easier for a self-contained script. If I address stability concerns, I can allow automatic `dt` calculations, so the user can select smaller `dz`. Explicit updates for interior nodes involve the formula: `T[i] += Fo * (T[i+1] - 2*T[i] + T[i-1])`. For the boundaries, I'll carefully derive the updates for the top and bottom, ensuring to check the signs and units. This approach could work well with convective cooling scenarios.**Clarifying boundary conditions**\n\nI'm thinking that if the interior temperature is higher than the boundary, then the interior will warm up, which seems correct. However, I need to confirm if this formula is derived from the linearized boundary conditions. It’s crucial to ensure that I'm applying the right theory here, as my conclusion depends on it being based on valid assumptions. I might need to revisit the derivation to clarify how everything fits together.**Considering temperature units**\n\nI'm focused on the ghost node central difference and think I might need to use query point interpolation with `np.interp(z_query, z_grid, T)` since the z grid is monotonic downward. If the query depth is outside the range, I should raise an error. Celsius seems more intuitive, but for physics models involving radiation, Kelvin is necessary. Since my PDE handles only differences, I could choose Celsius for output, but Kelvin might be safer if radiation is added later.**Deciding on temperature units**\n\nI think the user wants the temperature in Celsius, so I can implement conversion functions that state inputs in Celsius while using Kelvin for internal computations if radiation is involved. To keep it simple, I’ll stick with Celsius for the calculations, noting that it works since only differences are involved. If we later add radiation, I’ll mention converting to Kelvin. I might add an optional `use_radiation=False` parameter to simplify things. I'll explore including radiation while managing complexity!**Calculating net heat flux**\n\nTo compute the net heat flux directly, I’ll focus on the top boundary node (i=0) using finite differences with a ghost node. The convective flux equation will be \\( q = h_{conv} (T_0 - T_{air}) + εσ(...) \\). Since heat flux is nonlinear, I could approximate it with \\( h_{eff} = h_{conv} + h_{rad} \\) and use \\( h_{rad} = εσ(T_{s_K} + T_{sur_K})(T_{s_K}^2 + T_{sur_K}^2) \\), referencing the current boundary temperature. I can then use the same boundary update formula with \\( Bi = \\frac{h_{eff} \\cdot dz}{k} \\) and set \\( T_{\\infty} = T_{air} \\).**Considering radiation and temperature**\n\nI think about how radiation relates to surrounding temperatures, wondering if `T_inf` should equal `T_sur`. If `T_sur` equals `T_air` or maybe a different wall temperature, that could be fine. This linearization seems approximate. For my script, I don’t need it to be exact; using `h_eff` might just complicate things. Keeping the code tight by excluding radiation could streamline things, although using eigenfunction expansion for complicated solutions is tempting, it complicates implementation.**Exploring 1D slab solutions**\n\nI'm looking at the solution for a 1D slab with half-thickness L and symmetry at the center. The equation involves a sum with \\(A_n\\) and \\(\\lambda_n\\), but our geometry has two convective surfaces, which complicates things. Implementing root-finding is trickier than using explicit finite differences. Since the user likely doesn't need an exact analytical solution, FDM seems fine. Also, should I note that if molten steel is well-mixed, conduction might not capture actual gradients? It’s okay if the user just needs a basic model.**Considering molten steel modeling**\n\nI’m thinking that if the steel is fully molten and stirred, a lumped-capacity model might be better. However, since the user is asking for a specific point, I’ll mention that my script assumes a vertical temperature profile. If the ladle is well-stirred, the temperature should be nearly uniform. It’s probably wise to provide a script along with a comment on how to adjust for a specific depth, `z_query` in meters. Also, if the point isn’t right beneath the surface, we’d need to consider a 2D approach.**Refining heat transfer script**\n\nI'm considering refining my approach to the heat transfer script, possibly simplifying it with function parameters. For my defaults, I'm thinking about values like density, heat capacity, thermal conductivity, and bath depth. \n\nI might set the bottom temperature to around 300°C for effective ambient but want to ensure it doesn't cool too aggressively. Maybe I'll use placeholders and suggest a top ambient value of 1000°C since the top would be near steel temperature. It's all about clarity and ensuring users can adjust the parameters as needed!**Designing temperature functions**\n\nI want to ensure that my script runs until `t_end` and outputs the temperature at the specified depth at that time. It could also be useful to output a temperature history since users might want to predict the temperature at certain points. I'll provide functions for both calculating temperature over time and for sampling at a specific query time. \n\nI might also consider plotting the temperature profile, so I need to think carefully about naming these functions!**Considering temperature profiling**\n\nI’m thinking about this code that simulates temperature over time. To optimize memory, I could store temperature profiles at intervals, but if `t_end` is large and `dt` is small, it could take up too much memory. So, maybe it makes sense to just return the final temperature at the queried point instead. It might also be useful to allow an option to save profiles at specific intervals with something like `save_every`. That could balance between detail and memory use!**Refining temperature prediction**\n\nI’m considering how to predict temperature at a specific point over time. The output can be the temperature at that point, `T_query`, and I could offer an option to plot it as well. It might be beneficial to adapt to time-varying boundary conditions, like changing surface temperatures, but using constants for `T_top` and `T_bottom` could simplify things. Perhaps the user would appreciate a script that predicts temperature at a depth for any given future time based on current conditions. I'll think about adding a function called `temperature_at_depth(t_end, z_query, ...)`.**Considering modeling choices**\n\nI’m thinking about validating `z_query` within the range [0, H], maybe using `np.clip` or throwing an error. Using `np.interp` at the end could work too. Should I incorporate the steel phase change? Well, it seems the ladle steel stays molten in relevant timescales, so maybe not. I could consider a more advanced script with a 1D radial coordinate and convective wall, especially since the user didn’t specify solidification. I'll mention a simple model and offer to extend it if needed.**Drafting a preface**\n\nI think it's a good idea to include a short preface to clarify the model I'm using. I could start with: \"Model: 1D heat equation in the steel bath.\" Then I can present the equation: \n\\[\n\\frac{\\partial T}{\\partial t} = \\alpha \\frac{\\partial^2 T}{\\partial z^2}\n\\] \nFinally, I’ll mention the convective cooling at the top and bottom of the system before sharing the code. Clarity is key here!**Exploring numerical methods**\n\nI'm considering using `scipy.integrate.solve_ivp` instead of doing manual time-stepping for my lumped model. This method allows me to discretize space using finite differences and integrate ordinary differential equations (ODEs) elegantly. I could represent spatial grid nodes as a state vector and compute dT/dt while incorporating boundary conditions. Using `solve_ivp` with methods like 'BDF' or 'RK45' would avoid stability concerns. Overall, it seems like a more elegant and less error-prone approach!**Considering solver options**\n\nI'm thinking about the availability of `solve_ivp` from SciPy; it might not be in all environments, but it's likely found commonly. If the user just wants a script, using NumPy could be more universally applicable. However, the stiff solver like `BDF` in `solve_ivp` would manage larger time steps and stable boundaries well—definitely a good balance! Ultimately, I might consider providing an explicit NumPy solver and offering a note about the SciPy version for those who might find it useful.