基于时间——高斯混合模型的刀具磨损预测

Prediction of Tool Wear Based on Time-gaussian Hybrid Model

连续刀具磨损过程的预测对实现自适应控制和优化制造工艺、提高生产效率和质量起着重要作用。为最大化描述磨损数据内在关系和提高预测模型精度,提出一种基于时间—高斯混合模型的刀具磨损建模方法,即采取将刀具磨损时间序列数据分成线性和非线性结构的策略,时间序列用来描述数据的线性相关趋势,非线性的异构部分则进行高斯过程回归建模。通过与现有的单一模型对比,结果表明该混合模型能对刀具磨损进行有效的预测,而且具有更高的预测精度。

The prediction of continuous tool wear process plays an important role in achieving adaptive control and op- timizing the manufacturing process, improving production efficiency and quality. To describe the inherent relationship among wear data and improve prediction accuracy, a method of tool wear modeling based on Time-gaussian mixture model is proposed ,which is a strategy to divide the tool wear time series data into linear and nonlinear structures. The time series is used to describe the linear trend, then The non-linear heterogeneity is modeled using Gaussiau process regression. DEFORM-3D software simulated the turning process and obtained tool wear data. Compared with the existing single model, the results show that the hybrid model can effectively predict the tool wear ,and has higher prediction accuracy.