Research on Decline Pattern Recognition of Hydraulic System

Abstract

This paper investigates the method of identifying decline patterns based on condition monitoring data. The paper firstly investigates the process of decline pattern recognition from the inevitability of system decline; then uses the time-domain feature extraction method to extract features for hydraulic system health characteristic parameters; finally uses the ML-KNN algorithm and the recurrent neural network-based algorithm to identify decline patterns on the experimental data, and compares the two algorithms. The results demonstrate that the recurrent neural network-based model has stronger generalization ability compared with other classification algorithms, and characterizes the data more accurately and fits the distribution of the data more realistically.