P ~ ~ U I I I U U ~ I I ~ I I I Istics a Char ~ m m m ~vloael

c rculation C ontrol Rotor

Charles B. Watkins Professor Howard University Washington, D.C.

Kenneth R. Reader Senior Aerospace Engineer David W . Taylor Naval Ship Research and Development Center Bethesda, Md.

Subash K. Di~tta Graduate Assistant Howard University Washington, D.C.

Numerical and emcl.mm.sm.~. nulation of unsteady almow throueh the control valve and slotted air duct of .I. a circulation control rotor are described. The numerical analysis involves the solution of the quasi-one-dimensional compressible fluid-dynamic equations in the blade air duct together with the coupled isentropic flow equations far flow into the blade through the valve and out of the blade through the Coanda slot. Numerical solutions are comoared with basic emerimental results obtained for a mockup of a circulstion control rotor and its pneumatic valving system. The pneudynsmic phenomena that were observed are discussed with particular emohasis a n the characteristic system time lam associated with the response of the flow variables to transient and periodic control valve inpub

Notation
A
cross-sectional area of duct A, = effective expansion area at valve exit A, = valve area cr = duct friction factor C, = slot discharge coefficient C, = valve discharge coefficient c, = specific heat at constant volume DM = hydraulic diameter of duct = total energy per unit volume e F = flux times area vector G = vector of nonhomogeneous terms in flow equations h = heat transfer coefficient H = average duct height , k = thermal conductivity L = length of duct riz = mass flow rate m = mass flow rate per unit area in duct m. = mass flow rate oer unit area com~uted from isentropic ~~, flow theory m, = mass flow rate per unit area through slut m, = mass flow rate per unit area through valve p = static pressure in duct p,, = external pressure at slot exit po = total pressure in duct
=

= duct static temperature To = duct total temperature T, = duct wall temperature t = time U = vector of dependent variables V = vector of diffusion terms v, = average duct velocity in spanwise or x direction v, = approximate average duct velocity in chordwise or y direction w, = slot width x = spanwise coordinate x, = coordinate at duct entrance x,., = coordinate at duct end y = chordwise coordinate y = ratio of specific heats p = density = angular velocity Subscripts I = variable computed from isentropic flow at valve exit p t = variable in pressure supply plenum or rotor hub Prefix 6 = peak-to-peak value of variable

Introduction
IRCULATION control (CC) rotor technology, applied to rotary-wing aircraft or stopped-rotor vertical takeoff and landing (VTOL) aircraft such as the X-Wing, offers distinct advantages over conventional rotor technology. For example, CC technology provides a straightforward means of imple23

Presented at the Second Decennial Specialists' Meeting on Rotorcraft Dynamics, NASA Ames Research Center, Moffett Field, Calif., Nov. 1984

C.B. WATKMS

1 IURNAL OF THE AMERICAN HELICOPTER 0

mentlng higher harmonic control to suppress vlbrattlon and blade stresses. Both types of aircraft use a shaft-driven rotor with blades having circulation control airfoils which generate lift through the Coanda principle. These CC airfoils employ arounded trailing edge with a thin jet of air tangentially ejected from a spanwise slot immediately above the rounded (Coanda) surface. The iet of air suppresses boundarv laver separation and moves the r k r stagnatidi streamline toward ihe lower surface, thcrcby increasing lift. I.ift is increased with increasing mass flow ratc of compressedair in the jet. Thc remaining control requirements arc ohrained by cyclic modulation o i the mass flow ratc with valves in the nbnrbtating system. Higher harmonic cyclic control can he similarly applied for reducing blade stresses and vibration. The design of CC rotors and their pneumatic control systems requires anunderstanding and apprkciation of thc phcnonlena lnvolvcd i n thc control and distnhution of airflow to the Coendn slots in these rotors. A capability for analytical prediction of rotor Coanda airflow is also essential. The term "pneumodynamics" has been coined to refer to the aerodynamic response characteristics of CC rotor system internal ducting airflow. In a recent paper, Watkins, et al.' describe the modeling techniques employed in the high frequency pneumodynamic analysis (HFPA) computer code developed for CC rotor pneumodynamic analysis. The present paper reviews the modeling techniqucs described in Ref. I and presents results obtained by applying the code to a simple experimental mockup of a model rotor hlade and its pneumatic valving system. The discussion of the results, from both the HFPA predictions and the experiment, focuses on the dynamic response of the flow variables to periodic and transient control valve inputs. These response characteristics, particularly the characteristic system time lags, have implications for the performance of actual rotor systems.

