Abstract: The thermal error of CNC machine tools is one of the main factors that reduce the accuracy of the machining operation. For the establishment of the thermal error model, combined with the cuckoo algorithm random Levi flight mechanism, the least squares support vector machine structure risk minimization and linear programming advantages, a thermal error modeling method is proposed based on the least squares support vector machine which optimized by cuckoo algorithm. When the least squares support vector machine transforms the low-dimensional nonlinear problem into the high-dimensional linear problem, a hybrid kernel function is constructed, and the cuckoo algorithm is used. The penalty factor γ, the kernel width parameter σ and the mixed kernel weight λ of the least squares support vector machine is optimized by cuckoo algorithm. Taking the GMC2000A machine tool as the experimental object, clustering analysis and modeling analysis of the thermal error data are carried out. The prediction comparison analysis concluded that the error model established based on the cuckoo algorithm to optimize the hybrid kernel least squares support vector machine has achieved good prediction results, and is significantly better than the predictive effect of BP neural network model and the unoptimized least squares support vector machine model.

Key words: cuckoo algorithm, least squares support vector machine, thermal error, modeling, prediction