مدل فرآيند وينر براي داده هاي فرسايش‌ بريده شده ازچپ‌

Expectation-Maximization Algorithm , Left-Truncated Data , Multi-variate Degradation , Ran‎dom Drift an‎d Diffusion , Remaining Useful Life Pre-diction , Wiener Process

Wiener Process Model for Left-Truncated Degradation Data

آمار

Given the growing need for accurate prediction of the remaining useful life (RUL) of equipment in complex engineering systems, modeling degradation, especially under conditions of left-truncated data, is of paramount importance. In this research, a novel model based on the Wiener process with ran‎dom drift an‎d diffusion is presented, which is capable of simulating degradation paths an‎d estimating the RUL with appropriate accuracy. The proposed model, by con-sidering the effects of ran‎domness in the drift rate an‎d dispersion, an‎d by using the inverse Gaussian distribution, covers the heterogeneity between units an‎d the correlation between parameters. Furthermore, this model is capable of analyzing left-censored data by accounting for the unknown initial time of degradation. To estimate the parameters, the Expectation-Maximization (EM) algorithm an‎d its developed version were used, an‎d by utilizing the first hitting time distribution, the RUL was calculated analytically. The results from real degradation data of GaAs lasers show that the proposed model has more accurate performance com-pared to simpler models.

5