Jonathan Borg

Wavelet-based knock detection with fuzzy logic

This paper presents a novel approach to the determination of the knocking condition of a spark-ignition engine using the discrete wavelet transform as a means of analyzing the engine-block vibration signal and a fuzzy inference scheme to generate an estimate of the knock intensity. The block vibration sensor responds to knock-induced pressure waves generated when a portion of the air-fuel mixture in the cylinder combusts spontaneously. However, the block vibration signal is masked by other excitation signals, which render the deter- mination of the knocking condition of the combustion a difficult task. The wavelet transform lends itself very useful to the time-frequency analysis of the vibration signal. Various characteristic features can be determined from the wavelet decomposition of the signal. These are in turn fused together to generate an estimate of the knocking in- tensity using fuzzy logic. The proposed scheme was developed and tested using engine combustion data processed off-line. Off-line data was also used to optimize the parameters of the fuzzy scheme. The sys- tem was subsequently implemented on the real-time engine controller for on-line testing.

Automatic on-line tuning of the knock detection system on an ECU

Abstract Knock detection occupies a very important role in the ignition control of a Spark Ignition (S.I.) engine. Production-grade knock detection schemes must be computationally efficient and comprise a set of parameters which require calibration during Engine Control Unit (ECU) development. This is typically achieved empirically, whereby a calibration engineer induces knock on the ECU and monitors the detection performance to manually adjust the tunable parameters. In this paper, we propose automatic, on-line calibration of a production-grade, frequency-domain knock detection system. Compared to the conventional method of manual calibration, this method allows for on-line tuning of the parameters based on scientific optimization schemes. It thus enables the implementation of specific objective functions which can meet desired criteria.