Recognition of Mill Load with KPCA and KNN Based on Shell Vibration Signals

Recognition of the status of ball mill load (ML) is very important. In practice, operators keep the ML at optimizing range using experience, which always lead to the mill running in the status of lower-load or over-load. A novel ML recognition approach combined with fast Fourier transform (FFT), kernel principal component analysis (KPCA) and K nearest neighbor (KNN) based shell vibration signal is proposed in this paper. At first, the power spectral density (PSD) of the shell vibration signal is obtained using FFT. Then, the mainly frequency spectral features of different frequency spectral segments are extracted using KPCA. At last, KNN are used to recognize the status of ML. The experimental result shows that the proposed approach can recognize the ML effectively.