A Process Monitoring Model Based on Data Whitening and Dynamic Radius Support Vector Data Description Method

Due to the large number of equipment and complex technical processes involved in industrial production, various faults are inevitable during the production process, which may affect the quality of the final product. This paper proposes a method that uses data whitening technology to process data, which can efficiently reduce the dimension of the original data and an eliminate redundant information. And then establish a dynamic radius model based on support vector data description (SVDD) method, which retains useful features and achieves accurate monitoring result of industrial process faults. We name the method as S-DRSVDD. To evaluate the new model, comparisons have made between it and other advanced models in a typical industrial fault dataset, Tennessee Eastman Process. The new model shows superiority from the perspectives.