There is an increase need for accurate numerical solutions to transport problems in porous and/or fractured formations. Considerable efforts have been taken in the last decades to come up with numerical techniques able to solve subsurface transport problems, but obtaining accurate simulations still remains a challenge, especially when transport is advection-dominated. Traditional Eulerian approaches aren’t the best to get these results, since they require a highly fine discretisation of the domain to overcome unstable numerical solutions and/or artificial diffusion. To overcome these issues, the random walk particle tracking method is introduced; it is a Lagrangian method, in which particles are moved throughout the domain multiplying the velocity components by a given time-step and then, repeating the process. When dealing with fractured media, microporous unsaturated soils and heterogeneous saturated porous media, the non-equilibrium sorption equations are used to model such physical processes. In this work one of the first solution proposed in the literature is implemented in Matlab (Kinzelbach, The random walk method in pollutant transport simulation, 1988). According to Kinzelbach there exist two phases: a moving one and a non-moving one. Each particle can jump from one phase to the other thanks to the definition of two different transition probabilities, which are determined by the mass transfer coefficient and the distribution coefficient. Then, each particle belonging to the mobile phase is displaced according to a formula that takes into account both the advective and the dispersion term. An experimental dataset is analysed, and the implemented Matlab code is tested on it. The dataset comprehends two different tracer test column experiments. In both of them two solutes were injected: a conservative one (NaI in both cases) and a non-conservative one (Eosin-Y for the former one, FITC for the latter one). In the first place the BTC of the conservative solute were reproduced, because they required only two input parameters (pore-velocity and dispersion coefficient). Once the pore-velocity and the dispersion coefficient were assessed, the BTC of the two non-conservative solutes were reproduced. Since no data values of mass transfer coefficient and distribution coefficient were found in the literature, two minimisation functions (the quadratic and logarithmic errors) were defined, in order to fit the experimental BTCs. After several trials, the code perfectly matched the BTC of FITC only. As a conclusion it can be stated that the model is able to reproduce the mass transfer of a non-conservative solute, but it should be tested on more and broader datasets, in order to test its robustness. Moreover, a higher CPU should be used, seen the computational burden of the code.
Negli ultimi anni è aumentata la necessità di ottenere soluzioni numeriche accurate riguardo il trasporto di soluti in formazioni porose e/o fratturate. Notevoli sforzi sono stati fatti per elaborare tecniche numeriche all’avanguardia, ma ottenere soluzioni accurate rimane ancora una sfida, specialmente quando il trasporto è dominato dall’avvezione. Il tradizionale approccio Euleriano non dà i risultati sperati, dal momento che richiede un’altissima discretizzazione del dominio per evitare soluzioni numeriche instabili. Per ovviare a queste problematiche, il metodo chiamato “random walk particle tracking” viene introdotto; esso è un metodo Lagrangiano secondo il quale le particelle vengono spostate in tutto il dominio moltiplicando le componenti della velocità per un determinato passo temporale e poi, ripetendo il processo. Quando si ha a che fare con terreni fratturati, terreni insaturi microporosi e di mezzi porosi saturi eterogenei, vengono utilizzate le cosiddette equazioni “non-equilibrium sorption” per modellare tali processi fisici. In questa tesi, una delle prime soluzioni proposte in letteratura viene implementata in Matlab ( (Kinzelbach, The random walk method in pollutant transport simulation, 1988). Kinzelbach propone un modello costituito da due fasi: una mobile e una immobile. Ogni particella può saltare da una fase all’altra grazie alla definizione di due diverse probabilità di transizione, determinate dai coefficienti di scambio di massa e di ripartizione. Quindi, ogni particella appartenente alla fase mobile viene mossa secondo una legge che considera sia il contributo avvettivo sia quello dispersivo. Dopo essere stato implementato il codice viene testato su due set di dati sperimentali, che comprendono i dati ottenuti in due diversi esperimenti in colonna effettuati con traccianti. In entrambi due soluti vengono iniettati: uno conservativo (Ioduro di sodio in entrambi i casi) e uno non conservativo (Eosina-Y e Fluoresceina). Inizialmente sono state riprodotte le BTC del soluto conservativo, visto che richiedevano soltanto due parametri di input (velocità al poro e coefficiente di dispersione). Una volta valutati velocità al poro e coefficiente di dispersione, sono state riprodotte le BTC dei soluti non conservativi. Poiché in letteratura non sono stati trovati valori relativi al coefficiente di scambio di massa e al coefficiente di ripartizione, sono state definite e minimizzate due funzioni (scarto quadratico e logaritmico) in modo da trovare i migliori fit dei dati sperimentali. Dopo molteplici tentativi soltanto le BTC relative alla FITC sono state riprodotte perfettamente. In conclusione, si può affermare che il modello è in grado di riprodurre lo scambio di massa che intercorre tra un soluto e il mezzo poroso in cui viene iniettato, ma dovrebbe essere calibrato su set di dati molto più ampi. Inoltre, visto il peso computazionale del codice, CPU più potenti dovrebbero essere utilizzate.
Random particle tracking method for modeling tracer tests column experiments
BURATO, GIANLUCA
2017/2018
Abstract
There is an increase need for accurate numerical solutions to transport problems in porous and/or fractured formations. Considerable efforts have been taken in the last decades to come up with numerical techniques able to solve subsurface transport problems, but obtaining accurate simulations still remains a challenge, especially when transport is advection-dominated. Traditional Eulerian approaches aren’t the best to get these results, since they require a highly fine discretisation of the domain to overcome unstable numerical solutions and/or artificial diffusion. To overcome these issues, the random walk particle tracking method is introduced; it is a Lagrangian method, in which particles are moved throughout the domain multiplying the velocity components by a given time-step and then, repeating the process. When dealing with fractured media, microporous unsaturated soils and heterogeneous saturated porous media, the non-equilibrium sorption equations are used to model such physical processes. In this work one of the first solution proposed in the literature is implemented in Matlab (Kinzelbach, The random walk method in pollutant transport simulation, 1988). According to Kinzelbach there exist two phases: a moving one and a non-moving one. Each particle can jump from one phase to the other thanks to the definition of two different transition probabilities, which are determined by the mass transfer coefficient and the distribution coefficient. Then, each particle belonging to the mobile phase is displaced according to a formula that takes into account both the advective and the dispersion term. An experimental dataset is analysed, and the implemented Matlab code is tested on it. The dataset comprehends two different tracer test column experiments. In both of them two solutes were injected: a conservative one (NaI in both cases) and a non-conservative one (Eosin-Y for the former one, FITC for the latter one). In the first place the BTC of the conservative solute were reproduced, because they required only two input parameters (pore-velocity and dispersion coefficient). Once the pore-velocity and the dispersion coefficient were assessed, the BTC of the two non-conservative solutes were reproduced. Since no data values of mass transfer coefficient and distribution coefficient were found in the literature, two minimisation functions (the quadratic and logarithmic errors) were defined, in order to fit the experimental BTCs. After several trials, the code perfectly matched the BTC of FITC only. As a conclusion it can be stated that the model is able to reproduce the mass transfer of a non-conservative solute, but it should be tested on more and broader datasets, in order to test its robustness. Moreover, a higher CPU should be used, seen the computational burden of the code.File | Dimensione | Formato | |
---|---|---|---|
Random walk particle tracking method for modeling tracer tests column experiments.pdf
accessibile in internet solo dagli utenti autorizzati
Descrizione: Elaborato finale
Dimensione
4.73 MB
Formato
Adobe PDF
|
4.73 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/10589/146788