模拟电路与电子系统故障诊断及元件参数在线辨识的代数方法

电路与电子系统的故障诊断是高可靠性工程的重要环节，在故障预测/预报、系统剩余寿命估计等需求中，原件状态/参数的在线估计/辨识是难以回避的问题。与基于信号处理的“分析学”方法相比，基于“代数学”的方法在处理测试噪声、降低对概率模型的依赖、在参数辨识中观察“动态”变化过程，以及在诊断、参数辨识的“灵敏度”与“鲁棒性”之间实现良好平衡方面具有显著优势。本项目通过研究1）基于代数分类的测试信息预处理方法实现对易测故障与难测故障的“分而治之”，以降低诊断难度和成本；2）基于代数“覆盖”的有限故障集表示，以及基于连通、凝缩与不动点原理的故障高准确度定位方法；3）在线参数辨识的代数方法；突破1）从代数角度提取参数容差型故障特征和噪声中提取特征与2）建立在线辨识时参数与状态特征间的映射等关键问题，为故障预测、剩余寿命估计及制定维护保障计划提供依据，为实现故障诊断与参数在线辨识的自动化与智能化提供科学方法。

Fault diagnosis plays an important role in high reliability electronic circuits and systems. When the fault prediction and the residual life estimation are required in such systems, the online status (or parameter) estimation (or identification) for the circuit components becomes a challenging and unavoidable problem.Therefore, we propose a series of fault diagnosis methods based on the algebraic theory. Compared with the "analysis methods" based on the conventional signal processing theory and technology, the "algebra methods" have: 1) a better noise tolerance 2) a lower dependency on probabilistic model selection 3) the ability of capturing the dynamic changes in component parameters and 4) a good balance between "Sensitivity" and "robustness" in parameter identification and diagnosis.In this research project, we develop the idea of "algebra methods" in order to provide reliable evidences for the fault prediction, the residual life estimation and the maintenance planing, and offer an effective approach for the intelligent and automatic fault diagnosis and online parameter identification.To achieve these goals, Two key issues are to be settled: 1) the feature extraction of parametric faults from the perspective of algebra theory, especially when the effect of noise is considered and 2) the mapping between the parameters (obtained online) and the state features of the system under test. Seeking breakthrough on the above issues, we will make an intensive study on: 1) the pretreatment of test information based on algebra classification (The faults are being classified into two types, the "easy to test" fault and the "difficult to test" fault, in order that the difficulty and cost of fault diagnosis can be reduced), 2) the representation of the limited fault set (based on the finite covering theorem), and the high accuracy fault location method (based on the connection, condensation and fixed point theorem in algebra theory) and 3) the algebra method of online parameter identification.