数据缺失下基于改进生成对抗填补网络的碳耗预测方法  

Prediction Method of Hobbing Carbon Consumption Based on Improved Generative Adversarial Imputation Net with Missing Data

作　　者：易茜[1,2] 柳淳 李聪波[1,2] 赵希坤 易树平[2] YI Qian;LIU Chun;LI Congbo;ZHAO Xikun;YI Shuping(State Key Laboratory of Mechanical Transmission,Chongqing University,Chongqing 400044;College of Mechanical and Vehicle Engineering,Chongqing University,Chongqing 400044)

机构地区：[1]重庆大学机械传动国家重点实验室,重庆400044 [2]重庆大学机械与运载工程学院,重庆400044

出　　处：《机械工程学报》2023年第11期264-275,共12页Journal of Mechanical Engineering

基　　金：国家自然科学基金(52005062);国家重点研发计划(2018YFB1701205)资助项目。

摘　　要：针对碳耗数据缺失导致碳耗预测模型预测精度低的问题,提出一种基于改进生成对抗填补网络的碳耗预测方法。以滚齿加工为对象,揭示加工过程的碳耗特性,分析其碳耗数据缺失机制;引入正则化机制构建生成对抗填补网络(GAIN)损失函数,提出基于正则化生成对抗填补网络(RGAIN)的碳耗数据填补方法;使用随机森林(RF)算法构造碳排放预测模型,实现数据驱动的加工碳耗预测。与其他数据填补及碳耗预测方法对比,该方法能有效降低滚齿碳耗数据缺失带来的预测误差。Aiming at the problem of low prediction accuracy of carbon consumption prediction model due to the missing data of carbon consumption,a prediction method of carbon consumption based on improved generative adversarial imputation net is proposed.Taking gear hobbing as an example,the carbon consumption characteristics of gear hobbing process are revealed,and the missing mechanism of carbon consumption data in gear hobbing process is analyzed.The loss function of generative adversarial imputation net(GAIN)is constructed by introducing regularization mechanism,and the carbon consumption data imputation method based on regularized generative adversarial imputation net(RGAIN)is proposed.Then,the random forest(RF)algorithm is used to construct a prediction model of hobbing carbon emission,and the dynamic prediction of hobbing carbon consumption is realized.Finally,the proposed method is compared with other data imputation and carbon consumption prediction methods.The results show that the proposed method can effectively reduce the prediction error caused by the missing carbon consumption data of gear hobbing,.

关 键 词：缺失数据 滚齿加工 碳耗预测 生成对抗网络 

分 类 号：TH122[机械工程—机械设计及理论]