摘要
利用具有高度非线性映射能力的 BP神经网络解决滑动轴承磨损预测的计算问题。在论述 BP算法及改进模型原理的基础上 ,利用它们对滑动轴承材料边界磨损系数的预测效果进行了比较 ,进而在小样本情况下通过 Baysian正规化 BP网络对滑动轴承材料边界磨损系数进行了预测 ,分析了影响预测效果的原因 ,在合理剔除奇异点后给出了对滑动轴承材料磨损预测的最佳 Baysian正规化 BP网络结构 ,为合理进行磨损试验提供了理论依据。对预测结果进行残差分析证明 ,该方法效果较为理想。
In this paper,the Baysian regularized BP network is applied to predicate wear status of sliding bearing. After obtaining the prediction result, the causes to affect the prediction quality and accuracy are provided. The odd points are pruned to preclude the deviation of the prediction results, so as to propose a new method to predicate the wear status of sliding bearing both from theory and practice case. It is proved that this method is effective for predicating the wear status of sliding bearing by the residual error analysis of the obtained computation results.
出处
《中国机械工程》
EI
CAS
CSCD
北大核心
2003年第19期1707-1709,共3页
China Mechanical Engineering