**Considering heat conduction model**\n\nI'm thinking about the accuracy and physical realism of this 1D conduction model for molten steel in a ladle. Sure, it’s a mathematically based script, but I realize the user didn’t ask for high-fidelity detail. I should clearly state my assumptions to prevent any misrepresentation. I can mention that this model predicts temperature at a specified depth, but if the melt is well-mixed, a lumped model may work better. I'm imagining possibly using an energy balance for estimating temperature too.**Refining simulation code**\n\nI’m thinking about the temperature profile model \\(T(z,t) = T_{bulk}(t) + (T_s(t)-T_{bulk}(t))\\frac{z}{H}\\), but it feels like using a simple linear approach might be too simplistic. I see that using PDEs would improve accuracy. Maybe incorporating measured temperatures as boundary conditions could help, like using thermocouple data for calibration. \n\nFor the final code, I could add features such as `steel_properties`, simulation functions, query points, and plotting options. Including functions for `T_top` and `T_bottom` could be useful too, allowing for flexibility based on user preferences.**Considering boundary conditions**\n\nI'm thinking about the side boundaries and how they’re not included in the 1D model. I could mention that the top and bottom boundaries are approximated using effective heat transfer coefficients. For insulation via refractory, setting `h_bottom` to a small value might work well.\n\nMaybe I can use `np.clip` to prevent temperatures from going negative, though I’m not sure that’s essential. I also need to clarify whether the \"point temperature\" should be measured from the top surface or the bottom. I think using `z_query` from the top surface could be more intuitive!**Evaluating modeling options**\n\nI'm considering how to query depth in a bath, where if the point is 0.4 m below a 1.2 m surface, `z_query` would be set to that value. There’s a potential issue of it lying between grid nodes, but I think `np.interp` can handle that. Including `plot_profile=True` would be beneficial, though not strictly necessary. It could be good to mention installing `numpy` and `matplotlib`, and that `scipy` isn’t required. Let's also think about using `solve_ivp` for variable time steps to simplify the code.**Considering model options**\n\nI'm thinking about using a Gaussian process or regression model based on temperature sensor data, but I remember the user's request is specifically for a mathematics-based approach, which rules out ML. Since we're working with the heat equation and PDEs, I might mention that the most general model is a 2D axisymmetric solver, but for the script, we can stick to 1D. I should also check if the boundary formulas for an explicit scheme are correct and ensure the code handles this properly.**Defining temperature boundaries**\n\nI'm considering the scenario where `h_top` is large and `T_top` is lower, which means the boundary cools strongly. It might be helpful to include `T_init` as a uniform initial temperature. If the user knows the current temperature field, we could replace this with an array. Also, I need to choose the orientation for `z`. I'll define `z=0` at the top surface and `z=H` at the bottom, making the depth measurement intuitive as it goes downward from the top.**Considering code structure**\n\nI’m thinking about how to craft this code. I want `predict_point_temperature(...)` to return the temperature at `z_query` after `t_end`. Then, I could create `simulate_profile(...)`, but maybe just one function is enough with detailed comments, like the heat equation and boundary conditions. \n\nUsing `plot=True` sounds good to visualize temperature history, though it might not be necessary. Additionally, I'm considering how to model temperature losses in metallurgy with refractory layers, but maybe that’s more detail than the user needs.**Noting thermal behavior**\n\nI’m considering a potential issue: If the bath is molten and well-mixed, the internal conduction PDE might show little variation if both k and alpha are large. However, if the h values are small, small gradients may occur. I could query temperature at a point, noting that molten steel’s strong convection typically results in minor temperature differences. \n\nI need to calculate thermal diffusivity as alpha = k/(rho*cp), around 5.7e-6 m²/s. For H = 1 m, diffusion time would be approximately 50,000 seconds, meaning time steps of around 0.1 seconds for dz = 1 cm. This results in about 18,000 steps for 30 minutes, which isn’t too expensive.**Considering time step limits**\n\nI'm thinking about whether I should include a `min_time_step` to prevent having too many steps in the calculations. I could compute dt based on `dz`, ensuring it doesn't go below one second. However, it’s crucial to remember that stability relies on having an upper bound rather than just a lower limit. I want to make sure the time steps remain manageable while still adhering to the necessary stability requirements for the model.**Assessing temperature modeling**\n\nThe user's mention of a \"certain point\" is significant. Since the prompt is a bit ambiguous, I think I should clarify limitations in my answer. I could phrase it by saying, \"The script below estimates temperature at a point specified by depth from the steel surface; we can extend it for a fully 3D point.\" I'm considering whether to provide a 2D axisymmetric model instead, as it might yield more detail without being overwhelming. Let's explore if we can implement a 2D explicit solver with convective boundaries!**Evaluating grid efficiency**\n\nI'm considering how to approach this node-by-node with if statements, and it doesn't seem too bad. However, there might be speed issues with loops in Python for larger grids. For moderate sizes, though, it should work. Vectorization is an option with array slices for different regions, but boundary corner nodes pose a challenge. I could create helper functions like `radial_operator` and `axial_operator`, but the explicit 2D approach could be slow. However, a 1D method is simpler and faster. Maybe 2D is overcomplicated here!**Considering script details**\n\nI think D is fine, and I don't want to over-engineer it. I can mention that if the point isn't along the depth axis, the user should provide the coordinates, and I'll generate a 2D axisymmetric version. I need to decide whether to include `matplotlib` to plot the temperature versus depth profile for validation. Including the plot makes sense since the user asked for a script. However, I wonder if I should keep the code lean.**Exploring boundary conditions and outputs**\n\nI see a potential issue with constant boundary temperatures `T_top` and `T_bottom`: if `T_bottom` equals the initial temperature, we won’t have any cooling at the bottom. If `dz` doesn’t divide `H` exactly, I’ll need to adjust by calculating `n_nodes` and updating `dz` accordingly. If `dz` becomes larger than `H`, that would cause an error — I'll make sure to validate that. I also need to decide between using `np.arange` or `np.linspace` for creating nodes. Maybe storing temperature history could be useful, but I’m concerned about memory usage with full `T_hist` at each step.**Considering history storage options**\n\nI’m thinking that I could potentially only store the query point history instead of the full temperature history. This might help manage memory usage while still providing the essential data that the user might need later. It’s a simpler approach that still captures relevant details without overwhelming the system. I’ll consider how this fits into the overall structure of the simulation and whether it meets the user's needs adequately.**Considering stability in calculations**\n\nI’m thinking about how to handle a situation where `dt` is not provided. I want to compute a stable value using `dt_stable = 0.4 * dz**2 / alpha`. If `dt` is specified, we'll check that `Fo <= 0.5`, otherwise, a warning is raised. To ensure robustness, if `t_end` isn’t a multiple of `dt`, I should adjust the final step size to `min(dt, t_end - t)`. This way, I can manage any potential issues with the final step.**Evaluating modeling choices**\n\nI'm reflecting on the use of `numpy` and `numba` for modeling, but I see potential issues with explicit finite differences at boundary nodes, particularly with convective boundary conditions derived from a 1D slab assumption. Since I’m focusing on depth, the top and bottom boundaries are fine, but I need to note that sidewall losses are indirectly included if a larger effective `h_top` is chosen. However, in a 1D model, sidewall losses aren't modeled separately, which could be a limitation.