基于模糊神经网络的齿轮剩余寿命预测模型研究

A Model for Remaining Useful Life Prediction for Gear Based on Adaptive Neuro-fuzzy Inference System

齿轮及其齿轮产品是机械设备的基础元件,齿轮传动形式是机械装备常见的传动形式之一,运行是否正常直接影响到整台机械设备。利用状态监测信息分析其退化性能,建立适当的模型研究退化特征的发展趋势、预测设备剩余寿命,为制定更合理的维修计划和更换策略、确定维修周期提供先决条件。基于设备性能退化数据建模进行剩余寿命研究,提出基于改进型自适应神经模糊系统的学习算法。此算法融合了多测点的实时监测信息,增加记忆单元于模糊层节点上,将上一时刻信息记忆并应用到此刻的输出上,有效地提高了网络模型的预测精度。最后以齿轮弯曲疲劳寿命预测为例,验证改进的预测模型随着迭代次数的增多,误差相比传统的自适应神经模糊系统降低很多。

Gear and gear products are the basic components of machine equipment;gear transmission is the familiar forms,whether the operation is normal or not directly affects the entire machine equipment.Through analysis its degradation performance using state monitoring information,a fit model is established to study the development trend of degraded features and to predict the residual life of equipment,which provide prerequisites for making a more reasonable maintenance plan and replacement strategy,as well as determining the maintenance cycle.In this paper,the residual life is researched based on the modeling of degradation data,an improved adaptive neuro-fuzzy system is proposed.This algorithm integrates the real-time monitoring information of the multiple measurement points,increases the memory unit on the Fuzzy Layer node,and applies the information memory of the last moment to the output of the present moment,thus effectively improving the prediction precision of the network model.Finally,the fatigue life prediction of gear bending is taken as an example to verify that the error of the improved model is much lower than the traditional adaptive neuro-fuzzy system as the number of iterations increases.