Precision prediction of tooth surface roughness in dry hobbing using limited sample data: An approach with potential for enhancing gear surface quality

Abstract

Tooth surface roughness serves as one of the critical characterization indices for evaluating machining quality in dry hobbing. In this research, a precision prediction method with limited sample data is proposed for predicting tooth surface roughness in dry hobbing. Firstly, based on the envelope forming characteristics of dry hobbing, the formation mechanism of tooth surface roughness and its influence on gear transmission performance are clarified by analyzing the cutting vibrations and machine tool error mechanism. Secondly, a prediction model based on Support Vector Machine (SVM) is developed for tooth surface roughness in dry hobbing. In this SVM-based model, the model parameters (C, ε, and γ) are determined through a hybrid optimization strategy combining grid search and random search. Finally, experimental tests on a representative helical gear with large helix angles, shallow tooth depths, and a large number of teeth were conducted to verify the proposed method. The results demonstrate a relative error of less than 7% between measured and predicted values, which validates the feasibility and reliability of the prediction method. Sensitivity analysis further reveals that tooth surface roughness value is negatively dependent on spindle speed and positively dependent on feed speed. This study thereby provides theoretical support for the optimal control of tooth surface roughness in dry hobbing.