基于CBA-SVM的模拟电路故障诊断方法研究

Research for analog circuit fault diagnosis based on CBA-SVM

支持向量机（SVM）由于在解决小样本、非线性等实际问题中良好的表现,现已广泛应用于电路故障诊断之中,但是,在核函数参数及惩罚因子参数的选取上存在盲目性的问题。对此,将全局寻优性能较好的蝙蝠算法引入到基于SVM的故障诊断方法之中,并且为了进一步提高寻优性能,对算法中的关键参数β进行混沌优化。为了验证所提方法,对某型雷达导引头测试系统自检模块中的比例积分电路进行故障诊断,实验结果表明相同条件下该方法比万有引力搜索算法、粒子群算法优化后的SVM方法在诊断精度方面提高了2%和7%,在诊断耗时方面分别快了近4s和1s,可以很好的完成故障诊断任务。

Because support vector machine（SVM）has good performance in solving small sample,nonlinear and high dimension problems,it has been widely used in analog circuit fault diagnosis.However,there exist the blind selection problem of the kernel function parameters and penalty factor parameters when applying SVM.The bat algorithm with better global optimization performance is introduced into the fault diagnosis method based on SVM.In order to further improve the optimization performance,chaos optimization is performed for the key parameterβin the algorithm.In order to verify the proposed method,the fault diagnosis of a proportional integral circuit in a self-test module of a radar seeker test system is carried out.The experimental results show that the diagnostic accuracy was improved by 2% and7%,and the diagnostic time consumption was almost 5 seconds and 3 seconds faster,respectively compared with the SVM optimized by GA and PSO under the same conditions.