Under thermodynamic conditions near the liquid-vapor saturation curve and the critical point, flow behavior deviates significantly from that predicted by the ideal gas assumption. To describe these peculiar flows, a specialized branch of fluid mechanics has emerged called Non-Ideal Compressible Fluid Dynamics (NICFD). NICFD investigates compressible flows that exhibit distinctive behaviors modeled by non-ideal gas equations. Examples include non-monotone Mach number variation in isentropic flow within converging- diverging nozzles, Mach-increasing non-ideal oblique shocks, and non-classical rarefaction shocks. NICFD phenomena can be observed in under-expanded jets, where the nozzle exit pressure exceeds the ambient pressure. Although most research focused on under -expanded jets under ideal flow assumptions, in non-ideal conditions, thermodynamic and gas dynamic non-ideal effects are expected to influence the behavior of flow features such as expansion fans and shock waves. This thesis aims to address the existing knowledge gap through a numerical investigation that compares flow characteristics in ideal and non-ideal compressible regimes. A new fully open-source computational framework SU2-COOL is developed to compute non-ideal flow fields with state-of-the-art thermodynamic models. The open-source computational suite SU2 is integrated with the thermophysical property library CoolProp through a newly developed C++ interface. To validate the newly developed framework, three test cases are considered. An isentropic flow through a converging-diverging nozzle, using nitrogen to represent ideal flow and siloxane MDM to represent non-ideal flow, with the objective of capturing non-ideal thermodynamic effects. The second case examines isentropic flow through a converging-diverging nozzle using siloxane MM, and focuses on capturing a non-ideal gas dynamic phenomenon—namely, the non-monotonic behavior of the Mach number. The third case addresses supersonic flow over a wedge with MD4M, highlighting the emergence of a non-classical rarefaction shock—a phenomenon characteristic of non-ideal compressible flows. All numerical results are compared with either experimental data or reference solutions computed using established numerical solvers. The performance of SU2-COOL is also evaluated. The framework demonstrates good scalability with increasing computational workload and number of CPU cores. The solver accuracy is verified using the three test cases. The computational cost is higher com pared to implementations using embedded cubic equations of state, due to the overhead associated with external calls to the thermophysical property library. This framework is employed for the numerical simulations presented in this thesis. To investigate non-ideal effects on the expansion fans originating from the nozzle exit lip, the governing equations for the Prandtl–Meyer function in non-ideal flows are solved numerically using Runge–Kutta methods. The numerical solutions are compared with results from the literature, experimental data, and simulations performed using SU2-COOL across various test cases. Various operating conditions characterized by different degrees of flow non-ideality, quantified by the total compressibility factor Zt , are considered using siloxane MM to investigate the impact of non-ideal effects on expansion fans, specifically in terms of Mach number and pressure variation. As flow non-ideality increases, the changes in Mach number and pressure across the expansion fan become less pronounced, indicating reduced flow expansion. Additionally, working conditions characterized by different Pt and Tt but sharing the same Zt are considered. For MM, when Zt exceeds 0.6, results corresponding to identical Zt values exhibit close agreement, with a maximum deviation below 5%. Under these circumstances, Zt serves as an effective similarity parameter to represent the flow non-ideality. Conversely, the pre-expansion fundamental derivative of gasdynamics, Γ1, does not demonstrate this behavior. Moreover, fluids with high molecular complexity and molar mass exhibit a similar phenomenon to MM, whereas those with low molecular complexity and molar mass do not. The influence of flow non ideality on Prandtl–Meyer expansions also provides an explanation for the behavior of Mach disks in non-ideal flows. Numerical simulations of inviscid under-expanded