NVH Signal Analysis via Pattern Recognition ANNs: Automotive Brake Creep Groan as Case Study

Automotive Noise, Vibration and Harshness (NVH) issues often cause costly customer complaints. In particular, NVH of the brake system is critical for subjective ratings of vehicle safety and passenger comfort. Recently, especially a stick slip induced low frequency phenomenon, the so called brake creep groan, has become increasingly relevant. In order to avoid this brake NVH problem within upcoming automobile fleets, simulative and/or experimental parameter studies throughout all industrial brake development stages are indispensable. To this end, reasonable and efficient data analysis methods are necessary too. This kind of signal assessment challenge is addressed here by means of a method which applies techniques of Artificial Intelligence (AI) or Artificial Neural Networks (ANNs) respectively. The basis for this is a large number of generically synthesised brake component acceleration spectra which represents data in the frequency domain with and without creep groan. This generic data is used to create specifically elaborated pattern recognition ANNs. Eventually, the proposed approach provides an integrated framework of conditioned ANNs which is supposed to detect and separate non linear signatures of different brake creep groan vibrations. In order to examine the method’s practical limitations, additional data sets of synthetic accelerations including generic noise have been considered, and moreover, gauged accelerations concerning two test rig setups have been taken into account. Although the devised creep groan analysis approach is designated for automotive brake development workflows, its principle could be appropriate for similar NVH problems or signal analysis tasks in other engineering fields alike.