Dynamic-weighted ensemble for fatigue crack degradation state prediction

Abstract

This paper proposes a prognostic framework for online prediction of fatigue crack growth in industrial equipment. The key contribution is the combination of a recursive Bayesian technique and a dynamic-weighted ensemble methodology to integrate multiple stochastic degradation models. To show the application of the proposed framework, a case study is considered, concerning fatigue crack growth under time-varying operation conditions. The results indicate that the proposed prognostic framework performs well in comparison to single crack growth models in terms of prediction accuracy under evolving operating conditions.

A long-term prediction approach based on long short-term memory neural networks with automatic parameter optimization by Tree-structured Parzen Estimator and applied to time-series data of NPP steam generators

“…These approaches are used in applications where the model of the degradation process exists and is not too complicated, e.g. models of fatigue crack growth [6], [7], of capacity degradation in Lithium-ion batteries [8], [9]. Alternatively, data-driven approaches utilize condition monitoring data collected from sensors to learn and predict the component or system behavior and degradation via statistical and artificial intelligent (AI) models, such as autoregressive integrated moving average (ARIMA) [10], artificial neural network (ANN) [11]- [14], neuro-fuzzy (NF) [2] and support vector machine (SVM) [15]- [17].…”

A framework for dynamic risk assessment with condition monitoring data and inspection data

“…Inspection data are collected by physical inspections performed by maintenance personnel [29]. They have been widely used for online reliability assessment.…”