It is desirable to predict product reliability accurately in the early design stage, but the lack of information usually leads to the use of independent component failure assumption. This assumption makes the system reliability prediction much easier, but may produce large errors since component failures are usually dependent after the components are put into use within a mechanical system. The bounds of the system reliability can be estimated, but are usually wide. The wide reliability bounds make it difficult to make decisions in evaluating and selecting design concepts, during the early design stage. This work demonstrates the feasibility of considering dependent component failures during the early design stage with a new methodology that makes the system reliability bounds much narrower. The following situation is addressed: the reliability of each component and the distribution of its load are known, but the dependence between component failures is unknown. With a physics-based approach, an optimization model is established so that narrow bounds of the system reliability can be generated. Three examples demonstrate that it is possible to produce narrower system reliability bounds than the traditional reliability bounds, thereby better assisting decision making during the early design stage.
Analysis of Large Heterogeneous Repairable System Reliability Data With Static System Attributes and Dynamic Sensor Measurement in Big Data Environment