Date of Award
Mechanical, Automotive, and Materials Engineering
Creative Commons License
This work is licensed under a Creative Commons Attribution-NonCommercial-No Derivative Works 4.0 International License.
Combustion knock is a limiting factor in the efficiency of spark ignition internal combustion engines. Therefore, optimization of design and control dictates that an engine must operate as close to the knock limits as possible without allowing knock to occur. This is the challenge presented for knock detection systems. In-cylinder pressure techniques are considered the most reliable method for knock detection; however, installation of pressure transducers in the combustion chamber is both difficult and expensive. This leads to the requirement of a low cost, non-intrusive alternative. Although the current vibration-based methods meet these requirements, their susceptibility to background noise greatly reduces their effectiveness. Thus, the goal of consistently achieving the optimal operating conditions cannot be achieved. This research involves the use of multivariate analysis of vibration-based knock signals to improve the detection system reliability through enhanced signal to noise ratio. The techniques proposed apply a relatively new philosophy developed by Genichi Taguchi for pattern recognition based on the statistical parameter Mahalanobis Distance. Application of these methods results in the development of a new knock detection strategy which shows a significant improvement in determining the presence of knock. The development and validation of this vibration-based system required the use of in-cylinder pressure data for initial classification of knocking and non-knocking operation. This necessitated an independent study to validate pressure transducer type and mounting location. Results of this study are detailed herein.
Tousignant, Todd, "Multivariate knock detection for development and production applications" (2005). Electronic Theses and Dissertations. 7794.