Taking advantage of adaptive learning classification of wavelet neural networks, this paper proposes multiple wavelet neural networks used in parallel to improve the wavelet network structure. It also determines the number of wavelet neuron and the appropriate initialization wavelet in each wavelet feature space. The X-ray energy spectrum of drugs and explosives are identified by multiple wavelet neural network. Experiments show that the identification of different types of drugs and explosives can be achieved by multiple wavelet neural network, which provides an effective method for the technology of X-ray energy dispersive used in detection and identification of drugs and explosives.