Prediction Method for Carbon Emission of Hobbing Based on Cross-Process Data Fusion

Abstract Accurate prediction of manufacturing carbon emissions is of great significance for subsequent low-carbon optimization. To improve the accuracy of carbon emission prediction with insufficient hobbing data, combining the advantages of improved algorithm and supplementary data, a method of carbon emission prediction of hobbing based on cross-process data fusion was proposed. Firstly, we analyzed the similarity of machining process and manufacturing characteristics and selected milling data as the fusion material for hobbing data. Then, the adversarial learning was used to reduce the difference between data from the two processes, so as to realize the data fusion at the characteristic level. After that, based on Meta-Transfer Learning method, the carbon emission prediction model of hobbing was established. The effectiveness and superiority of the proposed method were verified by case analysis and comparison. The prediction accuracy of the proposed method is better than other methods across different data sizes.