基于RBF神经网络的Bootstrap数据扩充方法及其在IRSS可靠性估计中的应用

A Bootstrap data expansion method based on RBF neural network and its application on IRSS reliability evaluation

针对高可靠性、长寿命产品可靠性试验数据样本较少难以进行有效可靠性评估问题,提出一种基于RBF神经网络的Bootstrap数据扩充方法,利用RBF神经网络获取原样本连续分布特性,邻域函数构建网络输入集。仿真表明,由该扩充方法获得的扩充样本分布特性更接近于其真实分布,并有效利用了原样本取值区间上、下限数据信息,拓展更宽的样本取值范围。将其应用于工业机器人伺服系统(IRSS)伪失效寿命分布可靠性评估中,扩充伪失效寿命数据,获得IRSS有效可靠性评估结果,表明方法的实际应用价值。

For high reliability and long-life products,their reliability test data samples are too few to carry out effective reliability assessment.A Bootstrap data expansion method based on RBF neural network was proposed address to the issue.In the method,RBF neural network was applied to catch the continuous distribution characteristics of original samples,and neighborhood function was used to create the network input set.Simulation results show that the distribution of extended sample obtained by this method is closer to its real distribution,and the extended sample get a wider range of values as well.The proposed method is applied to the pseudo-failure life distribution reliability evaluation of industrial robot servo system(IRSS),and the effective reliability evaluation results of IRSS obtained,which indicated the practical application value of the proposed method.