Research And Application On Fault Diagnosis With DDAG-Based FDA Method

In the industrial process, the effective way of process monitoring and fault diagnosis is the key to ensure production safety, improve product quality and increase economic benefit. As the development of industry technology and computer’s wide application, a large of field data can be used to guide production. Thus, how to use the deep-level information to improve the performance of the process monitoring has become a major research topic in the field of process control.After studying the development and applications of multivariate statistical analysis, the thesis describes the FDA technique deeply based on injection molding process. The thesis consists of the following two parts:A novel fault diagnosis method, FDA approach based on decision directed acyclic graph (DDAG) is presented in order to solve discrimination failure of the FDA method for large class numbers. DDAGFDA method combines of the normal operation data and specific faults and proceeds fault diagnosis using DDAG and the limit calculated by KDE. The proposed method is proved to be effective and feasible by the application in diagnosing of the process data collected from an injection process.In order to realize a variety of statistical methods in the injection molding process, a fault diagnosis platform based on FDA method is developed for the injection system using C#as the research and development tool. The whole structure of the platform includes the configuration database and historical database. On the off-line modeling platform, the data can be selected to calculate the corresponding model parameters. And the fault diagnosis platform can perform fault diagnosis for both historical data and online data. At the same time, it can show the result to the field maintenance engineer as the reference and theoretical basis for maintaining and repairing machines.