A Complex Device State Feature Selection Method based on Improved Hybrid Algorithm

In this study, the authors aim to analyze how to improve equipment health state prediction by reducing the dimensionality of complex equipment condition monitoring features.In this paper, a hybrid algorithm that improves Filter and Wrapper based on mutual information is used, and simulated annealing and immunization algorithms are added to the search strategy for implementation.Experimental results show that the hybrid algorithm has advantages over the Filter and Wrapper algorithms alone, which not only improves the efficiency of feature selection while solving the feature redundancy problem well, but also improves the accuracy of state prediction greatly.This study is an effective means of fault prediction and health management intelligence, which can solve the problems of low accuracy and long delay of complex equipment health state prediction, and has a reference role in cost reduction and efficiency improvement for maintenance management departments.