基于感知机的故障树最小割集算法

Minimum Cuts of Fault Tree Based on Perceptron Algorithm

在人工神经元网络原理的基础上,提出了一种计算复杂系统最小割集的方法。该方法根据逻辑“与”和“或”门具有线性可分割的特点,选用了基于M-P模型算法的感知机网络作为计算模型,理论推导出故障树中的逻辑关系与感知机模型中神经网络基本单元之间的转换规则,利用该规则可将所建故障树转化成便于编程求解的由神经网络基本单元组成的神经网络树。实例计算结果表明,感知机网络模型适合表达故障树中的逻辑关系,神经网络树所反映的基本事件与顶事件之间的映射关系便于编程求解,并可快速准确地获得复杂系统故障树的最小割集。

An algorithm is provided for minimum cuts of a fault tree about a complex system based on the rationale in artificial neural network principle. According to the characteristic that the logic of ＂and＂ and ＂or＂ gates can be separated by a line, a perceptron net based on M- P mode is selected as a model for calculation. A conversion rule is derived theoretically between the logic of a fault tree and the neural unit by which the fault tree can be converted to a neural tree. The neural tree is composed of neural twits, and is easy to calculate by programing in the computer. The results of exemplified calculation demonstrate that the logic of a fault tree can be described ccrnpafibly by the perceptroon net. The mapping described by the neural tree between basic and top incidents is easy to program, and minimum cuts of a complex fault tree can be accurately and quickly obtained.