中文摘要：

HAZOP分析方法自提出至今在石油化工领域一直被广泛应用,为代替传统HAZOP人工头脑风暴分析的过程,提出基于因果依赖图（CDG）的HAZOP推理方法,即CDG-HAZOP。首先建立系统的多级流模型,将系统生产过程抽象为流过程,用简单的功能图形符号来表示复杂的装置,然后利用多级流模型中的告警分析算法代替人工头脑风暴分析,推理偏差的可能原因和结果。CDG-HAZOP推理机制将传统HAZOP分析的重点由人工分析大量资料转移到多级流模型（MFM）的建立和校验上,能节约人工成本。将CDG-HAZOP推理方法用于催化裂化装置的反应-再生系统,对该系统可能发生的偏差进行原因和后果推理。

英文摘要：

HAZOP analysis has been widely used in chemical and petroleum industry since its inception.A HAZOP reasoning method based on CDG was proposed to replace the ＇brain storm＇ of traditional HAZOP analysis.Firstly a multilevel flow model was built,by using simple function model symbols to represent complex devices applied for production industry.Then the alarm analysis in MFM was used to replace the ＇brain storm＇,which could infer cause and effect of deviation.Manual analysis was the key point of traditional HAZOP study.However,the CDG-HAZOP analysis greatly simplified it by transferring focus on model establishment.The HAZOP reasoning based on CDG was utilized to infer causes and effects of deviations in the reaction-regeneration system of the catalyst cracker.