基于贝叶斯网络的GO法模型算法

GO法是评价具有多状态时序特性的复杂系统可靠性的有效方法,但GO法操作符众多、算法复杂并且缺乏工具软件支持,制约了GO法的工程应用。针对该问题,本文提出一种基于贝叶斯网络的GO法新算法。首先，定义常用操作符到贝叶斯网络节点映射规则；然后，给出GO模型映射转换为贝叶斯网络的可编程流程；最后，利用贝叶斯网络成熟工具支持,定量求解映射后的贝叶斯网络模型。新算法操作符映射规则统一,模型映射转换流程简单直观,便于工程人员掌握和应用。此外,除了能得出传统的定量结果,新算法使得GO法还具有故障推理和诊断能力。

Algorithm based on Bayesian networks for GO methodology

The GO methodology is an efficient technique for modelling reliability of complex system with multi-states and time serial characteristics. However, it contains variant operators with complex logic, and the algorithm is complicated without tool supporting, which impact the application of the GO methodology in engineering. Therefore, a new algorithm based on Bayesian networks (BNs) is proposed. Firstly, the mapping regulation is defined from GO operators to the node of BNs. Then, the algorithm of mapping a complete GO model to corresponding BNs is proposed. Finally, the model is resolved with mature BNs toolbox. With the proposed unified operator mapping rules and simple concise model transformation procedures, the new algorithm is concise and easy to apply in real engineering scenarios. Based on the mapped BNs, conventional analysis can be performed directly, moreover, the new algorithm adds a new capability of fault diagnosis to the GO methodology.