This paper considers a case study of using non-intrusive passive acoustic sensors for on-line monitoring of hydrocyclone performance. Acoustic emissions generated within a hydrocyclone treating fine silica slurries were collected using a single 190-k-Hz piezoelectric sensor mounted externally on the conical section of the hydrocyclone. The investigation indicates that the spectral characteristics of the signals are particularly sensitive to several key operating and process parameters of the hydrocyclone. The use of multivariate statistical analysis techniques allows the quantitative relationships between various operating parameters of the hydrocyclone and the statistical and spectral characteristics of the acoustic signals to be established. This suggests that the method is useful for on-line control and monitoring.