National Technical University of Athens (NTUA)
Design, retrofit and predictive optimal control of a conventional marine diesel engine for diesel - methanol dual - fuel operation
Abstract
dc:descriptionThe escalating urgency of the global climate crisis, combined with the stringent regulations imposed by the International Maritime Organization (IMO), has prompted major shifts in the maritime industry, necessitating an immediate transition to carbon-neutral propulsion. In this context, methanol has emerged as a viable alternative fuel owing to its established production infrastructure, liquid state at ambient temperature, and sustainable production methods. As the deployment of new methanol-fueled vessels accelerates, the worldwide fleet requires viable methods to convert the numerous existing internal combustion engines still in operation. This Doctoral Thesis investigates the complex engineering challenges associated with modifying high-speed marine diesel engines for Diesel-Methanol Dual-Fuel operation, establishing a comprehensive research framework that encompasses experimental evaluation, high-fidelity multi-scale modeling, and advanced real-time optimal control. The research commences with the development and validation of a standardized retrofit methodology for conventional marine engines equipped with mechanical rotary fuel pumps, a prevalent yet understudied architecture in the dual-fuel conversion domain. A conventional high-speed diesel engine was modified for port-premixed methanol application by installing a custom low-pressure injection system, designed specifically for this application, and an independent electronic management unit. An extensive testing campaign was performed on a fully instrumented testbed to evaluate the performance and the operational limits of the retrofitted engine. It highlighted the non-linear trade-offs between Brake Thermal Efficiency and regulated emissions across various Methanol Energy Substitution Ratios. The findings indicated a significant reduction in Nitrogen Oxides (NOₓ) and soot; however, they also revealed a greater propensity for Volatile Organic Compounds and unburned methanol at partial loads, necessitating meticulous combustion control. A comprehensive modeling framework has been established to evaluate dual-fuel combustion, performance, and emission characteristics, and to support the development of control strategies. Initially, a high-fidelity 3-D Computational Fluid Dynamics (CFD) model was developed and utilized to perform a parametric study to elucidate the intricate dual-fuel combustion mechanisms and the kinetic pathways responsible for the generation of regulated and unregulated emissions. This study focused on fuel injection characteristics, particularly diesel injection timing and fuel ratios. Subsequently, a novel zero-dimensional (0-D) phenomenological combustion model was developed to predict the complex combustion process with relatively low computational cost. This model utilizes a modified Triple Wiebe function to accurately depict the multi-stage heat release rates of dual-fuel combustion and was calibrated against the CFD data. This formulation was subsequently integrated into a physics-based Mean Value Engine Model (MVEM), providing a control-oriented platform capable of accurately predicting the system's steady-state and transient response. A pivotal innovation of this dissertation is the development of a vessel-to-engine simulation ecosystem designed to evaluate propulsion performance in realistic maritime conditions. The research couples the MVEM with a dynamic 4-Degree-of-Freedom model of a planing craft, including hydrodynamic resistance and stochastic wave loads based on JONSWAP spectra. Furthermore, to address the absence of typical load cycles for this category of vessels, a novel data-driven methodology was developed utilizing Automatic Identification System (AIS) recordings. The research employed Conditional Adversarial Autoencoders to generate synthetic, statistically representative transient operational profiles that serve as rigorous disturbances for control validation. The framework evaluates various observer architectures, including Extended Kalman Filters and Moving Horizon Estimators, to facilitate the real-time estimation of propeller load torque from stochastic onboard measurements for practical application. Finally, the Thesis culminates in the design, implementation, and rigorous experimental validation of real-time optimal control strategies tailored for common or separate ECU architectures of retrofitted propulsion systems. The research formulates a constrained optimal control problem designed to arbitrate the trade-offs between thermal efficiency, emission compliance, and mechanical limitations. Three distinct control paradigms were synthesized and evaluated: Adaptive Linear Model Predictive Control, Nonlinear Model Predictive Control, and data-driven Soft Actor-Critic Reinforcement Learning. Furthermore, Dynamic Programming was employed to derive non-causal global optima, serving as a rigorous benchmark to quantify the "optimality gap" of the proposed real-time strategies. The comparative analysis reveals that while NMPC exhibits superior capability in handling hard constraints related to knock and misfire, the developed RL agents offer a robust, model-free alternative capable of adapting to the stochastic nature of marine environmental disturbances. Collectively, this work establishes a validated, holistic technical framework for the reliable operation of retrofitted methanol-fueled vessels, significantly contributing to the operational viability of the maritime energy transition.
Degree
thesis:*- Grantor dc:publisher
- National Technical University of Athens (NTUA)
- Year dc:date
- 2026
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Karystinos, Vasileios
- Καρυστινός, Βασίλειος
Subjects
dc:subject × 27- Μηχανές εσωτερικής καύσης
- Ενισχυτική εκμάθηση
- Ναυτικές Μηχανές
- Πολυκριτηριακή - πολυστοχική βελτιστοποίηση
- Μοντελοποίηση μηχανων διπλού καυσίμου
- Μεθανόλη
- Μηχνανες διπλού καυσίμου
- Αυτόματος έλεγχος
- Diesel-Methanol Dual-Fuel (DMDF)
- Marine Engine Retrofit
- Optimization
- Engine Modeling
- Model predictive control
- Internal combustion engines
- Reinforcement learning
- Control systems
- Άλλες Επιστήμες Μηχανικού και Τεχνολογίες
- Επιστήμες Μηχανικού και Τεχνολογία
- Μηχανική και Τεχνολογίες, άλλοι τομείς
- Επιστήμη Μηχανολόγου Μηχανικού
- Μηχανολογική μηχανική
- Μηχανολογία
- Other Engineering and Technologies
- Engineering and Technology
- Engineering and Technologies, miscellaneous
- Mechanical Engineering
- Mechanics
Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/62515