jets in non-ideal flows were conducted to investigate near-field flow characteristics up to the Mach disk. A mesh adaptation technique is employed to accurately capture flow features with high resolution while min imizing computational cost. The in-house Darwin code, which performs error estimation and grid modification, is used. It incorporates both isotropic and anisotropic mesh adap tation strategies and supports various flow variables. The numerical framework combining SU2 and Darwin is validated using a test case from the open literature. The computed distribution of Pitot pressure along the nozzle centerline shows good agreement with ex perimental data up to the Mach disk. Numerical simulation in ideal flow with Nitrogen is performed with different error indicators, including isotropic gradient, isotropic Hessian matrix, and anisotropic indicators based on density, Mach number, pressure, tempera ture, and entropy. Solutions using the isotropic gradient generally capture flow features effectively, except for those based on pressure, since the pressure gradient approaches zero near constant pressure boundaries or contact surfaces. In comparison, solutions based on the isotropic Hessian matrix tend to over-resolve the contact surface, and capture the Kelvin–Helmholtz instability along with associated vortex structures. Solutions employ ing anisotropic mesh adaptation capture flow features with the fewest elements and the lowest computational cost compared to their isotropic counterparts. Assessment of the same error indicators is then performed for non-ideal flows with MM. In non-ideal flows, the temperature change across the contact surface is significantly greater than that across the oblique shock, whereas in ideal flows, temperature differences across the contact sur face and the shock are of the same order of magnitude. Considering both the resolution of flow features and computational efficiency, anisotropic indicators based on density and Mach number are recommended for the numerical simulation of inviscid under-expanded jets in both ideal and non-ideal flows. Simulation under working conditions with different Zt is performed for MM, and it is found that the Mach disk moves upstream and has a larger radius with lower Zt for higher flow non-ideality. To capture flow features downstream of the Mach disk and the mixing process between the jet and the ambient, numerical simulation of under-expanded jets is carried out, account ing for the effect of viscosity and thermal conductivity, using the same mesh adaptation method adopted for inviscid flows. The same verification test case is considered, and the recovery of pitot pressure along the nozzle centerline is successfully captured due to turbulence momentum transfer. The contact surface is smoother in viscous flow, as ex pected. Assessment of error indicators is performed again for ideal and non-ideal flows. Solutions based solely on Mach number as the error sensor fail to capture flow features accurately. The Mach number exhibits strong variations near the no-slip wall boundary, thus masking the remaining flow features. To address this limitation, a combination of multiple error indicators is employed. Due to the rapid variation of flow variables near the wall boundary, solutions using anisotropic mesh adaptation result in significant mesh refinement in that region, leading to increased computational cost. Isotropic mesh adap tation based on the gradient of density is recommended, as it offers a favorable balance between flow feature resolution and computational efficiency. The optimal error indicators differ between inviscid and viscous flows due to the influence of viscosity on the variation of flow variables. The influence of flow non-ideality on the Mach disk location and radius in viscous flows is consistent with the observations made in inviscid flows. The results presented in this thesis highlight the impact of flow non-ideality on the flow of under-expanded jets in both ideal and non-ideal compressible regimes. Flow charac teristics are captured using the in-house mesh adaptation tool Darwin coupled to the SU2-COOL framework. Expansion fans and Mach disks exhibit distinct characteristics in non-ideal flows, thereby illustrating the influence of flow non-ideality. This work lays the foundation for future research, paving the way for NICFD studies on under-expanded jets.
