City Research Online

Abstract

The substitution of small-medium sized Diesel engines for dual-fuel operation is one of the strategies for meeting emission regulations. For these engines, fuel flexibility is an important criteria to ensure a reliable operation. However, this leads to some disadvantages related to the injection system complexity because of the need for multiple injectors or a design with multiple nozzles or needles. A single and simple injector capable of precisely delivering fuel across the entire operation range (dualfuel and Diesel-only modes) is the suitable solution to mitigate those drawbacks and advance the marine engine towards a more environment-friendly operation. However, an optimization process of an injector design is a dependent problem where one universal injector design cannot cover all specific demands of the engine and operation strategies for most of the cases. In this context, the main objective of the current thesis is to enhance and develop numerical tools to support the injector concept design, addressing the operation conditions relevant for fuel flexibility in marine engines. More specifically, the current work proposes models with affordable computational cost to predict in-nozzle cavitation and spray characteristics under non-reacting conditions. Therefore, a thermodynamic closure based on the National Institute of Standards and Technology data base and the Perturbed Chain Statistical Associating Fluid Theory equation of state was incorporated in the computational fluid dynamics model and validated against the experimental data from the Engine Combustion Network. Then, this model has been applied to investigate the potential of needle-tip design and underscore the in-flow conditions of each operation mode. Afterwards, this model was used to generate data for neural networks training, where the needle-tip design, conicity and needle lift are the input parameters to predict performance attributes as single values and in-nozzle cavitation as images. This CFD-assisted machine learning surrogate model demonstrates how those areas could work together for optimization tasks. Finally, an additional deep neural network based on experimental data was trained to predict the spray penetration and cone angle reaching 95% accuracy. This model included 10 geometrical parameters and 3 operation conditions as input features, which differentiate it from current ones found in the literature. Overall, the numerical solutions developed in this work were tailored to support the optimization process of marine Diesel injectors, which contributes to reducing the development cycle and decision-making in the early design phases.

Publication Type:	Thesis (Doctoral)
Subjects:	T Technology T Technology > TJ Mechanical engineering and machinery V Naval Science > VM Naval architecture. Shipbuilding. Marine engineering
Departments:	School of Science & Technology > Department of Engineering School of Science & Technology > School of Science & Technology Doctoral Theses Doctoral Theses

Export

Downloads