AI-based acoustic quality inspection: case study for the assurance of functional sounds in automotive manufacturing.

Functional sounds of a vehicle, such as the horn or the acoustic vehicle alerting system (AVAS) are highly relevant for the safety of a vehicle and its environment. Thus, the functionality of these sounds must be assured during production. Due to their limited reliability and high costs, neither conventional acoustic testing methods nor manual inspections by production associates are optimal. Therefore, this case study investigates whether AI-based approaches are suitable to automate these inspections. This case study is oriented on the CRISP-DM framework and applies acoustic scene classification models based on spectrograms. The investigations illustrate that the AI models can distinguish between OK/NOK with high accuracy and are therefore applicable for quality assurance.