Electronic Science and Technology ›› 2022, Vol. 35 ›› Issue (1): 29-34.doi: 10.16180/j.cnki.issn1007-7820.2022.01.005

Previous Articles     Next Articles

Research on Load Forecast of Ball Mill Based on PCA-PSO-SVM

FENG Xianding,WEI Jingtao,WU Zhangyong,QIAN Jie,PU Youshang   

  1. Faculty of Mechanical and Electrical Engineering,Kunming University of Science and Technology,Kunming 650500,China
  • Received:2020-08-21 Online:2022-01-15 Published:2022-02-24
  • Supported by:
    National Natural Science Foundation of China(51165012)


As the main equipment in the grinding production, the internal load of the ball mill cannot be directly detected during operation. Therefore, it is difficult to effectively control the load in real time, which leads to the reduction of grinding efficiency. To solve this problem, the grinding sound signal of the ball mill is collected through the grinding experiment, and the signal is analyzed by the Welch power spectrum, and the relationship between the grinding sound spectrum information and the load of the ball mill is investigated in the proposed study. PCA is used to extract the features of the power spectrum to provide external feature information for load prediction of the ball mill. PSO is adopted to optimize the relevant parameters of SVM and the load prediction model of PA-PSO-SVM ball mill is established. Research shows that the prediction root mean square error of the ball mill prediction model is 1.144 3, the average absolute error is 0.912 5, and the average percentage error is 2.797 9%, which proves the effectiveness and stability of the model for predicting the load of the ball mill.

Key words: ball mill, grinding sound signal, signal processing, Welch power spectrum analysis, PCA, PSO, SVM, load forecasting

CLC Number: 

  • TP312