Thesis

Back to list

Computer-based Classification of Running Characteristics of an Internal Combustion Engine with Airborne Sound Measurements in End-Of-Line Quality Control

Authors Fuchs, A.
Year 2016
Thesis Type Master's thesis
Topic Audio Signal Processing
Abstract The engine isn’t running smoothly, it’s knocking! – A motorcar mechanic is intuitively capable of assessing the running characteristics of an internal combustion engine by listening to its sounds. In the course of evolution the human sense of hearing has evolved to a measuring instrument with exceptional adaptability, fault tolerance and hierarchical structuring ability. After training, the ear can thus be used for quality assessment. However, for quality assurance at the end of the production line, automation and integrity are of central importance, which is why the condition on that specific day and suchlike of a specially trained individual must not interfere with the assessment. Thus, it is tempting to use the human hearing expertise as a template for automatic classification based on acoustic measurements of the running engine in order to test and attest the mechanical quality. By using modern hardware and efficient algorithms, an easily used system is developed, which is - in a best-case scenario - sufficiently robust to deliver reliable results in the production process. From existing measurement data from an internal combustion engine, characteristic sound emissions in the duty cycle of the engine are specified in different operation conditions. With the appropriate algorithms, the alignment on the duty cycle can be performed solely with the microphone signal itself. It is therefore possible to define automatically computable characteristics, which contain the desired information and form the input of the computer-based classification. By algorithms of supervised machine learning, the statistically optimal processing of features is guaranteed for classification. As a result, there is an automatic, cost-effective and precise discrimination between different running-characteristics of the engine.
Supervisors Höldrich, R., Sontacchi, A.