融合字典学习的滑动轴承摩擦故障趋势自记忆预测

发布时间：2018-12-19 11:31

【摘要】：将自记忆模型引入到滑动轴承的接触摩擦故障发展趋势预测中,针对一般形式的多变量自记忆模型的不足,提出一种基于字典学习的滑动轴承摩擦故障趋势的多变量自记忆预测方法。首先,构建多变量字典,利用稀疏编码筛选出影响系统演化的主要变量。然后通过误差平方和最小原则更新不同作用形式的变量字典,确定变量影响系统演化的最佳作用形式和影响系数,最终得到多变量的系统动力核函数,解决了系统动力核与多变量关系难处理问题。最后,引入自记忆函数,将所得的系统动力核反演成一个微分-差分方程,由此得到滑动轴承的多变量自记忆预测模型。应用到实例中,有效地预测了故障的发展趋势,为滑动轴承摩擦退化趋势预测提供了一种新的途径。
[Abstract]:The self-memory model is introduced into the development trend prediction of contact friction fault of sliding bearing, aiming at the deficiency of the general multivariable self-memory model. A multivariable self-memory prediction method for sliding bearing friction fault trend based on dictionary learning is proposed. Firstly, the multivariable dictionary is constructed and the main variables that affect the evolution of the system are screened by sparse coding. Then the dictionary of variables in different forms of action is updated by the principle of the sum of square error, and the best form of action and the coefficient of influence of variables affecting the evolution of the system are determined. Finally, the dynamic kernel function of the multivariable system is obtained. The relationship between dynamic kernel and multivariable is difficult to deal with. Finally, the self-memory function is introduced to invert the dynamic kernel of the system into a differential-difference equation, and a multivariable self-memory prediction model of sliding bearing is obtained. The application of this method can effectively predict the development trend of the failure and provide a new way for predicting the friction degradation trend of sliding bearing.
【作者单位】：军械工程学院七系;西南交通大学机械工程学院;
【基金】：国家自然科学基金资助项目(51205405,51305454)
【分类号】：TH133.31

【相似文献】