基于置信规则库的并发故障诊断方法

针对具有不确定性的并发故障诊断问题,提出基于属性权重和权衡分析的置信规则库(belief rule base, BRB)诊断方法。以属性权重大小来表示属性与特定故障模式之间的相关性,设计能够反映故障模型约束的基于差分进化的优化算法,通过相邻故障模式置信度与预设阈值的权衡分析完成对并发故障诊断。该方法仅需构造单个置信规则库来有效处理各种不确定性信息,与已有研究方法相比极大地降低了建模复杂度。诊断结果不仅能得到故障的并发情况,还可分辨故障的主次关系,并且建模和推理过程开放,可解释性强。最后以船用柴油机的并发故障诊断作为实例,验证了所提方法能够有效的诊断出并发故障并且模型具有较好的稳定性。

Concurrent fault diagnosis method based on belief rule base

To solve the problem of concurrent fault diagnosis with uncertainty, a belief rule base (BRB) diagnosis method based on attribute weights and tradeoff analysis is proposed. In the proposed method, the attribute weights are used to represent the correlation between the attributes and the specific failure mode, and an optimization algorithm via differential evolution (DE) which can reflect fault model constraints is designed. The concurrency fault mode diagnosis is completed by the trade-off analysis between the adjacent fault mode belief and the preset threshold. This method only needs to construct a single BRB to effectively deal with various uncertainty information, which greatly reduces the modeling complexity compared with the existing research methods. The diagnosis results can not only get the concurrency of the faults but also distinguish the main fault and the minor fault. The modeling and reasoning process is open and interpretable. Finally, the concurrent fault diagnosis of a marine diesel engine is taken as an example and proves the proposed method can effectively diagnose the concurrent faults with good stability.