R. von Mises - Mathematical theory of compressible fluid flow (798534), страница 27
Текст из файла (страница 27)
Thus the specifying condition isEquations (2) through (5) are four equations for the four unknowns u,p, p, and T, provided that σ is expressed in terms of these variables. T h eusual assumption, made in the Navier-Stokes theory, is that σ is proportional to the rate of expansion du/dx, as mentioned at the end of Sec. 3.1:1χχ\v)σχ= Mo — ·dxHere μο is a function, presumed to be known, of some or all of the variablesu, p, p, and T. T h e valueΜ = f Mo,however, rather than μ itself, is usually called the (physical) viscosity, orcoefficient of viscosity, for the following reason.011.2EQUATIONS FOR S T E A D Y137MOTIONUnder the assumption of one-dimensional flow, the y- and z- directionsare interchangeable. Then a' and σ' must be equal, and the condition (3.7),yσ'χ + <r' + σ'ζ=ζ0, givesy(7)—2xaN o w consider an element of quadratic cross section dx = dy.
Equilibriumconsiderations of the half-cell (Fig. 50) with cross section ABCrequire·σ 'Μ-•ΧF I G . 50. T h e viscous stresses on a rectangular fluid element in one-dimensional flow.that on the face represented by i Cdirection AC of magnitude(<r' dx dz) +-dy dz) = y |(σ'va shearing force must result in theχ(σ* -σ)νdx dz,and since the area of this diagonal surface element is (dx y/2) dz, theshearing stress isT—2\ Χ ~σν)σ— ϊΧ 'σOn the other hand, the rate of shear η, i.e., the time rate of change of theangle CAD, where DA is perpendicular to A C in the x,y-p\&ne, is determined by the differences between the velocities at A, C, and D.
Point Cadvances against A by (du/dx) dx units of length per second, and D is leftbehind by the same amount. ThusV=AC\dxdxand, using (6), the ratio of shearing stress to rate of shear is |μ = μ.For its numerical value, see Sec. 5.02. Equations for steady motionT h e partial differential equations ( 2 ) , ( 3 ) , and ( 5 ) , in which the independent variables are χ and t, become ordinary differential equations in χwhen a state of steady motion, d/dt = 0, is considered.
In the Euler rule138III. ONE-DIMENSIONAL FLOWof differentiation, (1.4), d/dt reduces to u{d/dx)and the three differentialequations are nowd (\duη. d(p — σ' )χΛ"s(i+,4i')t>-"i-«-*here Κ represents the heatfluxk(dT/dx).T h e first equation shows that the mass flux(9)m = pu,or rate of flow of mass across a unit cross section normal to the .τ-axis is aconstant. Using this, the second equation gives(10)= constant =mu + ρ — σχdm,say, and the third equation yieldsm (f+ ^ΖΓΪΤ)+<V°z) ~ Κ = constant =-say.
From (10) we have ρ — σ' = m(dmay be written— u), so that the last equationχ(11)m("I+ γΖΓχT)"dm,K=m(C2"C i u )'Finally, ρ may be replaced in (10) by use of the equation of state ( 4 ) :=ρgRpT=mgRT/u.When Eqs. (10) and ( 1 1 ) are solved for σ'χ=μ du/dx and Κ = k dT/dx respectively, there results0Mo du.Τ- — = u + gRm dxuCi,m dx2Dn(12)7 — 1These are t w o simultaneous ordinary differential equations for u and T.T h e solutions of this system, depending upon the constants Ci, C , and malready introduced, and upon t w o additional constants of integration,represent all possible patterns of one-dimensional steady flow of a perfectgas with viscosity and heat conduction.22I t is convenient to replace u and Τ by dimensionless variables ν and Θ;since d has the dimensions of velocity [see Eq. (10)] and C the dimensionsof velocity squared, we introduce2. _C2„ _u2Λ_ gRTρ11.3STEADYFLOW W I T H O U THP]AT139CONDUCTIONThen the system (12) takes the form= 2 r - V2v+Qm dx(14)1 k dO-β--τ-Ο77=-« +—c.V 2y -gRmdxy — 1T h e solutions v and θ of this system will depend on four arbitrary constants: ra, c, and two constants of integration.
