数据驱动的复杂系统非预期故障诊断通用过程模型

提高对复杂系统非预期故障诊断能力是故障诊断领域的难点。结合非预期故障诊断内涵及基本原理,构建了一种用于复杂系统非预期故障诊断的通用过程模型。该模型采用四层递进结构,包括四个主要模型,即预期(已知)故障检测模型、预期(已知)故障识别模型、非预期(未知)故障检测模型和非预期(未知)故障识别模型。分析了各模型所包含的关键问题及其相应的实现算法,包括检测统计量的构建及评估、故障特征方向提取、故障识别器设计及故障贡献率计算。该通用过程模型规范了复杂系统非预期故障的诊断流程,明确了数据驱动的实现原理。以卫星姿态控制系统为例,验证了非预期故障诊断通用过程模型的有效性。

A general process model for unanticipated fault diagnosis of complex system based on data driven

The unanticipated fault diagnosis for complex system is a difficult problem. The objective of this paper is to advance a general process model for unanticipated fault. The model adopted a four-layer structure, i.e. anticipated fault detection layer, anticipated fault isolation layer, unanticipated fault detection layer and unanticipated fault isolation layer. Several key problems and the corresponding algorithms in each layer are analyzed, including the establishment of the fault detection statistic, the extraction of the fault character, the design of the fault isolation criterion and the construction of the fault contribution rate. The general process model regularized the diagnosis process of unanticipated fault for complex system and defined the data-driven fault diagnosis framework. Finally, the effectiveness of the proposed general process model is validated by a satellite control system.