In term of dependency of professional knowledge in signal processing, it is necessary and urgent to perform intelligent fault diagnosis for modern industry based on machine learning. Therefore, an intelligent diagnosis method based on support vector machine is proposed for structure fault diagnosis. Firstly, the collected raw signal is translated into divided into frequency domain; and the signal in time domain and frequency domain are divided into segments according to the sampling frequency and the rotating speed of machine. Secondly, those segments are extracted eigenvalues in time domain and frequency domain, respectively. Thirdly, the extracted eigenvalues are inputted into support vector machine to perform fault diagnosis. Inhere, the cross-validation is utilized to find the optimal parameters (kernel function and penalty factor) for support vector machine. And then, the diagnosis model is trained with the obtained optimal parameters from the cross-validation. Finally, the test data is inputted into the trained diagnosis model to test the effectiveness of the proposed method. The experiment results show that the effectiveness of the proposed method when it is applied into structural fault diagnosis.