A data-driven stochastic approach to model and analyze test data on fatigue response

A stochastic approach to model and analyze test data on the fatigue respons e of materials and laminated composites is developed. The developed approac h is 'data-driven' in nature. It has been customary to describe the fatigue response of metallic and laminated composite materials using a suitable pa rameter that can serve as the indicator and descriptor of damage accumulati on. In the present methodology, the fatigue response of the material is qua ntified by interpreting the corresponding material parameter to be an embed ded Markov process. The true probability distributions of the fatigue respo nse parameter are extracted from sample test data based on an analytical ap proach, and they are used in the formulation. To this end, the Maximum Entr opy Method is incorporated into the formulation. A recursive stochastic mat rix equation is developed based on the test data using the theory of reliab ility and Fokker-Plank-Kolmogorov equation. Application of the methodology to a composite laminate is demonstrated. (C) 2000 Elsevier Science Ltd. All rights reserved.