多维特征数据驱动的数控车削加工过程碳排放预测研究

Research on Multidimensional-Feature Data-Driven for Carbon Emission Prediction of CNC Turning Process

针对数控车削加工过程碳排放影响要素繁多、动态特性复杂的特点,提出了一种基于多维特征数据驱动的数控车削加工过程碳排放预测方法。首先,对数控车削加工过程碳排放特性进行分析,建立了包含原材料消耗、辅助材料消耗、能源消耗和废弃物回收处理的总碳排放量计算模型,确定了碳排放不同维度的影响要素;其次,针对不同影响要素的类型,提出了数控车削加工过程碳排放数据的采集、预处理方法,利用岭回归方法对数据主要特征进行选择提取;再次,以提取的特征数据为自变量,提出了一种基于改进的果蝇-差分进化优化BP神经网络算法的数控车削加工过程碳排放预测模型;最后,通过实验对所提方法和模型的有效性进行了验证。

Aiming at the multi-influence factors and complex dynamic carbon emission characteristics of CNC turning process,a multidimensional-feature data-driven prediction method for carbon emission of CNC turning process was proposed.Firstly,the carbon emission characteristics of CNC turning process were analyzed,the calculation model of total carbon emission including raw material consumption,auxiliary material consumption,energy consumption and waste recycling was established,and the multidimensional influencing factors of carbon emission were determined.Secondly,aiming at the different types of influencing factors,the carbon emission data collection and data preprocessing methods of CNC turning process were proposed,and the main features of the data was extracted by ridge regression method.Thirdly,taking the extracted feature data as independent variables,a carbon emission prediction model for CNC turning process based on improved FOA-DE optimization BP neural network algorithm was proposed.Finally,the effectiveness of the proposed method and model was verified by experiments.