1720 MHz OH masers have been detected towards a number of supernova remnants (SNRs) at the shock interface where the SNR slams into the interstellar medium. Models indicate that these masers are shock excited and can only be produced under tight constraints of the physical conditions. In particular, the masers can only form behind a C-type shock. Jets from newlyformed stars plow into the surrounding gas, creating nebulous regions known as Herbig Haro (HH) objects. Signatures of C-type shocks have been found in many HH objects. If conditions behind the shock fronts of HH objects are able to support 1720 MHz OH masers they would be a usefull diagnostic tool for star formation. A survey toward HH objects detected a number of 1720 MHz OH lines in emission, but future observations with arrays are required to confirm the presence of masers. / Physics / M.Sc. (Astronomy)

The Eulerian description of dispersed two-phase flows results in a system of partial differential equations describing characteristics of the flow, namely volume fraction, density and velocity of the two phases, around any point in space over time. When pressure forces are neglected or a same pressure is considered for both phases, the resulting system is weakly hyperbolic and solutions may exhibit vacuum states (regions void of the dispersed phase) or localized unbounded singularities (delta shocks) that are not physically desirable. Therefore, it is crucial to find a physical way for preventing the formation of such undesirable solutions in weakly hyperbolic Eulerian two-phase flow models. This thesis focuses on the mathematical analysis of an Eulerian model for air- droplet flows, here called the Eulerian droplet model. This model can be seen as the sticky particle system with a source term and is successfully used for the prediction of droplet impingement and more recently for the prediction of particle flows in air- ways. However, this model includes only one-way momentum exchange coupling, and develops delta shocks and vacuum states. The main goal of this thesis is to improve this model, especially for the prevention of delta shocks and vacuum states, and the adjunction of two-way momentum exchange coupling. Using a characteristic analysis, the condition for loss of regularity of smooth solutions of the inviscid Burgers equation with a source term is established. The same condition applies to the droplet model. The Riemann problems associated, respectively, to the Burgers equation with a source term and the droplet model are solved. The characteristics are curves that tend asymptotically to straight lines. The existence of an entropic solution to the generalized Rankine-Hugoniot conditions is proven. Next, a way for preventing the formation of delta shocks and vacuum states in the model is identified and a new Eulerian droplet model is proposed. A new hierarchy of two-way coupling Eulerian models is derived. Each model is analyzed and numerical comparisons of the models are carried out. Finally, 2D computations of air-particle flows comparing the new Eulerian droplet model with the standard Eulerian droplet model are presented.

La simulation directe des ondes de souffle générées par une explosion maîtrisée, ou accidentelle, est un problème délicat du fait des différentes échelles spatiales en jeu. De plus, en environnement réel (topographie, zone urbaine, …), l’onde de souffle interagit avec les obstacles géométriques en se réfléchissant, se diffractant et se recombinant. La forme du front devient complexe, rendant difficile voire impossible une estimation a priori des effets des explosions.Ce travail de thèse contribue à la mise au point d’une méthode de calcul rapide des ondes de souffle en présence d’obstacles. Il repose sur des modèles hyperboliques simplifiés de propagation d'ondes de choc extraits de la littérature, où seul le front incident est modélisé. Ceci permet une réduction significative du coût des simulations : les 5 équations d'Euler 3D sont réduites à un problème 2D à 2 équations. L’analyse du problème de Riemann met en évidence l’absence de solution de ces modèles lors de la diffraction sur un coin convexe dans certaines configurations fréquemment rencontrées en pratique. L’extension des modèles aux ordres supérieurs ne permet pas de corriger ce défaut. Nous levons cette limitation au travers d'une modification ad hoc. L’effet de souffle consécutif à une explosion est ensuite introduit à partir d’une loi expérimentale pression/distance. Du point de vue numérique, un algorithme Lagrangien conservatif de suivi de front est développé en 2D. Les tests montrent que ce nouveau modèle se compare favorablement à l’expérience, avec une réduction de plusieurs ordres de grandeur du temps de calcul en comparaison des méthodes de résolution directe des équations d’Euler. / The direct numerical simulation of blast waves (accidental or industrial explosions) is a challenging task due to the wide range of spatial and temporal scales involved. Moreover, in a real environment (topography, urban area …), the blast wave interacts with the geometrical obstacles resulting in reflection, diffraction and waves recombination phenomena. The shape of the front becomes complex, which limits the efficiency of simple empirical methods.This thesis aims at contributing to the development of a fast running method for blast waves propagation in presence of obstacles. This is achieved through the use of simplified hyperbolic models for shock waves propagation such as Geometrical Shock Dynamics (GSD) or Kinematic models. These models describe only the leading shock front. This leads to a drastic reduction of the computational cost, from 5 Euler equations at 3D to a 2D problem with 2 equations. However, the study of the Riemann problem shows that the solution of these models does not always exist in the case of the diffraction over a convex corner. We propose an ad-hoc extension of GSD in order to remove this limitation. The blast effects are also recovered through an empirical law available in free field. From a numerical point of view, a 2D conservative Lagrangian algorithm has been implemented and validated. First comparisons with experimental data show the good behaviour of this new model at nearly free computational cost compared to direct Euler methods.

