The aim of the work is to perform a Topology Optimization (TO) coupling fluid-thermal problems in order to verify if the same result could be obtained from 2D and 3D computational domains. Starting from a fully fluid domain, porous regions have been added to it according to the imposed constraints of the Multi-Objective Optimization (MOO). The two Objective Functions (OFs) were in contrast to each other: the dissipated mechanical power had to be minimized while a maximization of the recovered thermal power was performed. OFs were summed up together by the Weighted Sum Method (WSM), where a proper normalization factor for each OF has been imposed; while the Adaptive Weighted-Sum (AWS) method was used to compute OFs’ weights a priori and to obtain the Pareto-optimal solution. The used solver is a steady-state one valid for non-Newtonian incompressible flows, where turbulence has been modeled by the κ − ε model. TO was conducted using the continuous adjoint method: physical constraints has been enforced using the method of Lagrange multipliers and the sensitivity corresponds to the augmented Lagrange function divided by the vector of design variables. Each cells’ domain were associated to a non-dimensional pseudo-density value η, allowing to recover governing equations for the fluid region or the solid one. Design variables were updated by the Method of Moving Asymptotes (MMA), while material properties and porosity field were computed by the Rational Approximation of Material Properties (RAMP) model. TO result consist in a computational domain composed by both fluid and porous regions. The latter force the fluid to flow in a different path, with different velocity and temperature, allowing to obtain distinct OFs values. The formed porous regions have been extracted by a python script and validation was conducted by CHT simulations, in which the bi-phase computational domain was composed by fluid and solid regions.
Lo scopo della tesi è quello di verificare se il risultato di un’ottimizzazione topologica (TO) per un problema fluido-termico sia lo stesso per configurazioni 2D e 3D. A partire da un dominio interamente fluido, delle regioni porose vengono aggiunte all’interno di esso a seconda dei vincoli imposti dall’ottimizzazione multi-obiettivo (MOO). Le due funzioni obiettivo (OF) risultano essere in contrasto l’una con l’altra, in quanto la potenza meccanica dissipata viene minimizzata mentre una massimizzazione della potenza termica recuperata è richiesta. Le due OF, opportunamente normalizzate, sono linearmente combinate in un’unica funzione obiettivo mediante il Weighted Sum Method (WSM), mentre il metodo dell’Adaptive Weighted-Sum (AWS) è stato utilizzato per la scelta a priori dei pesi e per ottenere la soluzione Pareto-ottimale. Il solver utilizzato è valido per un flusso stazionario, incomprimibile e non-Newtoniano, mentre gli effetti della turbolenza sono stati modellati secondo il modello κ − ε. La TO è stata eseguita secondo il metodo dell’aggiunto continuo, i vincoli fisici sono stati imposti secondo il metodo dei moltiplicatori di Lagrange e la sensibilità corrisponde al rapporto della funzione di Lagrange rispetto al vettore delle variabili di design. Tale vettore è composto da valori di pseudo-densità, associati a ciascuna cella del dominio, consentendo di recuperare le equazioni governanti la regione fluida o solida. Le variabili di progetto sono aggiornate secondo il Method of Moving Asymptotes (MMA) in base al gradiente, mentre le proprietà dei materiali così come il campo di porosità sono calcolati dal Rational Approximation of Material Properties (RAMP). Il risultato della TO consiste in uno dominio composto sia da regioni di liquido che porose: queste ultime costringono il fluido a fluire in modo differente consentendo di ottenere valori delle OF diversi. Le regioni porose formatesi dalla precedente ottimizzazione sono state estratte da uno script python e la validazione dei risultati è stata condotta tramite simulazioni CHT in cui nel dominio di partenza è presente sia un dominio solido che liquido.
Bi-objective topology optimization of conjugate heat transfer systems: a comparison between 2D and 3D cases
Celli, Alessandro
2022/2023
Abstract
The aim of the work is to perform a Topology Optimization (TO) coupling fluid-thermal problems in order to verify if the same result could be obtained from 2D and 3D computational domains. Starting from a fully fluid domain, porous regions have been added to it according to the imposed constraints of the Multi-Objective Optimization (MOO). The two Objective Functions (OFs) were in contrast to each other: the dissipated mechanical power had to be minimized while a maximization of the recovered thermal power was performed. OFs were summed up together by the Weighted Sum Method (WSM), where a proper normalization factor for each OF has been imposed; while the Adaptive Weighted-Sum (AWS) method was used to compute OFs’ weights a priori and to obtain the Pareto-optimal solution. The used solver is a steady-state one valid for non-Newtonian incompressible flows, where turbulence has been modeled by the κ − ε model. TO was conducted using the continuous adjoint method: physical constraints has been enforced using the method of Lagrange multipliers and the sensitivity corresponds to the augmented Lagrange function divided by the vector of design variables. Each cells’ domain were associated to a non-dimensional pseudo-density value η, allowing to recover governing equations for the fluid region or the solid one. Design variables were updated by the Method of Moving Asymptotes (MMA), while material properties and porosity field were computed by the Rational Approximation of Material Properties (RAMP) model. TO result consist in a computational domain composed by both fluid and porous regions. The latter force the fluid to flow in a different path, with different velocity and temperature, allowing to obtain distinct OFs values. The formed porous regions have been extracted by a python script and validation was conducted by CHT simulations, in which the bi-phase computational domain was composed by fluid and solid regions.File | Dimensione | Formato | |
---|---|---|---|
2023_10_Celli.pdf
accessibile in internet solo dagli utenti autorizzati
Dimensione
2.88 MB
Formato
Adobe PDF
|
2.88 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/210568