In condizioni termodinamiche prossime alla curva di saturazione liquido-vapore e al punto critico, il comportamento del flusso si discosta significativamente da quello previsto dall’ipotesi di gas ideale. Per affrontare questo problema, è emersa una branca specializzata della meccanica dei fluidi nota come Fluidodinamica Comprimibile Non Ideale (NICFD). La NICFD studia i flussi comprimibili che mostrano comportamenti distintivi solo quando modellati utilizzando equazioni di gas non ideali. Esempi includono il numero di Mach non monotono del flusso isentropico su un ugello convergente-divergente, un’onda d’urto obliqua non ideale dopo la quale il numero di Mach aumenta e un’onda d’urto di rar efazione non classica. I fenomeni NICFD possono essere osservati nei getti sottoespansi, dove la pressione di uscita dall’ugello supera la pressione ambiente. Sebbene la maggior parte della ricerca si sia concentrata sui getti sottoespansi in ipotesi di flusso ideale, gli ef fetti della non idealità del flusso non possono essere trascurati, in particolare in relazione a caratteristiche del flusso come ventagli di espansione e onde d’urto. Questa tesi si propone di colmare l’attuale lacuna di conoscenza attraverso un’indagine numerica che confronta le caratteristiche del flusso in regimi comprimibili ideali e non ideali. È stato sviluppato un nuovo framework computazionale completamente open-source SU2-COOL per migliorare l’accuratezza della soluzione in campi di flusso non ideali. La suite com putazionale open-source SU2 è integrata con la libreria di proprietà termofisiche CoolProp tramite una nuova interfaccia C++. Per validare il framework di nuova concezione, ven gono considerati tre casi di test. Il primo riguarda il flusso isentropico attraverso un ugello convergente-divergente, utilizzando azoto per rappresentare il flusso ideale e silos sano MDM per rappresentare il flusso non ideale, con l’obiettivo di catturare fenomeni termodinamici non ideali. Il secondo caso esamina anch’esso il flusso isentropico attraverso un ugello convergente-divergente utilizzando silossano MM, ma si concentra sulla cattura di un fenomeno gasdinamico non ideale, ovvero il comportamento non monotono del nu mero di Mach. Il terzo caso riguarda un flusso supersonico su un cuneo con MD4M, evidenziando l’emergere di uno shock di rarefazione non classico, un fenomeno caratter istico dei flussi comprimibili non ideali. Tutti i risultati numerici vengono confrontati con dati sperimentali o con soluzioni di riferimento calcolate utilizzando solutori numerici consolidati. Vengono inoltre valutate le prestazioni di SU2-COOL. Il framework dimostra una buona scalabilità con l’aumento del carico di lavoro computazionale e del numero di core della CPU. L’accuratezza della sua soluzione è verificata utilizzando i tre casi di test descritti in precedenza. Tuttavia, il costo computazionale è maggiore rispetto alle implementazioni che utilizzano equazioni di stato cubiche incorporate, a causa del sovrac carico associato alle chiamate esterne alla libreria delle proprietà termofisiche. Questo framework viene utilizzato per le simulazioni numeriche presentate in questa tesi, grazie alle sue prestazioni e alla sua accuratezza favorevoli. Per studiare i ventilatori di espansione che originano dal labbro di uscita dell’ugello in condizioni non ideali, le equazioni che governano la funzione di Prandtl-Meyer in flussi non ideali vengono risolte numericamente utilizzando i metodi di Runge-Kutta. Le soluzioni numeriche vengono confrontate con i risultati della letteratura, i dati sperimentali e le simulazioni eseguite utilizzando SU2-COOL in vari casi di prova. Diverse condizioni op erative caratterizzate da diversi gradi di non idealità del flusso, quantificati dal fattore di comprimibilità totale Zt , vengono considerate utilizzando il silossano MM per studiare l’impatto degli effetti non ideali sui ventilatori di espansione, in particolare in termini di numero di Mach e variazione di pressione. All’aumentare della non idealità del flusso, le variazioni di numero di Mach e pressione attraverso il ventilatore di espansione diventano meno pronunciate, indicando una ridotta espansione del flusso. Inoltre, vengono consid erate condizioni di lavoro caratterizzate da diversi valori di Pt e Tt , ma che condividono lo stesso valore di Zt . Per MM, quando Zt supera 0,6, i risultati corrispondenti a valori identici di Zt mostrano una stretta concordanza, con una deviazione massima inferiore al 5%. In queste circostanze, Zt funge da parametro di similarità efficace. Al contrario, la derivata fondamentale pre-espansione della gasdinamica, Γ1, non mostra questo compor tamento. Inoltre, i fluidi con elevata complessità molecolare e massa molare mostrano un fenomeno simile a MM, mentre quelli con bassa complessità molecolare e massa molare no. L’influenza della non idealità del flusso sulle espansioni di Prandtl-Meyer fornisce anche una spiegazione del comportamento dei dischi di