You are on page 1of 11

Shock Waves (2009) 18:425435 DOI 10.

1007/s00193-008-0172-8

ORIGINAL ARTICLE

Aerodynamic force measurement using 3-component accelerometer force balance system in a hypersonic shock tunnel
S. Saravanan G. Jagadeesh K. P. J. Reddy

Received: 18 March 2008 / Revised: 23 July 2008 / Accepted: 18 September 2008 / Published online: 14 October 2008 Springer-Verlag 2008

Abstract A new three-component accelerometer force balance has been designed, calibrated and tested in hypersonic shock tunnel (HST2) of Indian Institute of Science. The newly designed balance is able to measure aerodynamic forces (within test time of one millisecond) on test models at angles of attack from 0 to 12 . Two models, a blunt cone with after body and a blunt cone with after body and frustum are used to establish the accuracy of the force balance. The tests were conducted for the above two congurations with a constant Mach number of 8 and total enthalpy of 2.0 MJ/kg. The effectiveness of the balance is demonstrated by comparing the forces and moments of measured data with AGARD models. The ow elds around the test model are simulated using a 3D axisymmetric NavierStokes solver and the simulated results were compared with the measured values. Measured and computed force data are matched within 10% for two different models tested here. The accuracy of the force balance is also estimated with the Newtonian theory and the values are approximately 10% for the axial component and 8% for the normal and pitching moment components. Keywords Accelerometer force balance Force measurements High enthalpy Shock tunnel and hypersonics

PACS 47.85.Gj 1 Introduction Force measurements in ground-based testing facilities are of major interest to the designers of hypersonic ight vehicles. Accurate experimental results are essential to support and validate the ight data and results from computational uid dynamics (CFD) codes. In particular, the prediction of fundamental aerodynamic forces presents an important problem for the vehicle performance (stability and control). Aerodynamic coefcients can be obtained by (1) conducting a wind tunnel test which will be costly and time consuming and would probably produce results which are more accurate than warranted for preliminary and intermediate design, (2) performing hand calculations using analytical expressions [1] and (3) developing a computer program based on analytical techniques which is efcient and produces accuracies on the order of 10%. These methods can provide fundamental measurements and the data obtained from them will be very useful for the research and development activities. Today, however, coupling of three main methods of hypersonic ight vehicle design, namely hypersonic ground based testing, CFD and ight testing are becoming economically achievable. The rst one, although costlier and time consuming initially, is the best approach for long-term use and is the one which will be addressed in this paper. Impulse facilities [2] are necessary for ground-based experiments that simulate ight enthalpies corresponding to ight Mach numbers exceeding 10. The penalty that we pay for simulation of hypersonic ight conditions in such tunnels is the extremely short test times (500 s10 ms). Unfortunately, conventional force measurement techniques do not possess the required response times, thus making force measurement in impulse facilities a non-trivial matter. However, progress has been

Communicated by R. Boyce. S. Saravanan (B) G. Jagadeesh K. P. J. Reddy Department of Aerospace Engineering, Indian Institute of Science, Bangalore 560012, India e-mail: saravan@aero.iisc.ernet.in G. Jagadeesh e-mail: jaggie@aero.iisc.ernet.in K. P. J. Reddy e-mail: laser@aero.iisc.ernet.in

123

426

S. Saravanan et al.

made in recent years to develop techniques which enable successful measurement of forces over short time scales. The techniques used to overcome this problem are stress wave measurement techniques [3,4], strain gauge balance [5] and acceleration measurement techniques [6]. However, design and calibration methodologies for these balances are different and some of the important types of force balances reported in literature for impulse facilities are briey discussed here. An attempt was made in 1960s to study and resolve some of the important technical issues of force measurements exclusively for impulse facilities. Naumann et al. [7] developed a free ying technique and measured the acceleration of a model in a shock tunnel using a complex model mounting system. This system releases the model just before the onset of hypersonic ow and the subsequent movement of the model was measured by a set of small piezo accelerometers. However, the complete measurement system was very complex and cumbersome to implement. Carbonaro [8] reported a six-component balance that was used in the VKI Longshot tunnel and the balance was based on an acceleration compensation of strain-gauged signals. It involved the measurement of model accelerations using miniature sensors. But, measurement systems needed inertial corrections which had to be optimized taking into account the accelerometer transverse sensitivity effects, time lags, signal ltering or smoothing and tuning for the inertia of vibrating masses. In addition, stress wave force balance systems had been developed at the University of Queensland, Australia, [3,4] using strain gauges for measuring forces in hypervelocity ows of millisecond duration. The balance was based on interpretation of transient signals obtained from strain gauges and the complete model had to be suspended in the test section using inelastic strings. Moreover, mounting technique of strain gauges on the model sting will be more complicated and this method needs an exhaustive nite element analysis. The procedure for static calibration of strain gauges was very tedious and calibration had to be repeated whenever it was mounted on a new model. The errors introduced in the actual measurements due to static calibration procedure limit the use of this balance in short-duration facilities. Srulijes et al. [9] measured the aerodynamic forces on the body using piezo-electric pressure transducers. The individual values of the pressure signals were resolved in particular directions and then integrated over the model surface to get the total aerodynamic forces on the body. Hubner et al. [10] used pressure-sensitive paints to measure the step changes in pressure in a shock tube. The characteristic response times of the pressure-sensitive paints (PSP) was around 36 ms, which is too long a response time for the short duration test facilities. Lam and Stollenwerk [11] used the free-ight technique in a shock tunnel for the determination of forces and moment in a complex aerodynamic conguration. Optical instrumentation was used to monitor the model excursion. The model ight path was monitored

