An improved SVM integrated GS-PCA fault diagnosis approach of Tennessee Eastman process

In modern industry, fault diagnosis and process supervision are very important in detecting machinery failures and keeping the stability of production systems. In this paper, a multi-class support vector machine (SVM) based process supervision and fault diagnosis scheme is proposed to predict the status of the Tennessee Eastman (TE) Process. After preprocessing the collected data, principal component analysis (PCA) is firstly used to reduce the feature dimension. Then, to increase prediction accuracy and reduce computation load, the optimization of SVM parameters is accomplished with the grid search (GS) method, which generates comparable classification accuracy to genetic algorithm (GA) and particle swarm optimization (PSO) while being more efficient than the latter two algorithms. Finally, to demonstrate the effectiveness of the proposed SVM integrated GS-PCA fault diagnosis approach, a comparison is made with other related fault diagnosis methods.