Considering the randomness and instability of the oil pressure in the lubrication system, a new approach for fault detection and diagnosis of diesel engine lubrication system based on support vector machine optimized by particle swarm optimization (PSO-SVM) model and centroid location algorithm has been proposed. Firstly, PSO algorithm is chosen to determine the optimum parameters of SVM, to avoid the blindness of choosing parameters. It can improve the prediction accuracy of the model. The results show that the classify accuracy of PSO-SVM is improved compared with SVM in which parameters are set according to experience. Radial basis function (RBF) is selected as the kernel function. Optimal parameters σ=47.34, trade-off factor C =10.98, test accuracy was 99.86%. Then, the support vector machine classification interface is fitted to a curve, and the boundary conditions of fault diagnosis are obtained. Finally, diagnose algorithm is achieved through analyzing the centroid movement of features. According to Performance degradation data, degenerate trajectory model is established based on centroid location. And normal faults and performance degradation faults of diesel engine lubrication system are diagnosed. The simulation model of the diesel engine lubrication system is established using the AMESim software, and normal fault and performance degradation fault are set on the model. Results show that classification accuracy of the proposed PSO-SVM model achieved is 95.06% and 97.04% in two verify samples, it can meet the needs of fault diagnosis; and two typical faults and performance degradation fault of diesel engine can be diagnosed based on the proposed diagnosis method through simulation model based on AMESim.