Multi-Channel Vibration Feature Extraction of Ball Mill Using Synchronized Wavelet Based Multi-Scale Principal Component Analysis

The trait of the ball mill is chaotic in nature due its complex dynamics associated during grinding. Grinding in ball mill generates high-intensity vibration and is too complex on account dependency of multiple variables. In this paper, the vibration signal is acquired using a low power ZigBee based three axes wireless MEMS accelerometer sensor mounted onto the mill shell. Firstly, the exact frequency bands of the mill are identified under variable impact loading using non synchronized and Synchronized Frequency Estimation method (SFE) methods. The synchronization between the mill speed and the sampling rate are put forward by SFE to convert the random non stationary data to quasi stationary data. The actual signal length is calculated using proposed SFE approach and further it is used as window size for wavelet decomposition. Further, to decorrelate the auto-correlated and cross-correlated signal and signal spaces both PCA and Wavelet are used. Finally, the combination of all this techniques, i.e., Synchronized Wavelet Based Multi-Scale Principal Component Analysis (SWMSPCA) is used to extract the vibration feature of the ball mill in the presence of variable density ores i.e., iron ore and limestone. Key-Words: Ball Mill, Accelerometer Sensor, ZigBee, Wavelet, PCA, Windowing, SFE, SWMSPCA, Fast Fourier Transform.