摘要
为提高切削力预测模型的准确性和稳定性,采用最优权值组合预测模型,以实验数据训练为基础,将人工神经网络模型、高斯过程回归模型及切削力经验公式这3个单一预测模型进行组合,对机械加工过程中的切削力进行预测。应用3种误差分析方法(均方根误差、平均绝对百分比误差及平均绝对误差)对组合预测模型进行评价,以此验证组合模型的稳定性和准确性。结果表明,相比于单一预测模型,组合算术平均模型所得结果与实验数据吻合良好,具有较高的精度和稳定性,对于切削力的预测具有参考价值。
To improve the accuracy and stability of the cutting force prediction model,adopted the optimal weighted combination prediction model,a model of combination forecasting with optimal weights was adopt. Three single prediction models which are artificial neural network model,Gaussian process regression model and cutting force empirical formula,were combined to predict the cutting force during machining based on experimental data. To evaluate the predictive stability and accuracy of the combined predicting models,three kinds of error analysis methods( root mean square error,average absolute percentage error,average absolute error) were used to evaluate several prediction models. The consequences show that the combination model is more consistent with the experimental data and has a higher accuracy and stability compared with the single forecasting model,which is pretty valuable for cutting forces forecasting.
作者
李康
鲁娟
马俊燕
周刚
黄文
廖小平
Li Kang;Lu Juan;Ma Junyan;Zhou Gang;Huang Wen;Liao Xiaoping(Institute of Mechanical Engineering, Guangxi University, Nanning 530004, China;Department of Mechanical and Marine Engineering, Beibu Gulf University, Qinzhou 535011 , Guangxi, China)
出处
《现代制造工程》
CSCD
北大核心
2019年第3期6-10,129,共6页
Modern Manufacturing Engineering
基金
国家自然科学基金资助项目(51665005)
关键词
切削力预测
组合预测模型
实验数据训练
高斯过程回归模型
经验公式
人工神经网络模型
cutting force prediction
combination forecasting model
experimental data training
Gaussian process regression model
empirical formula
artificial neural network model