**Setting expectations for modeling**\n\nI think it might be helpful to mention in the preface that if someone wants the temperature at a specific 3D point, they should provide details like the ladle radius, bath depth, and whether the steel is stirred. This way, I can offer a 2D axisymmetric solver. It’s a good approach for setting the right expectations upfront, ensuring users know what information they need to provide for accurate results.**Considering model alternatives**\n\nI'm thinking about an alternative approach and might note that if the steel is strongly agitated, using a lumped model might be better, but only if the user wants a simpler option. It feels like mentioning it could distract from the main script, so I’ll add a sentence at the end instead. I'm also checking if explicit formulas could lead to unphysical results with large Biot numbers, but it seems stable with a small time step. I wonder about the trade-offs in solver types.**Evaluating dependencies**\n\nI want to avoid adding dependencies beyond standard scientific libraries, but I might consider noting that if the user has SciPy installed, there's a more efficient version I could provide. However, since they just asked for a script, minimizing dependencies isn't their main concern. Using `scipy.integrate.solve_ivp` seems natural in a mathematical model context, as the explicit Euler method isn't as robust. It'll be interesting to compare the code quality between these options.**Evaluating numerical methods**\n\nI'm thinking about how to determine temperature at a certain point over time, using interpolation on the solution across different time evaluations. The BDF solver can deal with stiffness well, and writing a complete script with solve_ivp isn't too complicated. I wonder if using BDF with many points in `t_eval` could be heavy, but it should be fine for small nodes. I could define `jac_sparsity` for performance enhancements, but it might not be necessary. Ultimately, I’m considering providing a primary SciPy-based approach with an explicit fallback.**Considering script options**\n\nI might provide a pure NumPy version to keep dependencies minimal, while mentioning that I can rewrite it using SciPy's solve_ivp if needed. However, since the user is asking for a script rather than a discussion, I think simpler is better. Maybe I can use `scipy.sparse.diags` to create a matrix for the discretized PDE, but that might complicate things unnecessarily. Instead, I should start with a concise introduction: “Here is a physics-based 1D transient heat-conduction script,” followed by the code and a brief note on assumptions regarding temperature depths and inputs.**Considering temperature units**\n\nI’m thinking about using Celsius for convection to ambient, but if radiation is involved later, it’ll be necessary to switch to Kelvin. Maybe it's best to note that. Just to be safe, I could convert all Celsius inputs to Kelvin at the beginning, run the model, and then convert the outputs back to Celsius. Since the model only uses differences, the results will remain consistent. This way, I'm keeping things clear and accurate!**Considering script improvements**\n\nI’m working on a code that includes arrays `z` and `T`. I think I can improve it by adding a plot of `Tq_hist` versus time, potentially using `matplotlib`. This might make the answer longer, but it could be really helpful. Since the user asked for a script, that's probably fine. I’ll also mention that `h_top` and `h_bottom` need calibration with plant temperature data—this avoids unrealistic predictions. Plus, we could improve accuracy with a “lumped” coefficient for cooling.**Exploring steel ladle modeling**\n\nI'm considering that the user might ask about the temperature of a specific point of steel in a ladle, where a thermocouple is inserted at a known radius/height. A 1D depth model won't capture the radial variation, so I’m thinking about incorporating sidewall losses using an effective `h_side`, even while keeping it 1D. \n\nI could also model the total heat loss with a well-mixed bulk model, or maybe consider a linear temperature profile. However, 1D conduction with top and bottom boundaries feels better, along with adjusting `h_top` and `h_bottom` if needed.**Clarifying script capabilities**\n\nI think it’s important to note that “this script models temperature versus depth and not full 3D point coordinates.” If the user needs a 3D model, I could definitely extend it. This clarification helps address any limitations in the current version. Additionally, I'll make sure the code remains readable and only utilizes `numpy` for clarity and efficiency. Keeping things straightforward is essential for a good user experience!**Considering solver options**\n\nI’m thinking about using `solve_ivp` with BDF, which would eliminate the need to select a time step. Still, the explicit code might be easier for users to inspect, and the mathematical basis remains solid. In the final answer, I could note: \"If you want, I can convert this to an implicit solver or a 2D axisymmetric ladle model.\" It's worth mentioning that \"Steel properties are approximate and temperature-dependent.\" I should also check for any potential bugs in the `Bi_bottom` formula, especially considering the bottom environment’s temperature.**Generating Python Script**\n\nThe user is asking for a mathematics-based Python script, likely without extensive explanations. I'll include a brief introduction mentioning the equations first, outlining that this script applies a 1D transient heat-conduction model. The specific equation is \\(\\partial T/\\partial t = \\alpha \\partial^2 T/\\partial z^2\\), focusing on temperature at a specific depth. \n\nI think using `solve_ivp` and `scipy` could enhance the approach, offering a better mathematical basis while keeping the script clear and adaptable. An explicit solver with comments seems like the best choice!**Finalizing Script Details**\n\nI’ll keep the current approach, ensuring we set `np.interp(z_query, z, T)` with `z` in ascending order and temperatures in Kelvin. I also need to add error handling for conditions like `if H <= 0 or t_end < 0 or dz <= 0: raise ValueError`.\n\nFor cases where `t_end = 0`, it should just return the initial temperature since the while loop skips. It’s important to ensure there are at least three nodes, especially if `dz` is small relative to `H`. I'll include a note about calibrating `h_top` and `h_bottom`, explaining these coefficients account for more than just convection, since they also reflect radiation and refractory losses in a ladle context.**Refining h Values**\n\nI believe using effective `h` values can accurately reflect the conditions we're simulating. I should include a comment in the code to clarify this, so anyone reading it understands the significance of these values. Capturing these nuances will enhance the code's clarity and usefulness. It’s great to think about how to communicate these technical aspects well, ensuring that the script is not only functional but also easy to interpret for future users!**Defining environment parameters**\n\nI’m thinking that the environment isn’t well-defined, so I’ll keep the height parameters as `h_top=150` and `h_bottom=25`. I should set `H` based on the bath depth, maybe around `1.8 m` or possibly `1.5`. I’m considering `t_end` of `30*60`. For the final temperature at depth, I think it might stay around 1550. I should also note the model ignores buoyancy-driven convection, which makes it an approximation, but still useful for a first-order estimate.**Enhancing code documentation**\n\nI think we should tell the user, “If you know the exact point coordinates, I can create a 2D axisymmetric version.” This could be helpful if they need more details. Now, let’s consider adding documentation for the code, specifically a function docstring with units and behavior, like this: \n\n```python\ndef predict_temperature_at_depth(...):\n \"\"\"\n Solve the 1D transient heat equation in a steel bath. Returns:\n times (s), temperature_at_query (°C), depth_grid (m), final_temperature_profile (°C)\n \"\"\"\n```\n\nI’m deciding whether to include the temperature history or just the final temperature. Since they want a prediction at a point, I'll return both. I'll also consider using `np.asarray` for the return values and maybe add functionality for script execution as a standalone.**Finalizing the script**\n\nI’m thinking this makes it a complete script! I should include a `matplotlib` import because it handles plotting. Even though the user can run the function with just NumPy, it will need `matplotlib` for visualizations. I wonder if it’s smart to include the plotting code only in the main block. Maybe I could even comment it out if `matplotlib` isn't installed, but honestly, it seems pretty common for users to have it.