Knock phenomenon is recognized as one of the major obstacles to further improvements in thermal efficiency of Spark-Ignition (SI) engines. It is well known that in order to ensure a proper knock safety margin, the combustion phasing is delayed up to the knock-limited value, causing a relevant reduction in the engine thermal efficiency and power output. Increasingly sensitive and fast knock detection techniques are thus required to face the need to operate as close as possible to the knock borderline, obtaining the maximum efficiency. The most common approach for on-board knock detection and control in series production SI engines consists in the monitoring of the knock sensor's vibration signal. Unfortunately, knock sensors typically provide noisy indications, so identifying the most effective vibration processing technique is a complex challenge in order to achieve adequate detection performance. In this paper, an accurate knock detection methodology, based on a novel processing of the knock sensor output, is proposed. The method first establishes a nonlinear black-box model on the vibration signal of the engine block in engine healthy-state, after an appropriate signal processing. The model, estimated by means of a sigmoid network nonlinearity estimator, is then used to implement a residual based knock detection on the future-state signal from the knock sensor. In detail, knock onset is diagnosed by characterizing the error signal between the model-produced and the measured vibration signals. The technique is validated based on vibrational and in-cylinder pressure data acquired on a four-stroke, multi- cylinder SI engine, installed at the test-bench. The results show that the proposed model-based methodology is effective and reliable in the detection and diagnosis of knock phenomena. If compared to a standard processing of the knock sensor output, it may lead to more sensitive and robust knock control strategies, allowing the setting of the optimum Knock-Limited-Spark-Advance (KSLA).
CITATION STYLE
Siano, D., & Panza, M. A. (2019). A nonlinear black-box modeling method for knock detection in spark-ignition engines. In AIP Conference Proceedings (Vol. 2191). American Institute of Physics Inc. https://doi.org/10.1063/1.5138870
Mendeley helps you to discover research relevant for your work.