A Bayesian Design-Use-Stage Data Fusion Framework for Reliability Assessment Under Uncertainty: A Case Study From Machine Tool Rotary Table Manufacturing Enterprise

Abstract

Reliability is a fundamental performance metric for products. However, for capital-intensive products, accurate reliability assessment remains challenging because of scarce, heterogeneous, and uncertain data. To address this issue, this article proposes a Bayesian design-use-stage data fusion framework for product reliability assessment. The framework integrates machining error data and lifetime data of similar products from the design stage, as well as precision testing data and field lifetime data from the use stage. In the design stage, an initial precision reliability model is first constructed based on component machining error data and the error accumulation mechanism. Then, the Weibull distribution is used to describe the uncertainty in the lifetime data of similar products. Finally, the lifetime data of similar products are fused through Bayesian updating to revise the initial precision reliability model. In the use stage, the reliability model obtained from the design stage is used as prior information. The Wiener process is used to describe the uncertainty and time evolution characteristics of precision testing data. Based on this process, the precision testing data are converted into approximate lifetime data. The approximate lifetime data are then used to update the prior model. Finally, field lifetime data are further fused to revise the updated model. A case study on a CNC rotary table is used to verify the effectiveness of the proposed framework. By integrating heterogeneous models constructed from multisource data into a unified Bayesian structure, the proposed framework enables continuous and stage based reliability assessment under uncertainty.