Fault Diagnosis and Model Learning for Hybrid Dynamic Systems with Incomplete Models

Fault diagnosis and model learning for hybrid dynamic systems with incomplete models is a particularly challenging problem. In this paper, model-based and data-driven methods are integrated to handle this problem. Firstly, ‘unknown fault’ is detected with model based approach of Particle Filters(PFs), ‘unknown fault’ is detected with the sum of unnormalized weights of samples in PFs, meanwhile, the data of observations are collected into Unknown-Dataset for further use in K-Means based cluster method. Secondly, a K-Means based cluster method is employed to handle Unknown-Dataset. Thirdly, a multi layer perceptron(MLP) is used to learning the models for each unknown model based on the results of K-Means. The proposed approach is testified on a simplified hybrid systems.