Deduction of vehicle related limits and subsequent classification of gearbox noises using Support Vector Machines to overcome limitations of current acoustic End-of-Line testing

The present paper deals with structure-borne noise of vehicle gearboxes, which can lead to disturbing noise in the passenger compartment. One of the significant noise sources of gearboxes is the gear contact - resulting in so-called tooth-meshing orders. The acoustic classification of gearboxes is still a challenging task. State-of-the-art acoustic analysis techniques partly fail in creating adequate classification rates caused by insufficient correlations between measured accelerations at the gearbox and the subjective ratings of the noise in the cabin. For that purpose, the authors applied a machine learning (ML) approach; Support Vector Machines (SVMs) are used to classify acoustic o.k. and noto.k. gearbox-orders of several vehicle measurements of 7-gear dual clutch gearboxes. The results show a proper classification-rate nearby 90 %. The results show.