R. von Mises - Mathematical theory of compressible fluid flow (798534), страница 66
Текст из файла (страница 66)
4. However, we do not enter into an explanation of the general pro1cedure recommended by Lighthill in the case of supersonic velocities.2. Replacement of Chaplygin's factor [ ψ η ( τ ι ) ]1W e refer to Eqs. (20.13) and (20.14) with ψ (τ)given by Eq. (20.8).ηFor r = η , the right side of Eq.
(20.14) coincides with that of Eq. (20.13)with q =1, where q = qi =1 is the velocity at infinity and η =l/q .2mW e have indicated earlier that if we start with one branch of the compressible flow ^ ( r , 0 ) , the main problem is to find its continuation over thedesired region. Lighthill noticed that for this purpose Chaplygin's factor[^ (ri)]n_ 1is inappropriate (leading to avoidable complications) and replaced it by another factor which retains the essential properties of theformer and is better adapted to the problem of continuation.
W h a t arethese properties? W e have seen thatlim ψηΜ/ψηΜ=q,nand require accordingly, /(n, n) denoting the factor in question, that/(1)lim ψ (τ)/(η,τι)η=\ n/2lim ( - )i.e., t h a t / ( n , τι) should behave like τ ι ~η / 2=as qmq,n—> oo. A second requirement is that(2)φ =g [Σ^ » ( r ) / ( n , τι)β- *] ,< ηwhich replaces E q . (20.14), should have the same circle of convergence asthe original f-series expansion (20.13) about f=0; and in fact shouldexhibit for τ = τ\ , a behavior similar to that of (20.13) for q = 1.T o arrive at an appropriate normalizing factor / ( η , τι) we introduce in-V. INTEGRATION THEORY AND SHOCKS354stead of r, the new variable, λ, the same as that in (17.17), defined bydX/dg=Λ/1 ~ M /q]it leads t o a normal form of the second-order2equation for ^ ( g , 0). In terms of τ we have(sVdX)=dr;dX/dg=V l-dr/dgMqj2g-V l=2gΜ2=Jl_2r/l2 r f-r/nl - τ'W e chose the limits of the integral so that X = 0 for r = r* :(3')tanh"W e shall use the variable8=(4)λ +σ,where σ is a constant defined by the requirement that slog (q/q )mforsmall r, i.e.e*lim — = 1e2(5)τ-*·0orΤΤThis gives with κ = y = 1.4 and ft = r "2σ =- f t tanh" J -e~ .2a= 6,1J log (ft -12Xlim — =log 2 =1) +2-1.17.ΔftI t is seen that s —> — oo, as r —> 0, Μ —> 0.
A s r increases from 0 tos increases from — oo to σ =r,t—1.17.W i t h s (or λ ) as an independent variable the stream-function equation(20.5) is transformed to(6)^tdsW+2^^t =ΘΘT2^oVwhere Τ is a function of s (or r or g ) , namely,Τ =-dΡV l-M22/ V l-ρds \M\2/'If we apply to Eq.
(6) the separation of variables ψ = ^ (s)e \n(6' ,λ_^_γnwe obtain*an equation which will be used later.I n order to obtain an equation with a term containing φ rather than21.3FLOW AROUND A CIRCULARCYLINDER355d\p/ds as in E q . (6), the dependent variable is changed by putting(7)ν(τ)ψ*.φ =A n elementary computation gives»>+%+where, with a prime denoting differentiation with respect to s,2V'—(9)V"F = —= Τ ,2V'——2-' - 0 - 4γ - [ ^ Τ - ( - 5 Γ · ·«T h e equation corresponding to E q .
(6') is thenwd ψ2, *η,. *ΓΛds2.*which suggests that asymptotically for large | η |, ψη~e .n8This, togetherwith E q . ( 7 ) , makes somewhat plausible the following result (which we givewithout proof). T h e function ψ (τ)is asymptoticallyηequal to e V{r),nsasI η I —> oo, for subsonic r (negative integers η excluded); or, more precisely,(10)φ (τ)η=[• (0]+oV(r)enas I η I —> oo , uniformly for 0 ^ r ^ rt— e, for complex η and | η + m | ^ δfor all positive integers m (δ, € arbitrarily small positive numbers).W e can now see that for η <τ< a simple and appropriate normalizingfactor is(11)/(n, n ) = e "In fact, from the asymptotic equality of en s i28in E q .,where=β(η).3and r for small r, as expressed( 5 ) , we see that e~ behaves like n *-8β ιin the limitη—> 0, aswas intended. Also it follows from E q .
(10) that the general term ofthe series in E q . (2) will behave for large η like F ( r ) c enr =η , s =n ( s - S l - t 0 )and forSi , this is indeed a behavior similar to that of the generalterm in E q . (20.13) for q =1.3. Flow around a circular cylinderT o gain insight into the problem of flow around an obstacle, we considerhere what is probably the simplest profile, the circle. W e shall give the4* N o t e that VA= A " , where Κ was introduced in (17.24')·-1356V. I N T E G R A T I O N T H E O R Y A N DSHOCKSy4F I G . 138. Flow past circular cylinder.main line of procedure and refer the reader for details to the literature thatwill be quoted.*L e t the radius of the circle be equal to one, let, as before, q„ = q = 1,and the corresponding η be subsonic.
