Remaining Useful Life Prediction of Analog Circuit Using Improved Unscented Particle Filter

Abstract

This work aims to predict the remaining useful life of various components in the analog circuit under non-linear operating conditions. However, the failure process is described by sensed information that can be used to predict the failure time and then the remaining useful life of the system components. This work proposes an Improved Unscented Particle Filter (IUPF) with relative entropy. At first, the mathematical model of second order Sallen-key filter is developed to quantify the relationship between the input and predicted fault. Then, the relative entropy is introduced in unscented particle filtering to reduce the computation time and also to maintain the accuracy in the prediction. The IUPF uses relative entropy to measure the distance between actual and observed particles. The least value of entropy is considered to reduce the error in the proposal distribution and also to adjust the particles in the next step. The performance of improved unscented particle filter is validated for Sallen-key band pass filter and low pass filter. The proposed prognosis method is highly efficient in remaining useful life prediction.