Soutenance de thèse de Francesco Galuppo “Design, optimization and control by condensation pressure manipulation of an ORC based waste heat recovery system in heavy-duty trucks”


La soutenance de thèse de Monsieur Francesco GALUPPO qui aura lieu le 7 janvier 2021 à 13h30 en visioconférence totale.

Title: Design, optimization and control by condensation pressure manipulation of an ORC based waste heat recovery system in heavy-duty trucks

Jury :

Mr Pierre DEWALEFF, Examinateur, Prof., Univ. de Liège

Mr Lars ERIKSSON, Rapporteur, Prof., Linköping Univ.

Mr Sotirios KARELLAS, Rapporteur, Prof., National Technical Univ. of Athens

Mr Vincent LEMORT, Directeur de thèse, Prof., Univ. de Liège

Mme Céline MORIN, Examinateur, Professeur, Univ. Polytechnique Hauts- de-France

Mme Melaz TAYAKOUT-FAYOLLE, Examinateur, Prof., Univ. Lyon 1

Mme Alina VODA, Examinateur, Maître de Conférences, Univ. Grenoble Alpes

Mme Madiha NADRI WOLF, co-directeur de thèse, Univ. Lyon 1

Mr Pascal DUFOUR, Directeur de thèse, Maître de Conférences, Univ. Lyon 1

Mr Thomas REICHE, Membre invité, Encadrant VOLVO GTT, Lyon 

Keywords: Waste heat recovery, Organic Rankine cycle, automotive, modeling, operating points selection, model-based control, weighting estimators, condensation pressure manipulation, modelbased real-time optimization, artificial intelligence.

Abstract: The new European regulation about the CO2 emissions of long-haul heavy duty trucks pushed the manufacturers to study and implement solutions that can limit the CO2 emissions by reducing the fuel consumption of the vehicle. Considering the limited efficiency of the internal combustion engines and the thermal losses that are encountered, the waste heat recovery technologies have gained much interest in the heavy-duty trucks industry; in particular thermodynamic bottoming cycles, as the Organic Rankine cycle, are considered suitable to reduce the fuel consumption and, consequently, the CO2 emissions. However, several obstacles are encountered in the implementation of a safe system that can ensure high standards of performance as long as possible, reducing the total cost of ownership of the vehicle. In the last decade, a large scientific community studied intensely this topic related to the working fluid selection, component modeling and testing, control and optimization. Another important aspect of this application derives from the fact that the heat is recovered from an heat source characterized by a transient behavior. This is taken into account in the system architecture definition, as well as in the implementation of suitable control strategies, intended to the maximization of the performance. In this thesis the exhaust gas and the engine coolant flow are investigated as heat sources; in both cases modeling and control strategies are implemented, in the latter an experimental campaign is performed. Although the heat source is often considered as the main secondary fluid in the Rankine system, the heat sink is often characterized by a transient behavior as well, that can limit the performance of the system and pushes to further control development in the cold side of the system. In this thesis, the control problems related to the working fluid conditions at the inlet of the expander (the controlled variable is called superheat) and the pump (the controlled variable is called subcooling) are addressed, studying the dynamics of the evaporator and condenser, in specific selected operating points, chosen by means of a specific algorithm. For both the heat sources that have been analyzed, the manipulation of the condensation pressure is shown to be not only recommendable, but it is necessary, in order to ensure component safety and high standards of performance of the system. The utilization of the condensation pressure as and additional manipulated variable leads to the definition of a MISO (Multiple Input Single Output) problem to control the subcooling at the inlet of the pump and, consequently, to improve performance. Another potential control improvement is identified in the utilization of a deep neural network, trained on-line, to replace the weighting estimators and determine, in an adaptive way, a representative model to design the controller.

Date/heure
Date(s) - 7 Jan 2021
13 h 00 min - 16 h 00 min

Catégories

Filed under: Soutenance