The present thesis work proposes a multisphere Euler-Lagrange numerical framework able to resolve the complex fluid dynamics and the catalytic heterogeneous chemistry in reactive non-spherical moving particle systems. The reactive Euler-Lagrange solver developed in previous thesis works, catalyticDPMFoam, has been extended to manage the non-sphericity of the solid particles, obtaining a new solver called catalyticDPMFoamMultiparticles. The existing framework has been modified by extending the classes representing the solid spheres in order to describe non-spherical particles as aggregates of spheres. The new solver finds applications in the simulations of gas-solid fluidized bed reactors involving non-spherical particles, as well as in the generation of packed beds with non-spherical catalysts that can be used to simulate fixed bed reactors. The multisphere particle tracking approach requires a method to generate the clumped spheres structures that approximate the real non-spherical particle to simulate. In this context, this work proposes two methodologies for the particles generation, which are based on the DEM approach and on the Ray-Tracing algorithm. These generators were validated in terms of final number of particles (that affects the computational time), and in terms of ratio between the volume of the multisphere body and the volume of the real non-spherical particle. The generated particles were simulated in different case studies that were analyzed in order to assess the reliability of the new solver. The capacity of the solver to replicate the correct void fraction and angle distribution of packed beds was tested against experimental data, analytical correlations and data from the literature. The comparison also involved the results obtained from the open-source codes LIGGGHTS and Blender. The new solver proved capable of giving precise results that are very close to the experimental data and in line with those obtained with the two open-source codes. At the same time, it is able to overcome the bottleneck of the two software. On the one hand, the new solver can process a larger number of solid entities with respect to the Blender code. On the other hand, it is able to represent the randomness of the experimental units that cannot be obtained with the LIGGGHTS code. The fluidization dynamics predicted using the catalyticDPMFoamMultiparticles solver were analyzed, highlighting the different behavior of differently-shaped particles. In particular, the pressure drops for beds of spherocylinders and disks obtained with the new solver were compared with the modified Ergun equation in the fixed bed regime, finding excellent agreement. The pressure drop during the fluidization was compared with the theoretical minimum fluidization pressure drop, resulting in a maximum error of 2.42% for the spherocylinders and 1.43% for the disks. The ability of the solver to manage reactive systems has been assessed by comparing a cylindrical reactor with 10,000 spherocylinders with a domain with the same number of volume equivalent spheres, simulated using the existing spherical solver. The domain was chosen small enough and the aspect ratio of the spheroylinders was sufficiently small in order to limit the effects of the different fluidization behavior on the conversion. A simple reaction with first order kinetics was simulated with the two the solvers obtaining almost identical conversions, which demonstrates the capability of the new solver to manage reactive systems.
Questo lavoro di tesi propone una metodologia di modellazione multisfera applicata al modello Euleriano-Lagrangiano capace di gestire la complessa fluidodinamica e la reattività catalitica eterogenea in sistemi reattivi gas-solido con particelle non sferiche. Il codice Euleriano-Lagrangiano sviluppato in precedenti lavori di tesi, catalyticDPMFoam, è stato esteso per considerare la forma non sferica delle particelle solide, ottenendo un nuovo solver chiamato catalyticDPMFoamMultiparticles. Il codice di partenza è stato modificato estendendo le classi che rappresentano la fase solida in modo tale da descrivere particelle non sferiche come aggregati di sfere. Il codice ottenuto trova applicazione nella simulazione di reattori gas-solido a letto fluidizzato con particelle non sferiche. Allo stesso modo, può essere impegato per generare letti impaccati con particelle non sferiche, necessari per simulazioni di reattori a letto fisso. L'approcio multisfera utilizzato richiede un metodo per generare l'aggregato di sfere che approssima la forma della particella non sferica da simulare. A tale scopo, questo lavoro di tesi propone due metodi per la generazione di particelle non sferiche, basate su un approcio DEM e sull'algoritmo di Ray-Tracing. Questi generatori sono stati convalidati confrontando il numero di particelle finale (che influenza il tempo di calcolo della simulazione), e il rapporto tra il volume della multisfera e il volume della particella originale. Le particelle generate sono state utilizzate per simulazioni di diversi case study che sono stati studiati per verificare l'affidabilità dei risultati ottenuti con il nuovo solver. Sono state testate le capacità del codice di replicare correttamente la frazione di vuoto e la distribuzione degli orientamenti delle particelle nei letti impaccati tramite confronti con dati sperimentali, correlazioni analitiche e risultati trovati in letteratura. Il confronto è stato inoltre esteso considerando i risultati ottenuti dai software open-source LIGGGHTS e Blender. Il codice sviluppato in questo lavoro di tesi ha dimostrato di fornire risultati precisi che sono molto simili ai dati sperimentali e in linea con i