Comparison of KPI related fault detection algorithms using a newly developed MATLAB toolbox: DB-KIT

Data driven fault diagnosis has been of real interest for the industry in recent years. Optimization of the existing fault detection algorithms with respect to significant monitoring indices are practically and economically preferred. In this paper, a recently developed Data Based Key-performance-Indicator-related fault detection Toolbox (DB-KIT) is described. The aim of DB-KIT is to provide a general platform for key performance indicator (KPI) related fault detection algorithms validation and comparison, and for further studies as well as educational purposes. Considering that MATLAB is probably one of the most popular programming environments in the fault diagnosis community, the related functions and a user-friendly graphical interface are programmed in MATLAB. Moreover, a demonstration of the 14 integrated linear and nonlinear algorithms in the toolbox is provided. The simulation is run with the Tennessee Eastman process dataset, which is a widely accepted benchmark challenge in the fault diagnosis field. Finally, the evaluation indices, including the false alarm rates, fault detection rates and algorithms efficiency, are further compared and discussed.