Συμβολή στο σχεδιασμό θερμο-ρευστοδυναμικών διατάξεων με χρήση αιτιοκρατικών και εξελικτικών μεθόδων βελτιστοποίησης

Abstract

This PhD thesis focuses on the development of optimization methods for the design of thermo-fluid systems based on criteria related to fluid mechanics and/or heat transfer. This thesis is concerned with the continuous adjoint method and evolutionary algorithms. The hybridization of gradient-based and stochastic methods is also presented. The present thesis developed the continuous adjoint method to incompressible flows with heat transfer, by focusing on the accuracy of the computed sensitivity derivatives. For this purpose, the formulation of the continuous adjoint equations for incompressible flows with the addition of the energy equation is presented. The first point of novelty of this thesis is the continuous adjoint method for incompressible flows with heat transfer, which is presented for the first time. The objective functions take into account viscous losses and exchanged heat. Apart from shape optimization, the present thesis is extended to topology optimization problems in fluid mechanics and heat transfer. In fluid mechanics, topology optimization is used for designing flow passages, connecting predefined inlets and outlets, with optimal performance based on selected criteria. In the present thesis, a topology optimization algorithm for incompressible, laminar and turbulent flow problems including heat transfer, was developed. The formulation of the primal and adjoint equations for laminar and turbulent flows with heat transfer, by introducing new porosity dependent terms, is presented for the first time in the literature. In turbulent flows, the formulation is developed for low-Reynolds number turbulence model. The topology optimization algorithm is used for the design of ducts/manifolds for minimum total pressure losses and/or maximum temperature rise between the outlet from and the inlet to the domain. Regarding the stochastic optimization methods, aiming at the solution of computationally demanding optimization problems, the present PhD thesis is concerned with the combined use of an Asynchronous Evolutionary Algorithm (AEA) together with the developed adjoint methods. The hybridization of the AEA with a local search method gave rise to a new asynchronous metamodel-assisted memetic algorithm (AMAMA), which performs better than AEA. This algorithm was also used for the optimization of heat transfer systems, such as geothermal power plants (Organic Rankine Cycle systems) and ground source heat pump systems.