The vector of nonhomogeneous terms. which inctuaes tne $kinfriction effects of blade rotation at angular veloc i mass flux factor cAx), wall heat transf er coefficie through the slot m,(x, t) is

1
For completeness, the vector of axial diffusion terms is included although it is insignificant in the present problem. It is

ro

The fluid dynamics equations, Eqs. (2) are presented here in conservation law form since this is the form employed for the numerical solution algorithm. The equation of state is

Theory
The modcling techniques cmploycd in the prcscnt research are descrihed in Rcf. I where cmphasis is on the devails of the theoretical formulation and numeAcal technique. The theory is presented here in abbreviated form; a more complete discussion 1s contained in Ref. 1. Basic Equations The HFPA code solves the quasi-one-dimensional, compressible flnid-dynamic equations for unsteady flow in a spanwise blowing air supply duct internal to a rotating CC rotor hlade, together with the coupled isentropic flow equations for flow into the hlade through the control valve and out of the hlade through the Coanda slot. The continuity, momentum and energy equations for flow in the duct have the conservative form representation

In the above relations, the conservative variables m and e are defined as m = pv,

and w,(x, t) and m,(x, I ) are the local (with respect to spanwise x location) slot width and blowing mass flux, respectively. The small average chordwise velocity component inside the duct can be crudelv anoroximated as the mean of the value of , chordwise velocity for flow in the duct near the slot (as determined from continuitv) and the value of zero at the wall ooposite the slot. ~ e n $ ,

..

where the vector of dependent variables is

in which HD(x) is the local effective height of the duct. The mass flux distribution through the Coanda slot is assumed to he well represented by one-dimensional isentropic flow theory with a discharge coefficient C,(x);

and the flnx times area vector is The above functional relationship is the usual expression for isentropic flow expanding through a nozzle from a plenum (in this case the duct interior) to ambient conditions (in this case the exterior of the rotor hlade). po(x, t) and TO(& t ) are the local (internal) total pressure and temperature, respectively, and p,,(x, t) is the external pressure at the exit of the Coanda slot. Dependency upon external pressure is eliminated when the flow is choked.

CIRCULL lrounoary LOB At the upstream boundaryx = x,, aplenum supplies blowing air to the duct through a valving system. The valve opening area has constant and rotor-azimuth-dependent components regulated by collective and cyclic components, respectively, of the aircraft control system. The mass flow and duct pressures are thus determined by these valve settings in concert with the ductlslot configuration. Figure 1 shows the idealized representation adopted for the cam type valve of interest in the present study. The controlled value of the valve opening is designated as A,@), and A, is the fixed effective opening at the duct entrance. Like the slot flow, one-dimensional isentropic flow theory with a discharge coefficient is used to represent flow through the valve. The valve mass flux is mu@)= Cvm,@oPt. Toptp P(x,, f) )
(5)

TOR MOD1

In Ref. 1 merical upstream boundary condition based on deriving for various valve types an auxiliary relationship, by ignoring the time dependent term in the momentum equation between the first two grid points (grid point 1 and 2 in Fig. 1). The approximation for the idealized valve of Fig. 1 is obtained from the steady momentum equation as

