Abstract: Aiming at the typical faults in the operation of belt conveyors， an accurate diagnosis method for double-layer fusion faults based on feature level and decision level is proposed. A fusion model for belt conveyor fault diagnosis information is established to extract the basic characteristics of the belt conveyor fault information and the wavelet packet feature and to realize the feature level fusion. Two classifiers including of the quantum particle swarm optimization kernel limit learning machine and the support vector machine are adopted to carry out the feature-level fault diagnosis. The D-S evidence theory is used to recombine the feature-level fault diagnosis results of the two classifiers to obtain decision-level fault diagnosis. The probability outputs of the two classifiers is used to construct the basic probability assignment function， which can effectively solve the construction of the basic probability assignment function in D-S evidence theory. The belt conveyor test bench was built and verified by MATLAB. The results show that the fault identification accuracy of the method can reach 97%， which improves the accuracy of fault diagnosis.

Key words: Belt conveyor, Accurate diagnosis of faults, D-S evidence theory, Nuclear limit learning machine, Support vector machines, Quantum particle swarm optimization