LS-SVM-based surface roughness prediction model for a reflective fiber optic sensor

Abstract

Reflective fiber optic sensors have advantages for surface roughness measurements of some special workpieces, but their measuring precision and efficiency need to be improved further. A least-squares support vector machine (LS-SVM)-based surface roughness prediction model is proposed to estimate the surface roughness, Ra, and the coupled simulated annealing (CSA) and standard simplex (SS) methods are combined for the parameter optimization of the mode. Experiments are conducted to test the performance of the proposed model, and the results show that the range of average relative errors is −4.232%–2.5709%. In comparison with the existing models, the LS-SVM-based model has the best performance in prediction precision, stability, and timesaving.