基于传感器信号融合和PSO-LSSVR的刀具磨损预测研究

Tool Wear Prediction Based on Sensor Signal Fusion and PSO-LSSVR

最小二乘支持向量回归模型能有效地构建特征与刀具实时磨损间映射关系,但受限于自身参数的复杂性,无法获得最优模型性能,同时由于单个传感器的局限性,难以全面反映刀具磨损的多维信息。为进一步提升刀具磨损预测精度,提出一种基于多传感器信号融合并利用粒子群算法优化参数的最小二乘支持向量回归模型的刀具预测方法。通过对采集的传感器信号进行小波降噪,提取可用于反映刀具磨损的多域特征,并通过核主成分分析法对多域特征进行降维融合,采用经过粒子群算法优化参数的最小二乘支持向量回归模型构建融合后特征与刀具磨损的映射关系。通过公开数据集进行的实验,表明该模型具有较高的预测精度,验证了所提出预测方法的有效性。

The least squares support vector regression can effectively construct the mapping relationship between features and real-time tool wear.However,it can't obtain the optimal model performance due to the complexity of its own parameters.And it’s difficult to fully reflect the multi-dimensional information of tool wear because of the limitations of a single sensor.In order to improve the accuracy of tool wear prediction,a tool wear prediction model based on multisensory information and LSSVR is proposed.In this method,multi-domain features are extracted from the denoised multi-sensor signals,then the multi-domain features are reduced and fused by using kernel principal component analysis.The LSSVR model is adopted and the parameters are optimized by particle swarm optimization.In the end the mapping model between dimension reduction features and tool wear is constructed by training the model.The experimental results show that the model has high prediction accuracy,which verifies the effectiveness of the proposed tool wear prediction method.