Engine Knock Estimation Using Neural Networks Based on a Real-World Database 980513
In this paper we present an advanced knock detection approach. The detection concept consists of a two-level feature extraction step followed by neural network detector. A knock tendency index is estimated that takes into account the statistical behavior of the knock phenomena. The configuration of the neural network is based on a signal database that was acquired under almost ‘on-road’ conditions. The experimental set-up consisted of several measurement sessions in a special vehicle test cell. In order to achieve a most realistic knock database the test engine was mounted on an in-production car.
Stefan Ortmann, Matthias Rychetsky, Manfred Glesner, Riccardo Groppo, Paolo Tubetti, Gianluca Morra
Darmstadt University of Technology, FIAT Research Center
International Congress & Exposition
Electronic Engine Control Technologies-PT-73, On- and Off-Board Diagnostics-PT-81, Electronic Engine Controls 1998: Diagnostics and Controls-SP-1357, SAE 1998 Transactions - Journal of Engines-V107-3