基于神经网络的小样本轴承球可靠度评估  

Reliability Assessment of Small Sample Bearing Balls Based on Artificial Neural Network

作　　者：张中益 沈雪瑾[1] 

机构地区：[1]上海大学机电工程与自动化学院,上海200072

出　　处：《工业控制计算机》2019年第5期66-67,70,共3页Industrial Control Computer

基　　金：国家"十二五"规划项目(D.50-0109-15-001)

摘　　要：为建立轴承球磨损量数学模型并结合模型对样本进行可靠度评估,设计了BP神经网络,对磨损量曲线进行拟合。分析表明,拟合曲线能够很好地表达磨损量的变化趋势。通过提取神经网络的权值、阈值参数,建立了磨损量数学模型,并结合模型和蒙特卡洛法对小样本数据进行了处理,进而基于性能退化量分布评估了轴承球的可靠度。结果表明提出的方法针对基于单一性能退化的小样本可靠度评估问题提供了有效的解决途径。A back propagation artificial neural network is designed to fit the curve of bearing balls' wear volumes in this paper.The analysis indicated that the fitting curve could express the trend of wear volumes' changing well.The model of wear volumes can be built through extracting weights and threshold from back propagation neural network,the data got from the model is processed by Monte Carlo methods.At last.the reliability of bearing balls was evaluated by the method which was based on the distribution of performance degradation.

关 键 词：轴承球 磨损量 BP神经网络 性能退化 小样本 可靠度 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]