On the random vibration-based progressive fatigue damage detection and classification for thermoplastic coupons under population and operational uncertainty

Abstract

The problem of vibration-based progressive fatigue damage detection and Fatigue State (FS) determination for thermoplastic coupons is experimentally addressed under significant population and experimental uncertainty. The study involves 13 coupons subjected to tension–tension fatigue testing with interruptions at 10,000-cycle intervals. Utilizing non-parametric random vibration analysis and a robust, uncertainty-tolerant methodology, the research employs Multiple Model Representations of uncertain dynamics through parametric ARMA(AutoRegressive-Moving Average)-type random vibration response modeling. Results, validated by ultrasonic C-Scan testing, reveal (a) the significance of uncertainties, (b) a resonant structural frequency indicating fatigue accumulation, (c) remarkable detection performance, achieving 100 % accuracy even at 10,000 cycles, and (d) FS determination with over 80 % accuracy across all coupons. This study offers promising avenues for automated fatigue damage detection and FS determination in thermoplastic structures, eliminating the need to interrupt their operational cycles.