基于定性定量模型的动态过程故障诊断新方法  

A NEW METHOD OF FAULT DIAGNOSIS FOR DYNAMIC PROCESSES BASED ON QUALITATIVE AND QUANTITATIVE MODEL

作　　者：朱六璋[1] 陈锋[1] 陈宗海[1] 

机构地区：[1]中国科技大学自动化系,合肥230027

出　　处：《模式识别与人工智能》2001年第2期150-155,共6页Pattern Recognition and Artificial Intelligence

摘　　要：给出了一种能够综合运用先验知识和启发性知识的基于定性定量模型的动态过程故障诊断新方法。首先描述了核心的模糊定性建模和仿真技术,接着设计一种新的知识观测器,其中正常和已知故障状态的知识用模糊定性模型表示,将观测的行为与模糊定性仿真预测行为比较,将匹配的结果诊断输出,若找不到匹配结果,进行故障假设和生成相应模型,并将其预测行为和观测行为比较直到正确诊断。Based on qualitative and quantitative models with which the empirical and heuristic knowledge is combined, a new method of fault diagnosis for dynamic processes is presented. An approach is described, which utilizes fuzzy sets and qualitative constraints to develop a fuzzy qualitative simulation algorithm that allows qualitative and quantitative extension to qualitative simulation. Besides, it can effectively eliminate the ambiguity during the qualitative reasoning and allow temporal information be considered in state transitions for filtering spurious behaviors. Such a fuzzy qualitative modeling and simulation can be used to improve qualitative simulation-based fault diagnosis, and it is used as the kernel part in the design of a novel knowledge-based ovserver for fault detection and identification. In the observer, the fuzzy qualitative models are used to describe the knowledge of the normal and faulty states, and the observed behavior is compared with predicted behaviors of fuzzy qualitative simulation under the supervision of the matching rules. If the observation matches the simulation prediction of certain model, a diagnostic result is given out under the model, otherwise the hypothesis process to some fault is carried out to generate the corresponding model for simulation. The hypothesis is based on layered causal functional structure of the system and the constraint suspension. The predicted behavior is repeately compared with the observed one for justification, until the comparison is matched or the set of faulty components is empty.

关 键 词：故障诊断 定性定量模型 定性仿真 人工智能 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]