Abstract: Concentrated on the distortion problem of milling force coefficient calculation caused by the different cutter tooth radii in thin-walled parts milling, actual milling force coefficient is constructed, and is predicted through kernel partial least square regression method when milling parameters change. To construct actual milling force coefficient, theoretical milling force coefficient is deduced for two-teeth spiral milling cutter, nominal milling force is defined based on the milling process with cutter of different tooth radii, and cutter tooth radius error is deduced. Prediction model of actual milling force coefficient about milling parameters and their compositional variables, is established in high dimension space through kernel partial least square regression method based on the prominent nonlinear analysis and prediction ability of nuclear analysis; kernel parameters of Gaussian kernel function and the number of kernel principal components, whose impact on prediction model is analyzed, are set in certain value range in kernel partial least square regression method. At last, milling force coefficients are analyzed respectively when cutter tooth radius error is considered or not, and kernel partial least square regression prediction method and partial least square regression prediction method are compared; according to above data, the proposed construction and prediction methods of milling force coefficient have high calculation accuracy and prediction ability.

Key words: nominal milling force, theoretical milling force coefficient, actual milling force coefficient, asymmetrical teeth, kernel partial least square regression