| In the past, reliability is usually quantified with sufficient information available. This is not only time-consuming and cost-expensive, but also too late for occurred failures and losses. For solving this problem, the objective of this dissertation is to predict product reliability in early design stages with limited information. The current research of early reliability prediction is far from mature. Inspired by methodologies for the detail design stage, this research uses statistics-based and physics-based methodologies by providing general models with quantitative results, which could help design for reliability and decision making during the early design stage. New methodologies which accommodate component dependence, time dependence, and limited information are developed in this research to help early accurate reliability assessment. The component dependence is considered implicitly and automatically without knowing component design details by constructing a strength-stress interference model. The time-dependent reliability analysis is converted into its time-independent counterpart with the use of the extreme value of the system load by simulation. The effect of dependent interval distribution parameters estimated from limited point and interval samples are also considered to obtain more accurate system reliability. Optimization is used to obtain narrower system reliability bounds compared to those from the traditional method with independent component assumption or independent distribution parameter assumption. With new methodologies, it is possible to obtain narrower time-dependent system reliability bounds with limited information during early design stages by considering component dependence and distribution parameter dependence. Examples are provided to demonstrate the proposed methodologies. |