Ce travail s'intéresse aux effets mécaniques créés par les décharges de surface et à leur efficacité comme actionneur. Une géométrie particulière d'électrodes permet de créer de manière pulsée un filament linéaire de plasma et de le chauffer très rapidement par effet Joule (à raison de plusieurs Joules en moins d'une microseconde). Ce chauffage rapide entraîne la formation d'ondes de choc qui peuvent interagir avec l'écoulement ambiant.Nous étudions le comportement électrique de la décharge de surface afin d'évaluer l'énergie déposée dans le filament de plasma par effet Joule. Pour ce faire, nous réalisons une étude paramétrique sur la configuration du circuit et nous déterminons les principaux paramètres qui pilotent la dynamique de la décharge. Différents modèles de résistance sont utilisés dans un code de simulation du circuit électrique, et leurs prédictions du courant et du dépôt d'énergie sont confrontées aux mesures expérimentales.Des mesures spectroscopiques dans différentes configurations de circuit donnent accès à certaines propriétés de la décharge comme la densité électronique, qui atteint des valeurs de 2x1018 cm-3. Le rayon du canal est également mesuré par imagerie rapide. Les ondes de chocs créées par la décharge de surface sont visualisées en imagerie Schlieren pour plusieurs configurations de circuit. Ces ondes de chocs créent une impulsion proportionnelle à l'énergie déposée dans la décharge. Nos développons un modèle de choc pour décrire la trajectoire du choc et pour calculer l'impulsion communiquée par la décharge de surface. Le modèle est en bon accord avec les mesures expérimentales et la décharge de surface a une efficacité mécanique de 0.12mNs/J pour notre configuration d'étude. Nous terminons cette étude en comparant cet actionneur potentiel avec d'autres actionneurs courants et proposons plusieurs pistes pour de futurs travaux. / In this study, we investigate the mechanical effects generated by pulsed surface discharges and their efficiency as an actuator. Using a specific electrode configuration, it is possible to create a short-lived, pulsed, rectilinear plasma channel and to heat it up rapidly (several Joules in less than a microsecond) through Joule heating. This fast energy deposition causes the formation of shock waves that can then interact with the surrounding flow.We study the electrical behavior of the pulsed surface discharge to assess the energy deposited in the plasma channel through Joule heating. To do so, we perform a parametric study on the circuit configuration and identify the main parameters driving the discharge dynamics. Several resistance models are implemented in a numerical description of the electrical circuit and their predictions of the current and deposited energy are compared with experimental measurements.Spectroscopic measurements in different circuit configurations give access to some of the plasma properties such as the electron number density that can reach values up to 2x1018 cm-3. Fast imaging also gives insight into the plasma channel radius. The shock waves generated by the pulsed surface discharge in different circuit configurations are visualized through Schlieren imaging. These shock waves generate an impulse that increases linearly with the energy deposited in the discharge. We develop a shock model to describe the shock trajectory and to compute the impulse imparted by the pulsed surface discharge. The model is in good agreement with our measurements and the pulsed surface discharge is found to have a mechanical efficiency of 0.12 mNs/J for our setup configuration. We conclude this study by comparing the proposed pulsed surface discharge actuator with other common designs and offer some directions for future studies.

