Figure 3
The two vertically stacked graphs are labeled “a” and “b”. In graph (a), titled “Influence of Physics–Data Weighting Parameter lambda on Degradation Trajectory”, the horizontal axis is labeled “Normalized Operational Time”, ranging from 0.0 to 1.0 with an interval of 0.2, and the vertical axis is labeled “Degradation State x (t)”, ranging from 0.0 to 0.8 with an interval of 0.2. Multiple curves are plotted corresponding to different lambda values: lambda equals 0.01, 0.05, 0.1, 0.5, and 1.0, along with a dashed black line labeled “Analytical Reference”. All curves start at (0, 0.0) and increase monotonically, closely overlapping with the analytical reference throughout the range. The trajectories show rapid initial growth that gradually slows, approaching saturation near 0.95 as time approaches 1.0, with only minor visible differences among lambda values. In graph (b), titled “N R M S E Variation Across lambda Values”, the horizontal axis is labeled “Physics–Data Weight lambda” and includes values from 0.01 to 1.0 with an interval of 0.2, while the vertical axis is labeled “N R M S E”, ranging from 0.0005 to 0.0040 with an interval of 0.0005. The plotted line with circular markers shows that the error decreases from around 0.0009 at lambda equals 0.01 to a minimum near 0.00035 at lambda equals 0.1, then increases steadily to about 0.0042 at lambda equals 1.0. Note: All numerical values are approximate.Sensitivity analysis of the physics–data weighting parameter λ. (a) Learned degradation trajectories for different λ values, showing preserved monotonic and bounded behaviour. (b) Variation of NRMSE with λ, indicating a balanced regime at λ = 0.1