A prior knowledge-integrated method of carbon emissions modeling and optimization for gear hobbing with small sample problem

With an increasing seriousness on global warming, low-carbon sustainable manufacturing has become the focus of the manufacturing industry. Selection of optimum process parameters is an effective method of decreasing carbon emissions, and carbon consumptions modeling in manufacturing process is an important link to realize parameter optimization and low-carbon control. To solve the problem of insufficient effective historical data in machining process, a prior knowledge-integrated method of carbon emissions modeling and optimization with small sample problem was proposed. Aiming at the gear hobbing process, knowledge from the manufacturing data was extracted, and a neural network prediction model integrated by prior knowledge (PKNN) was established, whose optimal parameters are solved by the augmented Lagrange multiplier method. Then, a multi-objective optimization model is proposed to take the minimum carbon emissions and cost as the optimization objectives, which is solved by a modified multi-objective gray wolf optimization (MOGWO) algorithm. The results demonstrate that the predictive model is effectively applicable to the conditions of small sample manufacturing data, and the combination of the predictive model and optimization algorithm can contribute to the low-carbon control of the gear hobbing process.