An integrated method of independent component analysis and support vector machines for industry distillation process monitoring

Abstract For the complex operation and multi-loop control in the industry distillation process, the diagnosis of the complex fault has become more and more difficult. An integrated method of independent component analysis (ICA) and support vector machines (SVM) is proposed to detect and diagnose industry distillation process faults. The ICA is used for feature extraction and data reduction from original features. And the ICA statistics I2, I e 2 and SPE are proposed as on-line fault detecting strategy. The principal component analysis is also applied in feature extraction process in comparison with ICA does. In this paper, the multi-classification strategy based on binary-tree SVM is applied to perform the faults diagnosis. Various scenarios are simulated using actual fault datasets of the butadiene industry distillation process, and the proposed method can effectively detect and diagnose faults when it compares to methods of original SVM and PCA–SVM in terms of diagnosis accuracy and time.