**Analyzing stability conditions**\n\nI'm thinking about `Fo` stability and boundary conditions. If `h` is large, I might need a smaller `dt`. When `Fo` is less than or equal to 0.5, things should be fine, but boundary conditions might impose limitations. To be on the safe side, I should likely choose `dt` based on `min(0.4 dz² / alpha, 0.4 * rho * cp * dz / h_max)`. Then there's the stability constraint `2Fo Bi <= 1`, which I'm gearing up to derive. So, I should ensure coefficients remain nonnegative. Good catch!**Considering coefficient behavior**\n\nI'm realizing that in an explicit scheme, negative coefficients don't automatically mean instability. Instead, they can lead to oscillations, which is an interesting nuance. I need to keep this in mind when analyzing my results and any effects on the overall stability of the system. It's a delicate balance, and I want to ensure I'm accurately accounting for these scenarios to avoid unexpected behavior in my computations. That's something to watch carefully!**Calculating boundary stability limits**\n\nI’m trying to derive a boundary stability limit from the equation. If the Biot number (Bi) gets large due to high heat transfer, such as with h=100 W/m²K, the condition leads to a Fo value of about 0.484. If h increases to 1000, it drops to a limit of 0.375. It’s sensible to choose a smaller dt to ensure stability. I'll calculate a robust time step limit, defaulting to a conservative approach for safety. Let's go ahead and compute it!**Verifying Biot number effects**\n\nI'm verifying with Bi=0 to confirm that Fo should be less than or equal to 0.5, which checks out. With positive Bi values, we tighten the requirements for stability. I will compute the maximum Biot number using the formula and set a stable dt. Choosing 0.45 instead of 0.5 might be smarter for safety. I need to decide whether to automatically reduce dt or raise an error if the user specifies a larger dt, since silently changing it could be unexpected. A warning or ValueError might be the best way to go.**Calibrating values**\n\nI need to replace `h_top` and `h_bottom` with the calibrated values from my plant data. It seems like just doing that is sufficient for what I'm trying to accomplish. I wonder if there’s anything else I need to consider for accuracy, but for now, focusing on these values should help streamline the process. Let’s make sure I have the correct data ready to input and keep things straightforward!**Validating parameters and temperature**\n\nI’m considering whether I should validate if `k <= 0`, which isn’t necessary but would be nice for ensuring positive physical parameters. The user might be looking for information at a certain depth rather than surface depth. I could phrase it in the code docstring about the `z_query` as the distance from the steel surface. \n\nAdditionally, it might be useful to mention \"bulk temperature\" versus \"point temperature.\" Since the bath could be mixed, perhaps the point temperature would be close to average. \n\nLastly, I think I should highlight that this is a first-principles model, not a fitted regression, and include a brief note on the Fourier number, `Fo = α dt / dz²`.**Improving code accuracy**\n\nI see that including some mathematical explanations in code comments or prose could really enhance understanding. It’s a good idea to use something like `while t < t_end - 1e-12` to help avoid any floating-point issues, which can be tricky. This way, I’m ensuring that the calculations will run smoothly without unintended errors from precision problems. Just thinking through this helps me feel more confident in the accuracy of the code!**Considering code documentation**\n\nI'm thinking about whether to include instructions for `pip install numpy matplotlib` at the top. It might not be necessary, but maybe I should mention it if it's missing. Code comments could be enough for guidance. Now, I'm also considering using `np.full(n_nodes, T_init + C2K, dtype=float)` for initialization, and that seems like a solid approach. It's all about providing clarity and making things easier for anyone using the code!**Considering temperature limits**\n\nI’m thinking about whether to include `np.maximum` to keep temperatures above 0 K, but it might not be necessary for short times and certain physical parameters. However, if the user inputs odd values, I could update it to `T = np.maximum(T_new, 0.0)`, though that might distort physical results if the model is unstable. I also wonder if using `np.interp` could cause issues if `z_query` is out of range. I'll validate that. It’s important to note that this uses effective heat-transfer coefficients, combining convection and radiation.**Evaluating temperature considerations**\n\nI'm thinking about using effective `h_top` to account for heat transfer, which is good. Maybe `T_bottom_env` should represent the refractory interface temperature instead of ambient temperature. It might be valuable to note at the end that if a true point inside the ladle is needed, I can generate an axisymmetric version. I also consider using `scipy.integrate.solve_ivp` for continuous point temperature but we decided to stick with pure NumPy. Meanwhile, I need to ensure I handle `h_bottom=0` with the convective boundary formula, then apply the Neumann zero-flux condition for bottom node updates.**Considering boundary conditions**\n\nI could mention in a comment to \"set `h_bottom=0` to model an insulated refractory bottom,\" which seems useful. But should I include a more realistic bottom boundary by setting `T_bottom_env` with a low `h_bottom` instead of zero? That would be smart. Now, I’m also thinking about how to format the answer. Since the code is lengthy, using a single code block might be the most effective way to present it neatly.**Considering thermal modeling**\n\nI’m thinking about the code and whether to include options for a 2D or radial model, and even calibration against sensor data. I wonder if I should mention the expected output, though it may not be necessary. For clarity, I could replace \"mathematics-based\" with \"physics-based\" to avoid confusion with machine learning. I’ll explore whether the user is looking for a script using measured variables to predict temperature at a point, noting that coefficients often relate more to conduction than convection.**Refining the model**\n\nI'm considering how to handle the sidewall in the model. If we lump it into effective top and bottom losses, that might work, but we could also mention explicitly using a 2D axisymmetric model if needed. I wonder if we should add physical constants like the Stefan-Boltzmann for radiation, but that might not be necessary. Looking into making the code more mathematically rigorous with `solve_ivp` could help, as it avoids some stability issues. Comparing script lengths, using `solve_ivp` might yield a shorter code than using pure NumPy.**Exploring implementation options**\n\nI'm considering whether to implement both methods: a `rhs` function along with using `solve_ivp` from SciPy, but that might complicate things a bit. Instead, I think I could focus on crafting a script that uses `solve_ivp` since it's concise and mathematically sound. I need to evaluate if having a SciPy dependency is okay, but most scientific Python setups include SciPy, and it's generally easy to install if it's not already there.**Considering a Python script**\n\nI'm thinking about the final point of historical data access, and how we can use `sol.y[:, -1]`. It seems good that the `solve_ivp` function has adaptive time steps, making `t_eval` less impactful. I wonder if the boundary updates would be stable with BDF? It feels like a better option. But if SciPy isn't available, I should suggest `pip install scipy`. Since the prompt didn't limit dependencies, using SciPy may align better with the mathematical approach. Still, I should consider the user's likely setup. Would a simpler pure numpy solution be more user-friendly for those without SciPy?**Deciding on solver approach**\n\nI'm weighing which approach would satisfy the user best. Pure NumPy seems like a solid choice since it’s universally runnable, and explicit time stepping isn’t an issue when applying to data. I might go with a default pure NumPy solution since it's easier to grasp and modify. I could also mention an implicit version if they need stability, but it doesn’t seem necessary. I'll finalize it as a 1D explicit finite-difference solver using NumPy and Matplotlib, mentioning assumptions and a future 2D upgrade. I'll also check if the `h_top` and `h_bottom` units and values make sense.