The Detection Of Alumina Clinker Based On Wavelet Packet Analysis And BP Network

Alumina is the main raw material of modern electrolytic aluminum production process.Growth characteristics of ore in China, so our country mainly produce alumina using mixture combination and sintering procedure.The production of using sintering procedure for alumina output is accounted for about50%of the total production of alumina in China, the sintering procedure is the core of clinker sintering technology, the core equipment of sintering procedure is the rotary kiln. In the procedure of rotary kiln production, the sintering procedure of alumina clinker had the disadvantage of parameters coupling is serious, the main technological parameters of detection will be hard and there is a big lag of time and so on.However, at present most of the rotary kiln for alumina clinker sintering is in a manual or semi-automatic control state in our country, so the identification condition of alumina clinker sintering is a big difficulty.The technology of ingredients for sintering procedure is a three-step method and the sintering temperature corresponding with it is hard controlled.For now, the judge of alumina sintering methods is mainly the method of artificial observation, the method are the shortcomings of low efficiency and bad accuracy, meanwhile it can not meet the efficient and safe production of modern industrial mode.Firstly, we established sound signal acquisition system to collect the sound signal produced by alumina clinkers collide the rotary kiln wall, and accessed to the different collision sound signal of sintered alumina clinker conditions using experiment. Based on the collision sound signal time domain and frequency analysis of the power spectrum analysis, time-domain statistical characteristics in the mean square value of the alumina sintered state is found to have a clear correspondence between the alumina sintering conditions associated with the characteristic frequency segment is1.75kHz-3.25kH. Based on wavelet packet analysis, first voice signal into eight different frequency range, and a percentage of the energy range of different frequency bands representing the total energy of the signal recognition clinker sintering conditions as feature vectors. Then we explored the the recognition of clinker sintering conditions based on BP neural network recognition. Finally, we developed wavelet analysis and BP neural network clinker detection system using LABVIEW and MATLAB mixed programming technology, and initially realized the recognition of clinker sintering conditions.