Application of LDA and SVM method in fault diagnosis of chemical process

Based on the variety and high dimensionality of the data and also the characteristic of repetitive risks in complicated chemical processes, a fault diagnosis method based on linear discriminant analysis (LDA) and support vector machine (SVM) was proposed combined with the grid search (GS) and K-fold cross validation (K-CV) theory. In this method, LDA is used to map the normal operation and five types of fault data by vectorization, compressing the dimensions of the feature space, extracting and reconstructing the feature information. Subsequently, the optimal parameters of SVM model are established for the processed data by using GS and K-CV to diagnose faults. In this work, the introduced LDA_SVM mixed model is compared with SVM and PCA (principal component analysis)_SVM fault diagnosis models, where the new method proved to be superior with fast convergence, high recognition rates and robustness. In this work, it is also showed that accuracy of diagnosis results for the six types of running modes in chemical process has reached 93.9% for the proposed method. © 2020, Editorial Board of "Journal of Chemical Engineering of Chinese Universities". All right reserved.