**Structuring the final answer**\n\nI’m considering setting `h_top = 150.0` and `h_bottom = 20.0` for sample usage, which feels good. Should I also include `steel_k = 30.0`, `steel_cp = 750.0`, and `steel_rho = 7000.0`? That seems alright. It might be useful to note that \"steel properties vary with temperature, but these values are common approximations.\" I think I'll structure the answer in four parts: a short introduction, an equation block, the full script, and a brief note on adapting it to actual ladle data. This could be concise yet informative!**Crafting code carefully**\n\nI’m planning to write about 100 lines of code, and I want to make sure it’s clear and understandable. I think it’s a good idea to include comments within the code to explain formulas and parameter names. That way, anyone reading it can follow along easily. I’m feeling motivated to ensure everything is well-explained and documented, so let’s focus on that as I start crafting the code!**Reviewing the Python code**\n\nI’m looking at a polished version of a Python function for predicting the temperature of molten steel in a ladle using a 1D transient heat-conduction model. It inputs various parameters such as depth, environment temperatures, and material properties. Validation checks ensure positive values for depth and time, while the thermal diffusivity is calculated using material properties. There’s also grid setup involving nodes along the bath depth, with a placeholder for stability limits. Let’s keep refining this!**Evaluating calculations**\n\nI should be cautious about using `Bi_max` computed with `dz` after rounding. If `t_end` is zero, the loop should skip and return the initial temperature, which is good. I see a potential issue: if `h_top` or `h_bottom` are zero, that gives `Fo_limit = 0.5`. That's fine, and I also realize `np.interp(z_query, z, T)` works even if `z_query = H`, which is great. Maybe I can include a note about the units for `rho`, `cp`, and `k` in the code comments.**Refining model parameters**\n\nI’m considering setting `cp=800.0` and `k=35.0`, but the user can adjust these values. Maybe I should suggest some typical approximate values and note that they can replace them with temperature-dependent properties if they have them. I think I should mention that the heating/cooling coefficients can be calibrated using measured ladle data. It’s also important to clarify that `h_top` may include radiation, while `h_bottom` represents overall refractory resistance. Let's keep things concise in the final response!**Structuring the code response**\n\nI’m thinking about using a short introduction, the code, and a one-line note at the end. I need to ensure that the code includes `numpy` and `matplotlib` imports. It’s important to mention that this model assumes the steel bath is homogeneous in the horizontal direction. Maybe I should include `z_query` as the depth from the top surface, with a comment like `# z_query measured downward from the free surface`. While I could provide a variant using `scipy` for robustness, it might not be necessary since the initial script meets the prompt requirements.**Enhancing the code explanation**\n\nI think it could be helpful to include a note stating: \"Set `h_top`, `h_bottom`, and `T_bottom_env` to values from your process; these are the main calibration knobs.\" This would be a practical guide for users. I should mention that after the code. I could also add a note that if users are measuring a point with coordinates (r,z), I can convert that to a 2D axisymmetric solver. I pondered adding an optional function for computing `bulk` temperature, but it might not be necessary. There's a potential issue relating to `Bi_max`, which uses `dz` and `k`.**Optimizing code efficiency**\n\nI'm considering optimizing some code where `Fo_limit` could drop below 0.5 if both parameters are large and `dz` is small. The code automatically selects a stable `dt`, which is great! I'm thinking about adding some transparency by printing `dt`, but I might hold off since the script shouldn't print by default. Also, it might be efficient to compute `Bi_top` and `Bi_bottom` outside the loop, since these constants don't change. That's definitely a good move for efficiency!**Refining temperature calculations**\n\nI'm adjusting the calculations for `Bi_top` and `Bi_bottom` to be defined in terms of heat transfer coefficients and depth. While working through the loop, I'll ensure `Fo` is calculated for validation. However, there’s a potential issue: if `dt` is based on `Bi_max`, it could vary due to the top and bottom heat coefficients. It might make sense to interpolate temperature at each depth step, though storing the times might not be necessary since history is useful. I'll consider returning temperature values in Celsius for clarity!**Noting limitations**\n\nI’ve made a note that the model doesn’t account for solidification or reaction heat. This seems important since those factors could influence the overall results significantly. I wonder how this might affect the conclusions drawn from the model. It’s a good reminder to consider all elements that might not be included in the analysis. I should keep this in mind as I proceed, just to ensure that I’m thorough in my evaluations.**Evaluating cooling parameters**\n\nI'm trying to avoid over-explaining. I need to consider the constant properties of the system. However, I see a potential issue if `T_bottom_env` is set at 25°C with `h_bottom>0`, as this could lead to excessive cooling of the bottom boundary. \n\nI’m thinking about calibrating `T_bottom_env` based on the actual refractory temperature or by adjusting `h_bottom=0`. It might be realistic to set `h_bottom=0` for cases with good insulation, yet it’s important to provide some flexibility. I wonder if `h_bottom=20` could work better since it balances reality with user adaptability.**Configuring environment settings**\n\nSince I'm working with a script, I think setting `h_bottom=10` and `h_top=150` could work well. The example can use those values along with `T_bottom_env=100`, but I realize the user will likely adjust them. \n\nI might consider creating separate variables for environment temperatures, like `T_slag_env` or `T_air_env`, though it doesn't seem necessary right now. I wonder if using `np.interp` on the final profile would oversmooth the query temperature, but maybe it’s workable. As for incorporating `pandas`, it doesn’t feel required, but tracking `point_history` at each time step sounds valuable for user flexibility.**Deciding on storage and libraries**\n\nI’m thinking about whether to reduce overhead by not storing `times` if the user only wants the final results, but I realize that having a history could be valuable. \n\nNext, I'm considering whether to use Python’s built-in `math` library instead of `numpy`, but I think `numpy` is still a good choice. It offers capabilities that could really enhance performance, so I should stick with it. This balance between performance and memory is interesting to navigate!**Considering modeling accuracy**\n\nI think the small time step might still incur some numerical diffusion, but it's acceptable for a simple model. It could be useful to mention alternative methods like Crank–Nicolson or BDF for more accurate predictions. A note could suggest an implicit solver to address performance concerns. Also, I should clarify that this isn't a CFD model. \n\nNow, for predicting temperature in a ladle, a 1D depth model seems like a reasonable first step, but we should be clear that this approach doesn't account for full 3D.**Finalizing temperature model**\n\nThe script predicts temperature at depth `z_query` from the steel surface. I plan to show the relevant equations and use a code block for clarity. At the end, I’ll suggest adjusting `h_top` and `h_bottom` to fit the user's ladle. If they need a full 2D axisymmetric model or something calibrated from data, I can generate that too. \n\nI’ll ensure the code only uses `matplotlib` in the example section. If the user doesn’t have it, they can remove the plotting lines. Including a comment about installing dependencies might be helpful, like: `# pip install numpy matplotlib`, but I want to keep the final answer concise.**Considering numerical methods**\n\nBefore finalizing, I'm wondering about using `solve_ivp` because it feels more mathematically grounded than explicit methods. But hold on—explicit methods are mathematically sound too. The user didn't specifically request high-order accuracy, so it's fine to stick with them.