where po and To are the total pressure and temperature, respectively, in the air supply plenum in the rotor hub and p(~,, 0 is the back pressure in the duct entrance downstream of the valve. When the flow is choked, the dependency onp(x,, t ) is eliminated. The other boundary condition, a t x = lend,is the stagnation of the flow at the end of the duct. The discharge coefficients C, and C,, used as correlation parameters for a given system, relate the geometric area of a system to the effective area. For simple configurations such as orifice plates or venturi tubes in a well conditioned system where the airflow is well behaved, the discharge coefficient value can be obtained from a standard engineering handbook. Finite-Difference Method In applying the theory of the previous section to the analysis of a rotor blade, the entire blade, including the transition duct (ducting from hub valve exit to airfoil portion of rotor blade), is discretized by dividing it into spanwise (radial) segments. The governing differential equations, Eq. (I), are solved by the implicit, "delta form,'' finite-difference procedure of Beam and Warming2 at the grid points located at segment boundaries, including the valve exit as shown in Fig. 1 and the duct end. The solution of the duct flow is actively coupled to the flow through the valve by means of the upstream boundary condition. With the exception of the upstream boundary condition for flow into the duct from the valve, the numerical analysis is rather standard and is described fully in Ref. 1.

Approximations such as in Eq. (6) were applied because the investigators were unsuccessful in applying the more standard and less approximate procedures for inlet boundary conditions of applying one-sided differencing or the method of characteristics. The presence of the valve poses special difficulties in the application of these standard techniques. In a recent communication, Anderson of United Technologies Research Center indicated that he was successful in applying the method of characteristics to derive inlet boundary conditions for the HFPA code for a sudden expansion or gate-type valve. The extension of his procedure to the cam valve type shown in Fig. 1 is not straightforward due to the 90' change in flow direction. Since this valve is the type considered in the present investigation, Eq. (6) was retained. Numerical Solution Procedure The basic approach followed in obtaining numerical solutions to control inputs, consisting of constant (collective) settings or constant settings with superimposed constant amplitude cyclic inputs, is to obtain the steady state or time periodic solutions by allowing them to evolve from a transient. The solution starts with the system essentially at supply pressure conditions. The initial valve position is assumed to be fully open as specified by the sum of its Fourier components supplied as input. The solution is then advanced with time as the flow in the blade adjusts itself to reach steady state flow conditions for the fully open valve. If cyclic components are to be considered, the cyclic input is imposed after fully open steady state flow conditions are achieved. The solution is thereafter advanced further with time until steady, periodic conditions are attained. A harmonic analysis of pressure, temperatures, and mass flow rates is performed on the results from the final cycle.

Experiment . The analytical formulation and the numerical methods described in the previous sections were validated by applying them to predict the results of a basic experiment3 The experiment was conducted using a mockup of a rotor blade pneumodynamic system incorporating a valve similar to the idealized valve of Fie. 1. The valve modulated the flow into a suhscale arculation control rotor hledc represented hy a stationary pluggd pipe with 3 lcn~thwisc slot. For sirnnlicitv. the correct rclativc motion of an actual rotor blade pnkumddynamic system was established by interchanging the normally rotating and nonrotating components. In the model, as shown in Fig. 2, the pipe was stationary and rotary motion was employed in the valving system. The model valveconsisted of astationq plenum supplying air to the stationary pipe inlet nozzle, which was partially obstructed by the surfaces of two rotating cams mounted contacting each other on the same shaft. The shaft is mounted perpendicular to the nozzle axis and the effective nozzle area formed by the area between the nozzle entrance and the cam surfaces varies as the cams rotate. The two cam surfaces each

AJt)
Fig. 1 Idealized valve

F THE AMI
.

lLICOPTEF

,.", =;

MLL

"nmcn"=,","=

ONE.

