Abstract: Heat treatment process data contains long-term operation experience of enterprises, which can reveal the relationship between process conditions, parameters and carbon emissions. However, the process data are often fragmented and stored separately during the production process. Therefore, a BRBP-MOSOA hybrid data-driven method based on historical process data sets is proposed to optimize the parameters of heat treatment process in order to reduce carbon emissions. Firstly, the heat treatment process history database is constructed; according to the carbon emission operation characteristics of heat treatment process, the carbon emission sources are identified and the emission model is built. Then, the historical process data set are used in training BRBP network, which is used to reveal the relationship between heat treatment process parameters, hardness and carbon emissions, as well as predicted the carbon emissions and hardness under specific process conditions. The process parameters optimization model is established by using MOSOA optimization algorithm, and the optimal heat treatment process parameters satisfying low carbon emissions are output through carbon efficiency evaluation. The case study shows that the comprehensive carbon efficiency optimization of heat treatment reaches 6.57%, which can guarantee the product performance and reduce the carbon emission of heat treatment. An enabling tool for low carbon operation of heat treatment process is provided.

Key words: data-driven, heat treatment, low carbon, optimization