An online sparse least square support vector regression and particle filter for remaining useful life prediction of capacitive-type pressure transducer

Abstract This paper proposes a novel online sparse least square support vector regression without bias for forecasting capacitive type pressure transducer remaining useful life prediction (RUL). The proposed approach is based on mechanism and statistical knowledge. The impedance variables are applied to calibrate the correlation between pressure capacitance capacity degradation and the resistance value to improve the prediction precision. Furthermore, a particle filter algorithm is proposed to estimate the pressure capacitance impedance fade parameters, and then the probability density function of the predicted RUL value is obtained by Gaussian. Lastly, experimental results consider #5 and #6. The pressure capacitance data set verifies the hybrid RUL prognostics strategy with high accuracy and robust stability.