论文部分内容阅读
为解决支持向量分类机多分类存在的样本重复训练、训练模型过多的问题,保证模拟电子系统在整体和局部多故障模式上的诊断正确率,提出基于最小偏差的最小二乘支持向量回归机多故障诊断方法.通过引进样本各维度拟合误差相对于平均拟合误差的偏差平方项,最小化维度间的拟合误差间距,得到能够输出多维变量及具有高分辨率的最小二乘支持向量回归机模型.将模型多维输出值与预设的各个多故障模式值相乘,所得结果集中最大值所对应的多故障模式即为最终诊断结果.仿真结果表明:提出的方法在简化训练过程的同时,能够保持良好的整体和局部多故障诊断正确率.
In order to solve the problem of repetitive training and training model of multi-classification of support vector machines, and to ensure the correctness of the diagnostic system in global and local multi-fault modes, a least square support vector regression Multi-fault diagnosis method.Through introducing the square of the variance of the fitting error of the sample in each dimension of the sample to the average fitting error and minimizing the fitting error distance between the two dimensions, a multi-dimensional variable least square support vector with high resolution can be obtained Regression model.The multi-dimensional output value of the model is multiplied by the preset value of each multi-fault mode, and the multi-fault mode corresponding to the maximum value of the result set is the final diagnosis result.The simulation results show that the proposed method simplifies the training process At the same time, good overall and local multi-fault diagnosis accuracy can be maintained.