Fault detection method based on modified local gravity and its application in multimodal process

The modern industrial process data with characteristics, such as multi-center and different structures, have challenged the traditional multivariate statistical process monitoring methods. To address this problem, a multimodal process fault detection method based on modified local gravity (MLG) is proposed in this paper. First, the method considers each data in the dataset as a particle, whose quality and gravitational constant are determined based on the distance information from the data to its nearest neighboring data, so as to obtain the partial force of the local variable. Second, the multimodal data are transformed into data that obey a single multivariate Gaussian distribution by calculating the sum of the partial forces on each local variable. Third, a principal component analysis (PCA) process monitoring model is established based on these local gravitational data. Finally, the effectiveness of the MLG-PCA method is verified with the numerical example and Tennessee Eastman process. The simulation results show that the fault detection rate of MLG-PCA is better than the rates of traditional methods.