Detecting Knock in Spark Ignited Engines

Abstract

In this paper we consider signal processing algorithms for detecting knock, i.e. auto-ignition of unburned fuel, in spark-ignited engines. To operate an engine with optimal efficiency, there is a need to operate close to the knock limit. This motivates the need for efficient knock detection algorithms that not only indicates knock, but also gives estimates on knock timing and size. The proposed detection algorithms are based on either incylinder pressure or ion current and both off-line and on-line methods are developed. The methods are tested on measured data, analysed, and compared with respect to detection performance, knock timing estimation accuracy, and robustness against model uncertainties.