Application of Psychoacoustic Filtering for Machine Fault Detection

A method for nondestructive machine fault detection, based on evaluation of acoustic machine signatures, is presented. Various mechanical defects of rotary machines can be reflected in altered acoustic signatures. Such phenomena can be often perceived by skilled human operators who can also explain the type of defect merely based on slightly changed acoustic signature. The proposed method is based on psychoacoustic modeling of human auditory perception. For the purpose of machine fault detection, the gammatone filterbank is applied in preprocessing of acoustic signals. Filtered signals are then rectified, and features are calculated as mean values of rectified signals. A set of features represents an extracted machine signature. In order to evaluate the current machine state, extracted features are compared to the database of normal machine states. If values of the extracted features exceed statistically determined upper and lower margins, a possible machine fault is indicated. The proposed approach is illustrated by a case study, where the quality of commercially produced compressors is predicted. Compressor states are estimated based on acoustic emission during operation, and features are extracted by psychoacoustic filtering as described above. Results show that major faults, that occur in a production, can be reliably detected.