Abstract: It is not completely accordant for fault reasons to match up corresponding symptoms of the marine nuclear power plant. A kind of fault diagnosis method was proposed, which is about invasive weed optimization algorithm combined with probability causal model. The probability causal model likelihood function was used as fitness function of the invasive weed optimization algorithm, after that the fault diagnosis of complex systems can be converted to optimization problem. The simulation results show that the method can not only be used for the process of diagnosis of uncertainty, but also for the purpose of multiple symptoms to multiple faults diagnose with high reliability and practicability.