Fault diagnosis system based on RBF-PCA

Subjects: Dynamic and Electric Engineering >> Engineering Thermophysics submitted time 2024-01-06

Yin Shengji

Abstract： With the promotion of "Industry 4.0" and "Intelligent Manufacturing 2025", intelligent equipment is becoming the future development direction of high-end equipment. In this process, the fault diagnosis and identification system, as an important research field, can suppress the unbalanced vibration of machinery in real time, so as to realize the autonomous health of equipment. Such a system can not only improve the operational efficiency and productivity of the equipment, but also reduce the cost and risk of equipment maintenance. Therefore, the research and development of such systems is of great significance to promote the development of intelligent manufacturing and high-end equipment manufacturing.
Condition monitoring and fault diagnosis of large rotating machinery is a necessary means for the production management of modern enterprises, through scientific monitoring and diagnosis, the efficient, safe and reliable operation of the equipment can be realized, and the sustainable development of the enterprise can be provided with a strong guarantee. This project takes the typical vibration faults of the rotor system of rotating machinery as the main body of research, considers the analysis of them, and establishes an intelligent identification system for the diagnosis of typical faults of the rotor system of rotating machinery. The main research contents and conclusions of this paper are as follows:
(1) This paper summarizes and explores the current mechanical intelligent fault diagnosis algorithm and the research method of using neural network for fault classification by consulting relevant literature, and finally decides to choose RBF neural network for fault diagnosis and classification.
(2) PCA dimensionality reduction technology is used to reduce the dimension of the data after extracting features, so as to reduce the dimension of the data to solve the problem of "dimensionality disaster" and make the data more reliable;
(3) The fault classification software is established by using the radial basis neural network in the artificial neural network, and the fault is identified by the classifier and verified by relevant experiments to complete the classification software design;
(4) A test bench was set up to collect data for relevant experimental verification, and the experiment showed that the diagnostic accuracy reached more than 85%.

Peer Review Status:Awaiting Review

 Hits 658  Downloads 172  Comment 0