A novel exponential model for tool remaining useful life prediction

Implementing proactive maintenance strategies based on condition prediction for cutting tools can reduce expensive, unscheduled maintenance events. This work proposes an novel exponential model to predict the Remaining Useful Life (RUL) of cutting tools. Firstly, a new monitoring indicator named second-order derivative of health index (SDHI) is constructed, and on this basis, a 3σ interval-based first predicting time (FPT) adaptive selection method is proposed to correlate the observable SDHI with the unobservable tool wear rate, automatically determines the abnormal tool wear state without human intervention. Secondly, the integration of the Bayesian inference mechanism with the expectation maximization (EM) algorithm enables the achieving of real-time iterative updates for model parameters. Thirdly, To reduce stochastic errors while predicting the RUL, particle filtering and probability density function (PDF) are applied to handle prediction uncertainty. The experimental findings obtained from the milling experiments demonstrate that the proposed model exhibits robust adaptability to various cutting conditions, thereby leading to enhanced RUL prediction performance.