多传感器融合下多工况刀具磨损状态预测的深度森林方法研究

准确监测加工过程刀具磨损状态有助于避免因刀具失效导致的产品质量问题。 建立不同工况的刀具磨损监测模型,往 往需要对每组工况调参以保证精度。 为减少调参并保证预测精度,结合深度森林的超参数少、参数对模型不敏感和训练过程自 适应等优点,利用深度森林建立了多传感器信号及多工况下自主特征选择的刀具磨损状态预测模型。 基于 3 组不同工艺参数 下 TC18 铣削过程的多传感器及磨损数据,以及预测与健康管理(PHM)学会 2010 年高速数控机床刀具健康预测竞赛的开放数 据,深度森林在 3 组工况的预测精度分别为 95. 35% 、96. 63% 和 97. 06% ,在 PHM 数据上为 98. 95% ,验证了深度森林对多工况 下刀具磨损预测的高精度和适用性,为在线监测技术提供了有力的指导。

Research on the deep forest method for tool wear state prediction under multiple working conditions with multi-sensor fusion

Accurate monitoring of tool wear during machining helps to avoid product quality problems caused by tool failure. To formulate tool wear monitoring models for different working conditions, it is necessary to adjust the parameters for each group of working conditions to ensure the accuracy. To reduce the number of parameter adjustment and ensure the prediction accuracy, the advantages of deep forest are combined, such as few hyperparameters, parameter insensitivity to the model and adaptive training process. A tool wear state prediction model with multi-sensor signals and autonomous feature selection for multi-conditions is established by using deep forest. The multi-sensor and wear data of TC18 milling process under three sets of different process parameters, and the open data in the predictive and health management (PHM) society 2010 high-speed CNC machine tool health prediction competition are utilized. For the three sets of working conditions, the prediction accuracy values of deep forest are 95. 35% , 96. 63% and 97. 06% , respectively, and 98. 95% on PHM data, which evaluate the high accuracy and applicability of deep forest for tool wear prediction under multiple working conditions. It provides a strong guidance for online monitoring technology in intelligent machining technology.