Location:Home > Engineering science > Control Science and Engineering > Research of Fault Detection Method for Batch Processes Based on Clustering
Details
Name

Research of Fault Detection Method for Batch Processes Based on Clustering

Downloads: []
Author
Tutor: ShiHongBo
School: East China University of Science and Technology
Course: Control Science and Engineering
Keywords: fault detection,Gaussian mixture model,Spectral Clustering,batch processes,multi
CLC: TP277
Type: Master's thesis
Year:  2014
Facebook Google+ Email Gmail Evernote LinkedIn Twitter Addthis

not access Image Error Other errors

Abstract:
The article first presents a comprehensive description of the background of the fault detection and the fault detection methods about traditional and Batches, including their advantages and disadvantages. Secondly, it describes the characteristic of batch data and bath process, a shortest length approach is used to solve the problem of unequal data, while a method of hybrid unfolding of a multi-way data matrix is used to eliminate data estimate problem. Finally, detailed analysis and introduction for the characteristic of multiple operation phases.A new novel of PCA-MGMM method is proposed in this article to handle multiple operation phases. At first, apply PCA sequentially with Gaussian Mixture Model (GMM) to achieve a low-dimensional representation of the original data space and to construct several clusters which represent different operation phases in the feature space, and then, the modified algorithm is adopted to automatically optimize the number of Gaussian components and estimate their statistical distribution parameters. Finally the online monitoring is guaranteed to be continuous by using a global probability index. The effectiveness and flexibility of the proposed method is validated through an empirical study on a real semiconductor process.A novel method of a new phase segmentation strategy is proposed in this article to handle batch processes with multiple operation phases. The method can make stage classification accurately and avoid to falling into local optimum, for this, the model precision is improved and the fault detection is accurate. At first, apply a method of Spectral Clustering with Figure divided guidelines to construct several clusters which represent different operation phases in the feature space, then establish models for each class respectively by PCA. Finally, monitor online by choosing the most suitable model. The effectiveness and flexibility of the proposed method is validated through an empirical study on penicillin simulation process. The specific experimental results illustrate the effectiveness of the approach.
Related Dissertations
Last updated
Sponsored Links
Home |About Us| Contact Us| Feedback| Privacy | copyright | Back to top