Abstract: Considering that the event tree and fault tree analysis bear some disadvantages， a new integrative risk analysis of HAZOPLOPA was proposed on the basis of Bayesian networks. The system fault tree was established and mapped into Bayesian networks by using GeNIe software. Then fault prediction and diagnosis could be conducted by Bayesian networks twoside reason function， and it could rapidly identify the system weakness， which was chosen to be the Bayesian risk fault node. Integrative risk analysis was used by combining HAZOP with LOPA， putting forward the corresponding independent protective layer， and the residual risk level could be determined according to failure probability of protective layer and semiquantitative risk matrix. This risk assessment model was then applied in the reactor regenerator system of RFCC equipment in Liaohe Petrochemical Industries Company， and the application results can verify the pertinence， objectivity and accuracy under the condition of uncertain information in complex system.

Key words: fault tree, Bayesian networks, layers of protection analysis, catalytic cracking, risk analysis