by tracking a target located at center of gravity of the model and aerodynamic coefcients were evaluated by monitoring the model ight path. Suspending the model and releasing mechanism makes this system very complex and moreover, launching a model at the proper time and orientation will dictate the level of accuracy during the measurements. Jessen and Gronig [12] developed a technique to measure the force components in impulse facilities with ow duration as short as one millisecond. Conventional force measurement techniques cannot be used in the shock tunnels (HST2) because of the short test time. To overcome the problem, the force measurement technique using accelerometers was proposed. The design of a three-component accelerometer balance at Cornell Aeronautical Laboratory [13], has been described by Vidal. Accelerometer balance system was used by Joshi and Reddy [14] and Raju and Reddy [15] to measure the aerodynamic forces acting on missile-shaped congurations ying at hypersonic Mach numbers in the hypersonic shock tunnel HST1. Although the above congurations have been used successfully to measure forces at a stagnation enthalpy of about 1 MJ/kg, elaborate calibration of accelerometer balance has not been carried out. Moreover, the signals were noisy and hence the data reduction procedure were rather complex. Recently, the high enthalpy research group at IISc developed miniature 3-component accelerometer balances [16,17] to measure drag, lift and pitching moment on various spiked blunt bodies, ying at a free stream Mach number of 5.75. But, none of these balances were tested at an enthalpy level more than 1.1 MJ/kg. The load acting on the model-balance system was small for those enthalpy. For high-enthalpy shots, the forces acting on a model-balance system will be sufciently large and as a result, rubber bushes used in the balance system may shear out. The free stream properties seen by the model-balance system at 2.2 MJ/kg of the present investigation are entirely different from 1.1 MJ/kg. In addition, they were not standardized with FEM which will be useful in verifying the accelerometer response and integrating the balance system with the model during the tunnel testing. In an effort to address this issue, Finite Element Analysis (FEA) is used in the present study for the design of the accelerometer balance system. In the present paper, a newthree component accelerometer force balance system is used to measure the aerodynamic coefcients for two different test models at an enthalpy of about 2 MJ/kg and Mach 8. In order to quantify the measured fundamental aerodynamic coefcients, the congurations of test model have been chosen to be very similar to AGARD models (HB1 and HB2). Here, the results of a shock-tunnel data for various angles of attack are compared with modied Newtonian theory, CFD and AGARD data for two different test congurations at a hypersonic Mach number of 8 in the IISc hypersonic shock tunnel HST2. The details of the balance design, its calibration

123

Force measurements over missile shaped body at Mach 8

427

and sample results along with the CFD and AGARD results are presented in this paper. 2 Facility description and instrumentation The shock tube (Fig. 1) has an internal diameter of 5.1 cm, length of 7.12 m and is divided into driver and driven sections by an aluminum diaphragm. A Mylar diaphragm divides the driven section from the wind tunnel, which is at even lower pressure (105 mbar). The shock wave velocity in the driven section of the shock tube can be measured by using two pressure transducers of 1,000 psi range (Model 113A24, PCBPiezotronics Ltd, USA), mounted 0.525 m apart at end of the driven section. The pressure behind the primary and the reected shock waves was measured using a pressure transducer (Model 113A22, PCB-Piezotronics Ltd, USA) located at end of the driven section. The wind tunnel portion of the HST2 shock tunnel consists of a truncated conical nozzle terminating into a 30 30 cm test section, which is attached to a dump tank of about 1m3 volume. The free stream Mach number could be varied in the reected mode by adding a convergent-divergent portion with different throat inserts, which produces range of Mach numbers from 8 to 13. All the experiments have been carried out only with the Mach 8 nozzle with a Reynolds number of one million per meter. The tunnel is capable of producing a reservoir enthalpy of 4 MJ/kg with an effective test time of 1 ms. All the data are sampled digitally, using a 12-bit multichannel system (NI PXI-6115 DAS, manufactured by NI Ltd) and data are recorded at 1 MHz/channel.

