基于SDG的化工过程多故障诊断  

SDG-based Multiple Fault Diagnosis of Chemical Process

作　　者：张卫华[1] 李传坤[2] 吴重光[1] 王春利[2] 

机构地区：[1]北京化工大学信息科学与技术学院,北京100029 [2]中国石油化工股份有限公司青岛安全工程研究院化学品安全控制国家重点实验室,青岛266071

出　　处：《系统仿真学报》2009年第21期6972-6977,共6页Journal of System Simulation

基　　金：国家高技术研究发展计划(2009AA04Z133)

摘　　要：目前,针对化工过程故障诊断的研究多集中在单"剧情",即在某一时刻只有一个故障源,故障传播的路径只有一条。其原因是系统发生单"剧情"的概率非常大。然而,多故障在化工过程中是确实存在的。多故障诊断的难点在于,系统内存在大量不可测节点,可测节点与不可测节点混杂造成剧情数目庞大。如何推理出海量剧情并进行剧情压缩,找到真正的故障传播路径,成为一个研究难点。基于深层知识定性模型的符号有向图(Signed Directed Graph,SDG)在推理、压缩剧情方面,完备性较好,推理结果便于压缩。At present, the research on fault diagnosis of chemical process is more focused on the single scenario. That is, there is only one failure at one time, and the propagation pathway of failure is only one. The reason is that the probability of single scenario is very great in chemical process. However, multiple faults in chemical process are really existent. The difficulty of multiple fault diagnosis is there are a large number of unmeasured variables; the mixing of measured variables and unmeasured variables cause excessive number of scenarios. How to infer all the scenarios and find out the real cause by compressing scenarios becomes a research difficult. Qualitative model based on deep knowledge of the signed directed graph （Signed Directed Graph, SDG） has high completeness in inference and compressing the amount of scenario inferred. The inference results are easy to simplify.

关 键 词：SDG 多故障 故障诊断 混合算法 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]