Figures & data
Figure 1. Schematic of the studied structure.
Figure 2. Calculation procedures (a) inversing the direct model and (b) finding an inverse model.
Figure 3. Example of the response of the structure (direct problem).
Figure 4. Detailed view of the response.
Figure 5. Standard architecture for the direct problem.
Figure 6. Example of the prediction and prediction error of the model for the direct problem.
Figure 7. Example of results after inversion (for an unknown heat flux sequence).
Figure 8. General architecture of the tested neural networks for the inverse model (DIM model).
Figure 9. Sum of prediction errors for various time delays.
Figure 10. Predicted heat flux and prediction errors during the test phase.
Figure 11. Detailed views of the predicted heat flux and the prediction error for the rectangular pulse.
Figure 12. Prediction error during the validation phase.
Figure 13. Detailed views of the predicted heat flux and the prediction error when noise is added to the temperatures.
Figure 14. Coupled neural models for the final validation.
Figure 15. Comparison of actual and estimated heat fluxes when coupled neural models are used.
Figure 16. Histogram of the RMSE for the 100 validation cases.
