Application of PSO-LSSVM and hybrid programming to fault diagnosis of refrigeration systems

Abstract

Fault detection and diagnosis (FDD) in refrigeration systems is of great importance for ensuring better equipment reliability and energy efficiency. Although numerous researches studied FDD algorithms and methodology, there is still a lack of mature commercial software in this field. This study presents a novel hybrid model by introducing particle swarm optimization (PSO) into least squares support vector machine (LSSVM) for parameter optimization to overcome the blindness of parameter selection, and proposes a novel idea of hybrid programming, where MATLAB is used to implement the FDD strategy and LabVIEW is employed for interface creation, to take the advantage of both sides. The hybrid programming is carried out through MATLAB script node, and an FDD platform for refrigeration systems is established. The strategy and the platform is validated using experimental data for a centrifugal chiller, where seven typical faults were investigated. The results show that the proposed PSO-LSSVM achieves an overall diagnostic accuracy of 99.70%, drastically improved (8.81%) from that of the LSSVM without optimization. The idea of hybrid programming is feasible for the establishment of an operable and highly integrated FDD platform with user-friendly interface and extendable functions. The practice of the idea also promotes the possibility of combining FDD with system control for better field applications.

Additional information

Funding

This project was supported by the National Natural Science Foundation of China (NSFC) under Grant No. 51506125.