|Tipo di tesi||Tesi di laurea magistrale|
|Autore||AGWA, ALI ALAA MOSTAFA|
|Titolo||Development of a model of Ignition Delay for GDCI engine using Machine Learning and comparison with Physical model|
|Titolo in inglese||Development of a model of Ignition Delay for GDCI engine using Machine Learning and comparison with Physical model|
|Struttura||Dipartimento di Ingegneria "Enzo Ferrari"|
|Corso di studi||Advanced Automotive Engineering (D.M.270/04)|
|Data inizio appello||2022-10-20|
|Disponibilità||Accessibile via web (tutti i file della tesi sono accessibili)|
With the aim to reduce and limit pollutant emission further towards carbon neutrality according to the latest EU ban on ICE cars by 2035.
With the aim to reduce and limit pollutant emission further towards carbon neutrality according to the latest EU ban on ICE cars by 2035. Aiming to reduce the emissions many researchers have been studying low temperature combustions, which provides an efficient combustion with low pollutant emissions such as NOx and CO. These concepts of low temperature combustions such as HCCI "Homogeneous charge compression ignition" and GCI "Gasoline compression ignition" engines reduced the emissions however came with other challenges such as requiring new calibration and control strategies. By relying on the natural phenomenon of autoignition, researches have shown that to optimize the GCI combustion multiple injections strategy is adopted such as in CI engines, The rise in temperature and pressure from the first two injections the pre and pilot, would result in better controlling the main injection ignition which is the injection responsible for the torque delivery which burns almost instantly when injected, the pre injection however have a delay between injection and injection. The control of this delay between is affected by many engine control parameters, which could cause misfire or change the combustion shape inside the cylinder. The aim of the thesis is to build a model and controller for the ignition delay using experimental data having a max error of 5 CA of error. the methodologies used was further optimizing a 0-D physical ignition delay model and then building a control using machine learning to estimate the ignition delay.