Abstract: At present, the malfunction diagnosis of dust collector is mainly based on manual experience judgment and combined with downtime check, which leads to some problems such as low diagnostic efficiency, and lack of science and automation.This paper analyzed four main fault types including filter core damage, ashing failure, filter clogging and ash discharge failure, then selected five diagnostic parameters including dust emission concentration, filtration resistance, inlet air volume, air leakage rate and air consumption to establish BP and RBF neural network prediction model.The example analysis shows that the BP neural network model has fast convergence speed, ideal prediction effect and can accurately judge the fault type of the dust collector.The average prediction model errors of filter core damage, ashing failure, filter clogging and ash discharge failure are 0.035%, 0.110%, 0.118%, 0.215% respectively.The prediction result is better than RBF neural network.