V CAM

ARE I N

2.00 DIA

ONE-PER-REV CAM TWO-PER-REV CAM

0.k

Fig. 2 Experimental model

have dierent wavy profiles, one with a single waveform amund the cam periphery, and the other with two distinct waveforms amund the cam periphery. Therefore, either a one-per-(cam) revolution or two-per-(cam) revolution nozzle area variation is achieved depending on which cam surface is positioned over the nozzle entrance. The experiment was originally performed3 to establish the feasibility of modulating the weight flow in a one-per-rev and two-per-rev manner and to show that the collective and cvclic components of the airflow were additive. The pipeincorpo;ated several pressure taps distributed alone its leneth. Dvnamic mea" suremeits of pressure were obtained for a range of cam rotational speeds. In addition, quasi-dynamic results were ohtained by positioning the cam at discrete increments of 30". Average mass flow data were collected from a venturi tube located upstream of the plenum. Figure 3 shows the measured valve area formed by the oneper-rev and two-per-rev cams spaced at a minimum distance of 0.01 inch from the nozzle inlet. The points shown were calculated from the measured gap between the cam profile and the nozzle; the area variations are intended to approximate sinusoids as shown. The two-per-rev cam profile is a rather crude representation and contains a significant higher harmonic content due to the deviation. The measured values were used in the numerical predictions.

I -

0
0

w
0
N N

REFERENCE SlNUSOlD MEASURED

TWO-PER-REV C A M
0.4

AZIMUTH POSITION (deg)

Fig. 3 Valve control area

Results and Discussion

Numerical results were obtained for the experimental configuration of Fig. 2 by discretizing the pipe into 16 segments of approximately equal length. The pipe was assumed to be adiabatic, and the friction coefficient was represented by the formulas for fully developed flow. The integration time-step for the calculations simulating the quasi-dynamic experiments sec. For the dynamic calculations, valve cycling was 5 x was imposed after an elapsed time of 0.25 sec. To ensure stability, it was found nccc<saryto reduce the time-step to 1200 the steps per revolution d u r i n ~ dvnaniic mode calculations. In general, the results were iualita<vely similar for the one- and two-per-rev cams. Therefore, the results presented here are, with one exception, for the one-per-rev cam since it is the closest to a pure sinusoid. Typical quasi-dynamic numerical and experimental results, indicating the dependency of mass flow rates on nozzle conhol area at a constant slot height of 0.042 in., are displayed for the one-per-rev cam in Fig. 4. In this figure the mass flow rate is presented as a function of total valve area at a set plenum pressure. As the valve area increases the data show that mass flow rate tends toward a constant value, indicating that the flow has become controlled by the slot opening. Figure 4 shows the sensitivity of the discharge coefficients C, and C, for a wide range of valve areas. Also shown is the effect of assuming constant values of 0.95 and 0.75 for both

D-- EXPERIMENT
0-..- COMPUTATION lCv'Cs=O.9SI *COMPUTATION lCv=Cs=0.751

0.000 1 0.0

0.4

0.8

1.2

1.6

2.0

VALVE AREA

Fig. 4 Mass flow variation

, ,nd slot discharge coefficients in . ..., , .u. t b.. A . lution. The results obtained for the higher value C, = C, = 0.95 agree more favorably with the experimental results, although the lower discharge coefficient is likely to be realistic physically. That a concept as simple as a discharge coefficient can be used to correlate the mass flow and pressure losses is very encouraging; this is discussed in depth in Ref. 4. For the remainder of the numerical results C = C = 0.95 were used , , without trying for a better correlation. As indicated in Ref. 1, strict interpretation of C, and C, is probably not desirable since the various deficiencies and approximations in the computer model can, to some degree, be absorbed by adjusting them. Moreover, the concept of a constant nozzle discharge coefficient is only a first approximation for the actual flow losses which occur over a range of mass flow rates in the analysis. Recent experience with the HFPA code indicates that more physically realistic discharge coefficients can be ohtained at the expense of decreased resolution of wave reflection phenomena by including second-order numerical damping in the HFPA algorithm. Figure 5. illustrates the numerical and experimental cycle variation of total pressure for a typical dynamic case. The numerical results of Fig. 5 were obtained with C = C, = , 0.95. The numerically computed pressure profiles have an azimuthal phase lag from the valve area profiles of Fig. 3 of approximately 35". For this configuration, the phase shift is principally due to the finite speed of wave propagation and is referred to as "sonic la^." A smaller portion of the delay is due to the "capacitancelag" effect ca"scd by the finite pipe volumc. To facilitate comparisonof the profile shapes the phase difference between the vdve m a setting and thepressuie response to it has been eliminated from both curves in Fig. 5 hccause the experimental phase angle results were not accuite. Table I contains data ohtained from a hdrmonic analysis of the dynamic numerical results of the case for which the pressures are plotted in Fig. 5. The numerical phase lag has been retained. A harmonic analysis of theexperimental pressure data obtained from the experimental curve presented in Fig. 5 is also shown for comparison, but without phase shift since, as mentioned, an accurate determination of experimental phase shift is not available. Magnitude and phase for the valve opening area, total mass flow rate, and total pressures at two locations m given from the numerical results where the phases are referred to the maximum valve opening at zero deg. (The experimental pressure phase shifts given in the table have an arbitrary reference). The magnitudes are all normalized with respect to the peak-to-peak values.
5.0

