Metodologie di diagnosi non intrusiva per la caratterizzazione del processo di combustione nei motori diesel

Abstract

This Thesis deals with advanced diagnostic methods able to perform complex measuring’s tasks regarding the combustion process in a diesel engine. Non-Intrusive measuring is a key factor to develop innovative modeling and control strategies on the purpose of improving the combustion process itself and so reducing the impact of these plants on the environment as pollutants and noise emissions. Modern measuring systems heavily rely on advanced Digital Signal Processing techniques leading to result unfeasible for conventional measuring chains. A consistent effort of the presented work has then been devoted to review the state of art of modern diagnostic’s methods for combustion engines available on literature. The last involves the analysis of different data mining approaches whose spread through heterogeneousness fields of application like clinical diagnostic, social sciences, finance. A well known but still promising procedure to perform Machine Condition Monitoring on engines regards the use of Instantaneous Angular Speed measurements. The experimental developing of a high performance IAS measuring system has been presented and its results about Transmission Error evaluation have been discussed. Thanks to probing via high speed digital counters, has been showed how to achieve very high sensitivity about angular displacement of a rotating shaft by simple yet proficient phase demodulation numerical conditioning. Another experimental campaign has been presented regarding the measurements of the acoustic emission of a small benched engine during stationary conditions. The acoustic signature of the combustion noise has been traced and the relations between the Rate Of Heat Release and airborne emissions have been then analyzed. Coupling results from both acoustic and IAS expertise, has been proposed an innovative multiparametric model of the engine by regressive analysis. The assessment of the model on different regressive structures has been performed regarding the quality of predictions obtained. Also Principal Component Analysis and Independent Component Analysis have been utilized on the aim of reducing the number of parameters used in system identification procedure. Overall identificated regressive models have been shown a tight attitude to be embedded as an effective tool for the developing of advanced combustion systems