基于AdaBoost与LCA的滚珠丝杠副状态识别方法

A novel approach for ball screw condition identification with AdaBoost and local class accuracy

作为数控机床的关键部件,滚珠丝杠副的工作状态关系到整个机床的性能与产品质量。因此,对滚珠丝杠副进行状态监测,准确识别其状态能够提高机床的可靠性与安全性,并降低生产成本。为了提高状态识别精度,提出基于AdaBoost与局部类精度(LCA)的滚珠丝杠副状态识别方法。首先,利用历史失效数据与AdaBoost算法生成一个包含多个分类器的分类器集合。然后,针对未知状态的滚珠丝杠副,根据当前监测信号的特征,利用LCA算法从分类器集合中选出最合适的分类器对其当前状态进行识别。实验结果表明,所提出方法能够有效地识别滚珠丝杠副状态,其状态识别准确率高于传统方法,达到96.3%。

The ball screw is one of the important components of a numerical control tool,and its working condition is critical for the performance of the machine tool and the product quality.Therefore,identifying the health state of the ball screw accurately can enhance the reliability and safety of the entire machine tool and reduce the maintenance cost.To improve the condition recognition accuracy,a novel approach is proposed to identify the ball screw performance based on AdaBoost and local class accuracy(LCA).First,a classifier ensemble is created with AdaBoost according to the failure history data.Second,once the new condition monitoring data is obtained for ball screw in unknow state,the LCA is utilized to select the most suited classifier to recognize the current state of the ball screw.Experimental results show that the proposed method can identify the ball screw condition effectively with an accuracy of 96.3%,which is higher than the traditional techniques.