Research On The Key Techniques Of FMS Fault Diagnosis

Flexible Manufacturing System(FMS), as an important part of CIMS, is of great significance to the development of intelligent manufacturing. Its trouble-free working time is the premise of guarantee the production efficiency. Therefore, a fault diagnosis system with intelligent algorithm is an important part of the FMS. Under the background of the national "Made in China 2025" plan and from a manufacturer of quantity to one of quality, the research of FMS fault diagnosis is of great significance. With the development of data mining technology, database technology, computer technology and signal processing technology, FMS fault diagnosis has a new solution. Based on the analysis of FMS fault characteristics and the problems of existing diagnosis methods, this thesis studies the key technologies of FMS fault diagnosis combined with the data mining technology. And this thesis develops a prototype system of remote fault diagnosis based on data mining.This thesis mainly carried out the following research:(1) According to the characteristics of FMS and its failure, from the viewpoint of system, the fault can be divided into processing system fault, the logistics system and control system failure, and sets up a integration architecture for fault diagnosis system;(2) According to the characteristics of the logistics system, puts forward a fault diagnosis strategy based on position/velocity control(i.e., movement) for the logistics system. Then according to the characteristics of the machining system failure, this thesis proposes a hierarchical diagnosis strategy, which divides the diagnosis of machining system into primary diagnosis and accurate diagnosis. And establishes the diagnosis model;(3) According to the characteristics of the primary diagnosis for the machining system, puts forward to a mathematical model based on Bayesian Networks(BN) for primary diagnosis. Based on the introduce BN theory, studied the structure of the Bayesian Network and its algorithm for diagnostic strategy, and the method shows effectiveness through examples;(4) To deal with the low recognition rate of BP Neural Network under the condition of small samples of early diagnosis for the accurate diagnosis of the processing system, a mathematical model of Support Vector Machine(SVM) is proposed. O n the basis of briefly introduce the concept of SVM, the extension strategy of SVM is mainly studied. And the recognition rate of SVM is higher than the BP Neural Network under the condition of small sample through experiment;(5) Using the diagnosis model proposed in this thesis, a prototype system of remote fault diagnosis based on data mining was established.