3 Pitot tube and the inviscid core The tunnel reservoir pressure history was measured just upstream of the nozzle. Pitot probe is less sensitive than the static probe to the ow alignment, which gives better accuracy even it is at higher yaw angle (up to 20 ). Hence, for the present investigation static pressure measurement was not used to determine the free stream Mach number. Therefore, the core ow in the test section has been characterized by using a pitot rake with three probes and the distance of 105 mm between the probes was selected based on test section width as shown in Fig. 2a. The chosen pressure tapping diameter was small enough that the pressure gradient over the surface of the tapping is negligible. The pitot pressures were measured with PCB type HM113A28 pressure transducers. The ratio of total pressure to pitot pressure was computed and then free stream Mach number could be easily determined from normal shock table. Pressure measurements were taken from the pitot rake mounted at various axial locations to determine experimentally the uniformity and size of the nozzle core. In this investigation, results havebeen highlighted only for the nozzle exit and measurement of other locations was discussed in detail elsewhere [18]. Figure 2b shows typical results of calibration tests at stagnation enthalpy of 2 MJ/kg. It is clear from the pressure trace that the nozzlestarting process takes about 0.6 ms. There exists a region of constant pressure after this starting process and then terminated by the arrival of a normal shock (i.e., disc shock at the nozzle exit). The mean pressure value during this constant period was used to evaluate the free stream Mach number. From the pitot pressure measurement, it is seen that the variation in Mach number is almost negligible across the nozzle. Similarly, when the rake was mounted 225 mm away from the nozzle exit along the axis of the test section it was found that the variation in Mach number was negligible over a distance of nearly 225 mm from the nozzle exit (results are not shown here). Indeed, for the current tunnel operating conditions, the viscous layer at the nozzle exit may occupy approximately one third of the exit radius. Therefore, the test jet is 200 mm in diameter and overall model lengths are limited to about 225 mm because of size constraints imposed by the test section dimensions and angle of attack. Hence, based on the above tunnel results, the location of model is chosen at exit of the nozzle and along with the model a single pitot tube is mounted permanently to measure the localized (point) pitot pressure of the airow (i.e., ow behavior). The Mach number distribution across the ow at two different axial locations in the tunnel is represented in the schematic diagram shown in Fig. 3. The variation in Mach number at exit of the nozzle is less than 1%, whereas it was 4.5%, when the Pitot rake was mounted 225 mm away from the nozzle exit. The pitot pressure measurement using a single probe (Fig 4), indicates the presence of a steady ow for about 1 ms, which in turn

Fig. 1 Schematic set-up and operation principle of a shock tunnel

123

428

S. Saravanan et al.

(a)

Fig. 3 Schematic diagram of Mach number distribution across the ow at two different locations in test section of hypersonic shock tunnel

2500

Stagnation pressure Pitot pressure x 50

2000

Pressure (kPa)

1500

(b) 40

(a) Flow establishment time (b) Test time and (c) Flow termination (a) (b) (c)

1000

30

Pitot pressure (kPa)

500

20

0 0 1 2 3 4

10

Average steady state test time

Time (s)
Pitot 1 Pitot 2(centre) Pitot 3

x 10

Fig. 4 The typical variation of pitot pressure recorded during the experiment along with the corresponding stagnation pressure variation

0 1 2 3

Table 1 Nominal test conditions in HST2 shock tunnel


4 5

Time (s)

x 10

M 8.0

P0 (Mpa) 2.01

H0 (MJ/kg) 2.01

(kg/m3 ) 0.0052

p (pa) 212

Re /m 1.05e6

T (K) 149

Fig. 2 a Schematic diagram of a pitot rake used for the nozzle calibration (Note that all dimensions are in mm) and b typical Pitot signals (pressure-time history) obtained at exit of the nozzle

is taken as the useful test duration of HST2 facility. Also, it is very clear that both the stagnation pressure in the shock tube and pitot pressure in the test section start decreasing after about 1 ms of test duration. The nominal ow conditions used in this study are summarized in Table 1.

4 Test model In the present investigation, two different models have been used, namely (a) blunt cone with after body and (b) blunt cone with after body and frustum, as shown in Fig. 5. The rst test model (blunt cone with after body) has been chosen

for its simple design and represents AGARD conguration, HB-1 (Hypervelocity Ballistic model). The rst part of the model is a blunt cone which has an apex angle of 41 and length of 40.6 mm. The second part of the model is the cylinder of 51 mm outer diameter and a length of 186 mm. The model has a spherical nose radius of 15 mm. The second test model (Blunt cone with after body and frustum) represents a simple AGARD conguration, HB-2. The geometry of the blunt coneremains the same and the length of the cylinder is reduced to 111 mm with the outer diameter of 51 mm. The frustum has an axial length of 75 mm with a semi-vertex angle of 5 . For both the congurations, the total length of the model has been kept the same in order to maintain the exact L/D (i.e., length to diameter) ratio of the model, such

123

Force measurements over missile shaped body at Mach 8

429

acceleration history within the observation window and verify the experimentally measured balance output with the simulated response. In addition, the specication of the accelerometer (i.e., sensitivity and range) and location of accelerometer with respect to center of gravity of the model have been chosen using nite-element analysis.

