LAUSR

Abstract Low carbonization is an inevitable pathway toward the sustainable development of gear machining. Reliable and reasonable prediction of hobbing parameters can effectively reduce energy consumption and carbon emissions. The initial value of process parameters is often generated in a large range, which has a negative effect on the subsequent parameter prediction and the improvement of carbon emission, cutting time and cost. In this paper, a three-stage parameter prediction approach is proposed based on the similarity retrieval method, e-support vector regression (e-SVR) and an improved Harris hawks optimization for reducing the cutting time, cost and carbon emissions under data-driven conditions. The first stage is responsible for searching past machining cases similar to hobbing process problems. The e-SVR and Harris hawks optimization hybrid approach (SVR-HHO) is applied to predict the hobbing parameters in the second stage. Finally, the hobbing process parameters obtained in the second stage are revised using the improved multi-objective Harris hawks optimization for reducing carbon emissions, cutting time and cost. Compared with several well-established algorithms, SVR-HHO is evaluated against 9 benchmark datasets, performing better than 7 of them. A practical case of hobbing prediction is used to perform the feasibility verification and comparison study for the entire approach. The proposed approach achieved a relative carbon emission reduction of 9.3%, when compared to multi-objective HHO. The proposed approach obtained the optimum carbon emission, cutting time and cost compared with other approaches; therefore, it can adequately resolve the problem of the low-carbon hobbing parameter prediction. The performance of prediction stability and running time is slightly weak. Therefore, improving it is a major challenge for future research.