Physics-Guided Ultrasonic Testing and a GAN-Augmented LSTM Framework for Multi-Phase Degradation Assessment of Ozone-Aged Synthetic Rubber

This study characterizes ozone-induced degradation in NR/CR rubber through an evidence hierarchy separating measurements, generated representations, and conceptual models. Five specimens were followed by Shore A hardness measurements over Days 0-16; the group mean increased from 53.39 to 61.45 Shore A (+15.10%). Five repeat ultrasonic acquisitions per endpoint showed decreases of 7.84% in normalized RMS and 15.20% in normalized energy from Day 0 to Day 16. These acoustic findings are acquisition-level descriptors subject to amplitude limiting and unverified specimen independence. Hardness-conditioned cGAN representations of Days 1-15 illustrate a potential pathway for temporal modeling; they do not establish the physical evolution of unmeasured acoustic states. LSTM and Bayesian-Weibull components remain exploratory temporal and model-conditional reliability frameworks. Missing intermediate acoustic measurements, independent RUL/failure labels, and complete computational lineage preclude validated RUL or quantitative posterior lifetime claims. The contribution is experimentally grounded degradation characterization with explicit boundaries on the use of generated information.