Predicting running-in wear volume with a SVMR-based model under a small amount of training samples

This paper proposes a support vector machine regression (SVMR) model to predict running-in wear volume, with field surface topography parameters and working conditions as the input variables, under the condition that the amount of training samples is very limited. Experimental results proved the effectiveness of the SVMR-based model with a small amount of training samples. Based on the established prediction model, the impacts of the field surface parameters on running-in wear volume have been analyzed. The results show that Sku has the largest influence on running-in wear volume, Sdq the second, and Svk the least.