Acoustic-based Condition Monitoring

Abstract

In this thesis, a system for monitoring valve switching operations based on structureborne sound signals was developed. The aim was to create a method that detects faulty or abnormal switching operations at an early stage, thereby contributing to preventive maintenance of industrial plants. Unlike conventional approaches, which often rely on airborne sound measurements with microphones, the system presented here is based on structure-borne sound sensors. The mechanical vibrations are not recorded directly at the valve, but at the associated pneumatic line. This indirect measurement position has proven to be practical and reliable, as the structure-borne sound impulses are well transmitted through the pipeline and can be evaluated reproducibly. At the same time, this method offers higher immunity to environmental influences and more robust installation in industrial environments. The system consists of several components: a sensor for detecting sound signals, software for time-accurate recording and processing of the data, and an evaluation module that automatically recognizes characteristic features of individual switching operations. Important signal characteristics which where examined included the duration and amplitude of the pulses, the dominant frequency components, and the background noise of the system. In addition, the influence of filters and different sampling rates was analyzed. The results show that these characteristics can be used to reliably distinguish between “good” and “abnormal” switching operations. Faulty valve movements could be reliably identified. A rule-based classification was implemented and successfully tested. A key finding of the tests was that up to 90 valves could be monitored simultaneously with just one sensor. Due to the good transmission properties, it is highly likely that significantly more valves will be able to be monitored with a single sensor in the future. The work thus contributes to the condition-based monitoring of valves on an acoustic basis. The developed system was implemented and validated in a test setup. However, further adjustments and system integrations are still necessary for practical application on an industrial scale, for example with regard to long-term behavior, real-time capability, and automated calibration.

Date of Award	2025
Original language	English
Supervisor	Roland Exler (Supervisor)