5.1 Finite element modeling and analysis The test model (nose cone-cylinder) exposed to hypersonic ow at 0 angle of attack is considered for FE analysis (FEA) to study the objectives described in previous section. Here, FEA has been focused to study only the drag component of the aerodynamic force experienced by the model by considering a drag accelerometer placed along the model axis and at a particular node. At the nodes, displacements are restrained along the bases of the rubber bushes. Axi-symmetric niteelement mesh was used for the test model. The nite element modeling consists of 9,237 three-dimensional 10-node elements which are tetrahedral in shape. Evaluation of a particular support arrangement and drag sensor location plays a vital role by providing the signal and retaining enough information about the frequency content of the loading. The accelerometers are treated as point sources with lumped mass and their masses are negligible compared to the model mass. In the present investigation, the pressure prole on the model surface in the experimental test ow domain can be obtained from the CFD technique using a commercial package CFX-Ansys 5.7. This code solves three-dimensional viscous ow NavierStokes equations in the conservative form. Viscous effects are incorporated in the numerical simulation. The ow is assumed to be laminar throughout the computational domain (The Reynolds number based on model length is estimated from the free stream properties and found to be 0.237 million). Boundary conditions used in this simulation are based on the experimental free stream conditions in the test section of the shock tunnel, used to solve the Navier Stokes equations. At the outlet of the computational domain, variables like pressure, temperature, density, force, wall heat ux etc., are extrapolated from the interior domain. At the wall, no-slip condition is used and can be treated as adiabatic. Other bounds of the computational domain, excluding inlet and outlet, are specied as symmetry planes to simulate axisymmetric ow situation. The uid velocity is assumed to be tangential at these symmetry planes. In the initial simulation, grid size of 9, 70, 474 is used for the computations of ow elds around blunt cone model with an after body and are. In the subsequent runs, the number of grid points are increased (grid independent study) and simulations are once again carried out to check variations in the static pressure. Figure 6 shows the surface static pressure predicted by three different meshes and zero on X -axis refers to junction

Fig. 5 Schematic diagram of test models; a 41 apex-angle blunt cone model with an after body, b 41 apex-angle blunt cone model with an after body and are for lifting purpose Table 2 Typical dimension of test models with AGARD Conguration Semi-apex angle (deg) Length to diameter

AGARD Saravanan et al. AGARD Saravanan et al. Cone-cylinder 22.5 Cone-cylinder- 22.5 frustum 20.5 20.5 4.9 4.9 4.44 4.44

that both the models can be accommodated onto the single force balance system. For a comparison of present models with AGARD, the typical dimensions of both the congurations are summarized in Table 2. The material used for the model was commercial duraluminum in order to reduce the mass.

5 Initial design of the accelerometer balance system An equivalent springmass system [13] was used in the preliminary design methodology for the accelerometer balance which was based on analytical solutions. Although accelerometer force balance have been used for many decades, there were no attempts to carry out the integrated analysis of the complete system to understand the effect of restraint characteristics on the measured acceleration history. Hence, the concept of nite-element analysis for the accelerometer balance system was introduced recently by Sahoo et al. [19,20] to study the dynamic behavior of the integrated model-balance system. The analysis was used for axisymmetric blunt-cone conguration. In the present analysis, this has been extended to three-dimensional modeling of integrated system for a cone-cylinder and cone-cylinder-frustum model congurations. The objective of the nite-element analysis is to evaluate the effect of rubber bushes on

123

430
9,70,474 elements
15000
30 25 20 15 10 5 0

S. Saravanan et al.

E=3 MPa

13,39,694 elements 33,19,220 elements

Static pressure, P (Pa)

10000

5000

0.05

0.05

0.1

0.15

0.2

Acceleration (m/s 2 )

Axial distance, X (m)

5 0

0.2

0.4

0.6

0.8

1.2

Time (s)

x 10

Fig. 6 Assessment of mesh convergence for a blunt-cone model with an after body and are

Fig. 7 FE simulated accelerationtime history for the integrated model-balance system with E = 3 MPa

of the bluntcone and cylinder of a frustum conguration. It is seen that there is no change in the static pressure, irrespective of grid size. These meshes produce very close matching throughout the model length. Finally, total number of around 13,39,634 nodes (grid points) has been chosen for the present study. Similarly, the other conguration (i.e.,cone model with an after body) was modeled using around 1.3 million nodes. A high resolution second order accurate scheme has been used for the computations with a local time step of 0.8. The target residuals to terminate the simulation have been set at 105 . About 600 iterations (time steps) have been used for the convergence and approximately 8 hours of CPU time was required for the simulation on an Intel Pentium 4, 2.8 GHz processor and 2 GB of RAM speed. The program was run on a Windows NT platform. The average pressure on model surfaces is preprocessed based on CFD technique and given as inputs to the nite element simulation (i.e., applied load). Linear transient dynamic analysis with uniform time step of 0.1 s (nite element analysis) is performed for 0 angle of attack using MSC/NASTRAN software. The result was analyzed using MSC/PATRAN. To start the FE analysis, property of the rubber material has to be known. The detailed parametric studies on stiffness of the rubber material had been carried out by Sahoo et al. [20] using FE analysis, which can ensure a sufciently soft suspension system (i.e., model-balance) to measure the aerodynamic forces within the observation time window. They have chosen aluminum and stainless steel as the model and balance materials, respectively, and nite element analysis were carried out to quantify the quality of the rubber bushes (i.e., Youngs modulas, E). They found that, for rubber bushes with higher values of E (>3 MPa), the stresswave induced oscillation becomes more predominant and also that the peak acceleration as well as the mean acceleration drops down drastically during the observation test time window. It was found that rubber bushes with 3 MPa

resembles the response of a free-oating body. Hence, in the present investigation we have used the same property of the rubber bushes in designing the balance system. FE analysis has been carried out for an integrated cone-cylinder with balance system and the corresponding accelerationtime history is shown in Fig. 7. It is found that there is no drop in the mean acceleration amplitude (Fig. 7) and at the same time, the rubber is exible to hold the model-balance system. Based on the FE analysis result, we can conclude that the stiffness of rubber material plays an important role on the sensitivity of the integrated system and we were able to select an accelerometer with a sensitivity of 10 mV/g and range of 500 g. 5.2 Calibration of the accelerometer force balance In general, any measuring system needs a calibration. The calibration of accelerometer has also been carried out by using a standard accelerometer calibration kit (PCB Piezotronics Type 086C01), which has a Piezo crystal oscillating at 1 g with a frequency range of 9,500 Hz. The accelerometer output was assumed to be linear with the applied acceleration over the range up to 500 g. But it is advisable to calibrate the Piezo crystal accelerometers before the experiments since harsh environmental changes and temperature variations will produce drifts in the sensitivity of accelerometers. A negligible difference has been found between the sensitivity values obtained from the calibration and the sensitivities given by the manufacturer. Hence, for the present study, the calibrated accelerometer sensitivities (8.4510.15 mV/g) have been considered for the computation. Different methods are available in practise to evaluate the system response function or transfer function which includes, ball impact, wire cutting and impulse hammer. But, for the

123

Force measurements over missile shaped body at Mach 8

431

present investigation, we have used an impulse hammer to get the transfer function. It is typically a linear dynamic system connecting input (applied load) and output (acceleration signal). In order to obtain transfer function, the model has to be tted with an accelerometer force balance system which is in turn mounted on a support system, very similar to shock tunnel testing. A gentle hit can be given at the nose portion of a test model using PCB impulse hammer. During the calibration, importance has been given to the drag component and hence, force signal generated by the hammer and drag acceleration signal from accelerometer balance system were recorded. These signals are used to determine the transfer function using a Fourier transform technique. Deconvolution is carried out in the frequency domain. The obtained transfer function from the calibration experiments are veried by means of recovering the force signals of either from same test or different hammer test. Hence, the transfer function [21] can be successfully used to measure the actual aerodynamic drag force acting on a model during the shock tunnel testing using the measured model acceleration from the balance system.

accelerometers (PCB, model 303A02), two exible rubber bushes and two stainless steel rings on which the model is suspended. The annular rubber bushes are xed between the steel rings and the central sting. In the balance system, two accelerometers are used to measure the acceleration in the normal direction and another accelerometer is mounted along the axis of the model to measure the drag force. The drag accelerometer is threaded directly into the nose portion of the model. Accelerometers whose frequency ranges were up to approximately 10 kHz were used. The normal force N (t), center of pressure location from base of the model (e N ) and axial force C(t) on the model are calculated using the measured accelerations from the following equations [22]: N (t) = m eN = I m b N1 + a N2 a+b N1 N2 b N1 + a N2

(1) (2) (3)

C(t) = m A 6 Force measurement technique and accelerometer balance for hypersonic ows A schematic diagram and photograph of the newly designed 3-component balance system is shown in Figs. 8a,b. This was used to measure the fundamental aerodynamic coefcients on two different test models. The balance is made out of stainless steel and consists 3 fast response piezo electric

where, N1 , N2 and A are the measured accelerations, a and b are the locations of the accelerometers from the center of gravity of the model, m is the weight of the model and I is the mass moment of inertia of the model. By knowing the angle of attack and center of pressure location from base of the model, the fundamental aerodynamic coefcients can be computed by using the following equations. Cl = Cm = Cd = N (t) cos q S N (t) q S C(t) sin q S

(4) (5) (6)

xcg + e N DB N (t) sin q S

C(t) cos + q S

where, xcg is the location of center of gravity from the base of the model, q is the free stream dynamic pressure, S is the reference area, D B is the base diameter of the cylinder and is the angle of attack.

7 Results and discussion The mass, moment of inertia of the model and center of gravity from the model base are estimated. The details of total mass, center of gravity location and mass moment of inertia for the two congurations of test model are given in Table 3.

Fig. 8 a Schematic illustration of the three-component accelerometer balance system and b Photograph of the three-component force balance system along with miniature accelerometers used for the present study

123

432 Table 3 Centre of gravity location and mass moment of inertia for different model congurations Model conguration Mass (kg) c.g from base of the model X c.g (mm) Mass moment of inertia of system, I (104 kg-m2 ) 10.816 4.181

S. Saravanan et al.

Blunt cone-cylinder Blunt cone-cylinderfrustum

0.365 0.43

115 72

7.1 Force measurements The results of FE simulations are compared with the experimentally measured acceleration from the drag accelerometer at a stagnation enthalpy of 2 MJ/kg with ow Mach number of 8, in Fig. 9. The response shown in Fig. 9, has initial transients similar to the input step load followed by local oscillations due to the mass distributed and constrained dynamics. The values are plotted for a test window of 1.4 ms as shown in Fig. 9. The plots show moderate agreement between the ltered response from FE simulations and the measured balance output in the steady region within the observation window. The deviation between the experiment and FEM is 17.4%. Based on the results shown in Fig. 9, this integrated model-balance modeling is considered to be suitable for the shock tunnel investigation. It is clear that the magnitude of the simulated acceleration amplitude remains unchanged with a gradual ow development over the model and matches with the measured balance output in the steady region within the observation window. It shows that we have chosen a location in the FEM simulation, where we get the maximum amplitude of the signal which will in turn essentially give the location of accelerometer along the model axis. Similarly, analysis
30 25

can be carried out for normal force accelerometers with respect to c.g of the model and for the present investigation, simulation was carried out only for the drag signal. From the results of FEM, we are able to choose the location of drag accelerometer. Two to three experiments for each angle of attack and model conguration are carried out to check the repeatability of the signals. The typical acceleration response obtained from the accelerometers mounted in the three-component balance system and in the model are shown in Fig. 10. Timehistory of pitot pressure is also shown in the same gure and it is seen that time history of force closely follow the time history of pitot pressure. It is seen that the output of the accelerometer balance system shows a gradual rise for about 0.5 ms and remains approximately constant for the test period of 0.4 ms. Good repeatability has been found during the shock tunnel testing. Since the pitot and drag sensors are located at the same point, the response time of the drag accelerometer matches well with the reference pitot signal whereas the lift accelerometers are mounted apart from the drag accelerometer [i.e., on either side of c.g of the model (Fig. 8)] and hence, the response of normal force accelerometers will not be the same as the drag accelerometer (shown in Fig. 10). Aft lift accelerometer remains constant for a period of approximately 0.4 ms. Time delay in measuring the normal force acceleration (i.e., response of the accelerometers) was estimated and found to be about 20 and 225 s, respectively, with respect to drag signal. However, this time delay can be theoretically estimated by knowing the location of accelerometers and ow velocity. All the experimental data of both the congurations are deduced by averaging results during the steady ow duration of 0.4 ms, whereas 1 ms test time is obtained from the pitot pressure measurement mounted in the test section. The
drag
1

FEM Experimental

front lift aft lift pitot

steady state test time

Drag acceleration (m/s2)

20

Voltage (Volts)

0.5

pitot

15 10 5 0 5

steady state test time of 0.4 ms


0.5 2
0 0.2 0.4 0.6 0.8 1 1.2

2.5

Time (s)

x 10

Time (s)

x 10

Fig. 9 Comparison of the measured drag accelerometer response with the FE simulated acceleration-time history with step inputs

Fig. 10 Typical signals from various accelerometers mounted in the model and accelerometer balance system for the test model ying at Mach 8

123

Force measurements over missile shaped body at Mach 8

433

discrepancy in test time may be due to hitting of either particle (i.e., primary diaphragm), or paper diaphragm on surface of the model. The deconvolved drag signal is plotted with the measured drag value. It can be clearly seen from Fig. 11 that the measured drag signal contains irregular and highfrequency noise and hence there is a necessity to nd out the dominant frequency of the ow. This dominant frequency is estimated by using a pitot pressure measurement and an efcient algorithm of fast Fourier transform (FFT) which is found to be 36 kHz. Hence, in order to get the useful data, a 10 kHz low pass lter has been used in the analysis and Fig. 10 shows the ltered drag signal. The deconvolved drag signal is recovered successfully (Fig. 11). The experimental values are compared with the theoretical values of aerodynamic coefcients at angles of attack by using modied Newtonian theory [1], taking into account of centrifugal forces. This theory can often be used analytically to provide reasonable estimation of the local pressure. Furthermore, one can easily integrate the pressure distributions to obtain adequate estimates of the forces and moments. The drag contributed by the skin friction component is considered in the modied Newtonian theory whereas base pressure is not included. However, base pressure could be obtained from CFD simulation. The value of base pressure from numerical simulation is approximately 1015 pa for both the congurations which is too low compared with the wave drag (i.e., pressure experienced at nose of the vehicle is 15 kpa, as shown in Fig. 6), due to the formation of shock waves. Hence, in the present investigation, base pressure is neglected for both the congurations. These results are shown in Figs. 12 and 13. For better estimates of the force coefcients, it is necessary to use CFD codes. Hence, the measured values are also compared with numerical values computed using a commercial CFD code (CFX- Ansys 5.7) and AGARD data [23]. The error band for the measured data is estimated for
30 20

both the congurations and shown in Figs. 12 and 13. The experimental results for the cone-cylinder model are in good agreement with theory as compared to frustum congurations. The measured drag coefcient (frustum conguration) value is less than theoretical value over the ranges of angle of attack. This difference could be probably due to the body shock which does not closely envelop the body and as a result the

(a)

0.7 0.6

Drag coefficient, C

0.5 0.4 0.3 0.2 0.1

Theory Expt. CFD AGARD

10

Angle of attack (degrees)

