【作者】 吕宁； 熊智华； 王雄；

【Author】 Lü Ning, Xiong Zhihua, Wang Xiong (Department of Automation, Tsinghua University, Beijing 100084, China)

【机构】 清华大学自动化系；

【摘要】 基于模型的SDG(Signed DiGraph,符号有向图)故障诊断方法因其具有完备性好、推理深度高等优点在过程工业安全工程中具有十分重要的意义。本文提出了一个综合性SDG故障诊断架构,以期能够实现在生产过程中及时发现故障并判明故障源。该综合性故障诊断架构按模型、推理和应用三个层次搭建,以传统定性SDG及概率SDG理论为基础,包含了从模型建立到故障诊断推理,从定性SDG方法到结合统计监控的SDG方法再到概率SDG方法等一系列实施方案。并以某石化公司的气体分馏装置为实际背景,利用该装置实时数据库中的实际生产工艺数据对该综合性诊断架构进行了实例验证,实验结果证明了该综合性故障诊断架构的有效性。更多还原

【Abstract】 SDG is one of the model-based methods for fault diagnosis. It has good maturity, in-depth reasoning ability and several other advantages, so that has a great significance in safety engineering of process industry. In this paper, an integrative framework for SDG based fault diagnosis is proposed to achieve the detection of fault symptoms and find out the fault origins as soon as possible in process industry. This framework, based on the theories of traditional qualitative SDG and probabilistic SDG, includes the procedure from SDG modeling to fault diagnosis inference and a series of methods including the qualitative SDG method, the hybrid method combining SDG with multivariable statistical process monitoring and the probabilistic SDG method. By this framework, the entire theoretical framework and application system for SDG are constructed and enriched in modeling, inference and application levels. The framework is validated through a fault diagnosis case on the gas fractionation unit of a certain petrochemical company.更多还原