Abstract: Slurry pump in grinding and classification process is used as the equipment to transport pulp and provide inlet pressure required for classification of hydro-cyclone.Particles in slurry transported by slurry pump are too large and the pulp concentration and particle size are unknown, which are easy to cause accidents such as rotor blocking and short circuit, and the accidents are difficult to predict.In order to solve this problem, a fault prediction method of slurry pump based on multi-scale characteristic analysis of stator current is proposed by making full use of the operation information of slurry pump reflected in stator current, combined with EMD adaptive decomposition method and LSTM learning algorithm.The industrial data verification shows that the method can effectively predict the fault of mine slurry pump, and can help the production and technical personnel to give the decision-making basis for switching the standby pump in advance.