This thesis addresses fluid dynamics problems in complex geometries, proposing the implementation of an efficient solver in order to run simulations in a three-dimensional domain, like the brain Ventricular System. Alzheimer's disease and other neurodegenerative pathologies have become more and more diffused, impacting on the medical community but also on the society. Even though some treatments have been experimented to slow down the progression of the disease, clinical research is ongoing to find effective treatment. Neurodegeneration is often correlated with the impairment of the waste clearance function played by the Cerebrospinal Fluid (CSF). Yet many features of CSF generation and flow in the Ventricular System are still subject to discussion. For this reason, CFD simulations may help in advancing research and integrate data from medical imaging. This thesis focuses on the implementation of a three-dimensional solver for the stationary Stokes problem, to represent the characteristics of a flow like that of CSF; then it is tested on benchmark and realistic mesh. To discretize and solve the problem we consider a Polytopal Discontinuous Galerkin (PolyDG) method, which guarantees geometrical flexibility in the choice of the computational mesh. The code builds upon the existing library Lymph3D, which relies on PETSc, METIS and fMETIS to perform matrix operations and mesh agglomeration. Lymph3D offers the advantage of agglomerating tetrahedral mesh in polyhedral elements, thus decreasing considerably computational time. Single terms of matrices and vectors are assembled on each tetrahedron but the most expensive steps as block matrix operations are performed on the polyhedral partition. As a result the novelty introduced by this thesis is an efficient strategy to solve a saddle point problem with polyhedral methods in three-dimensional domains. Efficiency of polyhedral agglomeration has been tested on a tetrahedral mesh, in order to show the advantage coming from the reduction of computational time. In the end polyhedral agglomeration is exploited to simulate the time-averaged CSF flow in the brain ventricular geometry of a patient affected by Alzheimer's disease, with suitable boundary conditions and parameters.
L'obiettivo principale di questa tesi è l'implementazione di un solutore efficiente per il problema di Stokes al fine di eseguire simulazioni di fluidodinamica in domini tridimensionali con geometrie complesse, come il sistema dei ventricoli cerebrali. Le malattie neurodegenerative, in particolare la malattia di Alzheimer, si stanno diffondendo sempre di più, con impatti non solo sul sistema sanitario, ma anche sulla società. Spesso l'insorgenza di queste malattie viene associata al malfunzionamento nei meccanismi di waste clearance associati al flusso del fluido cerebrospinale, ovver il processo di smaltimento di sostanze e proteine tossiche. Questo fluido presenta molte caratteristiche ancora discusse e oggetto di ricerca, come il suo processo generativo. A questo proposito le simulazioni di fluidodinamica possono aiutare nel campo della ricerca clinica integrando i dati acquisiti con strumenti medici come la risonanza magnetica. Questa tesi si propone di implementare un solutore tridimensionale per il problema di Stokes stazionario, per simulare il flusso del fluido cerebrospinale. La vera novità introdotta da questa tesi riguarda l'uso di metodi poliedrali per discretizzare e risolvere un problema di fluidodinamica computazionale tridimensionale, in particolare il metodo PolyDG permette una certa flessibilità geometrica nella scelta della mesh computazionale, avvantaggiando l'accuratezza delle simulazioni. Il codice nasce da uno sviluppo della libreria esistente Lymph3D, che fa uso di PETSc, METIS e fMETIS. Lymph3D offre il vantaggio di agglomerare mesh tetraedriche in poliedri, riducendo il costo computazionale che deriva dal numero di gradi di libertà sui singoli elementi. I blocchi di ciascuna matrice sono assemblati per ciascun elemento tetraedrico ma le operazioni matriciali più complesse e computazionalmente costose sono eseguite per ogni elemento poliedrico agglomerato. L'efficienza dell'agglomerazione in poliedri è stata testata, dimostrando che i tempi computazionali si abbassano considerevolmente quanto minore è il numero di poliedri che viene usato, mantenendo un comportamento degli errori stabile. Infine un reale vantaggio della strategia risolutiva implementata è stato osservato simulando il fluido cerebrospinale su una geometria del sistema ventricolare cerebrale di un paziente malato di malattia di Alzheimer. Abbiamo imposto parametri e condizioni al bordo realistiche e, grazie all'approccio PolyDG, è stato possibile portare a termine una simulazione tridimensionale grazie all'agglomerazione di una mesh tetraedrica molto fine in un numero sostenibile di poliedri.
A 3D polyhedral solver for fluid dynamics with applications to the brain ventricles
Raimondi, Francesca
2023/2024
Abstract
This thesis addresses fluid dynamics problems in complex geometries, proposing the implementation of an efficient solver in order to run simulations in a three-dimensional domain, like the brain Ventricular System. Alzheimer's disease and other neurodegenerative pathologies have become more and more diffused, impacting on the medical community but also on the society. Even though some treatments have been experimented to slow down the progression of the disease, clinical research is ongoing to find effective treatment. Neurodegeneration is often correlated with the impairment of the waste clearance function played by the Cerebrospinal Fluid (CSF). Yet many features of CSF generation and flow in the Ventricular System are still subject to discussion. For this reason, CFD simulations may help in advancing research and integrate data from medical imaging. This thesis focuses on the implementation of a three-dimensional solver for the stationary Stokes problem, to represent the characteristics of a flow like that of CSF; then it is tested on benchmark and realistic mesh. To discretize and solve the problem we consider a Polytopal Discontinuous Galerkin (PolyDG) method, which guarantees geometrical flexibility in the choice of the computational mesh. The code builds upon the existing library Lymph3D, which relies on PETSc, METIS and fMETIS to perform matrix operations and mesh agglomeration. Lymph3D offers the advantage of agglomerating tetrahedral mesh in polyhedral elements, thus decreasing considerably computational time. Single terms of matrices and vectors are assembled on each tetrahedron but the most expensive steps as block matrix operations are performed on the polyhedral partition. As a result the novelty introduced by this thesis is an efficient strategy to solve a saddle point problem with polyhedral methods in three-dimensional domains. Efficiency of polyhedral agglomeration has been tested on a tetrahedral mesh, in order to show the advantage coming from the reduction of computational time. In the end polyhedral agglomeration is exploited to simulate the time-averaged CSF flow in the brain ventricular geometry of a patient affected by Alzheimer's disease, with suitable boundary conditions and parameters.File | Dimensione | Formato | |
---|---|---|---|
Raimondi_Tesi_definitive.pdf
accessibile in internet per tutti
Descrizione: final version of the thesis
Dimensione
6.98 MB
Formato
Adobe PDF
|
6.98 MB | Adobe PDF | Visualizza/Apri |
Raimondi_Executive_Summary_definitive.pdf
accessibile in internet per tutti
Descrizione: final version of the executive summary
Dimensione
1.34 MB
Formato
Adobe PDF
|
1.34 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/10589/235699