本文对系统共因失效分析中涉及的概念，诸如共因失效的“根本原因”（root cause）和“耦合机制”（coupling mechanism）等，从载荷的随机性导致零件失效相关性这一观点给予了解释。同时还提出和讨论了“固有相关性”、“偶然相关性”、“绝对共因失效”和“相对共因失效”等概念。借助多维应力-强度干涉分析和对干涉模型进行的离散化处理，建立了根据零件/系统失效数据预测系统共因失效概率的简单、实用的离散化的模型。

从可靠性数学的角度，以环境载荷与零件性能的特征及其相互关系为背景，探讨了k/n（F）冗余系统共因失效的原因与机理，并致力于冗余系统共因失效概率预测的精确方法。分析表明，对于各零件处于同一载荷环境的系统，环境载荷的随机性是导致系统共因失效的最基本的原因。在一般情况下，环境载荷和零件性能都是随机变量，因而各类冗余系统都不同程度地存在共因失效这种失效相关性。通过机理分析，原则上可以说任何冗余系统失效概率（无论有无共因失效存在）都可以采用由环境载荷-零件性能干涉分析建立的具有普适性的模型进行预测。

根据两级循环载荷作用下剩余疲劳寿命分布规律的实验规律，以描述剩余寿命分布变化的数学模型为基础，提出了一个根据载荷历程作用下结构/零件状态变化预测随机载荷下疲劳可靠度的方法。该方法的基本思想是，在随机载荷作用下，结构/零件的剩余寿命分布参数（均值和标准差）将不断变化。在每一级或每一个载荷循环作用下，剩余寿命分布参数的变化量取决于当前材料状态（该载荷开始作用时的材料性能参数）及载荷参数。以这样的变化规律为基础，就可以根据已知的材料或零件的原始P-S-N曲线，借助剩余寿命分布和载荷循环数-疲劳寿命干涉分析计算随机载荷作用下的疲劳可靠度。

The present paper is mainly concerned with pipe system failure probability estimation. For this purpose, a multiple component dependent failure (MCDF) model is introduced to deal with dependency among pipe segment failures. The MCDF model is directly derived from load-strength interference analysis with the underlying notion that stochastic environment load brings about failure dependency. The influence of pipe segment failure dependency on pipe system failure probability is addressed and a modified system failure probability model is presented. It is also shown that, with regard to a pipe system, the multiple failures of pipe segment leaks might lead to a pipe system disabling leak or even to a break. The probability of pipe system disabling leak or break induced by the imultaneous failures of multiple segment leaks is estimated by means of the MCDF model. The estimation results indicate that pipe system failure probability depends not only on the mean values of the respective segment failure probability random variables, but also on other distribution parameters (e.g., the standard deviations) of them. Pipe system failure probability estimated by means of the modified model is sometimes several times lower than that estimated by the conventional system failure probability model.

General failure event data from various sources are often used to estimate the failure probability for the system of interest, especially when s-dependence exists among component failures, where common cause failure plays an important role. Failure event data from different sources must be reasonably explained, and correctly applied, so that the information about load environment, and component/system property can be used correctly. In estimating the probability for s-dependent system failure, both the load distribution, and component strength distribution are much more important than component failure probability index.