TU Berlin

Chair of Electronic Measurement and
Diagnostic Technology
Signal Processing for Knock Detection

Page Content

to Navigation

Signal Processing for Knock Detection

Today's spark-ignition engine works more and more on knocking limit because of higher requirements on consumption, exhaust-gas limits and engine power. Therefore knock detection is very important to operate the engine on the knocking limit. The knock detection works by analyzing the structure borne noise. Some engine components produce structure borne noise in the same time and/or frequency area of knocking. In this case it is very difficult to detect knocking.
The project "Disturbing Noise in Knock Detection Control Systems" deals with the problem of this disturbing noise. Measurements were carried out on a 1.6 l 4-cylinder in-line engine with various modifications to some engine components to influence the disturbing noise. The measurements were planed with the method Design of Experiments to reduce the number of possible variations. With DoE models the influence of disturbing noise can be investigated.
The first step was to analyse the signals in the time-frequency-domain with the Wigner-Ville-Distribution to find out and illustrate the influence of the different components on the structure borne noise. The next step was to model the dependencies of relevant noise and knocking parameters on the modified engine components. In the last step two different methods of signal processing are used, to reduce the influence of disturbing noise in the knock detection. Therefore the Blind Source Separation is used to separate the disturbing noise from the knocking. The other method is the nonlinear filter MOSH to reduce the influence of the disturbing noise. With both methods the influence of disturbing noise can be reduced.



Quick Access

Schnellnavigation zur Seite über Nummerneingabe