Figure 8
The image contains two graphs. The first graph on the left shows the training loss history plotted on a logarithmic scale, with the x-axis representing the number of epochs and the y-axis representing the total loss. The loss decreases steadily over 3,000 epochs. The second graph on the right shows the long-term extrapolation of the volume-averaged property retention over a 10-year period. The x-axis represents time in days, and the y-axis represents the volume-averaged property retention. The PINN-predicted degradation curve remains smooth and physically consistent, extending short-term measurements without oscillations or divergence. All values are approximated.(a) Training loss history plotted on a logarithmic scale, showing steady reduction of the total loss over 3,000 epochs as the PINN simultaneously fits the data and satisfies the governing physics. (b) Long-term (10-year) extrapolation of the volume-averaged property retention . The PINN-predicted degradation curve remains smooth and physically consistent, extending short-term measurements without oscillations or divergence