\n\nI’m also thinking whether the code could be simplified with `numpy.linalg`, though that might not be really necessary. It might be a good idea to check that `Fo_limit` isn't too small and verify that `dt` isn't tiny either. Lastly, I want to double-check the update formulas to ensure the top boundary sign aligns with `T_top_env_K`.**Considering temperature modeling**\n\nI’m thinking about how to explain that, for a vigorously stirred molten bath, the temperature is often uniform, which is a good first-order estimate. If I add this detail, my answer might get a bit longer, but that's okay. I could also mention that if the user needs temperature at an off-center point, they can provide the coordinates for a 2D axisymmetric model. I want to think about using `T_bulk(t)` from energy balance and adding a small vertical gradient. I should mention \"effective heat-transfer coefficients\" for calibration, too. Overall, the PDE is solid, but using actual convective values might not be enough, and calibration will be necessary.**Considering temperature adjustments**\n\nI’m pondering whether to set `T_bottom_env` lower than `T_init` but not at ambient temperature. The bottom refractory surface can get really hot, possibly around 1000°C. It might make sense to refer to the refractory interface temperature instead of the outside temperature. I can still use any `T_bottom_env` in the script. I think I should mention \"effective surrounding temperature or interface temperature.\" I’m also considering a simple linear interpolation of `h_top` with time for slag removal or pour time, though that might not be necessary.**Considering function adjustments**\n\nI'm thinking about how I might name the function — possibly `steel_ladle_temperature_point`, but that's not strictly necessary. I’m debating whether to return just `temperature_at_point` instead of an array, maybe adding a parameter like `return_profile=True`. To improve clarity, I could shorten the example output. Additionally, using a finite difference model in radial coordinates could be useful, but it seems easier to stick with vertical since I lack info. I’ll also add an initial comment about the governing equation and boundary conditions.**Exploring code comments and performance**\n\nI’m considering presenting the discrete equation in comments, like `T_i^{n+1} = T_i^n + Fo(...)`, but the code comments may already be clear enough. I want to ensure `T_top_env_K` and `T_bottom_env_K` remain unchanged over time. If the user wants temperatures in Celsius only, they can skip the K conversion since the code handles it. I’m also pondering memory and performance if `t_end` is long and `dt` is small. Even with 45,000 steps for an hour, this doesn't seem huge. I could include a note about adjusting `dz` for faster runs or higher spatial resolution, maybe even a comment about using `np.interp` for exact grid points.**Enhancing code functionality**\n\nI’m realizing that if the user wants a point not on the depth axis, we need to take note. I’m also thinking about improving the code by making `T_env` values time-dependent functions instead of constants. For instance, I could create functions like `def T_top_env_func(t): return ...` and `def h_top_func(t): return ...`. However, it may not be strictly necessary since, in a ladle, the top conditions do change as the process progresses, which is an important consideration.**Summarizing modeling approach**\n\nI’m considering how to frame the environment terms in relation to time, though it’s not essential. I think a \"mathematical model\" section could work well, including parts like PDE and BCs, and mentioning variables like `Fo` and `Bi`. I could even enhance the code with discrete formulas for interior and boundary updates through comments. There might be a question about setting `T_bottom_env` to 25, given insulation issues, but the user's calibration will ultimately take precedence.**Structuring the script**\n\nI'm thinking the answer could focus on a \"script only\" format, but adding a brief intro and notes might enhance clarity. I want to use SI units for variable names and provide outputs in Celsius. I don't think adding `from typing import Tuple` is necessary. For the script's organization, I’ll include sections like parameters, validation, grid, time step, simulation, and example usage. This should improve readability. A 1D model with convection should suffice for a mid-point analysis in a ladle, but I’ll encourage using a 2D model if near the wall or bottom.**Considering code approach**\n\nI’m exploring whether a `solve_ivp` version would be shorter and allow for a future 2D extension, but I think the explicit code is already straightforward. It would be good to include a note like, \"This is a first-order approximation,\" to set expectations. A potential issue arises if both `T_top_env` and `T_bottom_env` are lower than `T_init`, as that would cool the entire profile, while having one higher could allow for local heating.**Considering function parameters**\n\nI’m mulling over whether the equation `Fo_limit = 1/(2*(1+Bi_max))` is too strict, especially if `Bi_max` takes its maximum from `h_top` or `h_bottom`, but what if one boundary isn’t that large? It feels safe, though. Maybe printing the chosen dt could help with debugging, even if it’s not essential. Still, I think the function should remain silent overall. So, I’ll keep that in mind for now.**Considering effective coefficients**\n\nI'm thinking about using a comment regarding effective coefficients for radiation or refractory losses near the h parameters. It seems beneficial to clarify that for `T_bottom_env`, we could say \"use 0 if adiabatic bottom.\" That makes sense! For an example, I’m considering using `h_bottom` values of 10.0 or 0.0 to illustrate the point. Just want to make sure these comments help others understand what's going on!**Finalizing model notes**\n\nI’m thinking it would be helpful to provide a minimal additional note below the code. To model a point at coordinates (r,z), I should recommend using a 2D axisymmetric heat equation instead of the 1D depth model. It's good to incorporate `h_top` and `h_bottom` as time-dependent, if desired, but I don't think it's necessary. I need to clarify that the model is a physics-based approximation, without claiming it’s \"mathematically exact.\" I'll note that the steel has constant properties, and there’s no phase change. If solidification is needed, adding latent heat might be relevant. Lastly, I should clarify the model only considers conduction within the steel, as internal convection is absorbed into coefficients. For molten steel, the temperature is nearly uniform if well-mixed.**Preparing model expectations**\n\nI’m considering that our model may better approximate a solid bath or effective diffusion in a mixed melt. It's a simplification, so I want to be honest about that. I should say something like, \"This is a simplified thermal model. For molten steel with strong stirring, actual mixing can exceed conduction; in that case, consider using a lumped model or fitting effective coefficients.\" I'll ensure we provide the script as requested. It would help to label the model at the top as a \"Simplified physics-based model\" to set proper expectations. Let's finalize the response with moderate detail.**Evaluating default parameters**\n\nI'm considering whether the default `dt` should be `None` or a numeric value. Setting it to `None` allows for auto-selection of stable values, which is useful since the user can override it if needed. If the user changes `dz`, I might think about updating `dt` automatically. I’m noticing some potential issues with the while loop; if `t_end` is very small, it could affect the `dt_step`. Preallocating arrays like `times` and `T_query_hist` might improve performance, but it isn't strictly necessary. Let's reflect on the final code style!**Ensuring proper formatting**\n\nSince I'm working on the final answer, it's crucial to check that the indentation and formatting are spot on. I’m thinking of adding a comment to clarify the meaning, something like: `# z=0 -> steel surface, z=H -> bottom of bath`. That should help anyone reading the code understand the context better. It's all about making it clear and easy to follow! So yes, I’ll include that comment for sure.**Evaluating temperature boundaries**\n\nI'm considering that surface boundary can be significantly influenced by radiation, where h_top acts as an effective value. This brings up a potential issue: in a ladle, the top boundary temperature might not be the ambient 25°C due to the slag cover, possibly making T_top_env closer to 1200°C. Since h_top is effective and environment temperature is a reference, the user can set it accordingly. It’s good to mention in the introduction that effective coefficients and boundary temperatures should be calibrated.