risultati ottenuti dai due software open-source. Allo stessto tempo, il codice sviluppato riesce a superare alcune limitazioni dei due software. Da un parte, il nuovo solver è capace di gestire un maggior numero di particelle solide rispetto a Blender. Dall'altra parte, esso è in grado di iniettare nel sistema particelle con posizioni e orientamenti random, in modo da rappresentare la casualità tipica delle prove sperimentali, che non può essere riprodotta utilizzando LIGGGHTS. Il comportamento di sistemi fluidizzati ottenuto con il solver catalyticDPMFoamMultiparticles è stato analizzato concentrandosi sulla diversa fluidizzazione di particelle con forme diverse. In particolare, le perdite di carico in letti di sferocilindri e cilindri, ottenute con il nuovo solver, sono state confrontate, in regime di letto fisso, con l'equazione di Ergun modificata, ottenendo un ottimo accordo tra la correlazione e i risultati del solver. Le perdite di carico nel regime di fluidizzazione sono state confrontate con le perdite di carico di minima fluidizzazione, ottenendo un errore massimo di 2.42% per gli sferocilindri e 1.43% per i cilindri. La capacità del solver di gestire sistemi reattivi è stata testata confrontando la simulazione di un reattore cilindrico con 10,000 sferocilindri con quella di un sistema analogo con particelle sferiche (simulate con il solver di partenza). Il reattore scelto ha un diametro sufficientemente piccolo, in modo da limitare l'effetto del diverso comportamento fluidodinamico di particelle con forme diverse. Una semplice reazione con cinetica del primo ordine è stata scelta per effettuare questo confronto e i risultati ottenuti per sferocilindri e sfere, in termini di conversione, sono risultati pressoché identici, confermando la capacità del solver di gestire sistemi reattivi.
Reactive multisphere modeling approach for the simulation of non-spherical moving particles in catalytic systems
Cipriano, Edoardo
2019/2020
Abstract
The present thesis work proposes a multisphere Euler-Lagrange numerical framework able to resolve the complex fluid dynamics and the catalytic heterogeneous chemistry in reactive non-spherical moving particle systems. The reactive Euler-Lagrange solver developed in previous thesis works, catalyticDPMFoam, has been extended to manage the non-sphericity of the solid particles, obtaining a new solver called catalyticDPMFoamMultiparticles. The existing framework has been modified by extending the classes representing the solid spheres in order to describe non-spherical particles as aggregates of spheres. The new solver finds applications in the simulations of gas-solid fluidized bed reactors involving non-spherical particles, as well as in the generation of packed beds with non-spherical catalysts that can be used to simulate fixed bed reactors. The multisphere particle tracking approach requires a method to generate the clumped spheres structures that approximate the real non-spherical particle to simulate. In this context, this work proposes two methodologies for the particles generation, which are based on the DEM approach and on the Ray-Tracing algorithm. These generators were validated in terms of final number of particles (that affects the computational time), and in terms of ratio between the volume of the multisphere body and the volume of the real non-spherical particle. The generated particles were simulated in different case studies that were analyzed in order to assess the reliability of the new solver. The capacity of the solver to replicate the correct void fraction and angle distribution of packed beds was tested against experimental data, analytical correlations and data from the literature. The comparison also involved the results obtained from the open-source codes LIGGGHTS and Blender. The new solver proved capable of giving precise results that are very close to the experimental data and in line with those obtained with the two open-source codes. At the same time, it is able to overcome the bottleneck of the two software. On the one hand, the new solver can process a larger number of solid entities with respect to the Blender code. On the other hand, it is able to represent the randomness of the experimental units that cannot be obtained with the LIGGGHTS code. The fluidization dynamics predicted using the catalyticDPMFoamMultiparticles solver were analyzed, highlighting the different behavior of differently-shaped particles. In particular, the pressure drops for beds of spherocylinders and disks obtained with the new solver were compared with the modified Ergun equation in the fixed bed regime, finding excellent agreement. The pressure drop during the fluidization was compared with the theoretical minimum fluidization pressure drop, resulting in a maximum error of 2.42% for the spherocylinders and 1.43% for the disks. The ability of the solver to manage reactive systems has been assessed by comparing a cylindrical reactor with 10,000 spherocylinders with a domain with the same number of volume equivalent spheres, simulated using the existing spherical solver. The domain was chosen small enough and the aspect ratio of the spheroylinders was sufficiently small in order to limit the effects of the different fluidization behavior on the conversion. A simple reaction with first order kinetics was simulated with the two the solvers obtaining almost identical conversions, which demonstrates the capability of the new solver to manage reactive systems.| File | Dimensione | Formato | |
|---|---|---|---|
|
tesi_Cipriano.pdf
non accessibile
Descrizione: corpo tesi
Dimensione
22.04 MB
Formato
Adobe PDF
|
22.04 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/10589/166573