Reliability assessment for load‐sharing systems with exponential components using statistical expansion as a correction

Abstract

Among recent system models, one specific type of system is generally used to model the dependence among components. Components are connected parallel in such systems as they fail one by one and are supposed to share the system work load. The model is thus referred to as the load‐sharing system model. Despite the availability of extensive reliability assessment methods for different systems, load‐sharing systems have not received enough attention from the scholars who have studied reliability assessment so far. Load‐sharing systems are generally designed for high levels of reliability. Therefore, tests for such systems can be expensive and time consuming. Limitation on resources always leads to small test sample sizes. This increases the difficulties associated with obtaining an accurate and robust system reliability assessment result. This paper proposes a novel assessment method for a certain type of load‐sharing system with components following exponential lifetime distributions. Based on the parameter estimation of the system reliability model, we introduce the Winterbottom‐Cornish‐Fisher asymptotic expansion method for implementing a correction of normal approximation. We demonstrate the accuracy of our method through a series of examples and simulation studies.

Correlation‐driven multivariate degradation modeling and RUL prediction based on Wiener process model

“…For specific, the failure of one component may cause/induce/accelerate the degradation/failure of other components. 13 One typical scenario is the load-sharing system, 14 where the failure of a component is likely to accelerate the degradation of the surviving components. 15 (ii) A system with F I G U R E 1 Diagram of the bogie structure and stripping failure on the wheel treads multiple failure modes, 16 in which failure modes dependencies are based on the physical, logical, functional, or environmental connections of the components.…”

Prediction intervals for load‐sharing systems in accelerated life testing

“…Since W * and W are independent, we obtain for all 0 Proof of Theorem 2. The one-sided confidence intervals given by (17) and 18…”

“…Most of these results are based on sequential order statistics. [6][7][8][9][10][11] Several approaches assume exponential or Weibull distribution [12][13][14][15][16][17] or other parametric distributions 18 for the lifetime of the components or the time between successive failures. Semiparametric approaches are given by Kvam and Peña 1 and Deshpande et al 19 Kong and Ye use a cumulative exposure model 20 as it was used in step-stress accelerated life testing and provide a method 21 for testing the exponential distribution for the time between successive failures.…”