**Considering script accuracy**\n\nI'm wondering if the user might want a math-based script to predict the temperature at a specific point of molten steel in a ladle. However, I have some concerns since, in this situation, there's no conduction possible due to the movement of the molten steel. This might mean that our script won't be physically accurate enough for their needs. It's important to ensure that any predictions we make align well with the actual physical behaviors in this scenario.**Considering code adjustments**\n\nI'm thinking of including the return of `temperature_point_history` in Celsius and `final_profile`. There's also the name `z_query` vs. `point_depth` that I should evaluate. Perhaps I could add a snippet on calibrating `h_top` and `h_bottom`, suggesting fitting them to thermocouple trajectories using least squares, but that might not be necessary. I want to ensure that the final answer remains direct, focusing on the code since the user requested a script. A note about constant properties will be helpful too.**Improving function stability handling**\n\nI want to ensure the function correctly handles `dt` if a user supplies a value. I should compute `Fo = alpha * dt / dz**2` and then establish `Fo_limit`. If `dt` exceeds that limit, raising an error is appropriate, as users might be frustrated if they want to ignore stability but get an issue instead. If `dt` is `None`, I could compute a stable `dt` and allow the user to specify `steps_per_second`, but that might be unnecessary for the current implementation.**Evaluating code stability**\n\nI’m considering a potential issue: some users might copy-paste code without `matplotlib`, but they can comment out plots if they don’t need them. `numpy` and `matplotlib` are common imports, so that's fine. Maybe I should think about including a `try/except ImportError`? No, I need to ensure the finite difference update at boundaries uses `Bi_top` and `Bi_bottom` correctly. I’ll compute those outside the loop since they depend on `dz`. If `h_top` or `h_bottom` isn't supported, that could be a problem. Additionally, I’ll make sure the error message guides users on `dt`. That seems sufficient!**Considering analytical solutions**\n\nI’m wondering if there’s a more sophisticated analytical solution I could mention briefly. Maybe I could say, \"If you prefer an analytical series solution instead of numerical finite differences, I can provide that as well,\" but that’s not strictly necessary. I'm also thinking about the final length of my answer in relation to OV 6. The code plus introduction and notes should be moderately sized. Oh, and there’s probably no need to discuss the difference between \"steel\" and \"molten steel.\"**Evaluating temperature predictions**\n\nI see that `predict_temperature_at_depth(...)` returns `Tq` in Celsius, which is good. Instead of using `T_point_hist[-1]`, I might want to print `Tq[-1]`. Plotting `Tfinal` against `z` and inverting the y-axis could work well too. \n\nI'm considering if the user wants to predict the temperature at a specific point in steel within a ladle. Our model can help with that, although a point in the steel bath might show little temperature difference. I think adding a `plots=True` flag for optional graphs could be helpful.**Reviewing temperature model code**\n\nI’m considering the option of using a cell center grid instead of boundary nodes, although it doesn’t seem necessary. The final answer can stay as is. I should finalize the comment wording at the top of the code, like: “# Physics-based 1D model for temperature in a steel ladle,” and include details on the boundaries. \n\nUsing comments about the heat-transfer coefficients is a good idea. I’ll also check for any confusing variable names; for instance, `T_top_env` and `T_bottom_env` might be better as `T_surface_env` and `T_base_env`. 🤔 Also, computing `Bi_max` with `max(h_top, h_bottom)` makes sense!**Considering validation of coefficients**\n\nI'm thinking it might be important to validate the positive values of `h` because having negative coefficients would be unphysical and make no sense. If both coefficients are negative, that could indicate an invalid scenario. \n\nSince our current approach doesn’t include validation, I see the merit in adding a check for positive values only. This way, we can ensure that the coefficients make sense and align with physical principles. It's just a little tweak that could enhance the model's reliability!**Considering plotting figures**\n\nI'm thinking it might be a good idea to include a call to `plt.show()` after plotting each figure. That way, I can ensure that the figures are displayed properly. This will help in visualizing the data effectively, right? So, I will make sure to add that at the end to wrap things up nicely! I want to make sure I get everything polished. Let's finalize the code before running it.**Considering script adjustments**\n\nIn a non-GUI environment, plotting may not work, and that's okay. I can suggest commenting it out if needed. The initial lines of code should include `import numpy as np` and `import matplotlib.pyplot as plt`. There’s no need for `from __future__ import annotations`, but it might be helpful to annotate return types for Python 3.9+. I can phrase it as: \"This script predicts temperature at a depth `z_query` from the surface. To enhance realism, consider calibrating effective coefficients with plant data.\"**Clarifying parameters and notes**\n\nThis is looking concise and useful! I should mention that if the user's point is at a radial coordinate, they can let me know. I’ll consider adding a brief list of parameters outside the code for clarity: `H` for bath depth, `z_query` for depth from the surface, `h_top` and `h_bottom` for heat transfer coefficients, and `T_top_env` and `T_bottom_env` for surrounding temperatures. I want to keep the code self-explanatory but might add a compact note about time history resolution since it could be coarse due to larger `dt`.**Reflecting on boundary updates**\n\nI’m considering whether there might be a subtle issue with the boundary update formulas that use `Bi_top` and `Bi_bottom` defined without a half-step for `dz`. That's standard for boundary nodes, so it seems fine. However, maybe the update should use `Bi = h * dz / k` and consider `2Fo`. There's also a possibility of enhancing accuracy by using a one-sided finite difference of order 2. But honestly, the current method seems good enough for now.**Reviewing numerical code improvements**\n\nI'm aiming for higher accuracy with the Crank–Nicolson method, which sounds promising! I realize there's a potential issue with using `Fo_limit` based on `Bi_max`, as stability also depends on the largest `Fo` interior. Since I'm using the same dt, that’s fine. I think being more concise in my final answer would help too. Adding a comment about constant properties is a good idea. I should also clarify unit usage and temperatures in Celsius while indicating that they're converted to Kelvin internally. That improves clarity!**Enhancing code clarity**\n\nI think noting that this can represent a refractory interface temperature or an effective sink temperature is helpful! To make the code clearer, I could add small functions like `c_to_k` and `k_to_c`. They would enhance readability, even if it adds more lines. Alternatively, I could use a constant `C2K = 273.15`, which is neat! I also want to check if there's an issue with printing after `predict_temperature_at_depth`. Additionally, including \"mathematics based\" in comments might not be necessary, but it could clarify things. I should remember that this is a simplified model, especially for molten steel!**Clarifying boundary conditions**\n\nI want to consider the order of boundary conditions and equations in my explanation. The mathematical model could be presented as: \n\\[\n\\frac{\\partial T}{\\partial t} = \\alpha \\frac{\\partial^2T}{\\partial z^2}\n\\] \nwith convective boundary conditions. I think it’s crucial to mention that `alpha` represents thermal diffusivity. It makes sense to include that in the code comments instead of the prose for clarity. Keeping it concise while ensuring that the details are understood is important. Overall, I’m feeling good about this structure!