A process Fault Detection method Based on PCA and linear regression

PCA is a common fault detection method, which has good performance in fault detection. But it is difficult to distinguish the specific fault location. This paper established a linear regression model through the measurement value of different points, and used R-squared to evaluate the model to eliminate models with poor fitting. In this paper, the above model was used to simulate the data set of Tennessee Eastman process, and some models obtained can detect the fault and reduce the range of failure. This paper provided a new fault detection method applied to train non-faulty samples.