www.as-se.org/ccse
Communications in Control Science and Engineering (CCSE) Volume 3, 2015
Qualitative and Experimental Analysis of Ball Mill Shell Vibration Production Mechanism Jian Tang*1,2,a, ZhuoLiu2,b, Zhiwei Wu2,b, Xiaojie Zhou2,c Research Institute of Computing Technology, Beifang Jiaotong University, Beijing, China
1
State Key Laboratory of Synthetical Automation for Process Industries, Northeastern University, Shenyang, China
2
powernature@126.com; bzhuoliu_mail@neu.edu.cn; czwwu_mail@neu.edu.cn; dxj_zhou@neu.edu.cn
*a
Abstract Ball mill is a type of rotating heavy mechanical device in grinding process with characteristics of continuous running and closing working. Its load parameters have direct relation with production quality and grinding process safety. Strong shell vibration and acoustic signals are normally used to measure ball mill load. However, multi-component and non-stationary characteristics of these signals are very difficult to be explained under different grinding conditions. In this paper, an integrated analysis method is given out. It is based on the qualitative production mechanism analysis of the shell vibration production and ensemble empirical mode decomposion results to a laboratory-scale ball mill vibration signals with certain domain expert experiences. Results show that the quanlitative analysis conclusion is consistent with the experimental decomposition results. This research makes a foundation for accurately quatitative and mathematical simulation of the shell vibration production mechanism. Further, it is valuable to construct soft sensing mode with clear interpretation and high prediction accuracy. Keywords Vibration Production Mechanism; Quanlitative Analysis; Ensemble Empirical Mode Decomposion; Ball Mill
Introduction Ball mill is a type of widely used heavy rotating mechanical devices. Accurate measure load parameters within ball mill on time is important for ensuring safety, product quality and quantity of mineral grinding process. Numerous approaches have been used to address this issue. These methods include: direct measuring approach using Sensomag instrument [1], mathematical calculation approach based on first principal model; load status identification approach based on expert experiences; indirect data-driven soft measuring approaches based on mill shell vibration and acoustic signals [2,3], or based on mill shaft vibration signal and other measured process variables [4]. Recently, the indirect data-driven soft measuring method based on mill shell vibration has been a new focus. However, detailed production mechanism of the mill shell vibration and reasonable interpretation of these soft sensor models are far more understand. In practice, millions of balls inside the ball mill arrange hierarchically. Impact forces and periods of different layers’ balls to mill shell are different, which cause strong mechanical vibration. Shell vibration is the main source of the acoustical signal. Thus, shell vibration and acoustic signals have characteristics of non-stationarity and multicomponent. The shell vibration signals under different grinding conditions have different characteristics. Discrete element method has been widely used in dry ball mill (only ball and material in the mill), which has also been used to analyse shell vibration of the dry mill. However, most of the mathematical simulation models aren’t suitable for wet ball mill (ball, material and water load in the mill) [5]. Empirical mode decomposition (EMD) can adaptively decompose the original signal into some intrinsic mode functions (IMFs) from high frequency to low frequency orderly. Newly proposed ensemble EMD (EEMD) can overcome the mode-mixing problem of EMD. Thus, the shell vibration signal can be decomposed into sub-signals adaptively. Motivated by the above problems, the qualitative production mechanism analysis of the mill shell vibration and EEMD decomposition results of a laboratory-scale ball mill vibration signal is given out. After integrated with certain domain experts’ experience, some reasonable conclusions are obtained.
36