A novel fault measure and early warning system for air compressor

Air compressor is a main power equipment, and improving its reliability and security is important to enhance the competitiveness of pneumatic companies. This study proposes a real monitored data-based fault measure and early warning system for air compressor, which can guide field workers in economically and safely operating the air compressor by finding and dealing with faults ahead of time. A data-driven multivariate state estimation technique (MSET) is utilized to construct the non-parametric fault early warning model under normal operating condition, principal components analysis (PCA) and back propagation (BP) method are combined to construct the dynamic memory matrix and simplify the model. With the pre-built PCA-BP-MSET model, optimal values of the observed vectors are estimated and a deviation function is defined to calculated the deviation sequence between the estimated and observed vectors, where the qualitative-quantitative combined AHP method is adopted to determine the weights of the monitored variables. With the deviation sequence, the sliding window statistical method is applied to determine the fault early warming threshold, which can be used as a benchmark to measure and warn the fault of air compressor. The fault warning information of an air compressor is given out once the deviation degree of some observed vector exceeds the threshold, thereby realizing the function of fault measure and early warning in advance. In addition, an automatically updating method is proposed to ensure the accuracy of the PCA-BP-MSET model in case of the changes of running states. For verification, a real fault case of an air compressor has been chosen to verify the effectiveness of the proposed method. Result shows that the PCA-BP-MSET-based method can catch the fault developing process timely, which is appropriate for the faults prevention and rapid troubleshooting of air compressor. (C) 2018 Elsevier Ltd. All rights reserved.