Of these, m occurs in theequations only as a factor of dx and therefore in the general solution onlyas a factor of x, and one constant of integration can be absorbed by translating the origin χ = 0 (since Eqs. (14) are unchanged in form if, insteadof χ, x' = χ + C is used as the independent variable). Thus, except forsimilarity transformations: x" = mx + C", the solution depends on onlytwo parameters.Before discussing the general equations (14), we turn to a special case.3. Steady flow without heat conductionG.
I . Taylor has shown that the system (14) can be integrated in closedform when k is set equal to zero.* This is not a realistic assumption, since itis known that the ratio μο/k varies over a small finite range (see Sec. 5 ) .I t will be seen later, however, that some principle features of the flow canbe found in the solution of (14) under the assumption k = 0.Eliminating θ from Eqs. (14), with k = 0, we find3(15)=(y+l)„-yV 2 U + C (7-1).m axExcept for the scale factor m and translation of x, the solution of thisequation depends on only one parameter, c.Suppose that c lies within the limits° 2W^T)=%(16)<C<(Ύ =1· ·4)Then there exist two real positive values V\ and v (v\ > v , say) for whichthe right-hand member of (15) vanishes:22(7 + l)v - yV2v + (7 - l)c = 0,(17)and it can be written as(170(7 +yV2~v +l)v -c(y -1) = (7 + 1 ) ( V £ -VviKVv-Vv ).For constant μ and for ν between v\ and v , the solution of (15) for χ as02* Equation (3.23) shows that this amounts to strictly adiabaticflow.2140III.
O N E - D I M E N S I O N A LFLOWΥ (PROPORTIONOL TOVV)TX"TU8F I G . 5 1 . Variation of velocity u (or y/υ) with position x, for k = 0.a function of ν takes the form(18)27 +Vv~iμlm0log (ΛΛΙ λ / ί )—log (Λ/ν —+\V )/2constant.H e r e * χ decreases from + <χ> to — co as ν increases from v tovi . Forz; > v\the argument of the first logarithm in (18) must be changed in sign, andthen χ increases from — °o to + <*> as ν increases from v\ to + <*>. For0 < ν < v the argument of the second logarithm must be changed, andthen χ increases from a finite value to +as ν increases from 0 to v .
Ifwe now restrict our attention to flows for which the state variables tendto finite limits as χ —> ± oo then these last two branches of the solutionmay be neglected and attention focused on the function v(x) defined by(18). Figure 51 shows V ^ a s a function of x\ since yj ν is proportional tothe velocity u, we also obtain the graph of u as a function of χ merelyby taking a different scale on the vertical axis.
Since pu = constant, \/vis also inversely proportional to p. T w o considerations are of major interestto us: the relation between the initial and final values of u , and the steepness of the descent from Ui to u .22002y2Since Λ ΑΙ and s/vin \/Ί), we have2are roots of (17), considered as a quadratic equation(19)7 + 1or, in terms of u, using (13),(190Ui + u227 + 1which gives an interpretation of the constant Ci . When k = 0, the second* For the remainder of this article it is assumed that m is positive, i.e., χ is chosento increase along the direction of the flow.11.3STEADY FLOW WITHOUT HEATCONDUCTION141of Eqs.
(14) shows that ν and θ satisfy(20)—^—7 - 1V2v -ν +-c = 0.Eliminating c from (20) and ( 1 7 ) , which is valid however only for ν = Vior v , and then eliminating \/2v from the same equations, we find2(21)Oi + 2vi - Λ/2^ = 0and— 2 — O + v> = c7 - 1{i = 1, 2 ) ,tso thatOly,(210+V2Λ1Vl77 — 1vV2th^θι +=027 7 = +V2v/V2^2,2νχ = — ^ — 07 — 1+2v .2When ν and θ are replaced by their values from ( 1 3 ) , ν =0 = p/Cip,the first equation multiplied by m = pu and d(22)pi + mill = pu/2Ciandgives+ mM ,22and the second, multiplied by C i , gives2(23)^+2^L_P}=^7 — 1 Pi+^L_PJ.27 - IP2Equation (23) is the same as the Bernoulli equation (see Eq.
