The strong correlation and coupling of multi-fault of diesel engine fuel systems brings considerable amount of uncertainty to its diagnosis process. Meanwhile, establishing the diagnosis model always calls for massive prior knowledge because of the foregoing reason. The decoupling and diagnosis of the multi-fault has developed into a great technical difficulty in the study on the failure diagnosis of the diesel fuel system. Regarding this problem, a diagnosis method is proposed based on Bayesian networks with simplified model structure and quantitative parameters. In the aspect of model structure, the attribute reduction method in the rough sets theory is utilized to evaluate the equivalent relation between the information about failures. And on this basis, the redundant failure characteristics are removed and the topological structure of the Bayesian networks diagnosis model is simplified ultimately; in the aspect of quantitative parameters, the independent model of the casual mechanism is adopted to analyze the strength of causal relation for the failure events. The coupling effects of multi-fault to the same symptom are decoupled to the ones under a single failure and the quantity of the conditional probability needed by the model is simplified into the linear form of the failure quantity. By adopting the diagnosis method proposed, the prior knowledge needed by the Bayesian networks diagnosis model of the diesel engine fuel system is reduced significantly, which decreases the complexity in establishing and applying this diagnosis model.