.,.,,

CIRCUL

."-

Table 1 .,.e higher ha, content of the mass flow rate output is a rather significant 20% while the one-ner-rev valve control input contains onlv approximately 10%.Thcrcforc, the mass flbw exhibits somenonlincar amvlification. The higher harmonic content of the pressure response'does not exhibirsimilar amplification. ~ h higher d harmonic content of all the variables for the fourth and fifth (n = 4 and 5) harmonics are in the range of measurement or numerical error and are included only to give some feeling for the "noise" levels in these results. The phase shift phenomenon is clearly illustrated by the data in the table. The first harmonic of the mass flow rate lags the valve opening by 39". The first harmonic pressure at xlL = 0.524 lags the valve opening by 36". At xlL = 0.804, this lag increases to 42" due to the finite wave propagation speed or sonic lag. Because of the specification of a zero numerical integration dissipation parameter in the numerical algorithm used for the cyclic calculations, computed results contain slight, stable spatial numerical oscillations which make it difficult to place confidence in the exact amount of the lag in pressure predicted between locations. Based on the speed of sound, the pure sonic lag between the two points shown should amount to only about one-half the lag shown. The remaining lag may be due to a physical phenomenon or to numerical uncertainty. Although the problem pf spatial oscillations can be reduced by employing numerical damping, as discussed previously, this damping suppresses multiple wave reflections and renders the code less able to predict resonance phenomena. Fortunately, the computed masiflow rate is the result of a spatial integration which has the effect of smoothing the spatial oscillations. It should he fairly accurate.
Peak-to-Peak Comparison Figure 6 summarizes the results from the dynamic calculations in the form of peak-to-peak total pressures versus frequency for the one-per-rev cam. In Fig. 6 the numerical solution is able to predict fairly well the trends and magnitudes of the pressure response. One possible source of error in the numerical calculation;is the a ~ s u m ~ t i ~constant discharge cucfficient ~n lur the wide rangc of area variation. Somewhat more remote are the possibilities for experimental error. Figure 7 is a plot of the dynamic results of the peak-to-peak total pressure near the beginning of the slotted portion of the pipe against plenum pressure for the one-per-rev cam at two different frequencies. The agreement between the measured and numerical results in this figure, in general, is seen to be fair. As shown, the peak-to-peak total pressure at the entrance of the pipe is dependent on the frequency of rotation of the cam. The data shows a droop in the peak-to-peak value of pressure at high plenum pressure and high frequencies; these trends with plenum pressure and frequency are predicted by the computer code. The peak-to-peak total pressure at the entrance of the pipe is shown to be dependent upon the frequency of rotation of the cam. Figure 8 displays the peak-to-peak static pressures versus rotational frequency for various plenum pressures for the oneper-rev cam. The experimental and numerical curves generally show good agreement and the computer code predicts the &ends. Figures 9 and 10 are plots of the ratio of pipe end-to-entrance pressure versus cam rotational frequency for both cams. The two-per-rev cam results are included since the two-per-rev cam extends the frequency range two-fold. Results are presented for two different plenum pressures for both one- and two-perrev cams. These curves demonstrate the presence of a quarterwavelength resonance phenomenon in the pipe. This phenomenon is better illustrated inFig. 9 forthe two-per-rev cam where a resonance peak occurs at a frequency of approximately 44 Hz for the numerical solution at the higher pressure and around 55 Hz for the corresponding experimental data. The differences in the resonant frequencies between the numerical and exper-

,..,,. ..,. .., ..,.,.,.