Mach nei flussi non ideali. Viene condotta una simulazione numerica di getti sottoespansi non viscosi in flussi non ideali per studiare le caratteristiche del flusso in campo vicino fino al disco di Mach. Viene impiegata una tecnica di adattamento della mesh per catturare accuratamente le carat teristiche del flusso ad alta risoluzione, minimizzando al contempo i costi computazionali. Viene presentato il codice Darwin, che esegue la stima dell’errore e la modifica della griglia. Incorpora strategie di adattamento della mesh sia isotropiche che anisotropiche e supporta diverse variabili di flusso. Il framework numerico che combina SU2 e Dar win viene validato utilizzando un caso di test tratto dalla letteratura. La distribuzione calcolata della pressione del tubo di Pitot lungo l’asse dell’ugello mostra un buon ac cordo con i dati sperimentali fino al disco di Mach. Successivamente, viene eseguita una simulazione numerica in flusso ideale con azoto con diversi indicatori di errore, tra cui gradiente isotropico, matrice hessiana isotropica e indicatori anisotropici basati su den sità, numero di Mach, pressione, temperatura ed entropia. Le soluzioni che utilizzano il gradiente isotropico generalmente catturano efficacemente le caratteristiche del flusso, ad eccezione di quelle basate sulla pressione, poiché il gradiente di pressione si avvicina a zero in prossimità di limiti di pressione costanti o superfici di contatto. Al contrario, le soluzioni basate sulla matrice Hessiana isotropica tendono a sovra-risolvere la superficie di contatto e a catturare l’instabilità di Kelvin-Helmholtz insieme alle strutture vorticose associate. Le soluzioni che utilizzano l’adattamento anisotropico della mesh catturano le caratteristiche del flusso con il minor numero di elementi e il minor costo computazionale rispetto alle loro controparti isotropiche. La valutazione degli stessi indicatori di errore viene quindi eseguita per flussi non ideali con MM. Nei flussi non ideali, la variazione di temperatura lungo la superficie di contatto è significativamente maggiore di quella lungo l’onda d’urto obliqua, mentre nei flussi ideali entrambe le variazioni sono dello stesso ordine di grandezza. Considerando sia la risoluzione delle caratteristiche del flusso che l’efficienza computazionale, gli indicatori anisotropici basati sulla densità e sul numero di Mach sono raccomandati per la simulazione numerica di getti non viscosi sottoespansi sia in flussi ideali che non ideali. Viene eseguita una simulazione in condizioni di lavoro con diversi Zt per MM e si scopre che il disco di Mach si muove a monte e ha un raggio maggiore. Per catturare le caratteristiche del flusso dopo il disco di Mach e il processo di misce lazione, viene eseguita una simulazione numerica di getti viscosi sottoespansi, insieme allo stesso metodo di adattamento della mesh adottato per i flussi non viscosi. Viene considerato lo stesso caso di verifica e il recupero della pressione del tubo di Pitot lungo l’asse dell’ugello viene catturato con successo grazie al trasferimento di momento tur bolento. La superficie di contatto è più liscia nel flusso viscoso a causa della diffusione molecolare. La valutazione degli indicatori di errore viene eseguita nuovamente per flussi ideali e non ideali. Le soluzioni basate esclusivamente sul numero di Mach non riescono a catturare accuratamente le caratteristiche del flusso, poiché il numero di Mach mostra forti variazioni in prossimità del bordo della parete antiscivolo. Per ovviare a questa limitazione, viene utilizzata una combinazione di più indicatori di errore. A causa della rapida variazione delle variabili di flusso in prossimità del bordo della parete, le soluzioni che utilizzano l’adattamento della mesh anisotropica comportano un significativo raffina mento della mesh in quella regione, con conseguente aumento dei costi computazional Di conseguenza, si raccomanda l’adattamento della mesh isotropica basato sul gradiente di densità, in quanto offre un equilibrio favorevole tra risoluzione delle caratteristiche del flusso ed efficienza computazionale. Gli indicatori di errore ottimali differiscono tra flussi non viscosi e viscosi a causa dell’influenza della viscosità sulla variazione delle variabili di flusso. L’influenza della non idealità del flusso sul disco di Mach nei flussi viscosi è coerente con le osservazioni effettuate nei flussi non viscosi. I risultati presentati in questa tesi evidenziano l’impatto della non idealità del flusso sulle caratteristiche di flusso di getti sottoespansi in regimi comprimibili sia ideali che non ideali. Le caratteristiche di flusso sono catturate con il metodo di adattamento mesh nell’ambito del framework SU2-COOL. I ventilatori di espansione e i dischi di Mach mostrano caratteristiche distinte nei flussi non ideali, illustrando così l’influenza della non idealità del flusso. Questo lavoro getta le basi per la ricerca futura, aprendo la strada alla ricerca di NICFD sui getti sottoespansi.
