Audio perception in the context of quality assurance for manufacturing and industry
The use of acoustic event detection in industrial quality control allows for contact-less quality assurance. This might be useful in predictive maintenance, end-of-line-, or in-line-testing. It may support the use of available information (i.e., sound) to reduce rejects and production losses. Faults such as broken ball bearings or defective engines are heard and recognized by experienced machine operators, but hardly or only costly recognized otherwise. Due to highly complex noise scenarios in fabrication plants it is challenging to detect audible errors in the human hearing range. However, industrial sound analysis solutions that are robust against unpredictable background noise in factories but sensitive to subtle differences in the observed sound are developed. Therefore, qualitative annotation of data is inevitable. The remaining question is about collecting adequate feedback and annotations from the experts, the machine operators recognizing their personal ear-catcher. In this workshop we address this challenge by providing a three-parted program. We start with a formal introduction to the topic, provide a hands-on activity to allow the experience of the way of "hearing quality", and end with a panel discussion about useful methods, (non-)existing standards and how-to's in order to support future annotation tasks and overall acoustic quality assurance processes in industry. The outcome of this workshop supports finding the most adequate way of annotation for experienced machine operators using sound source description for quality assurance.