In a porous medium featuring heterogeneous permeabilities, a wide range of fluid velocities may be recorded, so that significant inertial and frictional effects may arise in high-speed regions. In such parts, the link between pressure gradient and velocity is typically made via Darcy's law, which may fail to account for these effects; instead, the Darcy--Forchheimer law, which introduces a nonlinear term, may be more adequate. Applying the Darcy--Forchheimer law globally in the domain is very costly numerically and, rather, should only be done where strictly necessary. The question of finding a prori the subdomain where to restrict the use of the Darcy--Forchheimer law was recently answered by using an adaptive model; given a threshold on the flow’s velocity, the model locally selects the more appropriate law as it is being solved. At the end of the resolution, each mesh cell is flagged as being in the Darcy or Darcy--Forchheimer subdomain. Still, this model is nonlinear itself and thus relatively expensive to run. In this paper, to accelerate the subdivision of the domain into low- and high-speed regions, we instead exploit the adaptive model from previous studies to generate partitioning data given an array of different input parameters, such as boundary conditions and inertial coefficients, and then train neural networks on these data classifying each mesh cell as Darcy or not. Two test cases are studied to illustrate the results, where cost functions, parity plots, precision-recall plots and receiver operating characteristic curves are analyzed.
In un mezzo poroso con permeabilità eterogenea, si possono osservare una vasta gamma di velocità del fluido, con conseguenti effetti inerziali e attriti significativi nelle regioni ad alta velocità. In queste aree, la relazione tra il gradiente di pressione e la velocità, di solito espressa dalla legge di Darcy, potrebbe non essere sufficiente a considerare tali effetti; pertanto, la legge di Darcy-Forchheimer, che include un termine non lineare, potrebbe risultare più appropriata. Tuttavia, l'applicazione globale della legge di Darcy-Forchheimer nell'intero dominio è computazionalmente onerosa e dovrebbe essere eseguita solo quando strettamente necessario. Recentemente, il problema di identificare "a priori" un sottoinsieme del dominio in cui applicare la legge di Darcy-Forchheimer è stato affrontato attraverso l'uso di un modello adattivo. Utilizzando una soglia sulla velocità del flusso, il modello seleziona localmente la legge più appropriata durante la risoluzione. Al termine della simulazione, ogni cella della griglia viene classificata come appartenente al sottoinsieme Darcy o Darcy-Forchheimer. Tuttavia, questo modello è intrinsecamente non lineare e quindi richiede risorse computazionali considerevoli. In questo lavoro, per accelerare la suddivisione del dominio in regioni a bassa e alta velocità, sfruttiamo il modello adattivo proposto in precedenza per generare dati delle partizione da una serie di parametri di input diversi, come le condizioni al contorno e i coefficienti inerziali. Successivamente, addestriamo le reti neurali su tali dati per classificare ogni cella della griglia come Darcy o non-Darcy. Per illustrare i risultati, abbiamo esaminato due casi di studio in cui abbiamo analizzato le funzioni di costo, i diagrammi di parità, i diagrammi precisione-richiamo e le curve caratteristiche operative del ricevitore.
Predicting nonlinear-flow regions in highly heterogeneous porous media using adaptive constitutive laws and neural networks
Giovannini, Chiara
2022/2023
Abstract
In a porous medium featuring heterogeneous permeabilities, a wide range of fluid velocities may be recorded, so that significant inertial and frictional effects may arise in high-speed regions. In such parts, the link between pressure gradient and velocity is typically made via Darcy's law, which may fail to account for these effects; instead, the Darcy--Forchheimer law, which introduces a nonlinear term, may be more adequate. Applying the Darcy--Forchheimer law globally in the domain is very costly numerically and, rather, should only be done where strictly necessary. The question of finding a prori the subdomain where to restrict the use of the Darcy--Forchheimer law was recently answered by using an adaptive model; given a threshold on the flow’s velocity, the model locally selects the more appropriate law as it is being solved. At the end of the resolution, each mesh cell is flagged as being in the Darcy or Darcy--Forchheimer subdomain. Still, this model is nonlinear itself and thus relatively expensive to run. In this paper, to accelerate the subdivision of the domain into low- and high-speed regions, we instead exploit the adaptive model from previous studies to generate partitioning data given an array of different input parameters, such as boundary conditions and inertial coefficients, and then train neural networks on these data classifying each mesh cell as Darcy or not. Two test cases are studied to illustrate the results, where cost functions, parity plots, precision-recall plots and receiver operating characteristic curves are analyzed.File | Dimensione | Formato | |
---|---|---|---|
2024_04_Giovannini_Executive_Summary_02.pdf
non accessibile
Descrizione: executive summary
Dimensione
521.73 kB
Formato
Adobe PDF
|
521.73 kB | Adobe PDF | Visualizza/Apri |
2024_04_Giovannini_01.pdf
non accessibile
Descrizione: tesi
Dimensione
2.61 MB
Formato
Adobe PDF
|
2.61 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/10589/218749