In order to monitor and identify the failure mode of the alloy head of the pick-cutting, a method based on BP neural network is proposed to deal with the multi-feature signals. The chapped, off, chipping and serious wear alloy head are tested in cutting process to extract the maximum, mean and variance of the three directional vibration characteristic signals and current signal of the cutting motor. The BP neural network is learned and trained by multi-feature signals to establish the recognition model and monitor and identify the failure mode on-line. The experimental results show that the results of BP neural network are consistent with the actual failure mode of the test samples. It provides new methods to realize online monitoring and identification of failure mode for shearer picks.