Research On The Detection Method Of Mill Load Based On Spectrum Analysis

The mill is the core equipment used to crush materials in industrial production. With frequent crash between the steel ball and materials inside the mill, it will end up completely crushing all the stuff in the ball mill. To ensure the efficiency and safety of the whole operation process, testing the working conditions of the ball mill comes as our priority. The ball mill load including the quantity of material and the material level is an important detection index of the grinding process, which directly affects the work efficiency. However, the working condition inside the ball mill is complex and changeable, it is difficult to ensure the stable operation all the ways, and it also brings great obstacles to the ball mill load detection. At the present stage, the load is estimated by the way of manual monitoring, some errors will be caused sometimes. And the traditional detection method cannot accurately detect the state of the mill load.From our research, we study further into the traditional detection method, which combines the Principle Component Analysis(PCA) and the Extreme Learning Machine(ELM). Then we have analyzed the defects in this traditional method from three aspects, including spectrum analysis, principal component extraction and model building, we finally put forward the corresponding solutions on our own. In order to combine with the characteristics of mill load measurement, we apply the kernel principal component analysis(KPCA) and the error minimized extreme learning machine(EM_ELM) to our mill load detection model, which is based on the spectrum analysis method. This method is combined with wavelet package for preprocessing, and uses maximum entropy method of modern power spectrum estimation method to transform the signals to frequency domain for analyses. This method can establish the relationship model between the internal and external state of mill load signal detection by using indirect detection method. Finally, we have combined the data of the ball mill working in the industrial field, the test results are then compared with the traditional PCA-ELM detection method.The results show that using the test method which is based on spectrum analysis with KPCA-EM_ELM method to detect has greatly improved the accuracy of measurements and the running time of the algorithm has also been reduced. It also improves the efficiency and accuracy. The method provides a theoretical basis for applying this detection method to the actual ball mill load detection system. And this method is of great significance and broad application prospects to improve the grinding efficiency and stable production of the ball mill.