Application Of Hidden Markov Models In Ball Mill Gearbox For Fault Diagnosis

The traditional fault diagnosis methods of bearing is based on the staticobservation, which often ignores the context information before failure occurs andcan not reveal the potential variation characteristics of system state.With theexpanding of HMM research applications in rotating machinery, HMM methodwas tried to in a ball mill gear for fault diagnosis in this paper. The observationand assessment of the ball mill gear reducer device in a dynamic environment. Inthe power and other mineral separation industries, ball mill is critical equipmentwhich grinds coal, To ensure its normal operation is improving the economicefficiency of enterprises indirectly. Therefore the fault diagnosis in ball mill gearis necessary and meaningful.Firstly, the simple exposition of that the research status in ball mill gear andthe feature of gear vibration signals provides a theoretical basis for the vibrationsignal feature extraction later. The basic theory and algorithm learning of HMMand the process of DHMM fault diagnosis in the mill gear were introduced. Thevibration signal was required a series transformations which are feature extraction,normalization, scalarization and quantization to get the sequence collections. Thenthe quantified sequence collections were trained to get the DHMM parameter, orthe Viterbi Algorithm which was used for the quantified sequence collections tocalculate the maximum probability, thereby the DHMM fault models library wasestablished or the type of fault was recognized. Experiments of five kinds of faultmodel diagnosis were carried out in this article. The results shows that theapplication of DHMM in a ball mill gear reducer for fault diagnosis has a highclassification accuracy. The multi-channel vibration signals were fused in ChapterIV, The basic theory and algorithm learning of SOM clustering coding wereintroduced. SOM clustering coding was used to achieve dimensionality reduction,coding, single and double channel DHMM models were established. Thecomparing between single and double channel fault diagnosis results shows thatthe effect of dual-channel fault diagnosis higher than the single channel. Finally, HMM fault diagnosis software was developed based on the Qt and matlab tofacilitate fast ball gear vibration signal analysis. The development environment,development tools and interface implementation were introduced.