Aiming at the problems of high acceleration test cost and the lack of systematic modeling and uncertainty quantification in existing virtual methods for the life verification of high-reliability MEMS sensor complete machines, this paper proposes a virtual assessment framework based on Physics of Failure (PoF). This method obtains local stress through multi-scale digital prototyping and thermal-vibration coupling simulation, and constructs a modular PoF model library to predict the life of key failure modes of critical devices. To quantify the uncertainty in manufacturing processes and service environments, a physical source-driven hybrid probability distribution modeling strategy is introduced. On this basis, a two-level competitive failure model consisting of "intra-device multi-mode fusion" and "inter-device first-passage failure" is established to realize the prediction of the Time-to-Failure (TTF) of the complete machine. Taking the MPU9250 as an example for verification, the independently conducted thermal-vibration combined physical test shows that the digital prototype has satisfactory accuracy (case temperature deviation shall not exceed ± 0.2℃, modal error < 5%), the predicted TTF of the complete machine is 12 721 h, and the weak link is accurately identified. The results show that the proposed method can effectively reduce the dependence on physical tests and improve the efficiency and credibility of life assessment.