基于运行工况和多分类支持向量机的柴油机共轨系统诊断方法

A fault diagnosis strategy based on operating conditions and multi-classification support vector machine was proposed for two typical faults of diesel common rail system with loose metering valve reset spring and worn injector needle valve coupling. The strategy was arranged as follows. Firstly, the influence of operating conditions on the fault diagnosis accuracy was taken into account and the data collected from real vehicles were divided into three sub-conditions according to the vehicle speed. Then the state parameters correlated highly with the fault condition were selected by cardinality test and mechanism analysis, and the feature parameters were extracted using principal component analysis, and the sub-condition set with the highest sensitivity to the fault was filtered according to the contour coefficient S. A hierarchical fixed-ratio sampling method was used to divide the training set, and the penalty parameter c and radial basis function (RBF) parameter g of the support vector machine were optimized by the particle swarm algorithm. Finally, the model was validated by using a real vehicle test data set. The experimental results show that the correct diagnosis rate of the method in the prominent working conditions can reach more than 90%, which meets the fault diagnosis requirements.