".

..,..,

0-

- - EXPERIMENT, DYNAMIC

COMPUTATION, DYNAMIC

AZIMUTH POSITION ldegl

Fig. 5 T t l pressure variation oa

nar-rad-auo !ad!d q uouamouaqd aJwuosag IzH1 A3N3nD3tlj lVN011V1011

06 I 08

01
01

OL

09

05

OD

OE

OZ

0
10.0

-lad-omt jueuosar am a3!mt 30 sapu&ba~j~tueuos& palaadxa aqt ie pau!eiqo slam etep pluamuadxa ou '01 3 r d UI umoqs ,old m a A~I-lad-auoa q I O ~ .arnssa~d a q a h ~ aq;ie pa3uniu -old arom qaum s! laaj~a a3ueuosaJ a q leqt alou 01 Zugsa~alu! s! 11 ')last! aauenua ad!d a q mo13 slaapu pealsu! pue uog -aaUaI sng ale[nm!s 01 alqeun s! lapom @uo!suam!p-auo-!senb aq.1. 'aauenua adrd am ur uon3as uorsuedxa arzzou am ro nua . . . .". aqTmo~j uo!i>aUaI we* JO i v n e ~ a ~ ' ~ i a u a ~ Jd~d3!lsualseleq= ~ a u o q e Ru!naq luamuadxa aqlnl anp Klay!l are vilnsal [quam! s s t p a 1 me3 A~I-1x1-om1a q q!m luais!suoi s! 01 .a!d u! a~nssard~aqs!q a q 103 ad aaueuosar pa3unouo1d arom aqL .sa!manbaq me3 A~I-lad-ow aa!mt m q ssa[ leqmamos aqt are sapuanbaq jmuosal asaqL 'arnssard ramol aql 103 ZH 08

JLY 1985

CIRCULP

lency on Ma!is Flow Rate Tects of Rota -. .. . . . .. rlgures 11 and iz snow tne errect or rotat~ona~ rrequency I the computed mean and peak-to-peak mass flow rates, re~ectively, the one-per-rev cam at several different plenum for .essures. No experimental data are available for comparison with the peak-to-peak data. The experimental lines shown for the mean mass flow rate represent an experimental average since no significant changes in average mass flow rate with frequency were obtained in the experimental data. Likewise :ry little change was observed in the numerical results. Figure 13 shows the effect of rotational frequency on the lase of the first harmonic of the mass flow results shown in gs. 11 and 12. The phase shift increases with increasing plenum pressure (and increasing mass flow rate) and increases almost linearly with frequency over most of the range examined. The linear trend indicates that within this range there is afixed time delay, almost independent of frequency, associated i t h a given set of system parameters such as plenum pressure ~dhlade internal geometry. At the higher pressure, capacince lag is controlling, while at the lower pressure the lag lould he principally sonic lag. No experimental phase shift data are available for comparison because it is very difficult, if not impossible, to measure dy-

".