(2.200 withgravity omitted) found in the case of a steady, strictly adiabatic, inviscidflow, and implies conservation of energy. Equation (22) may be interpretedas expressing the conservation of momentum. Finally, the continuity equation (9) yields(24)pitti =pu,22or conservation of mass. Equations ( 2 2 ) , ( 2 3 ) , and (24) also follow directlyfrom ( 9 ) , ( 1 0 ) , and (11) on setting σχ= Κ= 0 for χ = ± o o .In studying the transition from U\ to u , we use Eq.
( 1 8 ) , in which y/v,2y/vi,and χ/v2can be replaced by u, U\ , and u without further change,2except in the additive constant. L e t e be any number satisfying 0 < e < \(Fig. 51) satisfyingand consider the t w o intermediate values u' and u"(25)wi -Then also U\ — u"=u' = t(ui-u)2= u" -u(0 <2€ <J).(1 — €)(^i — u ), and the change from u' to u" is2(1 — 2e) times the total velocity change from Ui to u .2T h e difference be-142III.
ONE-DIMENSIONAL FLOWtween the abscissas χ corresponding to these values u is/ofi\Τ(26)L'"= χ0-χ2/1\ μ= — — - log ( - 7 +1\e0UI+U1)/ mUI—U22.For given values of the flux m and the ratio u /ui,2the right-hand sidetends to zero as μ decreases, no matter how small e may be. L e t p* denote0the value of the density at that point of flow where u = (ui + u ) /2 and u*2the velocity at the point where ρ = (pi + P2)/2; then the last factor in (26)may be written in two ways:^27)MOU\ +muiU2_μο P2 +Pi _m p2 — pi— u22μp*(ui_02μ— u)u*(p220— pi) 'If, for example, we take e = 0.05 and use the standard values (see Sec. 5)μο ~ 5 Χ 1 0 " slug/ft sec and ρ* ~ 0.0025 slug/ft , then (26) and (27) give73χ" — x' ~ 0.001/(ui — u ). Thus, if the total velocity drop UI — u amounts22to 10 ft/sec, then 90 per cent of this drop is effected within a distance of0.0001 ft, or about 0.03 mm. This is a significant result: the thickness of thelayer within which occurs the major part of the transition from U\, pi, pi toP2 tends toward zero with μ, and is actually extremely small in airu>2, p ,2under normalconditions.If we use a = yp/p as the expression for the sound velocity, as is usual2in discussing the adiabatic flow of an inviscid fluid, the M a c h number isgiven in terms of our dimensionless variables by(28)M = l = 4-=4r = ^227Θayp/pygRTUsing the first equation ( 2 1 ) , we obtain2(OQ)^y1==2»,yM?;1_V2^11'22Since v > v , Eqs.
(29) imply Μι > Μ .x=yM221_ 1V2v~2Moreover, vi and v satisfy ( 1 7 ) .2If the right-hand member of ( 1 5 ) , (7 + l)v — y\/2v + (7 — l ) c , is considered as a function of\/2vits derivative is (7 +l)\/2v —7, whichvanishes only for \/2v = 7 / ( 7 + 1 ) . Since the zero of the derivative mustlie between the zerosΛ/2νι and y/2v2of the function,\/2v~ < 7 / ( 7+2\/2v[ follows. If these inequalities are introduced into ( 2 9 ) , we find thatM2>1 and M22< 1. The transition flow represented by (18) begins in asupersonic and ends in a subsonic state. T h e inflection point of the curve ofν against χ occurs when \/2v = 7 / ( 7 +1 ) . This value of \/2Λ) wouldcorrespond to Μ = 1 if (29) held for all the flow; for υ between ν and v ,λ2however, the left-hand member of (17) is negative [see ( 1 7 ' ) ] and we findΜ> 1 at the inflection point.1) <11.4THE COMPLETE143PROBLEMA n interpretation of c may be found by combining Eqs.