A numerical study of under-expanded jets in ideal and Non-Ideal Compressible Flows
YAN, PENG
2026
Abstract
Under thermodynamic conditions near the liquid-vapor saturation curve and the critical point, flow behavior deviates significantly from that predicted by the ideal gas assumption. To describe these peculiar flows, a specialized branch of fluid mechanics has emerged called Non-Ideal Compressible Fluid Dynamics (NICFD). NICFD investigates compressible flows that exhibit distinctive behaviors modeled by non-ideal gas equations. Examples include non-monotone Mach number variation in isentropic flow within converging- diverging nozzles, Mach-increasing non-ideal oblique shocks, and non-classical rarefaction shocks. NICFD phenomena can be observed in under-expanded jets, where the nozzle exit pressure exceeds the ambient pressure. Although most research focused on under -expanded jets under ideal flow assumptions, in non-ideal conditions, thermodynamic and gas dynamic non-ideal effects are expected to influence the behavior of flow features such as expansion fans and shock waves. This thesis aims to address the existing knowledge gap through a numerical investigation that compares flow characteristics in ideal and non-ideal compressible regimes. A new fully open-source computational framework SU2-COOL is developed to compute non-ideal flow fields with state-of-the-art thermodynamic models. The open-source computational suite SU2 is integrated with the thermophysical property library CoolProp through a newly developed C++ interface. To validate the newly developed framework, three test cases are considered. An isentropic flow through a converging-diverging nozzle, using nitrogen to represent ideal flow and siloxane MDM to represent non-ideal flow, with the objective of capturing non-ideal thermodynamic effects. The second case examines isentropic flow through a converging-diverging nozzle using siloxane MM, and focuses on capturing a non-ideal gas dynamic phenomenon—namely, the non-monotonic behavior of the Mach number. The third case addresses supersonic flow over a wedge with MD4M, highlighting the emergence of a non-classical rarefaction shock—a phenomenon characteristic of non-ideal compressible flows. All numerical results are compared with either experimental data or reference solutions computed using established numerical solvers. The performance of SU2-COOL is also evaluated. The framework demonstrates good scalability with increasing computational workload and number of CPU cores. The solver accuracy is verified using the three test cases. The computational cost is higher com pared to implementations using embedded cubic equations of state, due to the overhead associated with external calls to the thermophysical property library. This framework is employed for the numerical simulations presented in this thesis. To investigate non-ideal effects on the expansion fans originating from the nozzle exit lip, the governing equations for the Prandtl–Meyer function in non-ideal flows are solved numerically using Runge–Kutta methods. The numerical solutions are compared with results from the literature, experimental data, and simulations performed using SU2-COOL across various test cases. Various operating conditions characterized by different degrees of flow non-ideality, quantified by the total compressibility factor Zt , are considered using siloxane MM to investigate the impact of non-ideal effects on expansion fans, specifically in terms of Mach number and pressure variation. As flow non-ideality increases, the changes in Mach number and pressure across the expansion fan become less pronounced, indicating reduced flow expansion. Additionally, working conditions characterized by different Pt and Tt but sharing the same Zt are considered. For MM, when Zt exceeds 0.6, results corresponding to identical Zt values exhibit close agreement, with a maximum deviation below 5%. Under these circumstances, Zt serves as an effective similarity parameter to represent the flow non-ideality. Conversely, the pre-expansion fundamental derivative of gasdynamics, Γ1, does not demonstrate this behavior. Moreover, fluids with high molecular complexity and molar mass exhibit a similar phenomenon to MM, whereas those with low molecular complexity and molar mass do not. The influence of flow non ideality on Prandtl–Meyer expansions also provides an explanation for the behavior of Mach disks in non-ideal flows. Numerical simulations of inviscid under-expanded jets in non-ideal flows were conducted to investigate near-field flow characteristics up to the Mach disk. A