Made available in DSpace on 2014-07-29T16:02:21Z (GMT). No. of bitstreams: 1 dissertacao gisele.pdf: 849978 bytes, checksum: 6707e4fe01d67af874b9cc826b72dfab (MD5) Previous issue date: 2007-11-29 / Transitional shock waves arises in solution of initial values problems for non linear systems of conservation laws that are not strictly hyperbolic. These waves are discontinuous solutions that posses viscous profile but do not conform to the Lax characteristic criterion, where inequalities between the shock propagation speed and the characteristic speeds must to be satisfied. These waves arise as transition between wave groups associated with distinct characteristic families. In this work we studied transitional shock waves for a system of two conservation laws with quadratic fux functions and positive defined viscosity matrix. In particular, we studied the transitional shock waves with viscous profile defined by orbits laying on straightlines. We show from examples, for systems with quadratic fux functions and viscosity matrix chosen in a convenience way, that is necessary to use transitional shock waves to solve the Riemann problem (initial data constant by parts) for these systems. / Ondas de choque transicionais aparecem nas soluções de problemas de valores iniciais para sistemas não lineares de leis de conservação não estritamente hiperbólicos . São soluções descontínuas que possuem perfil viscoso mas não satisfazem o critério de entropia de Lax, onde certas desigualdades entre a velocidade de propagação do choque e as velocidades características são satisfeitas. Estas ondas aparecem como transição entre grupos de ondas associados com diferentes famílias características. Neste trabalho estudamos as ondas de choque transicionais para um sistema de duas leis de conservação com função de fluxo quadratica e matriz de viscosidade definida positiva. Em particular estudamos os choques transicionais com perfil viscoso definidos por orbitas sobre um segmento de reta. Mostramos através de exemplos, para sistemas com funções de fluxo quadráticas e matrizes de viscosidade escolhidas de modo conveniente, que e necessário usar as ondas de choques transicionais para resolver o problema de Riemann (dados iniciais constantes por partes) para estes sistemas.

Diferentes tipos de sólitons têm sido observados em meios ópticos não-lineares, e seus comportamentos individuais descritos pela equação não-linear de Schrödinger e pela equação não-linear de Schrödinger generalizada, em diferentes dimensões e geometrias. Entretando, há situações onde muitos sólitons são gerados formando uma densa rede de sólitons. Nestes casos, é impossível desprezar as interações entre os sólitons e temos que considerar a evolução da estrutura como um todo. A teoria das ondas de choque dispersivas em meios fotorrefrativos e a teoria da difração não-linear de intensos feixes de luz propagando-se em meios fotorrefrativos com um fio refletor incorporado a esse meio foi desenvolvida, e verificamos que está em excelente acordo com nossas simulações numéricas. No caso da formação de sólitons em condensados, fizemos cálculos numéricos realísticos dentro da aproximação de campo médio usando a equação de Gross-Pitaevskii, incluindo também um potencial de confinamento, um potencial móvel e um potencial dipolar. A maioria dos resultados puderam ser comparados com experimentos recentes. / Different kinds of solitons have already been observed in various nonlinear optical media, and their behavior has been explained in the frameworks of such mathematical models as the nonlinear Schrödinger and generalized nonlinear Shrödinger equations for different dimensions and geometries. However, there are situations when many solitons are generated so that they can comprise a dense soliton train. In such situations, it is impossible to neglect interactions between solitons and one has to consider the evolution of the structure as a whole rather than to trace the evolution of each soliton separately. The theory of optical shock waves in photorefractive media and the theory of nonlinear diffraction of light beams propagating in photorefractive media with embedded reflecting wire was developed and agrees very well with our numerical simulations. In the condensate soliton formation case, we did numerical calculations in the mean field approach using the Gross-Pitaevskii equation, adding a trap potential and a moving potential and a potential of the dipole-dipole interaction. The main results were also checked by recent experiments.

