Abstract: Fatigue stress concentration factors as fatigue limit of material fatigue resistance not only reflected the extent of fatigue stress concentration but also reflected for notch sensitive degree. The newly developed technical support vector machine(SVM) into the domain of fatigue stress concentration factor is introduced. Support vector machine theory is introduced, LIBSVM is used, and RBF is chosen as kernel function to build a mode. The input parameters are tensile strength, yield strength, fatigue strength, stress concentration factor，notch root radius, samples size and notch fatigue limit; Fatigue stress concentration factor is output. The calculations of SVM model, experimental formula Neural and Peterson are compared. The results show that the relative maximum predicting error is 0.7% under the condition of using a small quantity of samples to build the mathematical model. The accuracy and applicability for the present model are thus verified.

Key words: Fatigue stress concentration factor, Mechanics performance, Predict, Support vector machine