中文摘要：

针对真空阀检测中可能出现的故障情形及相应的调试方法,提出一种基于遗传PNN神经网络的真空阀自动调试系统。将不同占空比电源驱动下出气口的压力作为故障特征值,并利用PNN网络进行故障分类,结合机械手臂进行相应的调试,从而实现故障自动检测与调试。在样本数有限的情况下,PNN神经网络通过遗传算法获得模式层最佳节点数目及相应的平滑参数,降低了网络的冗余度并提高了网络的精度,再将通过PNN分类并调试成功的数据加入到样本集中重新训练网络,直到网络精度增长率达到一定范围,使得网络的精度和模式层节点数达到最优值。测试结果表明,该方法能够有效的对真空阀进行故障分类与调试,能大幅提高检测的自动化程度和精度。

英文摘要：

The automatic debugging system based on genetic PNN was proposed for the fault and corresponding debugging method of vacuum valve. The pressure of the outlet port under the power with varying duty ratio can be extracted as the input for PNN for fault classification,and then debugged with manipulator to achieve automatic diagnosis and debugging. As the training samples were restricted,the PNN improved with genetic algorithm firstly get the optimum nodes of the pattern layer and corresponding smoothing parameters,then added the correct testing data through the trained PNN into the training samples,and trained the PNN again till the accuracy satisfied certain value. Thus,the accuracy and nodes of the PNN can be optimal. The results show that this method can not only diagnose and solve the fault for vacuum valve,but also improve the accuracy and automaticity of the system.