多源信息拟合摩擦系数的回归集成模型

Multi-source information fitting regression integrated model of coefficient of friction

机器系统运动部件摩擦系数(coefficient of friction,COF)的实时监测是一项具有挑战性的难题,智能感知和数据技术的发展为利用摩擦学关联信息对摩擦系数进行预测提供了可能性。该文利用摩擦磨损试验过程中的声音、振动等多源摩擦关联信息,形成时间截面化的摩擦信息数据集,针对摩擦系数拟合问题建立了K折交叉验证双层堆叠的回归集成模型,定义了范围性评估的评价指标,并通过多种载荷试验数据对模型进行了检验。结果表明所建立模型能够有效提炼摩擦信息的关联特性,从而实现对摩擦系数的准确拟合及预测,该方法对不同载荷条件数据具有通用性。

Real-time monitoring of the friction coefficient of the moving parts of a machine system is a challenging problem.The development of intelligent perception and data technology provides the possibility to use tribological correlation information to predict the friction coefficient.This paper uses multi-source friction information such as sound during the friction and wear test to form a time-sectioned friction information data set,establishes a K-fold cross-validation double-stacked regression integration model,defines the evaluation indicators for scope evaluation,and the model was tested with a variety of load test data.The results showed that the model can effectively refine the correlation characteristics of friction information,so as to accurately fit and predict the friction coefficient,and has universality for data under different load conditions.