A Data-Driven Design Approach for Carbon Emission Prediction of Machining

Abstract

Abstract The issue of carbon emission reduction for manufacturing industry attracts increasing attention. As a major contributor in the manufacturing industry, machining has generated large amounts of carbon emissions through the resource consumption, energy consumption, and waste disposal. The carbon emission prediction of machining is a priori technology for its reduction, and has been established as one of the most crucial research targets. The purpose of this study is to design a carbon emission prediction model of machining through a data-driven approach. First of all, the multiple sources and impact factors of carbon emissions in machining are studied, and the relationship between these factors is also studied to describe the carbon emissions. Then, a data-driven approach is designed to predict the carbon emission of machining, which consists of data collection and preprocessing, feature extraction, prediction model establishment and model validation. The ridge regression, BP neural network based on Genetic Algorithm (GA-BP), root means square error (RMSE) and mean relative percentage error (MPAE) are respectively employed to fulfill the above tasks in the design approach. Finally, an experimental study of a real turning machining is proposed to verify the feasibility and merits of the designed approach.