mesh adaptation technique is employed to accurately capture flow features with high resolution while min imizing computational cost. The in-house Darwin code, which performs error estimation and grid modification, is used. It incorporates both isotropic and anisotropic mesh adap tation strategies and supports various flow variables. The numerical framework combining SU2 and Darwin is validated using a test case from the open literature. The computed distribution of Pitot pressure along the nozzle centerline shows good agreement with ex perimental data up to the Mach disk. Numerical simulation in ideal flow with Nitrogen is performed with different error indicators, including isotropic gradient, isotropic Hessian matrix, and anisotropic indicators based on density, Mach number, pressure, tempera ture, and entropy. Solutions using the isotropic gradient generally capture flow features effectively, except for those based on pressure, since the pressure gradient approaches zero near constant pressure boundaries or contact surfaces. In comparison, solutions based on the isotropic Hessian matrix tend to over-resolve the contact surface, and capture the Kelvin–Helmholtz instability along with associated vortex structures. Solutions employ ing anisotropic mesh adaptation capture flow features with the fewest elements and the lowest computational cost compared to their isotropic counterparts. Assessment of the same error indicators is then performed for non-ideal flows with MM. In non-ideal flows, the temperature change across the contact surface is significantly greater than that across the oblique shock, whereas in ideal flows, temperature differences across the contact sur face and the shock are of the same order of magnitude. Considering both the resolution of flow features and computational efficiency, anisotropic indicators based on density and Mach number are recommended for the numerical simulation of inviscid under-expanded jets in both ideal and non-ideal flows. Simulation under working conditions with different Zt is performed for MM, and it is found that the Mach disk moves upstream and has a larger radius with lower Zt for higher flow non-ideality. To capture flow features downstream of the Mach disk and the mixing process between the jet and the ambient, numerical simulation of under-expanded jets is carried out, account ing for the effect of viscosity and thermal conductivity, using the same mesh adaptation method adopted for inviscid flows. The same verification test case is considered, and the recovery of pitot pressure along the nozzle centerline is successfully captured due to turbulence momentum transfer. The contact surface is smoother in viscous flow, as ex pected. Assessment of error indicators is performed again for ideal and non-ideal flows. Solutions based solely on Mach number as the error sensor fail to capture flow features accurately. The Mach number exhibits strong variations near the no-slip wall boundary, thus masking the remaining flow features. To address this limitation, a combination of multiple error indicators is employed. Due to the rapid variation of flow variables near the wall boundary, solutions using anisotropic mesh adaptation result in significant mesh refinement in that region, leading to increased computational cost. Isotropic mesh adap tation based on the gradient of density is recommended, as it offers a favorable balance between flow feature resolution and computational efficiency. The optimal error indicators differ between inviscid and viscous flows due to the influence of viscosity on the variation of flow variables. The influence of flow non-ideality on the Mach disk location and radius in viscous flows is consistent with the observations made in inviscid flows. The results presented in this thesis highlight the impact of flow non-ideality on the flow of under-expanded jets in both ideal and non-ideal compressible regimes. Flow charac teristics are captured using the in-house mesh adaptation tool Darwin coupled to the SU2-COOL framework. Expansion fans and Mach disks exhibit distinct characteristics in non-ideal flows, thereby illustrating the influence of flow non-ideality. This work lays the foundation for future research, paving the way for NICFD studies on under-expanded jets.| File | Dimensione | Formato | |
|---|---|---|---|
|
2026_06_Yan.pdf
accesso aperto
Licenza:
Tutti i diritti riservati
Dimensione
33.02 MB
Formato
Adobe PDF
|
33.02 MB | Adobe PDF | Visualizza/Apri |
I documenti in UNITESI sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/20.500.14242/384636
URN:NBN:IT:POLIMI-384636