Answers are grounded in this paper and cite their sources.
Composite structures are susceptible to fatigue damage under complex cyclic loading, and accurate prediction of stiffness degradation is important for structural condition assessment and life management. In this study, constant‐amplitude and two‐stage variable‐amplitude fatigue tests were conducted on four carbon‐fiber composite laminate configurations. A physics‐constrained sequential‐learning framework, termed PI‐LSTM‐VA, is developed for full‐curve stiffness degradation prediction. The model predicts a nonnegative degradation intensity and integrates it over an endpoint‐independent logarithmic cycle coordinate. A parameter‐supervised switch module represents additional posttransition degradation while preserving stiffness continuity. On four held‐out variable‐amplitude specimens, specimen‐mean MAE and RMSE were 0.008238 and 0.010016, respectively. RMSE decreased by 21.8% relative to the shared constrained baseline and by 39.2% relative to direct LSTM. Repeated specimen partitions confirmed lower average error, although dispersion increased and some predictions worsened. Separate five‐fold constant‐amplitude holdout tests yielded specimen‐mean RMSE of 0.015918 across 38 specimens and supported applicability within the investigated domain.
No takes yet. Share an insight, caveat, or question.
Wang et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: