betaFeedback

WorksPhysics and Astronomy

Nearly homogeneous and isotropic turbulence generated by the interaction of supersonic jets

Abstract

This study reports the development and characterization of a multiple-supersonic-jet wind tunnel designed to investigate the decay of nearly homogeneous and isotropic turbulence in a compressible regime. The interaction of 36 supersonic jets generates turbulence that decays in the streamwise direction. The velocity field is measured with particle image velocimetry by seeding tracer particles with ethanol condensation. Various velocity statistics are evaluated to diagnose decaying turbulence generated by the supersonic jet interaction. The flow is initially inhomogeneous and anisotropic and possesses intermittent large-scale velocity fluctuations. The flow evolves into a statistically homogeneous and isotropic state as the mean velocity profile becomes uniform. In the nearly homogeneous and isotropic region, the ratio of root-mean-squared velocity fluctuations in the streamwise and vertical directions is about 1.08, the longitudinal integral scales are also similar in these directions, and the large-scale intermittency becomes insignificant. The turbulent kinetic energy per unit mass decays according to a power law with an exponent of about 2, larger than those reported for incompressible grid turbulence. The energy spectra in the inertial subrange agree well with other turbulent flows when normalized by the dissipation rate and kinematic viscosity. The non-dimensional dissipation rate is within a range of 0.51–0.87, which is also consistent with incompressible grid turbulence. These results demonstrate that the multiple-supersonic-jet wind tunnel is helpful in the investigation of decaying homogeneous isotropic turbulence whose generation process is strongly influenced by fluid compressibility.

Introduction

Compressible turbulence plays a crucial role in flows in engineering and physics fields [3, 14]. For sufficiently large velocity fluctuations, turbulent motion begins to cause fluid compression and expansion, by which fluid density varies. The compressibility effects significantly alter the evolution of turbulence. The growth rate of turbulent shear flows is suppressed at a high Mach number [12]. When a Mach number defined for velocity fluctuations is very high, turbulent motion also locally generates shock waves called shocklets [41]. Velocity, pressure, density, and temperature drastically vary across shock waves, and the propagation of shocklets also influences turbulence.

Theoretical studies of turbulence often consider canonical turbulent flows under relatively simple conditions. These studies also mainly concern statistical behaviors of turbulence because of the difficulty in constructing theories and models for an instantaneous flow field based on the Navier–Stokes equations. One of the simplest problems is homogeneous isotropic turbulence (HIT) [5], for which various theories and models have been developed in previous studies, e.g., non-dimensional energy dissipation rate [42], longitudinal structure functions [10], and decay in absence of external force [21]. Laboratory experiments are as crucial as numerical simulations in providing data that support the construction and validation of theories and models. Previous studies have developed various facilities and methods for generating HIT in a subsonic regime for which the assumption of incompressibility is valid. A grid installed in a uniform mean flow in a wind tunnel can generate nearly homogeneous and isotropic turbulence (e.g., Comte-Bellot and Corrsin [17]; Uberoi and Wallis [69]; Valente and Vassilicos [70]; Kitamura et al. [36]; Djenidi et al. [22]). Similarly, experiments with water flumes utilize a grid to generate HIT (Tan and Ling, [65]; Suzuki et al., [63]). Most grids consist of many intersecting bars, which stir a uniform mean flow. Another type of turbulence-generating grid is an active grid (Makita, [43]), which utilizes rotating bars with winglets. The random motion of winglets generates HIT with a high Reynolds number (Mydlarski and Warhaft, [48]; Larssen and Devenport, [40]; Zheng et al., [83]). A similar method to randomly stir fluid is utilized in a multi-fan wind tunnel, by which the interaction of fan-induced flows generates HIT (Ozono and Ikeda, [51]; Takamure and Ozono, [64]). A mean flow advects the HIT generated in wind tunnels and water flumes. Because of negligible shear in the mean flow, turbulence decays as it evolves. The decay of grid turbulence is often compared with the theories of freely evolving HIT (Krogstad and Davidson, 2010).

Previous studies have also developed facilities to generate turbulence without a mean flow. They often utilize stirring devices in a closed chamber. Birouk et al. [8] reported the generation of HIT with electrical fans placed inside a chamber. Other studies also developed similar multi-fan facilities (Semenov, [61]; Zimmermann et al., [85]; Ravi et al., [57]; Xu et al., [79]; Bradley et al., [11]; Yamamoto et al., [80]). Synthetic jet actuators are also widely used to generate HIT by the jet interaction (Hwang and Eaton, [30]; Variano et al., [71]). Similarly, the interaction of random jets can generate HIT: Bellani and Variano [6] and Carter et al. [15] used two opposing planar arrays of randomly actuated jets while Pérez-Alvarado et al. [52] used a single array. These facilities were used to investigate various phenomena related to turbulent flows, such as particle dispersion (Hwang and Eaton, [31]) and evaporation of droplets (Marié et al., [44]).

Fewer experimental studies have been reported for HIT in compressible flows than in incompressible flows. Experiments of compressible turbulence have been extensively conducted for turbulent shear flows, such as jets Karthick et al. [34], mixing layers Goebel et al. [27], and wakes Bonnet and Chaput [9], which have significant application in aerospace engineering. Compressible grid turbulence has been investigated in shock tube facilities. Grid turbulence is generated when the mean flow induced by a planar shock wave passes a grid installed inside a shock tube. Most studies of grid turbulence in shock tubes focus on the interaction between the shock wave and grid turbulence [29, 1, 25]. Exceptionally, Briassulis et al. [13] conducted comprehensive velocity measurements of grid turbulence in a shock tube. They compared the statistics of velocity fluctuations for various flow parameters, such as the Mach number of the mean flow and the Reynolds number based on the mesh size. The turbulent kinetic energy in grid turbulence is known to decay according to a power law given by ak(x−x0)−na_k(x-x_0)^{-n}, with a coefficient aka_k, a virtual origin x0x_0, and a decay exponent nn. The experiments by Briassulis et al. [13] suggest that compressibility effects cause the decrease of nn. Later, Fukushima et al. [25] also observed the decrease of nn in compressible grid turbulence in a shock tube. Zwart et al. [86] reported compressible grid turbulence generated with a perforated plate installed in a high-speed wind tunnel. They found that nn in compressible grid turbulence exceeds values in incompressible cases when the Mach number is high. These results imply that the compressibility effects on grid turbulence are not universal and may depend on various factors, including the generation process of turbulence, as also discussed in Zwart et al. [86]. Indeed, even in an incompressible fluid, different decay laws of HIT were theoretically derived for the decay of Saffman turbulence and Batchelor turbulence, which differ in the shape of an energy spectrum at a low wavenumber range Davidson [20]. The low-wavenumber spectral shape depends on the turbulence generation process. Therefore, different compressibility effects are possible for the decay of HIT generated in different facilities.

Recent studies have developed other facilities for compressible turbulence. Yamamoto et al [81] developed a compressible-turbulence chamber with opposing arrays of piston-driven synthetic jet actuators. The piston-driven actuators, initially proposed by Crittenden and Glezer [19], are capable of generating supersonic synthetic jets at a sufficiently high operation frequency ([68, 60, 53]). The interaction of supersonic synthetic jets generates turbulence strongly influenced by compressibility effects. Another recently developed facility for compressible turbulence is the variable density and speed-of-sound vessel. The experiments use a heavy gas (sulfur hexafluoride SF6SF_6) as a working fluid. Even though a stirring device generates a flow with a velocity lower than 100 m/s, the Mach number can be high because of the low speed of sound of SF6SF_6. These new facilities will be helpful in the future investigation of compressible turbulence.

As also found for the decay exponent of grid turbulence discussed above, flow-dependent compressibility effects on turbulence have been reported for other velocity statistics. In fully-developed incompressible turbulence, the skewness and flatness of a longitudinal velocity gradient, e.g., ∂u/∂x\partial u/\partial x, have universal relations with a turbulent Reynolds number ([62]). Compressible turbulence tends to have larger absolute values of skewness and flatness than incompressible turbulence with a comparable Reynolds number ([23, 77]). One of the critical parameters in compressible turbulence is the turbulent Mach number, defined as MT=urms2+vrms2+wrms2/aM_T = \sqrt{u_{\rm rms}^2+v_{\rm rms}^2+w_{\rm rms}^2}/a, where urmsu_{\rm rms}, vrmsv_{\rm rms}, and wrmsw_{\rm rms} are the root-mean-squared (rms) velocity fluctuations in three directions, and aa is the speed of sound. MTM_T is often treated as a dominant parameter determining compressibility effects ([41, 58]). However, a comparison of different compressible turbulent flows in experiments and direct numerical simulations (DNS) has shown that the compressibility effects on the velocity gradient statistics are not solely determined by local values of MTM_T [80]: the deviation of the velocity derivative flatness from incompressible values occurs in supersonic turbulent jets and HIT generated by supersonic synthetic jets at much lower MTM_T than in compressible HIT sustained by a solenoidal forcing. The dilatational fluid motion causes additional kinetic energy dissipation in strongly compressible turbulence. The dilational and solenoidal dissipation ratio is not also determined solely by MTM_T [23]. These behaviors are possibly explained by a non-local property of wave motion [80]: even if a particular location in turbulence has low MTM_T, much higher values of MTM_T can be observed in other regions, where strong pressure waves, such as shock waves, are generated and propagate into a flow region of interest. These studies imply that the compressibility effects on turbulence significantly depend on the generation process of turbulence. Investigating compressible turbulence in various facilities is crucial in extending our knowledge of turbulence behavior under compressibility effects. The further development of compressible turbulence facilities is demanded for this purpose.

The present study develops a new wind tunnel facility to investigate compressible decaying HIT. In this wind tunnel, the interaction of many supersonic jets generates nearly homogeneous and isotropic turbulence, which continuously decays in a test section without significant influences of mean shear. Velocity measurements with particle image velocimetry (PIV) and temperature and pressure measurements characterize the turbulent flow generated in the multiple-supersonic-jet wind tunnel, which is useful for investigating decaying HIT in a compressible regime. The paper is organized as follows. § 2 describes the multiple-supersonic-jet wind tunnel and measurement methods. § 3 presents the measurement results to discuss the fundamental turbulence statistics compared with other turbulent flows. Finally, the paper is summarized in Sect. 4.

Experimental setup and measurement methods

Multiple-supersonic-jet wind tunnel

The multiple-supersonic-jet wind tunnel generates nearly homogeneous and isotropic turbulence by the interaction of supersonic round jets. The development of the facility is motivated by the experiments of incompressible HIT generated by the interaction of many subsonic jets [6, 15, 52]. Figure 1 shows a picture of the wind tunnel with three main components: an air supply system, a plenum chamber, and a test section. Compressed air is supplied to the cylinder-shaped plenum chamber. The front surface of the plenum chamber is equipped with 36 Laval nozzles, which generate supersonic round jets. The interaction of 36 jets generates turbulence which eventually decays in the streamwise direction in the test section. This section describes the details of the wind tunnel.

A multiple-supersonic-jet wind tunnel

Figure 1. A multiple-supersonic-jet wind tunnel.

Compressed air is stored in two air tanks with a volume of 220 L (Anest Iwata, SAT-220C-140). Compressors (Fujiwara Industrial Co., Ltd., SW-L30LPF-01) are connected to each tank via a moisture separator (SMC, AFM40-02-2A). The outlet port of each tank is connected to the back surface of the cylinder-shaped plenum chamber by a flexible hose with an inner diameter of 29.5 mm. Here, the air is supplied to the chamber through a moisture separator (SMC, AF60-10-A), a pressure regulator (SMC, AR60-10BG-B), and a pilot kick 2-port solenoid valve (CKD, ADK11-25A). The pressure regulator is used to adjust the plenum pressure.

Figure 2 shows a schematic of the plenum chamber and test section. The plenum chamber consists of a 180 mm length pipe and two round plates with a 330 mm diameter fixed on flanges welded to the pipe. The pipe size is JIS 20A, with an inner diameter of 206.3 mm and a thickness of 5 mm. The pipe, plates, and flanges are made of stainless steel. Rubber seals designed for JIS 20A pipes fill the space between the plates and flanges. The flexible hoses are connected to two threaded holes on the back plate with a thickness of 6 mm. A pressure sensor (SMC, ISE30A-C6L-N-M) connected to a port on the front plate monitors the pressure in the plenum chamber. The plenum pressure is also measured with another pressure sensor (SMC, PSE-540A R06), whose signal is recorded by an oscilloscope (Yokokawa, DL850E) at a sampling rate of 1 kHz.

Schematic of a multiple-supersonic-jet wind tunnel

Figure 2. A schematic of a multiple-supersonic-jet wind tunnel. All dimensions are in mm.

A nozzle plate with 6×66 \times6 Laval nozzles is installed on the front plate. Figure 3 shows a picture of the nozzle plate manufactured with stereolithography 3D printing with a printing resolution of 10 μm10\ \mu\mathrm{m}. All 36 nozzles have the same geometry. The holes along the perimeter of the nozzle plate are used to fix the plate to the plenum chamber. The inner diameters at the jet exit and the throat are 4.31 mm and 4.12 mm, respectively. The nozzle inlet diameter is 6.3 mm. The constant outer diameter is 7.5 mm. The center-to-center distance between the two nozzles is 12 mm. The inner diameter of the converging section varies according to a square of a sinusoidal function. The internal geometry of the diverging section is determined with an analytical method for nozzle design [24]. Each nozzle with these dimensions generates a fully expanded supersonic jet with a Mach number of 1.36 in atmospheric air when the gauge pressure in the chamber is 200 kPaG. The base of the nozzle plate has a square shape of 110×110 mm2110 \times110\ \mathrm{mm}^{2} and a thickness of 5 mm. The front plate of the chamber has holes with a diameter of 9 mm at the locations corresponding to the nozzles. The nozzles are inserted and fixed from the inner side of the chamber plate with bolts and nuts. Air leakage is prevented with seal washers used with the nuts and a hand-cut PTFE gasket sheet with a 1 mm thickness, which fills the space between the nozzle plate and the chamber plate.

Nozzle plate with 36 Laval nozzles

Figure 3. A nozzle plate with 36 Laval nozzles.

The test section has 100×100 mm2100 \times100\ \mathrm{mm}^{2} square cross-section and 1000 mm length. The test section is made of an aluminum frame and acrylic plates. For the side and bottom walls, optical-grade acrylic sheets (Nitto Jushi Kogyo, CLAREX) with a 1.5 mm thickness are used for optical measurements and visualizations. The top wall is a conventional acrylic plate with a 3 mm thickness. The plenum chamber is placed on a urethane pipe-holder designed for JIS 20A pipes (Nichiei Intec, N-040219). The test section and pipe holder are fixed on another aluminum frame.

All experiments are conducted to generate the supersonic jets with a plenum pressure of 200 kPaG. In each run of the experiments, the air tanks are filled with compressed air with a gauge pressure of 900 kPaG. Once the valves are opened, the compressed air flows into the plenum chamber, increasing the plenum pressure. The pressure regulators between the air tanks and the plenum chamber are adjusted to realize the desired value of the plenum pressure. Figure 4 presents the temporal variation of the gauge pressure in the plenum chamber, PCP_C, during the experiment. In this plot, t=0t = 0 corresponds to the maximum value of PCP_C. After the valve is opened, the pressure immediately increases to 200 kPaG. The pressure hardly varies with time for about 3 s, and then begins to decrease because the pressure in the air tanks decreases.

Figure 4. Temporal variation of gauge pressure in the plenum chamber, PCP_C. t=0t = 0 corresponds to the maximum value of PCP_C.

The plenum pressure decreases from 200 kPaG by 5% at t=2.9t = 2.9 s. The flow in the test section is statistically steady for about 2 s, for which the measurements are conducted in this study. The streamwise, vertical, and spanwise directions are denoted by xx, yy, and zz, respectively, and the origin of the coordinate system is at the center of the jet exits of the nozzle plate, as shown in Figure 2.

Particle image velocimetry

Velocity measurements are conducted with two-dimensional and two-component PIV. The present study uses the DANTEC PIV system, which consists of a double-pulse Nd:YAG laser (Dantec Dynamics, Dual Power 65-15) and a high-speed camera (Dantec Dynamics, SpeedSense 9070) with a camera lens with a focal length of 105 mm (Nikkon, AI AF Micro Nikkor 105mm F2.8D). The system is controlled with a synchronizer (Dantec Dynamics, 80N77) and PIV software (Dantec Dynamics, Dynamic Studio). Light sheet optics mounted on the laser produce a thin light sheet with a thickness smaller than 1 mm. The laser unit is placed under the wind tunnel, and a mirror in front of the laser sheet optics is used to illuminate an xx–yy plane from the bottom of the test section. The camera captures particle images from the side of the test section. The time interval of two laser pulses is adjusted depending on the measurement location, as discussed below. Each laser pulse has a width of less than 4 ns. The particle-pair images are captured at 15 Hz. The images are processed with an adaptive PIV algorithm (Theunissen, [66]) and universal outlier detection (Westerweel and Scarano, [78]) implemented in Dynamic Studio.

The present PIV uses condensed ethanol droplets generated by the Laval nozzles as tracer particles. Liquid ethanol is seeded and evaporated in the air tanks before the experiments. When fluid with evaporated ethanol passes the Laval nozzles, fluid expansion in the diverging sections causes a temperature drop, resulting in the condensation of ethanol. This seeding technique has widely been used in supersonic wind tunnels, which also use divergent nozzles to produce supersonic flows. Previous studies have confirmed that generated particles have diameters less than 1 μm1\,\mu\mathrm{m}: Pizzaia and Rossmann (2018) reported that the diameter of ethanol droplets was approximately 0.050.05–0.2 μm0.2\,\mu\mathrm{m}; Kouchi et al [37] used acetone droplets produced by a Lavar nozzle for PIV measurement, by which the droplet diameter was estimated as 160 nm. These studies have shown that the condensation droplets are small enough for PIV measurements of supersonic flows. The diameter can be estimated from light scattered by droplets. By visibly observing particles illuminated by white LED (Sigmakoki, SLA100A), the droplet diameter is confirmed to be in the Rayleigh-scattering regime and smaller than visible light wavelength, 360–830 nm. The present study investigates turbulence generated by the interaction of supersonic jets. As the jets mix with ambient fluid, the mean velocity in the measurement area is in a subsonic regime. Kouchi et al [37] conducted accurate PIV measurements of supersonic flows with condensation particles. Therefore, for the present subsonic velocity range, the ethanol droplets are useful as tracer particles of PIV.

Because how the ethanol is injected in the air tanks affects the number density of tracer particles in the test section, we describe the detailed procedure of the experiments. The air tanks have three ports: an inlet port connected to the compressor, an outlet port connected to the plenum chamber, and an exhaust port used to release compressed air after experiments. The exhaust port can be opened or closed by a hand-operated valve, while the inlet port can be opened by disconnecting a gas hose from the compressor. The following procedure is repeated for the two air tanks. First, ethanol is sprayed into the air tank via the inlet port by using an airbrush. Sprayed ethanol quickly evaporates in the tank. When ethanol is injected, the exhaust valve is kept partially opened to prevent the reversal flow from the inlet port. 70 ml of liquid ethanol is injected into each tank. The exact amount of ethanol in the tank is difficult to estimate because the air partially leaks from the tank while spraying the ethanol.

However, we have confirmed that increasing the amount from 70 ml does not influence particle images captured by the PIV system, whereas smaller amounts result in fewer particles. After the compressor is connected to the inlet port of the tank again and the exhaust valve is closed, the air is stored in the tank until the gauge pressure reaches 900 kPaG. Even for a supersaturated case, the mass fraction of ethanol at this stage is smaller than 1%, for which the ethanol/air mixture and condensation are expected to have negligible influences on the turbulent flow in the test section [16] (Pizzaia and Rossmann 2018). We have also confirmed that ethanol does not change the average and standard deviation of dynamic pressure measured with a Pitot tube. Upon opening the solenoidal valves between the tanks and the plenum chamber, the camera starts capturing particle images. Velocity vectors are calculated by analyzing the images captured while the plenum pressure is constant.

The PIV measurements are conducted at six streamwise locations. Table 1 summarizes the measurement locations and parameters. The center of the measurement area is (x,y)=(xC,yC)(x, y) = (x_C, y_C). These locations are chosen to investigate turbulence generated by the jet interaction and do not include the region where each jet evolves before the interaction. This is because our primary interest is in a nearly homogeneous and isotropic region where turbulence decays. Although the supersonic jets are issued in the test section, the mean velocity in the measurement locations is subsonic because the high-speed jets mix with surrounding fluids with a small mean velocity. The table also shows the time interval Δt\Delta t between two pulses of the PIV. At xC=0.150x_C = 0.150 m, the flow is highly inhomogeneous in the cross-section, and the mean velocity along the centerline is higher than the surroundings. Therefore, Δt\Delta t is smaller at xC=0.150x_C = 0.150 m than in the downstream region. These Δt\Delta t values are determined based on the particle displacement during Δt\Delta t. The spatial resolution, defined as the spacing between points to calculate velocity vectors, is about 1 mm in both directions. As shown below, the present PIV resolves the scales close to the small-scale end of the inertial subrange and does not resolve the smallest scale of turbulent motion. This resolution is sufficient to evaluate most velocity statistics dominated by large-scale motion. In each wind tunnel operation, about 35 snapshots of velocity vectors are captured on average.

0.150 m0.267 m0.360 m0.465 m0.605 m0.746 m
xCx_C0.150 m0.267 m0.360 m0.465 m0.605 m0.746 m
yCy_C0 m0 m0 m0 m0 m0 m
Δt\Delta t3 μ\mus4 μ\mus4 μ\mus4 μ\mus4 μ\mus4 μ\mus
NSN_S722797700726651595

Table 1. Locations and parameters of PIV measurements.

The experiments are repeated many times for statistical analysis. The total number of acquired vector images, NSN_S, is also shown in Table 1. The snapshots of velocity vectors are measured at 15 Hz, for which the time interval between two snapshots is 0.067 s. The mean velocity in the fully developed turbulent region is about 40 m/s. The characteristic length scale of the flow is roughly estimated as the spacing between the nozzles, 0.012 m. The flow time scale defined with these velocity and length scales is 0.0003 s, much shorter than the sampling interval. Thus, the snapshots of velocity vectors are independent samples that are not correlated with each other. Statistics are evaluated with ensemble averages. Two-dimensional velocity vectors are denoted by (u,v)(u,v). For a variable ff, the nnth snapshot is denoted by f(n)(x,y)f^{(n)}(x,y). The average of ff, ⟨f⟩\langle f\rangle, is evaluated as

⟨f⟩(x,y)=1NS∑n=1NSf(n)(x,y).(1)\langle f\rangle(x,y)=\frac{1}{N_S}\sum_{n=1}^{N_S}f^{(n)}(x,y). \tag*{(1)}

Fluctuations are defined as f′(n)(x,y)=f(n)(x,y)−⟨f⟩(x,y)f^{\prime(n)}(x,y)=f^{(n)}(x,y)-\langle f\rangle(x,y), and the rms fluctuations are evaluated as frms(x,y)=⟨f′2⟩1/2=(⟨f2⟩−⟨f⟩2)1/2f_{\mathrm{rms}}(x,y)=\langle f^{\prime2}\rangle^{1/2}=\left(\langle f^2\rangle-\langle f\rangle^2\right)^{1/2}. In addition, skewness and flatness of ff are defined as S(f)=⟨f′3⟩/⟨f′2⟩3/2S(f)=\langle f^{\prime3}\rangle/\langle f^{\prime2}\rangle^{3/2} and F(f)=⟨f′4⟩/⟨f′2⟩2F(f)=\langle f^{\prime4}\rangle/\langle f^{\prime2}\rangle^2, respectively.

Pitot tube measurement

Pitot tube measurements are conducted to investigate the flow evolution along the centerline of the test section. The present study uses an L-shaped standard Pitot tube with a 7 mm diameter (Testo, 0635 2145) and a pressure sensor (SMC, PSE543A-R06). The signal from the pressure sensor is recorded with an oscilloscope (Yokokawa, DL850E) at a sampling rate of 10 kHz. Total pressure PTP_T is measured at the tip of the tube, while static pressure PSP_S is measured on the side of the tube, which is separated from the tip by 30 mm in the streamwise direction. The top wall of the test section is replaced with many small plates between which the Pitot tube is inserted into the test section. The measurement location is along the centerline of the test section. As shown in Sect. 3.1, the streamwise distribution of static pressure PS(x)P_S(x) is well fitted by an exponential curve PS,F(x)=aP(bP)xP_{S,F}(x)=a_P(b_P)^x, where the parameters can be obtained with a least squares method for measured PSP_S. Because the Pitot tube measures PT(x)P_T(x) and PS,F(x+ΔPitot)P_{S,F}(x+\Delta_{Pitot}) with ΔPitot=30\Delta_{Pitot}=30 mm, the pressure difference ΔP(x)=PT(x)−PS(x)\Delta P(x)=P_T(x)-P_S(x) is calculated with the measured PT(x)P_T(x) and the fitting curve PS,F(x)P_{S,F}(x). A time average is calculated for the time series pressure data over 2 s, for which the plenum pressure is almost constant.

Because the measurements are conducted in turbulence, the velocity fluctuations also affect ΔP\Delta P, whose measured values are not related solely to the dynamic pressure due to the mean flow. Turbulence increases the total pressure measured by the Pitot tube. The following relation expresses the pressure difference measured by a Pitot tube in turbulence [4]:

ΔP=12ρ(⟨u2⟩+⟨u′2⟩+⟨v′2⟩+⟨w′2⟩).(2)\Delta P=\frac{1}{2}\rho\left(\langle u^2\rangle+\langle u'^2\rangle+\langle v'^2\rangle+\langle w'^2\rangle\right). \tag*{(2)}

The velocity variances are evaluated with the PIV as ⟨u′2⟩+⟨v′2⟩+⟨w′2⟩≈⟨u′2⟩+2⟨v′2⟩\langle u'^2\rangle+\langle v'^2\rangle+\langle w'^2\rangle\approx\langle u'^2\rangle+2\langle v'^2\rangle. Here, the Pitot tube measurement is conducted along the centerline, where ⟨v′2⟩≈⟨w′2⟩\langle v'^{2}\rangle\approx\langle w'^{2}\rangle is expected to be valid. The density is also estimated from static pressure and temperature measurements, as described below. Thus, the mean velocity is calculated from ΔP\Delta P measured by the Pitot tube as

⟨u⟩=2ΔPρ−(⟨u′2⟩+2⟨v′2⟩).(3)\langle u\rangle= \sqrt{\frac{2\Delta P}{\rho} - \left(\langle u'^{2}\rangle+ 2\langle v'^{2}\rangle\right)}. \tag*{(3)}

Another factor that may affect the Pitot tube measurement is the compressibility effect. However, based on the mean velocity measured with PIV, we have estimated that this effect results in underestimating the mean velocity only by less than 2%. Therefore, the compressibility correction for mean velocity is not applied to the Pitot tube measurement.

Temperature measurement

Temperature is measured with a fine sheathed K-type thermocouple (J Thermo, TJK-LS1501GP), whose sheath has a 0.15 mm diameter and a 100 mm length. The response time of this thermocouple is less than 1 s. The output voltage of the thermocouple is processed and recorded by a data logger (Hioki, LR8431) at a sampling rate of 100 Hz. Because of the response time, the temperature recorded over 1 s after 1 s from the start of the wind tunnel is averaged to estimate the mean temperature. The thermocouple is inserted from the top of the test section in the same manner as the Pitot tube. Temperature measurements are conducted at different streamwise locations.

Estimation of the dissipation rate of turbulent kinetic energy

The decay of turbulent kinetic energy along the centerline is investigated for a nearly homogeneous and isotropic region corresponding to the downstream region. The turbulent kinetic energy per unit mass, kT=(⟨u′2⟩+⟨v′2⟩+⟨w′2⟩)/2k_T = \left(\langle u'^{2}\rangle+ \langle v'^{2}\rangle+ \langle w'^{2}\rangle\right)/2 is evaluated as (⟨u′2⟩+2⟨v′2⟩)/2(\langle u'^{2}\rangle+2\langle v'^{2}\rangle)/2 with the PIV data. The governing equation for kTk_T can be derived from the conservation equation of momentum for a compressible fluid, which is rewritten in the form of

∂ui∂t+uj∂ui∂xj=−1ρ∂p∂xi+1ρ∂τij∂xj.(4)\frac{\partial u_i}{\partial t}+u_j\frac{\partial u_i}{\partial x_j}=-\frac{1}{\rho}\frac{\partial p}{\partial x_i}+\frac{1}{\rho}\frac{\partial\tau_{ij}}{\partial x_j}. \tag*{(4)}

Here, p(x,y,z,t)p(x,y,z,t) and ρ(x,y,z,t)\rho(x,y,z,t) are local values of pressure and density, respectively, and τij\tau_{ij} is the viscous stress tensor. Multiplying Eq. (4) by ui′=ui−⟨ui⟩u_i'=u_i-\langle u_i\rangle and rearranging terms yield the following governing equation for kTk_T:

∂kT∂t+⟨uj⟩∂kT∂xj=−⟨ui′uj′⟩∂⟨ui⟩∂xj⏟Pk−12∂⟨ui′ui′uj′⟩∂xj⏟DT+⟨ui′ui′2∂uj′∂xj⟩⏟Θk−⟨1ρ∂uj′p∂xj⟩⏟DP+⟨pρ∂uj′∂xj⟩⏟ΘP+⟨1ρ∂ui′τij∂xj⟩⏟DV−⟨τijρ∂ui′∂xj⟩⏟ε.(5)\begin{aligned} \frac{\partial k_T}{\partial t}+\langle u_j\rangle\frac{\partial k_T}{\partial x_j} ={}&-\underbrace{\langle u_i'u_j'\rangle\frac{\partial\langle u_i\rangle}{\partial x_j}}_{P_k} -\underbrace{\frac{1}{2}\frac{\partial\langle u_i'u_i'u_j'\rangle}{\partial x_j}}_{D_T} +\underbrace{\left\langle\frac{u_i'u_i'}{2}\frac{\partial u_j'}{\partial x_j}\right\rangle}_{\Theta_k}\\ &-\underbrace{\left\langle\frac{1}{\rho}\frac{\partial u_j'p}{\partial x_j}\right\rangle}_{D_P} +\underbrace{\left\langle\frac{p}{\rho}\frac{\partial u_j'}{\partial x_j}\right\rangle}_{\Theta_P} +\underbrace{\left\langle\frac{1}{\rho}\frac{\partial u_i'\tau_{ij}}{\partial x_j}\right\rangle}_{D_V} -\underbrace{\left\langle\frac{\tau_{ij}}{\rho}\frac{\partial u_i'}{\partial x_j}\right\rangle}_{\varepsilon}. \tag*{(5)} \end{aligned}

The first and second terms of the left-hand side represent the temporal variation of kTk_T and the advection due to mean flow, respectively. On the right-hand side, PkP_k is the production term, DTD_T is the turbulent diffusion term, DPD_P is the pressure diffusion term, DVD_V is the viscous diffusion term, and ε\varepsilon is the dissipation term. These terms also appear in the governing equation for an incompressible fluid. The remaining terms represent direct influences of fluid compressibility: Θk\Theta_k is the change in the turbulent kinetic energy per unit mass due to fluid expansion or compression, and ΘP\Theta_P is the energy conversion between the kinetic and internal energies by pressure work. Even though the interaction of the supersonic jets generates turbulence near the nozzles, the velocity fluctuations in the decay region are subsonic, implying that Θk\Theta_k and ΘP\Theta_P are not dominant in the variation of kTk_T once the nearly homogeneous and isotropic turbulence has fully developed. As long as the Reynolds number is not too small, the viscous contribution to the spatial transport of kinetic energy is negligible, and $D_V is much smaller than other terms in turbulence [75, 73]. In addition, the flow is assumed to be statistically steady and homogeneous in the yy and zz directions, and the diffusion terms in these directions are negligible. The turbulent and pressure diffusions in the streamwise direction are also negligible compared with the transport due to the mean flow in decaying HIT [73]. Then, Eq. (5) is simplified as

⟨u⟩∂kT∂x=−ε.(6)\langle u\rangle\frac{\partial k_T}{\partial x}=-\varepsilon. \tag*{(6)}

This relation is often used to evaluate the dissipation rate in grid turbulence because ⟨u⟩\langle u\rangle and kTk_T are much easier to measure than ε\varepsilon [35, 67]. The decay of HIT depends on the spectral shape at large scales, which can differ depending on how turbulence is generated [20]. The compressibility effect on the decay properties may also emerge through the generation process of turbulence, although this influence does not appear in the governing equation of kTk_T. Such influences may appear in the virtual origin and decay exponent in the power law decay of kTk_T [13, 86].

Equation (6) is used to estimate the turbulent kinetic energy dissipation rate from the PIV data. The decay of kTk_T in HIT is well represented by a power law given by

kT=ak(x−x0)−n,(7)k_T=a_k(x-x_0)^{-n}, \tag*{(7)}

where aka_k is a coefficient related to the level of velocity fluctuations, x0x_0 is the virtual origin, and nn is the decay exponent [45, 20]. A least squares method is applied to the streamwise distribution of kTk_T obtained by the PIV data. Then, ε\varepsilon is evaluated as

ε=⟨u⟩an(x−x0)−n−1.(8)\varepsilon= \langle u\rangle an(x-x_0)^{-n-1}. \tag*{(8)}

Results and discussion

Pressure and temperature profiles

Figure 5(a) presents the streamwise variation of the mean static pressure (gauge pressure), PSP_S. The error bars represent the standard deviations of time series pressure histories, which indicate the degree of temporal pressure fluctuations. The pressure increases in the downstream direction and approaches the atmospheric pressure. The pressure distribution is well approximated by an exponential curve PS=aP(bP)xP_S = a_P(b_P)^x, where aP=−15.0 kPaGa_P = -15.0\ \mathrm{kPaG} and bP=6.61×10−5b_P = 6.61 \times10^{-5} are obtained with a least squares method. As the results are shown as gauge pressure, this pressure variation is slight compared to the atmospheric pressure and hardly affects the mean fluid density in the test section. Figure 5(b) presents the mean temperature TT along the centerline. The error bars also represent the standard deviation. The streamwise variation of TT is slight, and the average of all data points yields 303 K.

Figure 5. Streamwise variations of (a) mean static pressure PSP_S and (b) mean temperature TT. The error bars represent the standard deviation.

These measurement results enable us to estimate fluid density ρ\rho with the equation of state for an ideal gas as ρ=P/RT\rho= P/RT with the gas constant R=287.05R = 287.05 J/kg K. Here, the exponential function PS=aP(bP)xP_{S} = a_{P}(b_{P})^{x} and mean temperature of 303 K are used to evaluate ρ(x)\rho(x). The mean temperature yields the viscosity coefficient μ\mu, calculated with Sutherland’s law as

μ=μ0(TT0)(T0+ST+S),(9)\mu= \mu_{0}\left(\frac{T}{T_{0}}\right)\left(\frac{T_{0}+S}{T+S}\right), \tag*{(9)}

with T0=273T_{0} = 273 K, S=110.4S = 110.4 K, and μ0=1.724×10−5\mu_{0} = 1.724 \times10^{-5} Pa s. As the mean temperature does not vary with xx, μ\mu is also treated as a constant. These fluid properties are used in the analysis of velocity statistics obtained with the PIV.

Velocity statistics

Figure 6 shows instantaneous fluctuating velocity vectors u′=(u′,v′)\mathbf{u}^{\prime} = (u^{\prime}, v^{\prime}) measured at xC=0.150x_{C} = 0.150 m and 0.465 m. The color represents the vector magnitude. For both measurement locations, the turbulent jets from the nozzles have fully developed and have interacted with each other, and the clear imprints of each jet are not observed in the velocity field. At xC=0.150x_{C} = 0.150 m, the velocity fluctuations reach about 150 m/s. As turbulence decays, velocity fluctuations become small from xC=0.150x_{C} = 0.150 m to 0.465 m. At both locations, the vector fields exhibit typical rotating patterns associated with vortices, e.g., (x,y)=(0.11(x,y) = (0.11 m, −0.01-0.01 m) in Figure 6(a), indicating that the present PIV well captures these turbulent structures.

Figure 6. Instantaneous fluctuating velocity vectors u′=(u′,v′)\mathbf{u}'=(u',v') measured at (a) xC=0.150 mx_C=0.150\ \mathrm{m} and (b) xC=0.465 mx_C=0.465\ \mathrm{m}. The color represents the vector magnitude ∣u′∣|\mathbf{u}'|.

The degree of statistical convergence is examined before the velocity statistics are discussed. For the measurement location of (xC,yC)=(0.360(x_{C}, y_{C}) = (0.360 m, 0 m), 700 snapshots of velocity vectors are acquired. These snapshots are divided into two datasets, each containing 350 snapshots. The velocity statistics are separately calculated for the two datasets whose differences arise from statistical errors. Figure 7 compares the streamwise distributions of the average and rms fluctuations of streamwise velocity, ⟨u⟩\langle u\rangle and urmsu_{rms}, at y=0y=0. The two datasets yield slightly different distributions. A statistical quantity ff calculated with datasets 1 and 2 is denoted by f1f_1 and f2f_2, respectively. For

Figure 7. Statistical convergence tests for mean velocity and rms velocity fluctuations with the PIV data measured at (xC,yC)=(0.360 m,0 m)(x_C,y_C)=(0.360\ \mathrm{m},0\ \mathrm{m}). Totally 700 vector snapshots are divided into two datasets, each containing 350 snapshots. The mean velocity and rms velocity fluctuations calculated separately for the two datasets are plotted as functions of xx at y=0y=0.

the distribution of ff along the xx axis at y=0y = 0, the statistical error arising from the finite number of samples is evaluated with f1(x,0)f_1(x,0) and f2(x,0)f_2(x,0) as an rms error:

E(f)=1xS−xE∫xSxE(f1−f2)2dx,(10)E(f) = \sqrt{\frac{1}{x_S-x_E}\int_{x_S}^{x_E}(f_1-f_2)^2 dx}, \tag*{(10)}

where xS≤x≤xEx_S \le x \le x_E is the measurement area. For Figure 7, the relative errors, E(⟨u⟩)/⟨u⟩E(\langle u\rangle)/\langle u\rangle and E(urms)/urmsE(u_{rms})/u_{rms}, are about 5%, which is a typical statistical error when the number of the snapshots is about 350. Thus, statistical errors for the present PIV measurement with NS>500N_S > 500 are expected to be smaller than 5%. The statistical errors of other velocity statistics are evaluated with this method for each PIV measurement location, and E(f)E(f) is shown in figures below.

Figure 8(a) shows the streamwise distribution of mean streamwise velocity ⟨u⟩\langle u\rangle. The statistical errors estimated with Eq. (10) are shown in gray for PIV data. The mean velocity obtained with the Pitot tube is also plotted for comparison. For the Pitot tube measurement, the error bars represent the mean velocities which correspond to the pressure differences of ΔP±(ΔP)rms\Delta P \pm(\Delta P)_{rms}, where (ΔP)rms(\Delta P)_{rms} is the rms fluctuations of ΔP\Delta P. For the Pitot tube measurement at x=0.09x = 0.09 m, the PIV data for turbulence correction, Eq. (3), is unavailable. Therefore, an extrapolation of the PIV data provides the velocity variance used for the correction. The mean velocity distribution hardly differs for the PIV and the Pitot tube, ensuring the measurement accuracy of both methods. The mean velocity rapidly decreases in the xx direction up to x≈0.3x \approx0.3 m and then stays at an almost constant value of 37 m/s.

Figure 8. Streamwise distributions of (a) mean streamwise velocity ⟨u⟩\langle u\rangle and (b) rms velocity fluctuations of streamwise and vertical velocities, urmsu_{rms} and vrmsv_{rms}, at y=0y = 0. The gray shaded area represents the statistical errors estimated by Eq. (10).

Figure 9. Vertical distributions of (a) mean velocity and (b) rms velocity fluctuations for streamwise and vertical velocities.

Figure 8(b) shows the distribution of rms velocity fluctuations of streamwise and vertical velocities, urmsu_{rms} and vrmsv_{rms}. The rms velocity fluctuations increase with xx for x≲0.2x \lesssim0.2 m, where turbulence is generated. They continuously decay in the xx direction after the peaks at x≈0.2x \approx0.2 m. Although urmsu_{rms} is much larger than vrmsv_{rms} in the upstream region, the difference between urmsu_{rms} and vrmsv_{rms} becomes small with xx. Their ratio urms/vrmsu_{rms}/v_{rms} is often used to evaluate the isotropy at large scales. The average of urms/vrmsu_{rms}/v_{rms} for x>0.4x > 0.4 m is 1.08. This value is close to those in grid turbulence, for which urms/vrms≈1.1u_{rms}/v_{rms} \approx1.1 has been reported [32, 36] (Krogstad and Davidson 2012). Thus, grid turbulence and turbulence generated by the supersonic jet interaction have a similar degree of isotropy.

Figure 9 shows the vertical distributions of mean velocities, ⟨u⟩\langle u\rangle and ⟨v⟩\langle v\rangle, and rms velocity fluctuations, urmsu_{rms} and vrmsv_{rms}. At x=0.16x = 0.16 m, the mean velocity varies significantly with yy. The yy dependence of the mean velocity becomes weak as xx increases. For x≥0.36x \ge0.36 m, ⟨u⟩\langle u\rangle and ⟨v⟩\langle v\rangle are almost uniform in both xx and yy directions, and the mean velocity gradient is negligible in the downstream region. The rms velocity fluctuations are also inhomogeneous in the yy direction at x=0.16x = 0.16 m, where urmsu_{\mathrm{rms}} and vrmsv_{\mathrm{rms}} are large away from the center (y=0y = 0). Large urmsu_{\mathrm{rms}} and vrmsv_{\mathrm{rms}} at x=0.16x = 0.16 m appear in the region of yy with the large mean velocity gradient ∂⟨u⟩/∂y\partial\langle u\rangle/\partial y, and the production of turbulent kinetic energy due to the mean shear still influences the flow at x=0.16x = 0.16 m. The rms velocity fluctuations become homogeneous in the yy direction for x≥0.27x \geq0.27 m. These results suggest the nearly homogeneous and isotropic turbulence with a uniform mean flow forms for x≳0.3x \gtrsim0.3 m. With the nozzle spacing LJ=12L_J = 12 mm and the jet exit diameter DE=4.31D_E = 4.31 mm, x≳0.3x \gtrsim0.3 m corresponds to x/LJ≳25x/L_J \gtrsim25 and x/DE≳70x/D_E \gtrsim70.

Figure 10 shows the streamwise distribution of skewness SS and flatness FF of velocity components. A variable with a probability distribution following a Gaussian function has S=0S = 0 and F=3F = 3, shown as horizontal lines in Figure 10. These Gaussian values of SS and FF have been obtained for various types of fully-developed isotropic turbulence ([30] (Krogstad and Davidson 2010)). For the present facility, SS and FF greatly differ from the Gaussian values in the upstream region of x≲0.15x \lesssim0.15 m. The flow is highly inhomogeneous and anisotropic in this region. Large FF is related to the intermittency of large-scale fluctuations. The intermittent behavior of the jet interaction may cause large values of F(u)F(u) and F(v)F(v), as also reported for the interaction of two incompressible jets ([84]). As turbulence develops, SS and FF approach the Gaussian values, and S≈0S \approx0 and F≈3F \approx3 are achieved for x≳0.2x \gtrsim0.2. Therefore, the intermittent velocity fluctuations become insignificant once the flow becomes statistically homogeneous in the cross-section.

Figure 10. Streamwise distributions of skewness and flatness of streamwise and vertical velocities at y=0y = 0.

The large-scale turbulent motion is characterized by integral scales, which can be evaluated with velocity auto-correlation functions. The longitudinal auto-correlation functions for streamwise and vertical velocities are respectively defined as

fu(x,y,rx)=⟨u′(x,y)u′(x+rx,y)⟩⟨u′2(x,y)⟩,(11)f_u(x,y,r_x)=\frac{\langle u'(x,y)u'(x+r_x,y)\rangle}{\langle u'^2(x,y)\rangle}, \tag*{(11)}
fv(x,y,ry)=⟨v′(x,y)v′(x,y+ry)⟩⟨v′2(x,y)⟩,(12)f_v(x,y,r_y)=\frac{\langle v'(x,y)v'(x,y+r_y)\rangle}{\langle v'^2(x,y)\rangle}, \tag*{(12)}

where rxr_x and ryr_y are separation distances. Because the PIV measures the velocity profiles in a finite area, the available ranges of rxr_x and ryr_y are limited. For evaluations of the correlation function with large distances, fu(rx)f_u(r_x) is calculated for (x,y)(x,y) corresponding to a location near the upstream end of the measurement area at y=0y=0. Similarly, fv(ry)f_v(r_y) is calculated for (x,y)(x,y) corresponding to the upstream bottom corner of the measurement area. In this way, rxr_x and ryr_y range up to the size of the measurement area.

The longitudinal integral scales are defined as the integrals of fuf_u and fvf_v:

Lu=∫0rx0fu drx,Lv=∫0ry0fv dry.(13)L_u=\int_{0}^{r_{x0}} f_u\,\mathrm{d}r_x,\qquad L_v=\int_{0}^{r_{y0}} f_v\,\mathrm{d}r_y. \tag*{(13)}

Here, rx0r_{x0} and ry0r_{y0} are often taken as the first zero-crossing point [13]. For the present PIV, the measurement area is insufficient to calculate the correlation functions up to the zero-crossing point. In addition, as also observed for hot-wire measurements in wind tunnels, an accurate evaluation of correlation functions for large rxr_x and ryr_y is problematic because of a limited number of large-scale samples. For these reasons, the integral scale is often evaluated by approximating the correlation function at large rr with an exponential function [47 Bewley et al.,] 2012; Griffin et al., [28]. The present study also adapts this methodology for the calculation of integral scales. An exponential function afexp⁡(−bfrx)a_f \exp(-b_f r_x) is obtained with a least squares method applied for measured fuf_u in the range of rxr_x with fu<0.5f_u < 0.5. The numerical integration of fuf_u is calculated with the measured fuf_u and the exponential function as

Lu=∫0r1/2fudrx+∫r1/2rendafe−bfrxdrx,(14)L_u = \int_{0}^{r_{1/2}} f_u \mathrm{d}r_x + \int_{r_{1/2}}^{r_{\mathrm{end}}} a_f e^{-b_f r_x} \mathrm{d}r_x, \tag*{(14)}

where r1/2r_{1/2} is defined as fu(r1/2)=0.5f_u(r_{1/2}) = 0.5 and rendr_{\mathrm{end}} corresponds to rxr_x where afexp⁡(−bfrx)a_f \exp(-b_f r_x) is equal to 10−510^{-5}. Because of the exponential decay of the correlation, larger values of rendr_{\mathrm{end}} do not affect the evaluation of the integral scale. The integral scale in the vertical direction LvL_v is also calculated by the same method applied for fvf_v.

Figure 11(a) shows the correlaton function fuf_u and its approximation based on the exponential function at (x,y)=(570 mm,0 mm)(x, y) = (570\ \mathrm{mm}, 0\ \mathrm{mm}). The separation distance rxr_x is normalized with the integral scale for comparisons with other turbulent flows. The exponential function accurately approximates the decay of fuf_u for large rxr_x. In addition, the profile of fuf_u agrees well with the correlation functions measured in other turbulent flows.

Figure 11. (a) Longitudinal auto-correlation function of streamwise velocity at x=570 mmx = 570\ \mathrm{mm} and y=0 mmy = 0\ \mathrm{mm}. The measured correlation is compared with an exponential function fu=afexp⁡(−bfrx)f_u = a_f \exp(-b_f r_x) and the correlation functions in incompressible turbulent jets [49, 76]. A least squares method determines af=0.843a_f = 0.843 and bf=−32.8 m−1b_f = -32.8\ \mathrm{m}^{-1} from the measured fuf_u. The separation distance rxr_x is normalized by the integral scale LuL_u. (b) The streamwise variation of integral scales LuL_u and LvL_v, which are calculated from the longitudinal auto-correlation functions of streamwise and vertical velocities, respectively. The inset shows the ratio between LuL_u and LvL_v.

Figure 11(b) presents the streamwise variation of the integral scales. In the inhomogeneous and anisotropic region, x<0.3 mx < 0.3\ \mathrm{m}, the streamwise integral scale LuL_u is larger than the vertical scale LvL_v, and large-scale velocity fluctuations are also anisotropic. On the other hand, the downstream region satisfies Lu≈LvL_u \approx L_v. This is also confirmed by the inset presenting Lu/LvL_u/L_v. The average of Lu/LvL_u/L_v for x>0.3 mx > 0.3\ \mathrm{m} is 1.03, and the integral scales hardly differ in both directions, suggesting that turbulent motion at large scales is statistically isotropic. For 0.3≲x≲0.60.3 \lesssim x \lesssim0.6, the integral scales increase as turbulence decays. However, in the most downstream location of the measurement, LuL_u and LvL_v decrease with xx. In DNS of decaying isotropic turbulence in a triply periodic domain, the decay is influenced by the boundary conditions when the domain size divided by the integral scale is smaller than about 3 [2]. The side length of the cross-section, 0.1 m, is about 3.7Lu3.7L_u at x=0.57 mx = 0.57\ \mathrm{m}. The decrease of LuL_u and LvL_v in the further downstream region may be caused by the confinement effects.

The dissipation rate of turbulent kinetic energy

Figure 12(a) shows the streamwise variation of turbulent kinetic energy per unit mass kT≈(⟨u′2⟩+2⟨v′2⟩)/2k_T \approx(\langle u'^2\rangle+ 2\langle v'^2\rangle)/2 along the centerline. The decay of kTk_T is investigated for the nearly homogeneous and isotropic region, x>0.32x > 0.32 m. A nonlinear least squares method (Levenberg–Marquardt method) applied for x>0.32x > 0.32 m yields the parameters in the power law, Eq. (7), as ak=21.0a_k = 21.0 m2^2/s2^2, x0=0.12x_0 = 0.12 m, and n=2.1n = 2.1. Equation (7) with these parameters is also shown in the figure. The decay of kTk_T is well approximated by the power law. The exponent n=2.1n = 2.1 is larger than typical values reported for incompressible grid turbulence at high Reynolds numbers, n≈1.0n \approx1.0–1.51.5 [35, 33, 67, 83]. As discussed in Sect. 1, Zwart et al. [86] also reported that nn in compressible grid turbulence tends to be larger than incompressible cases. This tendency is qualitatively consistent with the large exponent in this study. However, these results contradict the decrease of nn observed in shock-tube experiments [13, 25], and the compressibility effects on the decay exponent should be addressed in future studies.

Figure 12. (a) The decay of turbulent kinetic energy kTk_T along the centerline. The red broken line represents a power law kT=ak(x−x0)−nk_T = a_k(x-x_0)^{-n} with ak=21.0 m2/s2a_k = 21.0\ \mathrm{m^2/s^2}, x0=0.12 mx_0 = 0.12\ \mathrm{m}, and n=2.1n = 2.1, which are obtained with a least squares method applied for x>0.32 mx > 0.32\ \mathrm{m}. (b) The streamwise variations of Taylor microscale λ\lambda, Kolmogorov scale η\eta, and turbulent Reynolds number ReλRe_\lambda, estimated with the power law decay of kTk_T.

The small-scale turbulence properties are often related to the kinetic energy dissipation rate. The Taylor microscale λ\lambda and Kolmogorov scale η\eta are defined as

λ=10μkTρε,(15)\lambda= \sqrt{\frac{10\mu k_T}{\rho\varepsilon}}, \tag*{(15)}
η=(μρ)3/4ε−1/4.(16)\eta= \left(\frac{\mu}{\rho}\right)^{3/4}\varepsilon^{-1/4}. \tag*{(16)}

The turbulent Reynolds number ReλRe_\lambda is define with λ\lambda as

Reλ=ρ2kT/3λμ.(17)Re_\lambda= \frac{\rho\sqrt{2k_T/3}\lambda}{\mu}. \tag*{(17)}

The dissipation rate estimated from the decay of kTk_T with Eq. (8) is used to evaluate λ\lambda, η\eta, and ReλRe_\lambda. Figure 12(b) shows their streamwise variations in the nearly homogeneous and isotropic region. λ\lambda and η\eta increase as turbulence decays in the xx direction. The orders of the length scales are λ=O(10−3 m)\lambda= O(10^{-3}\ \mathrm{m}) and η=O(10−5 m)\eta= O(10^{-5}\ \mathrm{m}). The spatial resolution of PIV is about 1.0×10−3 m1.0 \times10^{-3}\ \mathrm{m}, and the scales greater than the Taylor microscale are resolved by the PIV. This resolution is sufficient to investigate the large-scale properties, such as mean velocity and rms velocity fluctuations. The turbulent Reynolds number decreases with the decay of turbulence. It ranges between about 800 and 400. These values of ReλRe_\lambda are larger than those of most subsonic grid turbulence in large wind tunnel facilities because of the large velocity fluctuations of supersonic jets.

One-dimensional energy spectra of streamwise and vertical velocities (EuE_u and EvE_v) are calculated with Fourier transform in the xx and yy directions, respectively. The wavenumbers in these directions are denoted by kxk_x and kyk_y. The energy spectra in turbulence with a high Reynolds number are known to have a universal shape in an inertial subrange in various turbulent flows [55]. Figure 13 presents EuE_u and EvE_v measured at several streamwise locations. The spectra and wavenumbers are normalized by the dissipation rate ε\varepsilon and kinematic viscosity ν=μ/ρ\nu= \mu/\rho. The present results are compared with incompressible turbulent flows. The spectra are evaluated for normalized wavenumbers between 10−310^{-3} and 10−110^{-1}, determined by the measurement area and spatial resolution. The present results quantitatively agree with other turbulent flows. They are also consistent with the power laws expected for the inertial subrange, Eu∼kx−5/3E_u \sim k_x^{-5/3} and Ev∼ky−5/3E_v \sim k_y^{-5/3}. Thus, the present PIV with condensation particles is capable of measuring the scale dependence of velocity fluctuations in the inertial subrange. In addition, the flow at a wide range of scales is in a statistically isotropic state, as attested by Eu≈EvE_u \approx E_v.

Normalized energy spectra of streamwise and vertical velocities at four streamwise locations

Figure 13. Normalized energy spectra of (a) streamwise velocity, Eu(kx)E_u(k_x), and (b) vertical velocity, Ev(ky)E_v(k_y), where kxk_x and kyk_y are the wavenumbers in the streamwise and vertical directions, respectively. The spectra and wavenumbers are non-dimensionalized with the kinematic viscosity ν=μ/ρ\nu= \mu/\rho, the energy dissipation rate per unit mass ε\varepsilon, and the Kolmogorov scale η\eta. The present results are compared with direct numerical simulation of forced HIT (Reλ=202Re_\lambda= 202) [74] and experiments of a boundary layer (Reλ=202Re_\lambda= 202) [59, 50] and grid turbulence (Reλ=520Re_\lambda= 520) [35].

The dissipation rate is often related to characteristic velocity U\mathcal{U} and length L\mathcal{L} of large-scale turbulent motion as

ε=CεU3L,(18)\varepsilon= C_{\varepsilon}\frac{\mathcal{U}^{3}}{\mathcal{L}}, \tag*{(18)}

where CεC_{\varepsilon} is a non-dimensional dissipation rate. CεC_{\varepsilon} plays a central role in theories and models of turbulent flows [72]. The velocity and length scales in the present experiment are evaluated as U=(⟨u′2⟩+⟨v′2⟩)/2\mathcal{U} = \sqrt{(\langle u^{\prime2}\rangle+ \langle v^{\prime2}\rangle)/2} and L=(Lu+Lv)/2\mathcal{L} = (L_u+L_v)/2, and CεC_{\varepsilon} is evaluated with ε\varepsilon in the nearly homogeneous and isotropic region. Figure 14 plots#+#+#+#+ CεC_\varepsilon as a function of ReλRe_\lambda for the present experiment and other studies of incompressible grid turbulence. In fully-developed grid turbulence with high ReλRe_\lambda in incompressible flows, CεC_\varepsilon takes values approximately between 0.4 and 1.1 while CεC_\varepsilon increases rapidly as ReλRe_\lambda becomes smaller than about 100. The present values of Cε=0.51C_\varepsilon= 0.51–0.87 are within a range reported for grid turbulence with high ReλRe_\lambda. The decay exponent of the turbulent kinetic energy is known to be influenced by the spatial variation of CεC_\varepsilon. For the present experiment, CεC_\varepsilon slightly increases in the streamwise direction. The increase of CεC_\varepsilon causes a larger decay exponent nn (Krogstad and Davidson, 2010). The spatial variation of CεC_\varepsilon is possibly relevant to the large values of nn in the present experiment and in compressible grid turbulence [86].

Dependence of non-dimensional energy dissipation rate $C_{\varepsilon}$ on the turbulent Reynolds number $Re_{\lambda}$

Figure 14. Dependence of non-dimensional energy dissipation rate CεC_{\varepsilon} on the turbulent Reynolds number ReλRe_{\lambda}. The present results are compared with experimental data of subsonic grid turbulence: Kistler and Vrebalovich (1966), Comte-Bellot and Corrsin (1971), and Zheng et al. (2021a) for passive-grid generated turbulence; Gamard and George (2000), Kang et al. (2003), Thormann and Meneveau (2014), Puga and LaRue (2017), Mora et al. (2019), and Zheng et al. (2021b) for active-grid generated turbulence.

Conclusion

The present study has reported the development and characterization of the multiple-supersonic-jet wind tunnel, which is designed to investigate decaying nearly homogeneous and isotropic turbulence in a compressible flow regime. The PIV, pitot tube, and thermocouple measurements characterize the generated turbulence, where velocity vectors, static and dynamic pressures, and temperature are measured. The present wind tunnel can sustain a steady mean flow over 2–3 s. It can also seed tracer particles in the test section by condensation. The accuracy of the PIV is confirmed by comparing the velocity statistics with other turbulent flows. In addition, the mean velocity measured by the PIV agrees well with the Pitot tube measurements, for which turbulence correction is applied with the PIV data of velocity variances.

The mean velocity in the upstream region is large along the centerline of the test section. However, as the flow evolves, the mean velocity profile becomes uniform in the test section. Then, the mean velocity gradient becomes negligibly small in the downstream region. Once the mean flow profile becomes uniform, nearly homogeneous and isotropic turbulence begins to decay. At this decay regime, rms velocity fluctuations are also uniform in the cross-section. The ratio of rms values of streamwise- and vertical-velocity fluctuations is about 1.08, close to the values reported for incompressible grid turbulence. The longitudinal integral length scales are also similar in the streamwise and vertical directions. These statistics have confirmed that the supersonic jet interaction generates nearly homogeneous and isotropic turbulence. In the inhomogeneous and anisotropic region near the jet nozzles, the flatness of velocity fluctuations is larger than 3, indicating that the large-scale velocity fluctuations are highly intermittent. The skewness and flatness of velocity fluctuations are close to the Gaussian values in the decay region, where the large-scale intermittency is insignificant.

The decay of turbulent kinetic energy per unit mass is investigated in the nearly homogeneous and isotropic region. The turbulent kinetic energy decays according to a power law, as reported for subsonic grid turbulence. The decay exponent nn for the supersonic jet interaction is about 2, which is larger than the values reported for incompressible grid turbulence. This large value of nn is consistent with experiments of compressible grid turbulence in a high-speed wind tunnel by Zwart et al [86]. In addition, the dissipation rate of turbulent kinetic energy per unit mass has been evaluated with the decay of the turbulent kinetic energy. The energy spectra of velocity fluctuations normalized by the dissipation rate and kinematic viscosity follow the universal power law of the inertial subrange, observed for other incompressible turbulent flows. The non-dimensional dissipation rate CεC_{\varepsilon} for the supersonic jet interaction is within a range of 0.51–0.87, which is also consistent with grid turbulence at a high Reynolds number.

The nozzle component in the present wind tunnel can be replaced with other components with different nozzle geometries, generating turbulent jets with different Mach and Reynolds numbers. This feature is helpful in varying the degree of the compressibility effects, which will be considered in future studies. These results have demonstrated that the multiple-supersonic-jet wind tunnel is helpful in the investigation of decaying homogeneous isotropic turbulence in a compressible flow regime.

Statements and Declarations

Funding

This work was supported by Japan Society for the Promotion of Science (KAKENHI Grant Numbers JP22K03903 and JP22H01398).

Competing interests

The authors have no competing interests to declare that are relevant to the content of this article.

Availability of data and materials

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Authors’ contributions

Takahiro Mori: Data curation (equal); Formal Analysis (equal); Investigation (lead); Methodology (equal); Resources (equal); Software (supporting); Validation (equal); Visualization (equal); Writing - original draft (supporting); Writing - review & editing (equal). Tomoaki Watanabe: Conceptualization (lead); Data curation (equal); Formal Analysis (equal); Funding acquisition (equal); Investigation (supporting); Methodology (equal); Project administration (lead); Resources (equal); Software (lead); Supervision (lead); Validation (equal); Visualization (equal); Writing - original draft (lead); Writing - review & editing (equal). Koji Nagata: Conceptualization (supporting); Data curation (supporting); Funding acquisition (equal); Investigation (supporting); Methodology (supporting); Project administration (supporting); Resources (equal); Software (supporting); Supervision (supporting); Validation (equal); Visualization (supporting); Writing - original draft (supporting); Writing - review & editing (equal).

Acknowledgments

Acknowledgments. The authors acknowledge Prof. K. Mori (Osaka Prefecture University), Mr. N. Iwakura (Nagoya University), and Ms. R. Nakayama (Nagoya University) for their help and valuable comments in developing the multiple-supersonic-jet wind tunnel. The authors also acknowledge Mr. K. Ishizawa (Nagoya University) for assistance in experiments.

