中文摘要：

为保证炼化装置生产的连续性和安全性,及时诊断和排除生产中出现的故障,提出一种基于动态贝叶斯网络（DBN）模型的故障诊断技术。采用危险性与可操作性分析（HAZOP）方法分析对炼化装置故障传播机理,详细分析系统内不同偏差发生的可能原因及可能后果,在此基础上建立复杂生产过程的故障传播动态贝叶斯模型。依据现场在线监测数据,对生产过程中出现的报警进行诊断,推理发生报警的最可能根原因。应用该方法诊断某催化裂化装置再生器温度低报警故障。结果表明,该方法诊断出的故障原因为其主风机系统故障,与事实相符。

英文摘要：

Timely fault diagnosis and exclusion are necessary for ensuring the continuity and safety of refining process production. A novel fault diagnosis technology based on DBN model was developed. Fault propagation mechanism in refining equipment was analyzed using HAZOP method. Possible causes and consequences of different deviations were studied in detail. A fault DBN model was built. Then,alarms in the production process were diagnosed according to monitoring data,and most likely reason（ s） were inferred. Finally,the effectiveness of DBN model-based fault diagnosis technology was checked by diagnosing temperature low alarm FCCU regenerator in China. It was diagnosed that cause was main air blower failure,which was consistent with facts.