TOR MODE
20

-151 0

10

30 a 60 W ROTATIONhL FREOUENCV 11111

70

80

SO

Fig. 13 Maw flow rate phase angle namic mass flow containing higher harmonics and obtain phase information. In summary, the comparison of the numerical calculations with experiment indicates that, even with a simple constant discharge coefficient model, the computer code satisfactorily predicts the system performance over a wide operating range. It is able to predict the trends and magnitudes of the total and static pressures as well as the mass flow rate for plenum pressures from 0.98 to 13.75 psig, for rotational frequencies from 15 to 80 Hz, and for valve areas varying from 0 to approximately 1 in2. Theoretical Transient Response For pneumatic control system dcsign, in addition to a knowledce of the pcriodic rcsponse of the system to pcriodic control in*, the ttansient dynamics are a ~ s d o intereit. In the present f study an independent numerical investigation of transient dynamics was conducted. This investigation simulated the pneumodynamic response of the model system to step inputs in control valve settings. This was accomplished by suddenly opening the control valve permitting air to flow into a hlade at atmospheric pressure, allowing steady flow to he established, and then suddenly shutting off the flow, allowing the pressure inside the pipe to return to atmospheric pressure. The results of this investigation into the system's hansient dynamics are shown in Figs. 14 to 17. Figure 14 shows the behavior of the mass flow rate in response to the two control actions for two plenum pressures. The rise or decay time for the mass flow is in the approximate range 0.002 to 0.007 sec and does not change significantly for the two plenum pressures. Thib suggests that in ihese cases thc charaiteristic iimc phenomena are principally related to thc time it takes for a pressure the length of the tuhe, since this is disturhance.to about 0.003 sec. This explanation is reinforced by examining Figs. 15 and 16 which are plots of the pressure distribution at various elapsed times after the sudden valve opening. Figure 15 shows the propagation of the incident pressure wave created by the valve opening and Fig. 16 shows the propagation of the wave after its reflection from the end of the tuhe. The sonic lag is apparently controlling the transient response to the step input. For a system where the capacitance lag is much greater than the sonic lag, the capacitance lag would tend to control the rise or decay. Figure 17 shows the propagation of the incident pressure wave created by sudden valve closing. To smooth these hansient results, second-order spatial numerical damping has been added to the numerical algorithm' for computing the incremental or "delta" solution for a given time level from the solution for the preceding one. This was done by computing the increment from the weighted average of the second-order (in time) Beam and Warming algorithm and the first-order (in time) algorithm of Lax5. The weighting factors applied to the Beam and Warming method and the Lax method were 0.98 and 0.02, respectively.

0.0001 0

10

'

----EXPERIMENTAL

30 40 50 60 70 ROTATIONAL FREQUENCY IHz1 AVERAGE M COMPUTATION

20

80

I 90

Cv=CS=0.95

Fig. 11 Mean mass flow rate

pol=

8.84 lblinz
.

$ 0.008
(L

9
0.006 .
-p

2
Y

5 0.004 4

- 3.04 PI Iblin -

'Iv .I

-0.

* 0 002 -

0-0. ppe= 0.982 lb,

w.

5 P

0.000 0

10

50 20 30 40 ROTATIONAL FREQUENCY IHzl

Cv=Cs=0.95 60 70 80

COMPUTATION

Fig. 12 Peak-to-peak mass flow rate

0.016

SUDDEN VALVE OPENING

A"
=

0.92 in2

SUDDEN VALVE OPEN

0.014

ppO = 13.75 lblIn2

SUDDEN VALVE CLOSING A = 0.001 in.? "

SUDDEN VALVE OPENING A = 0.92 in2 " pL = 8.84 lblln2 ,

I
0

IDDEN VALVE CLOSING A" = 0.001 i n 2

10

15

20

25

30

35

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

SPANWISE LOCATION (in.) Fig. 15 Theoretical incident transient pressure wave

TIME 1x10 2aaci

Theoretical transient mass flow rate Numerical investigation of the response of the flow variables to step inputs in the control valve area indicates that, for the cases examined, the response time is controlled by the characteristic time for wave propagation (sonic lag).

Conclusions
The results presented indicate that quasi-one-dimensional unsteady flow theory can he applied to predict, with reasonable accuracy, the pneumodynamic response of an idealized circulation control rotor model to cyclic control valve inputs. The results also show that, for a given set of system parameters, the phase lag in the response of the system to cyclic control input is a fixed time delay almost independent of frequency. Higher harmonic content of the mass flow rate output exhibits some amplification due to nonlinear effects, but the higher harmonic content of the pressure response does not. A resonance phenomenon is predicted by the numerical results at approximately the frequency for quarter-wave acoustic resonance in a pipe. The predicted resonant frequencies are slightly different than the experimentally observed resonant frequencies. A possible explanation for this is the inability of the quasi-one-dimensional theory to predict wave reflection from the expansion in the entrance section of the experimental model.