**Refining update formula**\n\nI'm considering the explicit update formula at the boundary, which relies on the ghost node temperature formula. I think there might be a factor of 2 in the boundary convective term, and it's good we derived that. \n\nAs I'm preparing the final answer, I wonder if I should add the note for users to install the necessary libraries. A line like `# pip install numpy matplotlib` would be helpful without cluttering things up. \n\nI want to keep it concise, mentioning the main calibration knobs and the option for a 2D axisymmetric version too.**Clarifying script expectations**\n\nI'm considering whether the user just wants the script without much explanation, but I believe a brief introduction and note would be fine. I should emphasize that the model is 1D vertical rather than full 3D. For completeness, I could mention, \"For a full 3D point, you need radial and axial coordinates (r,z) and a 2D axisymmetric PDE.\" \n\nI think including a note about the temperature uniformity when the bath is well-stirred is helpful. I also need to clarify the term `T_top_env`, since the top surface is likely a slag cover. Using \"effective boundary temperature\" instead could be clearer, I think.**Optimizing the code**\n\nI'm considering whether to include `num_steps = int(np.ceil(t_end/dt)` to preallocate arrays, but it seems unnecessary. I noticed a potential issue with using `np.interp` since `T` is in Kelvin. Subtracting Celsius to Kelvin gives Celsius, which is important to clarify. It might be better if `T_init` is in °C. \n\nAlso, there's a possible improvement I could make: I could vectorize the update of the interior nodes with `T_new[1:-1] = T[1:-1] + Fo_step*(T[2:] - 2*T[1:-1] + T[:-2])`. That sounds good!**Considering temperature prediction**\n\nI’m thinking about how the user might want to predict the temperature at a point in a steel ladle using actual sensor data from a previous time series. The function I have takes `t_end` and model parameters without needing data input right now, which is okay since it’s about generating a script. I’m considering potential issues, like how rounding can affect grid spacing. Using `ceil` instead of `round` might be safer for ensuring accuracy in different situations.**Tweaking calculations for nodes**\n\nI’m considering how to set up my calculations for grid spacing using H and dz. If I set H to 1.535 and dz to 0.01, I might end up with 154 for both `ceil` and `round`. However, using `round` could lead to an unintended 153. To ensure I don’t end up with a coarser grid, switching to `ceil` seems smarter. I’ll set `n_nodes` with `ceil` and ensure the actual dz is calculated afterward. This adjustment should help maintain accuracy in my calculations!**Refining time steps**\n\nI'm considering setting the default `dt` based on the actual `dz`, which seems like a good idea. Another improvement could be raising an error if `dt` is provided and too large. I worry that `Bi_top` and `Bi_bottom` might not be symmetric if `h_top` and `h_bottom` differ. \n\nI think the time step stability limit formula should use `Bi_max = max(h_top, h_bottom) * dz / k` with the actual `dz`. It feels polished now! Maybe we should define `t_points = np.linspace(0, t_end, 200)` for equally spaced outputs.**Optimizing data storage**\n\nI’m considering whether we should store history at specific output intervals instead of every step to save memory. Since the function only records one value per step and one profile at the end, managing 22,500 data points seems fine. If `dt` is small and `t_end` is long, we might hit 45,000 points, but that’s okay. Maybe we can limit history by choosing every `n_output` steps, but the user might prefer the full curve. Since the model isn’t huge, storing everything should be fine. I also want to clarify that `h_top` and `h_bottom` are “effective” values, noting that the top surface radiates, which could be important!**Considering boundary conditions**\n\nI'm thinking about using both convection and radiation at the top boundary, which feels more realistic. Should I include optional radiation with an effective coefficient based on the current surface temperature? It adds complexity, but it might improve the model's realism without becoming unmanageable. It’s worth exploring to see if that increase in detail enhances the accuracy of our results. I like the idea of balancing realism with complexity carefully!**Considering radiation and code complexity**\n\nThe user didn't mention radiation, but maybe it would be nice to include it. The simpler code works well as is, and adding parameters for radiation could complicate things. Still, we could introduce a flag for it, but that might make the code harder to follow for the user. It's likely better to keep it simple and make a note about extending the top boundary for radiation if needed. Then I’ll check for any potential issues with the calculation parameters.**Evaluating interpolation code**\n\nI’m considering whether to return `final_point_temperature` as a separate value since that could be convenient for the user. But, it feels less necessary because they can use `T_query_hist[-1]`. Keeping the return signature simple is probably best. I wonder if we should mention that `z_query` can be chosen as `H/2` for center depth, but maybe that’s overkill. I should note in comments about the trade-off with `dz` for detail and stability. Also, the repeated `np.interp` calls seem fine since it's simple linear interpolation.**Considering code optimization**\n\nI’m contemplating whether to compute `idx = int(z_query / dz)` for manual linear interpolation to possibly speed things up. While that might work, I think `np.interp` keeps the code clearer, which is great. If optimizing becomes necessary, we could precompute weights for fixed query depths, but that might not be needed. I should mention commenting out the plot section for headless environments, though it's not crucial. Finally, I believe including a markdown code fence with Python syntax highlighting is a good idea. I'll also start with a brief introduction before presenting the code.**Generating 2D axisymmetric models**\n\nIf the user is looking for a 2D axisymmetric model for a specific point (r, z), or even if they need a version that fits parameters based on plant measurements, I'm ready to generate that. I think it’s great that I can provide different model types to meet their needs. It shows flexibility in the approach, so whatever they prefer, I’ll make it work! Let’s get started on that if they want it.**Reviewing temperature equations**\n\nI’m looking at the equations involving temperature differences and boundary conditions, and they seem standard, which is great! I want to mention the assumption about constant properties and no phase change. One potential issue is that using `z_query` from the surface could confuse users if they're thinking of measurements from the bottom. I’ll clarify this in code comments and documentations, especially around conversions to Kelvin. Also, maybe I should consider renaming `h_bottom` to `h_base`, but I think everything looks good for now!**Clarifying code notes**\n\nI'm thinking it would be useful to mention \"Set `h_bottom=0` to make the bottom adiabatic\" in the notes. For a well-stirred bath, the model will likely remain consistent across depth, which gives a good expectation for results. I want to make sure the final answer doesn’t have any contradictory statements. \n\nThe code name could be `predict_temperature_in_ladle`, but `predict_temperature_at_depth` might be more accurate. I'll add an example print statement for clarity and consider using depth from the surface, noting if it's at the centerline too.","format":"openai-responses-v1","index":0},{"type":"reasoning.encrypted","data":"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","format":"openai-responses-v1","id":"rs_0cff022bce7f611a016a35257520bc8196b94c43c2274f65c3","index":1}]},"native_finish_reason":"completed"}],"created":1781867892,"model":"openai/gpt-5.4-mini-20260317","object":"chat.completion","service_tier":"default","system_fingerprint":null,"usage":{"completion_tokens":42190,"prompt_tokens":25,"total_tokens":42215,"completion_tokens_details":{"accepted_prediction_tokens":null,"audio_tokens":0,"reasoning_tokens":40324,"rejected_prediction_tokens":null,"image_tokens":0},"prompt_tokens_details":{"audio_tokens":0,"cached_tokens":0,"cache_write_tokens":0,"video_tokens":0},"cost":0,"is_byok":true,"cost_details":{"upstream_inference_cost":0.18987375,"upstream_inference_prompt_cost":0.00001875,"upstream_inference_completions_cost":0.189855}},"provider":"OpenAI"}

generate a mathematics based python script to predict temper...