Abstract: At present in the process of mine production, the shearer rocker arm high failure rate, slow fault identification, brought certain safety hazard to underground production, in order to solve this problem, this paper to the transmission system fault gear and bearing as the research object, starting from the principle of fault, construction of shearer rocker arm fault identification model, through the depth migration learning, determine the global parameters of the convolution layer network, test the model of fault identification accuracy reached 95%, to the safe operation of shearer and mine safety production play a certain role.