Aero engines remaining useful life prediction based on enhanced adaptive guided diferential evolution

Abstract

Remaining Useful Life (RUL) prediction is a key process for prognostic health management in almost all engineering real- world applications, especially which are in hazardous and challenging environments where the failures and disastrous faults cannot be avoided such as space vehicles and aircraft. This paper proposes a predictive approach based on our proposed algorithm Enhanced Adaptive Guided Diferential Evolution (EAGDE) is used to optimize the parameter selection of Support Vector Machine (SVM) to give high RUL prediction accuracy. The advantages of the proposed approach (EAGDE–SVM) are verifed using the popular benchmark C-MAPSS which describes the degradation of the aircraft turbofan engine datasets. The experimental study compares EAGDE–SVM with the basic SVM with randomized parameter selection and with an optimized SVM using three diferent optimization algorithms. Also, the EAGDE–SVM is evaluated against three popular classifer models that have been used in the comparisons of recent research. Diferent evaluation criteria of classifcation, prediction, and optimization aspects have been used, the obtained results show that the EAGDE is capable to achieve the lowest classifcation error rates and RUL high prediction accuracy through fnding the optimum values of the SVM param- eters with high stability and fast convergence rate.