En phase d’approche, le bruit rayonné par l’entrée d’air des turboréacteurs est principalement dû aux interactions entre le rotor et le stator. Cependant les ondes de choc (ou ondes en N) générées par le rotor en régime transsonique peuvent devenir une source de bruit dominante durant le décollage et la montée de l’appareil. L’étude des ondes en N nécessite de se concentrer sur deux processus majeurs : 1) la génération des chocs par un rotor parfait (dont toutes les aubes sont identiques) et par un rotor réel (en tenant compte des irrégularités géométriques des aubes), et 2) la propagation de ces ondes en N à travers la nacelle, produisant du bruit dont le spectre se compose des harmoniques de la fréquence de passage des aubes pour un rotor régulier, et des harmoniques aux fréquences multiples de la rotation du rotor (FMR) pour un rotor irrégulier. Plusieurs approches analytiques et numériques ont été développées durant les 40 dernières années.Cette thèse relate dans un tout premier temps les principales théories de la propagation des ondes de choc ainsi que les modèles majeurs de génération de FMR. Une attention particulière est portée sur les liens entre les équations générales de la mécanique des fluides et ces modèles de propagation non linéaire afin de mettre en évidence les différentes hypothèses formulées dans ces modèles. Dans un deuxième temps, les principales méthodes semi-analytiques de génération et de propagation des chocs seront évaluées et comparées en les appliquant à des configurations de turboréacteurs. En outre, un nouveau modèle de génération de FMR basé sur des considérations géométriques est élaboré par l’intermédiaire d’une campagne d’essais comportant d’une part des mesures de signaux de pression dans la nacelle et d’autre part les mesures des angles de calage des aubes pendant le fonctionnement du moteur. Le deuxième volet de la thèse concerne le développement d’une méthodologie de simulation numérique basée sur l’utilisation du code elsA de l’ONERA en résolvant les équations d’Euler (approche CAA). L’objectif de cette approche est de s’affranchir des limitations des modèles de propagation semi-analytiques et de tenir compte de la géométrie réelle de la nacelle ainsi que d’un écoulement réaliste. Des ondes de choc régulières et irrégulières sont directement injectées dans un plan proche de la soufflante et se propagent en remontant l’écoulement. Ces ondes de choc sont injectées par l’intermédiaire d’une condition limite de non-réflexion qui nécessite d’imposer le champ conservatif. La signature des chocs peut provenir d’un RANS, de mesures ou d’un signal analytique. Étant donné que les mesures ou le signal théorique ne permettent d’obtenir que la pression, une méthode de reconstruction du champ conservatif à partir des variations de pression induites par le choc a été élaborée. Cette méthode d’injection est tout d’abord appliquée à un conduit annulaire infiniment mince et validée par la méthode de propagation semi-analytique de McAlpine & Fisher. Ensuite, les effets de propagation 3D sont étudiés en augmentant l’épaisseur du conduit. Enfin, la méthode CAA est appliquée à des configurations de turboréacteurs modernes et des ondes de choc régulières et irrégulières sont propagées numériquement. Les résultats sont comparés aux solutions RANS ainsi qu’aux mesures disponibles. / Whereas the sound radiated from the inlet of turbofans is mainly due to rotor–stator interactions in approach flight, the shock waves (or N-waves) emitted by the rotor at transonic rotation speeds can be a dominant noise source during takeoff and climb. The study of N-waves needs to take account of two main processes: 1) the generation of N-waves for a perfect rotor (in which all blades are identical) and for a real rotor (considering small geometrical blade dispersion), and 2) the N-wave propagation through the inlet duct producing the blade passing harmonics for a perfect rotor, and the multiple pure tones (harmonics of the rotation frequency) for a real rotor. Several analytical and numerical approaches have been investigated for the past 40 years.This thesis first intends to relate the main propagation theories and to address the foremost MPT generation method hypotheses. The links between fluid dynamics equations and practical non-linear theories currently adopted are emphasized and discussed. In a second step, the main relevant semi-analytical methods are cross-checked by applying them to representative turbofan configurations. Moreover, a novel model of irregular N-wave generation based on geometrical considerations is investigated thanks to test data related to in-duct pressure signatures and blade stagger angle measurements during the engine operation. Then, a second part of the work investigates a numerical strategy based on elsA ONERA code, solving the full Euler’s equations (CAA approach). The objective is to prevent from the limitations of 2D analytical models and to take into account actual inlet geometry and realistic convection flow. Regular and non-regular shock waves are directly injected in a plane close to the fan and propagated through the inlet. These shock waves are injected through a non-reflective boundary condition which requires the conservative field. The initial shock description near the fan is provided either by a RANS computation or by experiment, or else from analytical model. As experiment or analytical signals only provide pressure signatures, a theory is set up to re-built the whole conservative field from the basis of a pressure shockwave. This injection method is firstly applied on an infinitely narrow annular duct and validated through the comparison with the McAlpine & Fisher analytical method. Then, the 3D propagation effects are pointed out by increasing the duct height. Finally, the CAA method is applied on actual intake geometry of modern turbofan demonstrators, and propagation of regular and irregular shock-waves are simulated. The numerical results are compared to RANS solutions and to available measurements.

