Prognosis of fatigue induced stiffness degradation in GFRPs using multi-modal NDE data

Abstract Prediction of expected life of a composite structure especially at the initial stages of degradation is challenging owing to inherent heterogeneity and lack of robust damage growth models. This paper focuses on prognostic study of matrix stiffness degradation in glass fiber reinforced polymers (GFRP) subjected to fatigue testing using data from multi-modal nondestructive evaluation (NDE) techniques, specifically the optical transmission and guided wave sensing. Combining information from multiple sensors exploits advantages of signal complementary and hence effectively improve damage growth modeling and prediction in composites. However, matrix stiffness inferred from two independent NDE techniques varies owing to differences in their sensitivity, measurement noise or model discrepancy, often leading to inconsistent and inaccurate reliability assessment. A joint likelihood updation technique is therefore proposed in existing particle filtering (PF) framework which enables dynamic optimization of Paris-Paris model parameters at every time step by discarding noisy or biased measurements. Comparison of stiffness prediction using multi-sensor data with prognosis results on single sensor or average measurement demonstrates the benefit of joint likelihood based prediction of residual stiffness. An additional advantage of the proposed approach towards reduction of particle count in existing particle filtering framework is discussed, thereby lowering prediction time and computation resources. Overall, multi-sensor NDE and prognosis methodology is discussed for reliable assessment of fatigue life in GFRP composites structures.