Abstract: A new method of fault diagnosis and detection based on GA and IDBN is proposed to identify the acoustic emission signals of rolling bearing faults under variable working conditions. The experimental results show that the GA-IDBN model is superior to the original deep belief network and support vector machine model in the recognition accuracy of the acoustic emission monitoring signals of the outer ring fault and the inner ring fault and cage fault of rolling bearings.And the accuracy can reach more than 95.5%. Moreover, GA-IDBN model is universal, and can recognize the running state of different channel and speed. It is proved that GA-IDBN model has a strong practical value in engineering.