(b)
0.6 0.5 0.4 0.3 0.2 0.1 0 0.1 0

Lift coefficient, Cl

Theory Expt. CFD AGARD

10

Angle of attack (degrees)


raw drag signal calibrated signal

(c)
Pitching moment coefficient, Cm
2

1.5

Theory Expt. CFD AGARD

Acceleration ms 2

10 0

10 20 30
2 4 6 8 10 12 14

0.5

0 0 5 10

Angle of attack (degrees)

Time (s)

x 10

Fig. 11 Deconvolved signal from the drag accelerometer

Fig. 12 Variation of aerodynamic coefcients for the 41 apex angle blunt cone model with after body ying at Mach 8 for various angles of attack: a drag coefcient, b lift coefcient and c pitching moment coefcient

123

434

S. Saravanan et al.

(a)

0.8 0.7

Drag coefficient, Cd

0.6 0.5 0.4 0.3 0.2 0.1 0

Theory Expt. CFD AGARD

10

Angle of attack (degrees)

(b) 0.7
0.6

Lift coefficient, C l

0.5 0.4 0.3 0.2 0.1 0

Theory Expt. CFD AGARD

0.1

10

Angle of attack (degrees)

(c)
Pitching moment coefficient, C
1.6 1.4 1.2 1 0.8 0.6 0.4 0.2 0 0
m

Theory Expt. CFD AGARD

10

Angle of attack (degrees)

characteristics in relation to various parametric changes, like are angle and are length. It is also noted for both the congurations that the scatter in measured values of Cl and Cm are due to the differences in center of pressure location and the effect of measuring error with respect to low values. In Fig. 12, it is noted that increase in normal force is not exactly linear with increasing angles of attack, whereas it was the drag force remains almost same. As angle of attack increases, the pressure acting on windward surface increases which will produce an increase in the normal force. For a chosen conguration of test model, we nd that the center of gravity is aft of center of pressure and hence, the vehicle is statically unstable. It means that that any perturbation which produces an increased angle of attack causes a nose-up (positive) pitching moment which further increases the angle of attack. As a result, the pitching moment coefcient exhibits a non-linear trend with angle of attack. This non-linear behavior is due to unstable conguration. However, both the axial and normal force contribute to the pitching moment. Thus, it is possible that the vehicle is statically stable when X c p = X cg . The variation of lift coefcient with angle of attack for both the congurations at Mach 8 is presented in Figs. 12 and 13. The variation of Cl is essentially linear with angle of incidence for both conguration and it can be clearly seen for angles less than 8 . At 8 angle of attack, where the ow on the leeward side of a body separates, ow models (computed) usually fail to represent the actual ow and hence, they lie below the measured values. The initial slope of the curve for the ared body is practically one and half times the values produced by simple cone-cylinder conguration. Similarly, the data for pitching moment coefcient is shown in the Fig. 13. The gure shows initially a non-linear trend at small angles of attack which is characteristic of the differential movement of boundary-layer reattachment around the are circumference and for higher angles of attack, it exhibits linear trend. The addition of ared tail is to produce a stabilizing moment at all angles of attack and the curve shows the linear trend which is a good indication of stable conguration. 7.2 Measurement uncertainty The uncertainty analysis has been carried out based on Mee [24]. The estimated uncertainties in the measured data are, Cd /Cd = 0.0783, Cl /Cl = 0.0806 and Cm /Cm = 0.0953. Uncertainties in the sensor sensitivities, setting of angle of attack, free stream conditions in the tunnel test section, the restraint offered by the rubber bushes to the free ight of the model during test time, outputs of the accelerometer balance system and data acquisition systems are some of the factors that contribute to the above force measurement uncertainties.

Fig. 13 Variation of aerodynamic coefcients for the 41 apex angle blunt cone model with after body and frustum ying at Mach 8 for various angles of attack: a drag coefcient, b lift coefcient and c pitching moment coefcient

Newtonian theory is not accurate. Indeed, the prediction of axial force coefcient is marginally better at lower angles of attack and difference becomes more when the angle of attack is increased. In addition, the are is known to act as a wedge near its junction with the cylinder and as a result, the ow changes to conical along its length. However, this theory can be used to provide a basis for the trend of basic aerodynamic

