Knock acoustic signal estimation using parametric inversion

Abstract

Knock in spark ignition engines goes against car manufacturers' efforts to reduce fuel consumption as well as exhaust gas emissions. This paper develops a signal-processing method to quantify knock. After discussing some classical techniques of knock energy estimation, an acoustical measurement technique is presented. An original signal-processing method based on a parametric compartmental model for both knock and apparatus is used. A special inversion technique is also proposed to obtain actual knock parameters. The knock-related parameters are computed through a two-step process. First, a deconvolution algorithm is used to obtain a sparse spike train signal. Then, an efficient inversion method follows. The whole process is applied to real data from a one-cylinder engine. Results are compared to those obtained from an existing technique. They demonstrate the usefulness of such a procedure in a common industrial application.