References

  1. [1]Agui JH, Briassulis G, Andreopoulos Y (2005) Studies of interactions of a propagating shock wave with decaying grid turbulence: velocity and vorticity fields. J Fluid Mech 524:143–195DOI
  2. [2]Anas M, Joshi P, Verma MK (2020) Freely decaying turbulence in a finite domain at finite Reynolds number. Phys Fluids 32(9):095109DOI
  3. [3]Anderson JD (1990) Modern compressible flow: with historical perspective, vol 12. McGraw-Hill New York
  4. [4]Bailey SCC, Hultmark M, Monty JP, et al (2013) Obtaining accurate mean velocity measurements in high reynolds number turbulent boundary layers using Pitot tubes. J Fluid Mech 715:642–670DOI
  5. [5]Batchelor GK (1953) The theory of homogeneous turbulence. Cambridge university press
  6. [6]Bellani G, Variano EA (2014) Homogeneity and isotropy in a laboratory turbulent flow. Exp Fluids 55(1):1–12DOI
  7. [8]Birouk M, Sarh B, Gökalp I (2003) An attempt to realize experimental isotropic turbulence at low Reynolds number. Flow, Turbul Combust 70(1-4):325–348DOI
  8. [9]Bonnet JP, Chaput E (1986) Large-scale structures visualization in a high Reynolds number, turbulent flat-plate wake at supersonic speed. Exp Fluids 4(6):350–356DOI
  9. [10]Bos WJT, Chevillard L, Scott JF, et al (2012) Reynolds number effect on the velocity increment skewness in isotropic turbulence. Phys Fluids 24(1)DOI
  10. [11]Bradley D, Lawes M, Morsy ME (2019) Measurement of turbulence characteristics in a large scale fan-stirred spherical vessel. J Turbul 20(3):195–213
  11. [12]Bradshaw P (1977) Compressible turbulent shear layers. Annu Rev Fluid Mech 9(1):33–52DOI
  12. [13]Briassulis G, Agui JH, Andreopoulos Y (2001) The structure of weakly compressible grid-generated turbulence. J Fluid Mech 432:219–283DOI
  13. [14]Canuto VM (1997) Compressible turbulence. Astrophys J 482(2):827
  14. [15]Carter D, Petersen A, Amili O, et al (2016) Generating and controlling homogeneous air turbulence using random jet arrays. Exp Fluids 57(12):1–15DOI
  15. [16]Clemens NT, Mungal MG (1991) A planar Mie scattering technique for visualizing supersonic mixing flows. Exp Fluids 11(2-3):175–185DOI
  16. [17]Comte-Bellot G, Corrsin S (1966) The use of a contraction to improve the isotropy of grid-generated turbulence. J Fluid Mech 25(4):657–682
  17. [19]Crittenden TM, Glezer A (2006) A high-speed, compressible synthetic jet. Phys Fluids 18(1):017107DOI
  18. [20]Davidson PA (2004) Turbulence: An Introduction for Scientists and Engineers. Oxford Univ. Pr.
  19. [21]Davidson PA (2009) The role of angular momentum conservation in homogeneous turbulence. J Fluid Mech 632:329–358DOI
  20. [22]Djenidi L, Kamruzzaman M, Antonia RA (2015) Power-law exponent in the transition period of decay in grid turbulence. J Fluid Mech 779:544–555DOI
  21. [23]Donzis DA, John JP (2020) Universality and scaling in homogeneous compressible turbulence. Phys Rev Fluids 5(8):084609DOI
  22. [24]Foelsch K (1949) The analytical design of an axially symmetric Laval nozzle for a parallel and uniform jet. J Aeronaut Sci 16(3):161–166DOI
  23. [25]Fukushima G, Ogawa S, Wei J, et al (2021) Impacts of grid turbulence on the side projection of planar shock waves. Shock Waves 31(2):101–115DOI
  24. [27]Goebel SG, Dutton JC, Krier H, et al (1990) Mean and turbulent velocity measurements of supersonic mixing layers. Exp Fluids 8:263–272
  25. [28]Griffin KP, Wei NJ, Bodenschatz E, et al (2019) Control of long-range correlations in turbulence. Exp Fluids 60:1–14
  26. [29]Honkan A, Andreopoulos J (1992) Rapid compression of grid-generated turbulence by a moving shock wave. Phys Fluids 4(11):2562–2572DOI
  27. [30]Hwang W, Eaton JK (2004) Creating homogeneous and isotropic turbulence without a mean flow. Exp Fluids 36(3):444–454DOI
  28. [31]Hwang W, Eaton JK (2006) Homogeneous and isotropic turbulence modulation by small heavy (St ∼ 50) particles. J Fluid Mech 564:361DOI
  29. [32]Isaza JC, Salazar R, Warhaft Z (2014) On grid-generated turbulence in the near-and far field regions. Journal of Fluid Mechanics 753:402–426DOI
  30. [33]Kang HS, Chester S, Meneveau C (2003) Decaying turbulence in an active-grid-generated flow and comparisons with large-eddy simulation. J Fluid Mech 480:129–160DOI
  31. [34]Karthick SK, Rao SMV, Jagadeesh G, et al (2017) Passive scalar mixing studies to identify the mixing length in a supersonic confined jet. Exp Fluids 58:1–20DOI
  32. [35]Kistler AL, Vrebalovich T (1966) Grid turbulence at large Reynolds numbers. J Fluid Mech 26(1):37–47DOI
  33. [36]Kitamura T, Nagata K, Sakai Y, et al (2014) On invariants in grid turbulence at moderate Reynolds numbers. J Fluid Mech 738:378–406DOI
  34. [37]Kouchi T, Fukuda S, Miyai S, et al (2019) Acetone-condensation nano-particle image velocimetry in a supersonic boundary layer. In: AIAA Scitech 2019 Forum, p 1821DOI
  35. [40]Larssen JV, Devenport WJ (2011) On the generation of large-scale homogeneous turbulence. Exp Fluids 50:1207–1223DOI
  36. [41]Lee S, Lele SK, Moin P (1991) Eddy shocklets in decaying compressible turbulence. Phys Fluids 3(4):657–664DOI
  37. [42]Lohse D (1994) Crossover from high to low Reynolds number turbulence. Phys Rev Lett 73(24):3223DOI
  38. [43]Makita H (1991) Realization of a large-scale turbulence field in a small wind tunnel. Fluid Dyn Res 8(1-4):53DOI
  39. [44]Marié JL, Tronchin T, Grosjean N, et al (2017) Digital holographic measurement of the Lagrangian evaporation rate of droplets dispersing in a homogeneous isotropic turbulence. Exp Fluids 58(2):11DOI
  40. [45]Mohamed MS, Larue JC (1990) The decay power law in grid-generated turbulence. J Fluid Mech 219:195–214DOI
  41. [47]Morikawa K, Urano S, Sanada T, et al (2008) Turbulence modulation induced by bubble swarm in oscillating-grid turbulence. J Power Energy Syst 2(1):330–339
  42. [48]Mydlarski L, Warhaft Z (1996) On the onset of high-Reynolds-number grid-generated wind tunnel turbulence. J Fluid Mech 320(1):331–368DOI
  43. [49]Namer I, Ötügen MV (1988) Velocity measurements in a plane turbulent air jet at moderate Reynolds numbers. Exp Fluids 6(6):387–399DOI
  44. [50]Nieuwstadt JF. T .and Westerweel, Boersma B (2016) Introduction to Theory and Applications of Turbulent Flows. Springer
  45. [51]Ozono S, Ikeda H (2018) Realization of both high-intensity and large-scale turbulence using a multi-fan wind tunnel. Exp Fluids 59(12):1–12DOI
  46. [52]Pérez-Alvarado A, Mydlarski L, Gaskin S (2016) Effect of the driving algorithm on the turbulence generated by a random jet array. Exp Fluids 57(2):20
  47. [53]Pham TD, Watanabe T, Nagata K (2023) Large-eddy simulation of a flow generated by a piston-driven synthetic jet actuator. CFD Lett 15(8):1–18
  48. [55]Pope SB (2000) Turbulent Flows. Cambridge Univ. Pr.DOI
  49. [57]Ravi S, Peltier SJ, Petersen EL (2013) Analysis of the impact of impeller geometry on the turbulent statistics inside a fan-stirred, cylindrical flame speed vessel using piv. Exp Fluids 54(1):1424DOI
  50. [58]Ristorcelli JR, Blaisdell GA (1997) Consistent initial conditions for the DNS of compressible turbulence. Phys Fluids 9(1):4–6DOI
  51. [59]Saddoughi SG, Veeravalli SV (1994) Local isotropy in turbulent boundary layers at high Reynolds number. J Fluid Mech 268:333–372
  52. [60]Sakakibara H, Watanabe T, Nagata K (2018) Supersonic piston synthetic jets with single/multiple orifice. Exp Fluids 59(5):76
  53. [61]Semenov ES (1965) Measurement of turbulence characteristics in a closed volume with artificial turbulence. Combust Explos ShockWaves 1(2):57–62DOI
  54. [62]Sreenivasan KR, Antonia RA (1997) The phenomenology of small-scale turbulence. Annu Rev Fluid Mech 29(1):435–472DOI
  55. [63]Suzuki H, Nagata K, Sakai Y, et al (2010) High-Schmidt-number scalar transfer in regular and fractal grid turbulence. Phys Scr 2010(T142):014069DOI
  56. [64]Takamure K, Ozono S (2019) Relative importance of initial conditions on outflows from multiple fans. Phys Rev E 99(1):013112DOI
  57. [65]Tan HS, Ling SC (1963) Final stage decay of grid-produced turbulence. Phys Fluids 6(12):1693–1699DOI
  58. [66]Theunissen R (2010) Adaptive image interrogation for PIV: Application to compressible flows and interfaces. PhD thesis, TU Delft, Delft University of Technology, Delft, The Netherlands
  59. [67]Thormann A, Meneveau C (2014) Decay of homogeneous, nearly isotropic turbulence behind active fractal grids. Phys Fluids 26(2)DOI
  60. [68]Traub LW, Sweet M, Nilssen K (2012) Evaluation and characterization of a lateral synthetic jet actuator. J Aircraft 49(4):1039–1050DOI
  61. [69]Uberoi MS, Wallis S (1967) Effect of grid geometry on turbulence decay. Phys Fluids 10:1216–1224DOI
  62. [70]Valente PC, Vassilicos JC (2011) The decay of turbulence generated by a class of multiscale grids. J Fluid Mech 687:300–340DOI
  63. [71]Variano EA, Bodenschatz E, Cowen EA (2004) A random synthetic jet array driven turbulence tank. Exp Fluids 37(4):613–615DOI
  64. [72]Vassilicos JC (2015) Dissipation in turbulent flows. Annu Rev Fluid Mech 47:95–114DOI
  65. [73]Wang M, Yurikusa T, Sakai Y, et al (2022) Interscale transfer of turbulent energy in grid-generated turbulence with low Reynolds numbers. Int J Heat Fluid Flow 97:109031DOI
  66. [74]Watanabe T, Nagata K (2023) The response of small-scale shear layers to perturbations in turbulence. J Fluid Mech 963:A31DOI
  67. [75]Watanabe T, Sakai Y, Nagata K, et al (2016) Large eddy simulation study of turbulent kinetic energy and scalar variance budgets and turbulent/non-turbulent interface in planar jets. Fluid Dyn Res 48(2):021407DOI
  68. [76]Watanabe T, Zhang X, Nagata K (2019) Direct numerical simulation of incompressible turbulent boundary layers and planar jets at high Reynolds numbers initialized with implicit large eddy simulation. Comput Fluids 194:104314DOI
  69. [77]Watanabe T, Tanaka K, Nagata K (2021) Solenoidal linear forcing for compressible, statistically steady, homogeneous isotropic turbulence with reduced turbulent mach number oscillation. Phys Fluids 33(9):095108DOI
  70. [78]Westerweel J, Scarano F (2005) Universal outlier detection for PIV data. Exp Fluids 39(6):1096–1100DOI
  71. [79]Xu S, Huang S, Huang R, et al (2017) Estimation of turbulence characteristics from PIV in a high-pressure fan-stirred constant volume combustion chamber. Appl Therm Eng 110:346–355DOI
  72. [80]Yamamoto K, Ishida T, Watanabe T, et al (2022a) Experimental and numerical investigation of compressibility effects on velocity derivative flatness in turbulence. Phys Fluids 34(5)DOI
  73. [81]Yamamoto K, Watanabe T, Nagata K (2022b) Turbulence generated by an array of opposed piston-driven synthetic jet actuators. Exp Fluids 63(1):1–17DOI
  74. [83]Zheng Y, Nagata K, Watanabe T (2021b) Turbulent characteristics and energy transfer in the far field of active-grid turbulence. Phys Fluids 33(11):115119DOI
  75. [84]Zhou Y, Nagata K, Sakai Y, et al (2018) Dual-plane turbulent jets and their non-gaussian velocity fluctuations. Phys Rev Fluids 3(12):124604DOI
  76. [85]Zimmermann R, Xu H, Gasteuil Y, et al (2010) The Lagrangian exploration module: An apparatus for the study of statistically homogeneous and isotropic turbulence. Rev Sci Instrum 81(5):055112DOI
  77. [86]Zwart PJ, Budwig R, Tavoularis S (1997) Grid turbulence in compressible flow. Exp Fluids 23(6):520–522DOI

Paper details

vlnc:W_ne6sadgd
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents