Mining machine cutting load classification based on vibration signal

Abstract

There are some errors and lags in the way of judging the cutting load type of the mining machine manually. In order to solve the above problem, a classification method of mining machine cutting load based on wavelet packet decomposition and sparrow search algorithm optimized BP neural network (SSA-BPNN) is proposed. The method comprises two parts of signal feature extraction and mode classification. In the part of signal feature extraction, the collected vibration signal of the mining machine rocker arm is decomposed by wavelet packet to obtain the energy of each subband and the total energy of the signal. After normalization, feature vectors representing different load types are obtained. The principal component analysis is used to reduce the dimensions of the feature vector. In the mode classification part, SSA is used to optimize the initial weight and threshold of BPNN. The feature vector is used as the input of SSA-BPNN to realize the load classification and recognition. Taking the MG500/1170-AWD1 mining machine as an object, the magnetic acceleration sensor is attached to the shell of the rocker arm of the mining machine near the bracket side. The vibration signals of the mining machine drum under three working conditions of no-load, cutting bauxite and rock are collected and tested. The experimental results show that the vibration signals under different cutting loads have some differences in the energy of each sub-band. This result indicates that the energy features obtained by wavelet packet decomposition can be used as feature vectors to distinguish different load types. Compared with BPNN, SSA-BPNN has faster convergence speed and higher recognition accuracy, and the recognition accuracy of load classification is 95.3%.

Keywords

• mining machine
• cutting load classification
• rocker arm vibration signal
• wavelet packet decomposition
• sparrow search algorithm
• bp neural network