References
'Watkins, C. B., Reader, K. R., and Duna, S. K., "Numerical and Experimental Simulation of Circulation Control Rotor Pneumodynamics," AIAA Paper No. 83-2551, AlAAlAHS Aircraft Design, Systems, and Operations Meeting, Fort Worth, Texas, Oct. 1983. 'Beam, R. M. and Warming, R. E., "An Implicit Factored Scheme for the Compressible Navier-Stokes Equations," AIAA Journal, Val. 16, No. 4, pp. 393-402, May 1978. 'Reader, K. R., "Evaluation of a Pneumatic Valving System for Application to a Circulation Cantml Rotor," Naval Ship Research and Develovment Center Rewrt 4070. Mav 1973. "~kader, R., "fhe ~ffects'ofdam and Nozzle Configurations on the K. Performance of a Circulation Contml Rotor Pneumatic Valving System," David Taylor Naval Ship Research and Development Center Repon ASED?91, Nnv , 1977. . . ... . . . . . .

'Lax, P. E., "Solutions of Nonlinear Hyperbolic Equations and Their Numerical Computation," Commu,tications on Pure and Applied Mothemarics, Vol. 7, 1954, pp. 159-193.

r(-0
\

c ~m
L"
0

~ m Z J U w T W xL"-", w m z & , P I XOU" = w n 4 u N .o-x N z Z U WYO\

00 P
\

W W

^ '^

-+xu = . o x

3
I

-o V

+ .

o
o
L"

VI

0 W W

. . . . . .
0

= 0
0

--C I -

SUDDEN VALVE OPENING

ppl = 13.75 1b/in,2

0----C---

"

VI

N 0

c o
0 0

G-

28 I

0.002125 0.003000 0.003750 6wc 0.004375 h......... 0.005000 0.006000

a Y C C 3 C x ~ ) ~ . ~ u r m m ~ ~ ~ L " 0 "
0

0x4

C-1C-I

n n
.A 4

----m N m m I m C I ~ o 0
.

-- L " =

. o

. . . .
O O O
"

m o o 0

c 0

c 0

z
CI

^0

0
W
P

C mC n z-1u
W T W TVICI",
Y C X U

\ W W

Wm=P C L , a OX0!=0 m W nu u

= . o x

2" . o n
\

- 2.::: z' - - - - - -. =. . -. . -. L

m
VI 0

m
0

-5

m
0

c o
0 0

; 1

0 0

Lo-E

SPANWISE LOCATION (in.)

N
2 -

c-10, o\wm P X C J
4 * k *

P ~ o U

P
0

" m - - N N . E N N m O 0 m u = m o o n . . . . 0 0 0 0 0 0 0 0

Y .A

u L

------- - m m m Y I L n m c c m m l -

Fig. 16 Theoreticnl reflected transient pressure wave

SUDDEN VALVE CLOSING p p k = 13.75 lblin2

0 9

r
N

. . "
Y .A
\

.a

w " .m

.eCI: .
0

3 1 . A

W .A 4

0 W n

m - m 0 = m N r O m m l l w Y o
0 0

- ----I N
0
0 0 0

z .
5
m

A C.-.

*---0.0000
........
0.0005 0.0015 0.0020 0.0040 0.0045 0.0070

t, (sec)

"0.

e D m W i l u m m m w = o o

* P

.E

" . u
0

I :

:
?

*
0
W

0,

U 0 . o m \ m c .

- P
W .A

----- :
. r 1 1 C I V I
0

-- m w m ( 3 m ~ ~ N I I O
0

fl

r
Z W

3n r

m V
0

m O
0

U O

= O
0 0

: - . .
P
0
4

: :
S

I :

N . m , L "

Q;
i 2
0

Table 1 Harmonic analysis of typical results

u
0 5 10 15 2 0 25 30 SPANWISE LOCATION (in.) 35

Fig. 17 Theoretical transient pressure wave