Mill Load Identification Method for Ball milling Process Based on Grinding Signal. Issue 21 (2021)

• Record Type: Journal Article
• Authors: Wang, Xiaoli; Sun, Kaixuan; Zhang, He; Xiong, Wei; Yang, Chunhua
• Is Part Of: IFAC-PapersOnLine. Volume 54:Issue 21(2021)
• Page Start: 7
• Page End: 12
• Publication Date: 2021
• Subjects: Mill load -- Grinding sound -- Signal processing -- Spectral subtraction -- Extreme learning machine; Automatic control -- Periodicals; 629.805
• DOI: 10.1016/j.ifacol.2021.12.002
• Languages: English
• ISSNs: 2405-8963

Abstract

Abstract: Mill load is a key parameter for the safety and optimal control of the grinding process in mineral processing. Grinding sound signal is usually used to detect the mill load indirectly. However, the relationship between grinding sound signal and the mill load is really complicated. In this study, a mill load identification method is proposed by combining the machine learning algorithm with the sound recognition technology. Firstly, a geometric spectral subtraction denoising method based on auto regressive (AR) spectrum estimation is proposed to preprocess the grinding sound signal. Then, the ensemble empirical mode decomposition (EEMD) method is used to decompose the grinding signal. Suitable IMF components are then selected to reconstruct the grinding signal and the box fractal dimension feature is extracted. Finally, an optimized extreme learning machine (ELM) method was proposed to identify the mill load. The simulation results using industrial data show that the proposed method has better overall recognition accuracy compared with other machine learning methods.