123

Force measurements over missile shaped body at Mach 8

435 7. Naumann, K.W., Ende, H., Mathieu, G., George, A.: Millisecond aerodynamic force measurement with side jet model in the ISL shock tunnel. AIAA J. 31-6, 10681074 (1993) 8. Carbonaro, M.: Aerodynamic force measurements in the VKI longshot hypersonic facility. New trends in instrumentation for hypersonic research, pp. 317325 (1993) 9. Srulijes, J., Gnemmi, P., Runne, K., Seiler, F.: High-pressure shock tunnel experiments and CFD calculations on spike-tipped blunt bodies. AIAA Paper 20022918 (2002) 10. Hubner, J.P., Carroll, B.F., Schanze, K.S., Ji, H.F.: PressureSensitive paint measurements in a shock tube. Exp. Fluids 28, 21 28 (2000) 11. Lam, L.Y., Stollenwerk, E.J.: Aerodynamic force measurement of free-ight models in a hypervelocity shock tunnel. AIAA 1966 0771 (1966) 12. Jessen, C., Gronig, H.: A six component balance for short duration hypersonic facilities. New trends in Instrumentation for hypersonic research, pp. 295305 (1993) 13. Vidal, R.J.: Model instrumentation techniques for heat transfer and force measurements in a hypersonic shock tunnel. Cornell Aeronautical Laboratory Report, WADC TN 56-315 (1956) 14. Joshi, M.V., Reddy, N.M.: Aerodynamic force measurements over missile conguration in IISc shock tunnel at Mach 5.5. Exp. Fluids 4, 338340 (1986) 15. Raju, C., Reddy, N.M.: Aerodynamic force measurements over missile conguration in IISc shock tunnel at Mach 3.85 and 9.15. Exp. Fluids 10, 175177 (1990) 16. Viren, M., Saravanan, S., Jagadeesh, G., Reddy, K.P.J.: Experimental investigations of hypersonic ow over highly blunted cones with aerospikes. AIAA J. 41-10, 19551966 (2003) 17. Viren, M., Saravanan, S., Reddy, K.P.J.: Shock tunnel study of spiked aerodynamic bodies ying at hypersonic Mach Number. Shock Waves. 12, 197204 (2002) 18. Saravanan, S.: Experimental investigation of the effect of nose cavity on the aerothermodynamics of the missile shaped bodies ying at hypersonic Mach numbers: Ph.D. Thesis, Indian Institute of Science, Bangalore, India (2007) 19. Sahoo, N., Mahapatra, D.R., Jagadeesh, G., Gopalakrishnan, S., Reddy, K.P.J.: Aerodynamic force measurements on 60 apex angle blunt cones ying at Mach 5.75 using a 3-component acceleromeer balance: AIAA Paper-2002-5204(2002) 20. Sahoo, N., Mahapatra, D.R., Jagadeesh, G., Gopalakrishnan, S., Reddy, K.P.J.: An accelerometer balance system for measurement of aerodynamic force coefcients over blunt bodies in a hypersonic shock tunnel. Meas. Sci. Technol. 14, 260272 (2003) 21. Mee, D.J.: Dynamic calibration of force balances for impulse facilities. Shock Waves 12-6, 443456 (2003) 22. Reddy, N.M.: Aerodynamic force measurements in the hypersonic shock tunnel. In: Proceedings of 14th International Symposium on Shock Waves, p. 358 (1983) 23. Ashby, G.C. Jr., Cary, A.M. Jr.: A parametric study of the aerodynamic characteristics of nose-cylinder-are bodies at a Mach number of 6.0 : NASA TN D-2854 (1965) 24. Mee, D.J.: Uncertainties analysis of conditions in test section of the T4 shock tunnel. University of Queensland, Centre for Hypersonics Research Report No 4/93, Australia (1993)

8 Conclusions The following are the main observations drawn from this work. We have developed a three-component accelerometer force balance system for short duration testing facilities and used it to measure the aerodynamic force coefcients for two different test models at ow Mach number of 8 in the HST2 hypersonic shock tunnel. The accelerometer force balance has a number of advantage in comparison with classical conventional balances for shock tunnel testing. By providing a are to the basic model conguration (blunt cone with afterbody), the contribution to Cn is quite signicant and in addition, the model is more statically stable. Investigations have shown that drag coefcient remains same for both the congurations at higher angles of attack whereas it decreased about 84% for zero and lower angles of attack for frustum model conguration. For moderate angles of attack, the lift coefcient of frustum increased by 8.230% compared with basic conguration. For a pitching moment, frustum conguration shows a gradual decrease at lower angles of attack and at higher angle of attack, it decreased by 34%.
Acknowledgments The support rendered by Dr. Reddeppa, Mr. Mahapatra, Mr. Vinayak Kulkarni, Mr. K. Nagashetty and Mr. Rajagopal during the course of this work is gratefully acknowledged. We would like to thank Mr. Ramesh Babu for his patience and care in the drawing.

References
1. Truitt, R.W.: Hypersonic Aerodynamics. The Ronald Press Company, New York (1959) 2. Stalker, R.J.: Development of a hypervelocity wind tunnel. Aeronaut. J. 76, 374384 (1972) 3. Sanderson, S.R., Simmons, J.M.: Drag balance for hypersonic impulse facilities. AIAA J. 29, 21852191 (1991) 4. Mee, D.J., Daniel, W.J.T., Simmons, J.M.: Three-component force balance for ows of millisecond duration. AIAA J. 34-1, 590 595 (1996) 5. Storkmann, V., Olivier, H., Gronig, H.: Force measurements in hypersonic impulse facilities. AIAA J. 36-3, 342348 (1998) 6. Takahashi, M., Komuro, K., Itoh, K.H.T., Ueda, S.: Development of a new force measurement method for scramjet testing in a high enthalpy shock tunnel. AIAA Paper 19994961 (1999)

123

You might also like