Design and Evaluation of Conventional and NN-Based Methods for Corrosion and Fatigue Growth Rate Prediction

Prediction of defect growth rate on the mechanical and civil structure is critical and important. An autonomous prognostic and health management system is required for safe and cost-effective maintenance of critical components and structures. In this study neural network based approach and classical non-linear dynamics based approach such as phase space reconstruction is deployed on an edge device Nvidia Jetson Nano for prediction of defect growth rate. The defects are induced on DMR249A high strength low alloy steel. The corrosion damage is induced by potentio-dynamic test while fatigue is induced by cyclic fatigue test both with in-situ acoustic emission. The results indicate that the method which utilises neural network is more efficient than the classical non-linear dynamics based approach.