基于粗糙集的船舶电力系统故障诊断

enetic Rough set-based acquisition of decision rules method

针对船舶电力系统故障诊断的现实需要,提出基于粗糙集理论的故障诊断方法。本文运用粗糙集理论分析多个传感器融合得到的船舶电站机组故障参数,建立机组的故障诊断知识库。加入一种可以有效保证遗传迭代中样本多样性的排挤策略,以此利用遗传算法求取信息系统的属性约简。根据识别矩阵与约简属性的交集所形成的矩阵的某行中若存在单个元素,则此必为该规则的值核属性这一性质,迅速求出决策规则的值核属性,并在求得此核值属性的基础上,利用一致性规则对此约简进行规则泛化。文章从决策规则本身出发,依据规则的一致性进行算法设计,方便快捷地实现了决策表的属性约简和属性值约简。最后通过实例仿真验证了此方法的可行性。

According to requires of marine power system fault diagnosis, the paper presents a new fault diagnosis method＇based on rough sets. By analyzing the fault parameters of marine power stations from muti-sensor fusion based on rough sets, we established fault diagnosis knowledge base. A crowding strategy that could ensure the diversity of samples in genetic iterative has been put for- ward for the large database with high dimension, and attributes reduction of the information system was accomplished through genetic algorithm and this strategy. According to the property that if there is single element in a certain line in the matrix come from the intersection of recogni- tion matrix and reductive attributes, then it is the core attributes of this rule, core attributes of this rule could be gotten effectively, and use the consistent rule to generalize this reduction. It starts from decision rules itself, design the algorithm with consistency of rules, accomplish attributes reduction and attributes value reduction of the decision table. Feasibility of this method was proved by the emulation experiment.