Thermal error modeling of gear hobbing machine based on IGWO-GRNN

• Published in: International journal of advanced manufacturing technology Vol. 106; no. 11-12; pp. 5001 - 5016
• Main Authors: Liu, Zihui, Yang, Bo, Ma, Chi, Wang, Shilong, Yang, Yefeng
• Format: Journal Article
• Language: English
• Published: London Springer London 01.02.2020; Springer Nature B.V
• Subjects: Accuracy; Algorithms; Artificial neural networks; CAE) and Design; Computer-Aided Engineering (CAD; Engineering; Error reduction; Fuzzy logic; Gear hobbing machines; Global optimization; Hobbing; Industrial and Production Engineering; Local optimization; Machine tool industry; Machine tools; Machining; Mapping; Mechanical Engineering; Media Management; Neural networks; Original Article; Performance prediction; Prediction models; Workpieces; Gray wolf optimizer; Generalized regression neural network; Gear hobbing; Thermal error prediction
• ISSN: 0268-3768; 1433-3015
• DOI: 10.1007/s00170-020-04957-z

It is an increasingly urgent to improve the machining accuracy of the gear hobbing machine. Thermal error is the main source of the machining error of the hobbing machine, and reducing thermal error is necessary to improve the machining accuracy of hobbing machine. In this paper, a novel thermal error prediction model for the hobbing machine was proposed based on the improved gray wolf optimizer (IGWO) and generalized regression neural network (GRNN). The fuzzy cluster grouping and mean impact value (MIV) were firstly combined to select the typical temperature variables and reduce the coupling between temperature variables, so the robustness of the thermal error model can be guaranteed. Then GRNN was used to establish the mapping relationship between temperature variables and thermal error. The IGWO considering the proportion of local optimization and global optimization was applied to optimize the smoothing parameter of GRNN. Finally, the proposed IGWO-GRNN was used to predict the thermal drift of the workpiece shaft of the dry cutting hobbing machine, and its predictive accuracy and generalization performance were compared with four existing algorithms. The results indicate that the prediction accuracy of IGWO-GRNN is at least 5.1% higher than other algorithms and its generalization performance is also promoted.