The limitation of single fault signal for roller bearing fault diagnosis and nonlinear relation of fault features were studied. Starting from theory of information fusion

the method based on feature fusion of empirical mode decomposition (EMD) entropy was proposed using nonlinear dynamics parameters of entropy as features to deal with roller bearing fault diagnosis problem. Firstly

EMD was conducted for original signal

and on the basis of the property of adaptive multi-resolution for the EMD

different entropies of the intrinsic mode function (IMF) signal reconstructed by using the EMD were calculated. Secondly

the information fusion of the state features was further implemented by using the kernel principal component analysis (KPCA) to extract the complementary feature. Finally

the support vector machine (SVM) was employed to diagnose the fault by using the extracted fusion features. The experiment of rolling bearing fault diagnosis shows that the proposed method combines EMD

information entropy theory and strong nonlinear processing of KPAC

so it can be used for roller bearing fault diagnosis.