改进 PSO-RBFNN算法在退化型产品寿命预测中的应用

针对部分高可靠性产品退化规律无法掌握的难题,提出了使用改进粒子群优化—基于神经网络函数(PSO-RBFNN)算法拟合样品退化轨迹、预测伪寿命值的方法。首先,通过改进PSO算法对RBFNN进行训练优化;然后,使用部分测量数据对训练后的RBFNN进行准确度测试;最后,通过RBFNN预测样品退化轨迹,估计出伪寿命值。使用某型电连接器的加速退化试验数据对提出的方法进行了试验验证,成功对该型电连接器进行了寿命预测,得出平均寿命为200 412 h。

Application of APSO-RBFNN Algorithm on Degradation Production Lifetime Predition

According to the problem that the degradation rule of some high-reliability production cannot be acqurled, the APSO-RBFNN algorithm, which was ued to fit the degradation path and predict the pseudo lifetime, was proposed. Firstly, RBFNN was trained and optimized through APSO. Then, the accuracy of trained RBFNN was tested with parts of measurements. Lastly, the RBFNN was applied to predict the degradation path of production and then evaluating the lifetimes. The proposed approach was methodologically explained and experimentally was evaluated using accelerated degradation data of some electrical connector. The lifetime of electrical conuector was successfully predicted and the average lifetime was obtained, 200412 hours.