Application of digital signal processing and multivariate data analysis to vibration signals from ball-mill grinding

During mechanical grinding high intensity vibration is emitted that contains information on the mill's operating state. Tests, using a laboratory ball mill, studied the influence of operating parameters on the vibration signature. The parameters studied were: mill speed; powder filling; pulp density; grinding period; and pulp temperature. Vibration signals were detected by an accelerometer, amplified by a vibrometer and recorded onto digital audio tape. A digital oscilloscope converted the signals into IBM-compatible data format. Digital signal processing and multivariate data analysis were carried out with specially developed software. The power spectral density was estimated from the time-domain vibration signal. Four new latent variables were used to describe changes in the vibration spectra. A relationship was established between operating parameters and latent variables which could form the basis of a quick and automatic method of monitoring operating parameters in ball mill grinding.