Research On Mill Load Detection Method Based On Audio Analysis

The ball mill is a material crushing equipment in the cement production industry.Due to the lack of reliable mill load detection methods,the ball mill has been in a working state with high power consumption and low efficiency for a long time.In addition,when the ball mill is working,it is impossible to install a load measuring device inside it.Therefore,the load detection of the ball mill usually adopts indirect measurement.The research shows that the mill load detection method based on the grinding signal is the most effective detection method at present.However,the source of noise is very complex in the ball mill and leads to the low accuracy of traditional grinding signal analysis and processing method.Therefore,it is of great significance to study a more effective method for mill load detection.This thesis investigates a mill load detection method by analyzing the audio signal of the ball mill.In the thesis,the Independent Component Analysis(ICA)method and the wavelet packet energy spectrum method are applied to the extracting and processing of grinding sound signals.And according to the processing results,a method of mill load detection based on the analysis of audio signal is proposed.First of all,it turns out that the original signal of the ball mill site is independent of each other by studying the working environment of the ball mill.The frequency spectrum and energy of grinding signal are different under different load states of the ball mill.Then,ICA algorithm is used to extract the original grinding signal from the audio signal collected by multiple audio sensors.In order to obtain each frequency band of the grinding signal,the original grinding signal is analyzed by wavelet packet energy spectrum with the wavelet basis function of sym10.And it can calculate the energy and sound intensity values of the grinding signal in each frequency band.By analyzing the grinding signal of ball mill under different load state,the most sensitive characteristic frequency band with the change of the mill load is obtained.Finally,the purpose of detecting mill load achieved by establishing the relationship with grinding intensity and mill load in the characteristic frequency band.Finally,the experiments validate the method proposed in this thesis.The verification results show that the algorithm based on independent component analysis proposed in this thesis can effectively extract the original grinding signal.And the signal-to-noise ratio of the grinding signal is enhanced compared with the traditional method.The processing method of grinding signal based on wavelet packet energy spectrum can effectively realize the processing of grinding signal under different mill loads.And it can get the most sensitive characteristic frequency band with the change of mill load.The load state of the mill can be effectively detected by the relationship with grinding intensity and mill load in the characteristic frequency band.The objective of mill load detection can be achieved within the margin of error.The methods in this thesis provide a theoretical basis for the load detection of the ball mill,and it has important significance for the development and application of the load detection theory of the ball mill.