基于量子万有引力搜索的SVM自驾故障诊断

针对自动驾驶仪在实际测试过程中故障样本较少的情况,提出一种基于量子万有引力搜索算法(QGSA)的支持向量机(SVM)故障诊断模型。SVM能较好地解决小样本、非线性问题,适用于自动驾驶仪的故障诊断。为进一步提高万有引力搜索算法(GSA)对参数寻优的收敛速度和收敛精度,将基于GSA的QGSA应用于SVM的参数寻优中,以解决SVM由于参数选取不当导致过学习或欠学习的问题,从而获得最优的分类模型。通过模拟实验分析,当训练样本数量为50时,基于QGSA的SVM故障诊断模型分类准确率便能达到96.530 6%,而基于遗传算法(GA)的SVM故障诊断模型分类准确率为92.040 8%,基于GSA的SVM故障诊断模型分类准确率为91.632 7%。仿真实验结果表明,基于QGSA的SVM故障诊断模型具有更好的故障诊断能力。

SVM fault diagnosis of autopilot based on quantum inspired gravitational search algorithm

With regard to the lack of the sample of faults in the test of autopilot, a model of fault diagnosis based on support vector machine (SVM) optimized by quantum inspired gravitational search algorithm is put forward. SVM does well in solving the few samples and nonlinear problem, which is suitable for the fault diagnosis of autopilot. To improve the convergence rate and accuracy of parameters optimizing based on gravitational search algorithm (GSA), quantum inspired gravitational search algorithm (QGSA) was applied to optimizing the parameters of SVM. SVM based on QGSA can solve the overfitting and underfitting resulted from the improper parameters. By this way, a model of fault diagnosis with better performance was built. The simulation experiment results show that the accuracy of SVM based on QGSA can achieve 96.530 6% using 50 training samples. However, the accuracy of genetic algorithm (GA)-SVM achieves 92.040 8% and the accuracy of SVM based on GSA achieves 91.632 7%. The simulation experiment results shows that SVM based on QGSA has much better performance than others.