Adaptive Remaining Useful Lifetime Prediction of Magnetic Head under Varying Stress Conditions

Current Remaining useful life studies mainly focus on the linear degradation case. A few nonlinear degradation models often rely on parameters learning process from a batch of samples obtained from the same population. This result in large estimation biases and uncertainties in term of RUL prediction under varying stress conditions. To address this problem, this paper proposed an adaptive RUL prediction method only using the observed head wear data. First, an exponential item is incorporated into a power law wear model to build a nonlinear wear model under varying stress conditions. Second, PF method was used to update the model parameters dynamically. Third, an adaptive strategy was developed based on expectation-maximization algorithm to update these initial values of parameters of degradation model recursively. Finally, a real-life data set was used to verify the effectiveness of our proposed approach, and the results show that the proposed approach can improve the prediction accuracy significantly.