Les mousses métalliques ont connu un essor important durant les dernières décennies. Leur capacité à supporter de très larges niveaux de déformation tout en transmettant de faibles contraintes les rend particulièrement adaptés à des solutions d'absorption d'énergie ou de protection contre des sollicitations intenses.Le comportement dynamique de ce type de matériau peut être influencé par les effets inertiels au niveau des parois ou des ligaments constituant son squelette (micro-inertie). Un modèle de comportement à base micromécanique a été développé pour prendre en compte les effets micro-inertiels sur le comportement macroscopique de mousses à porosités fermées. Le modèle proposé repose sur la procédure d'homogénéisation dynamique introduite par Molinari et Mercier (2001). Par cette approche, les effets micro-inertiels apparaissent sous la forme d'un terme supplémentaire dans le tenseur des contraintes, appelé contrainte dynamique. À partir de comparaisons avec des données extraites de la littérature, il est ainsi démontré qu'inclure les effets micro-inertiels permet d'obtenir une meilleure description de la réponse des mousses sous choc.L'influence d'une épaisseur de mousse localisée entre un explosif et une enveloppe cylindrique a ensuite été étudiée en suivant deux approches. La première, qui s'appuie sur les travaux de Gurney (1943), repose sur des considérations énergétiques. La seconde méthode permet d'aboutir à une description plus détaillée des tailles et vitesses de fragments. Elle repose sur la combinaison d'un modèle éléments finis pour décrire la propagation de l'onde de choc dans la mousse et l'expansion de l'enveloppe et d'un modèle de fragmentation de type Mott (1947). / Metallic foams have known a growing interest in the last decades. Their ability to undergo very large strains while transmitting only reasonable stress levels makes them particularly suitable for energy absorption applications and protection against intense solicitations. The dynamic behavior of metal foams is linked to inertial effects appearing at the walls and ligaments of the material microstructure (micro-inertia). A constitutive model has been developed to take micro-inertial effects into account when describing the macroscopic behavior of closed-cell foams submitted to dynamic loadings. The proposed approach was developed using the dynamic homogenization procedure introduced by Molinari and Mercier (2001). Within this framework, micro-inertial effects appear as an additional stress term, called dynamic stress. Comparisons with data from literature have showed that including micro-inertia effects allows one to achieve a better description of the foam response under shock loading.The influence of a foam layer placed between an explosive and a cylindrical casing has been investigated by following two approaches. The first one is based on energetic considerations, following the work of Gurney (1943). The second method allows one to obtain a more detailed description of fragment sizes and velocities. It relies on the combined use of a finite element model and a description of the shell fragmentation based on the work of Mott (1947).

Normal shock problems in high enthalpy flows are of special interests to aerodynamicists and fluid dynamicists. When the shock Mach number become hypersonic and increasing further, the gas passing through the shock is compressed resulting in increase in temperature and pressure. As the Mach number increases the internal degrees of freedom of the diatomic molecules are activated to an increasing extent when it crosses the shock resulting dissociation especially for high enthalpy flows. Hence dissociation of diatomic molecules must be taken into account in the determination of some of the aerodynamic parameters. This thermal and chemical process can be divided into three types such as nearly frozen, non-equilibrium and nearly non-equilibrium depending on the rates of reaction and excitation. For typical re-entry conditions of spacecrafts into a planets atmosphere, dissociation reactions of the molecules is dominant in the stagnation flow. Further in the stagnation region of the flow field one of the most important parameter that characterizes the flow field is the shock stand-off distance. This parameter is often employed for validation purposes of numerical methods as well as for non-reactive and reactive gases. For high Mach number flows the shock is very close to the body hence experimental determination of shock stand-off distance is very difficult and there would be relatively large errors. Therefore the theoretical determination of this parameter is of great significance in the discussion of this physical phenomenon. There are some works which presents how the dissociation behind shock affects the shock stand-off distance. Thus the dissociation behind the shock is a very important process which has great impact in aerodynamic flight and design. In this present work we studied how dissociation of diatoms occur behind a normal shock. Treanor and Marrone (1962) proposed CVD(coupled vibration-dissociation) model for diatoms by assuming diatom as a harmonic oscillator with a cut-off level. But actually diatoms are not harmonic oscillator, because spectroscopic data of energy level spacing is not like harmonic oscillator. For this reason, Treanor, Rich, and