In order to solve the problem that the artificial selection of the Convolutional Neural Network tructure in the rolling bearing fault diagnosis is uncertain and the diagnosis accuracy is low, based on the CNN, this paper proposes an applied Genetic Algorithm (GA) GA-CNN, a new method for fault diagnosis of rolling bearings that adaptively selects CNN network structure. This paper first uses Hilbert-Huang Transform (HHT) to extract features of rolling bearing fault signals, and then input the fault features into CNN improved by GA and three groups of CNN with artificially randomly selected network structures for feature recognition. Finally, a conclusion is drawn by comparing the experimental results. GA automatically selects the best structure of CNN network, avoiding the uncertainty of manual selection of CNN network structure, thereby reducing the time required for parameter selection and improving the accuracy of rolling bearing fault diagnosis. Experimental verification shows that the GA-CNN-based rolling bearing fault diagnosis method proposed in this paper greatly improves the efficiency of fault diagnosis and has higher accuracy compared with manual random selection of CNN network structure.