Novel MSMOTE Combined with LSDA for Imbalanced Fault Diagnosis

The data obtained from the current process industry is featured with high-dimensional, massive, nonlinear, and imbalanced, which poses a huge challenge to fault diagnosis in chemical processes. Facing data with these features, this article proposes a novel imbalanced fault diagnosis method based on Synthetic Minority Over-Sampling Technique Considering the Majority (MSMOTE) combined with Local Sensitive Discriminant Analysis (LSDA) (MSMOTE-LSDA). In the proposed MSMOTE-LSDA, MSMOTE is a novel method of over-sampling for minor samples to address the issue of data imbalance; LSDA is used for Dimensionality Reduction (DR), extracting critical features. Finally, AdaBoost.M2 classifier is used for fault classification to diagnose different types of faults. To verify the effectiveness of the proposed method, the Tennessee Eastman (TE) process with different types of fault data is chosen for simulation experiments. The simulation results show that the MSMOTE-LSDA method proposed in this paper has excellent fault diagnosis results in the imbalanced state.