Rehm(1968) used anharmonic oscillator model for diatoms to study vibrational relaxation. Taking the anharmonicity of diatom, Philip Morse(1929) gave a formula for potential energy levels for diatoms, which is known to express the experimental values quite accurately. Unlike the energy levels of the harmonic oscillator potential, which are evenly spaced , the Morse potential level spacing decreases as the energy approaches the dissociation energy and then it is continuous. So it is quite accurate to take Morse oscillator theory for diatomic dissociation instead of harmonic oscillator with a cut-off level. We have used Morse oscillator theory to derive a dissociation-recombination reaction rate equation for diatom. To derive the rate equation we have used the transition probability between different vibrational energy levels . The rate equation is numerically solved to get the different flow variables behind the shock. The result of the present work has been compared with some of the previous work. Some of the flow variables are well matching with the previous work and some has discrepancy near the shock but well matching after few distance from the shock. We have also studied under what conditions the post shock flow shows self-similar behavior in its scaling relations. It is shown that as far as there is no dissociation, we could expect to obtain self-similar solutions. However, when there is dissociation, the non-equillibrium nature of the phenomenon disrupts the self-similar nature of the flow.

The motion of shock waves is important in many fields of engineering and increasingly so with medical applications and applications to inertial confinement fusion technologies. The flow structures that moving shock waves create when they encounter a change in area is complex and can be difficult to understand. Previousresearchers have carried out experimental studies and many numerical studies looking at this problem in more detail. There has been a discrepancy between numerical and experimental work which had remained unanswered. One of the aims of this project is to try and resolve the discrepancy between numerical and experimental work and try to investigate what experimental techniques are suitable for work of this type and the exact way in which they should be applied. Most previous work has focused on sharp changes in geometry which induce immediate flow separation. In this project rounded corners will also be investigated and the complex flow features will be analyzed.Two geometries, namely a sharp 172 degree knife-edge and a 2.8 mm radius rounded corner will be investigated at three experimental pressure ratios of 4, 8 and 12 using air as the driver gas. This yields experimental shock Mach numbers of 1.28, 1.46 and 1.55. High-speed schlieren and shadowgraph photography with varying levels of sensitivity were used to qualitatively investigate the wave structures. Particle image velocimetry (PIV), pressure-sensitive paint (PSP) and traditional pressure transducers were used to quantify the flow field. Numerical simulations were performed using the commercial package Fluent to investigate the effect of numerical schemes on the flow field produced and for comparison with the experimental results. The sharp geometry was simulated successfully using an inviscid simulation while the rounded geometry required the addition of laminar viscosity. Reynolds number effects will be only sparsely referred to in this project as the flows under investigation show largely inviscid characteristics. As the flow is developing in time rather than in space, quotation of a distance-based Reynolds number is not entirely appropriate; however, Reynolds number based on the same spatial location but varying in time will be mentioned. The density-based diagnostics in this project were designed to have a depth of field appropriate to the test under consideration. This approach has been used relatively few times despite its easy setup and significant impact on the results. This project contains the first quantative use of PIV and PSP to shock wave diffraction. Previous studies have almost exclusively used density-based diagnostics which, although give the best impression of the flow field, do not allow for complete analysis and explanation of all of the flow features present. PIV measurements showed a maximum uncertainty of 5% while the PSP measurements showed an uncertainty of approximately 10%.The shock wave diffraction process, vortex formation, shear layer structure, secondary and even tertiary expansions and the shock vortex interaction were investigate. The experimental results have shown that using one experimental technique in isolation can give misleading results. Only by using a combination of experimental techniques can we achieve a complete understanding of the flow field and draw conclusions on the validity of the numerical results. Expanding the range of the experimental techniques currently in use is vital for experimental aerodynamic testing to remain relevant in an industry increasingly dominated by numerical research. To this end, significant research work has been carried out on extending the range of the PSP technique to allow for the capture of shock wave diffraction, one of the fastest transient fluid processes, and for applications to low-speed flow (< 20 ms−1).