Font Size: a A A

Research On Fault Diagnosis Methods For Chemical Industry Process Based On Tennessee Eastman Process

Posted on:2011-01-10Degree:MasterType:Thesis
Country:ChinaCandidate:X M WangFull Text:PDF
GTID:2178360305990555Subject:Control theory and control engineering
Abstract/Summary:PDF Full Text Request
Chemical industry plays a very important role in industrial production, it involves almost all aspects of people's lives. In recent years, along with the development of science and technology, especially computer technology, production units of chemical industry become bigger and bigger, technological processes become more complex, amounts of investment become larger and larger, automatic level becomes higher and higher. Therefore, the security and reliability of chemical process is more important. There is are many methods to improve system's, fault diagnosis technology is a very effective and important method to improve the security and reliability of chemical process. So research on fault diagnosis has vitally important value for chemical industry process.The thesis introduces the methods of fault diagnosis, Tennessee Eastman model and multivariate statistical methods in detail, and introduces the basic principle of multi-variable squared prediction error method, Hotelling T2 statistical method for fault detection and contribution map to identify fault. Main advantages and disadvantages of methods of principal component analysis and kernel principal component analysis are analyzed for fault diagnosis, and simulations are applied to Tennessee Eastman chemical process. Aimed at the shortcoming of principal component analysis which can not be applied to the nonlinear process, the modified kernel principal component analysis method is developed. An integrated fault detection method based on wavelet denoising and kernel principal component analysis is developed for the complex and nonlinear chemical industry process, it can solve the noise and random disturbances problem effectively. Aimed at the main disadvantage of large calculation quantity and long time-consuming process of kernel principal component analysis method, an integrated method based on feature vector selection and kernel principal component analysis is developed. Aimed at the disadvantage of poor fault identification capability of kernel principal component analysis method, according to calculate the contribution of each original variable for Hotelling T2 and SPE based on the gradient algorithm of kernel function, and identify contribution degrees of these correlative fault variables, a new idea is proposed. The application study of Tennessee Eastman chemical process proves the feasibility and superiority of the poposed methods, and achieves the expected results.
Keywords/Search Tags:fault diagnosis, Tennessee Eastman process, fault detection, multivariate statistical, principal component analysis, kernel principal component analysis, feature vector selection
PDF Full Text Request
Related items