Physical-Principle Based Extension of Attributes for Plant Fault Detection

Process monitoring is of importance to maintain process safety, reliability, performance and cost efficiency. This work presents a hybrid fault detection approach that combines physical-principle process knowledge with data-driven method. Process knowledge is presented as extended attribute variables added into original process dataset. This extended dataset is expected to have better fault detection performance than the original dataset by using the same common data-driven fault detection approach. In this paper, the method was applied to the kernel PCA based fault detection and tested on a benchmark problem of a chemical plant.