基于aiNet算法优化SVM模型的惯性器件故障预报

在标准支撑矢量机算法中,其模型结构参数和核函数中的参数一般凭经验通过交叉验证的方法选择确定,缺乏理论基础,影响支撑矢量机的学习效果.针对这种局限性,文中利用人工免疫算法对支撑矢量机的参数进行优化.将待优化参数作为抗体,经过抗体克隆、变异和抑制等操作,找到最优抗体,即对应最优化参数的支撑矢量机模型.然后基于优化后的支撑矢量机利用惯性器件的历史数据,对其进行故障预报.仿真结果显示:该算法的故障预报误差小于标准支撑矢量机的预报误差.证明了免疫aiNet算法优化支撑矢量机模型参数的有效性,及优化模型在惯性器件故障预报中的有效性.

Inertia Device Fault Prediction Based on Support Vector Machine with aiNet Algorithm

In standard support vector machine algorithm(SVM),the structure parameters and the parameters in kernel functions can only be determined through cross certification with experience.However,it lacks theoretical foundation and influences the learning effect of SVM.To overcome the limitations,the paper presented a novel support vector machine algorithm based on immune algorithm.The immune algorithm is used to optimize parameters of SVM,which were regarded as the antibodies.The satisfied antibody was found after these operations: clone,mutation,and restrain.Then,the optimized SVM was put into use in fault prediction of missile inertia device,using history data.The simulated experiment revealed that prediction error of the algorithm is smaller than standard SVM.And it proved that aiNet algorithm was useful for optimizing parameters of SVM,and the optimized SVM was effective in inertia device fault prediction.