Python code
Cold Plate code
Python scripts from 3_SourceCode/ that train the POD-NN surrogate and make the figures. Run 01_train_model.py first; the plotting scripts read what it saves in 2_AnalyzedData/. Packages are listed in requirements.txt.
_utils.py
3_SourceCode · 7.5 KB
Shared helpers: loads sim_*.npz and model_setup.json, builds, trains and saves the POD-NN.
01_train_model.py
3_SourceCode · 1.5 KB
Trains the POD-NN surrogate and reports test-set R², RMSE and MAE.
02_pod_modes.py
3_SourceCode · 1.3 KB
Plots the first 6 POD mode shapes on the yz-mid plane.
03_pod_reconstruction.py
3_SourceCode · 2.4 KB
Original field vs 1-PC and 4-PC reconstructions.
04_pod_variance.py
3_SourceCode · 1.4 KB
Cumulative variance explained vs number of POD modes.
05_mse_history.py
3_SourceCode · 908 B
Training and validation loss per epoch.
06_time_crossover.py
3_SourceCode · 3.8 KB
CFD vs POD-NN wall-clock time.
07_single_prediction.py
3_SourceCode · 3.8 KB
Prediction, truth and error for 5 test samples.
08_mean_error_field.py
3_SourceCode · 1.8 KB
Mean absolute error at each point over the 400 test samples.
09_dataset_size_sensitivity.py
3_SourceCode · 4.8 KB
Test error vs training-set size, split into POD and NN error.
10_pod_alpha_sweep.py
3_SourceCode · 3.5 KB
Sweeps each POD coefficient and plots the fields.
11_pearson_correlation.py
3_SourceCode · 6.2 KB
Pearson correlation between POD coefficients and inputs/outputs.
12_rth_latent_space.py
3_SourceCode · 3.7 KB
Thermal resistance over the PC3-PC4 plane.
requirements.txt
3_SourceCode · 47 B
Python packages the scripts need.