An incremental data-driven approach for carbon emission prediction and optimization of heat treatment processes

In manufacturing enterprises, the variety of heat treatment processes, insufficient data in the early stages of production, and the continuous increase of samples over time make it difficult to predict and optimize carbon emissions dynamically. Therefore, this paper proposes a low-carbon optimization method for heat treatment processes based on incremental data-driven approaches. Using life cycle assessment (LCA) theory, carbon emission sources are accurately analyzed and quantified, and a full life cycle carbon emission model is established. The key process parameters affecting part performance and carbon emission were screened through mechanism analysis, and the incremental data were fused by the Elasticity Weight Consolidation (EWC) algorithm to establish an EWC-BPNN heat treatment carbon emission prediction model. Then the model is solved by the Multi-objective Seagull Optimization Algorithm (MOSOA), and the result achieves a good balance between the carbon emission and hardness of the heat treatment process. Finally, using real data collected in enterprises for four consecutive years, it is verified that the predictive model can be effectively applied to the working conditions of multi-variety, small batch, and incremental heat treatment data.