T h e complex potential w(z) of theincompressible flow is well known:x(12)/\w(z) =z,1+ ->.dw-χβ1f = - τ - = qe= 1 - - ,dzz12zζ = (1 -f)l= wo(t)w(z)= (1 -f)* +(1 -iT*.Consider the upper half of the z-plane. I t is seen (Fig. 138) that the #-axisfrom point 5 to oo and from - co to 1 is mapped onto the cut betweenf = 0 and f = 1; the image of the profile streamline 1 2 3 4 5 appears inthe hodograph as the circle of center q = 1, q = 0 through the origin,and a few more streamlines are sketched roughly indicating how the flowregion in the upper z-plane outside the obstacle is mapped onto the insideof the circle in the g ,g -plane.xxyyT h e expansion of w ( f ) must be made separately for | ξ \ < 1 and | f | > 1.T h e line | f | = 1 is an arc of circle in the hodograph plane, and its image11 — 1/z 1 = 1, or x — y = |, separates the flow region above the x-axis0222* T h e example of this section requires only some idea of analytic continuation andthe residue theorem.21.3FLOW AROUND A CIRCULAR C Y L I N D E Rinto three regions Ri,R,with corresponding images in the hodoRz,2357graph.1 we must take ζ =Expanding for | f | <+ (1 — f )— (1 — f ) ~ * in Rz (then for f = 0 we obtain ζ =Rz); for I Π1, ζ = i Γ * (1 ->F o r the regions Ri,w22in Ri and ζand 2 =x=— 1 in.β , ft we then obtain the expansions2Ml =(13)1 / f ) " * in β_ i1 in RΣ( »-1)Γ(η -I )^Σ(η +Ί)Γ(» -I)I (i)n=o»_oΓ(5)=ζη !n!If we consider also the lower part of the 2-plane, namely, y < 0, the regionsRi and Rz are continued symmetrically below the x-axis, and we denoteby Ri the region symmetric to R .T h e hodograph image of this lower2part covers a second time the same hodograph circle as in Fig.
138 and isnot shown. T h e formulas (13) remain correct and we have to add the formula(130=-u>2.One may verify that each of the four series is the analytic continuation ofits t w o neighbors. This expansion into series of w(z)= tA>(f) is the firststep.N e x t we seek a corresponding compressible flow and begin b y constructcorresponding to Wi and w , according to Eqs.ing the series W\ and W ,22(2) and ( 1 1 ) ,(14)WX=*Σ(-N1)F1 ν2/ = °(15)W2='r(J)ΣT(N"Η"^Ur)e-^ °\+i^'N(N"+),Γ(h )t U r ) e(n- ^+m.nln=oW e shall see immediately that W2is not the analytic continuation of Wiand that it can never be so, no matter h o w / ( n , n ) might have been chosen.T o find the continuation of Wi for r >ri is a mathematical problem whichcan be approached in various ways. A comparatively simple solution (seeN o t e 4 ) is based on a generalization of the representation of the hypergeometric function F(a, b, c; x) b y a "Barnes Integral".T o explain this idea we first return for a moment to the incompressibleflow problem.
Denote by Β the integrale- ffa C 'S!(1)r<5'-* -' - '*)r<)(f)358V. I N T E G R A T I O N T H E O R Y A N D S H O C K SlcF I G . 139. Integration paths used in connection with Barnes integral.along the path C in the complex y-plane indicated in Fig. 139. W e thenapply the residue theorem to the above integral taken around an appropriate closed circuit (/) to the right of the imaginary axis, such as the oneindicated in the figure. W e use the fact thatν =Γ(*>) has simple poles at0, 1, 2, · · · with residues ( — l) /n\.—n, ft =nThen the poles of(ν — 1)Γ(—J >), which are at ft = 0, 2, 3, · · · are inside (/) and those ofwhich are at J — η, η = 0, 1, 2, · · · are outside (/).
I t can be—shown that, as | ν \ —> <χ>, q <1, the integral around the semicircle to theright converges to zero and therefore —B equals the limit of the sum of theresidues at the poles which are inside (/), or(13"}=Bf k ) £in-1)Ti-nh)% 'Ifl<1-T h e series to the right in E q .
( 1 3 " ) is identical with the expansion (13) ofWi . In this way W\ is now expressed by the integral B, and we can use thisrepresentation to find the continuation of w for | ξ \ >x1. W e shall nowexplain this main point directly for the compressible flow problem.If, with fψη(τ)β~ ,η81qe~ \ we replace, according to Eqs. ( 2 ) and ( 1 1 ) , q=l^(16)nbywe are led to consider, instead of B, the integral100Λ +Β' = ^-±φ(„ -1)Γ(ν-£)r(-„)(-l)V,(r)e-' "(+ W )dvalong the same path C as in Fig.
139. W e now apply the residue theorem tothe integrand in (16) and the circuit (/). W e find that for r < r i , (corresponding to q < 1) — B equals the limit of the sum of the residues of thef21.3FLOW AROUND A CIRCULARCYLINDER350integrand at ν = 0 , 2 , 3 , · · · and that this sum is exactly the series W\ . T ofind then the analytic continuation we integrate around the closed circuit( i 7 ) (indicated in the figure) to the left, where for | ^ ) —> oo and τ >τ ι , theintegral around the left semicircle tends to zero.
Inside this circuit are nownot only the poles of T(v — \), but also those of ψΛτ),of ν for fixed τ (0 <r <considered as a function1). T h e former are at \ — η (η= 0, 1, 2, · · · )and the limit of the corresponding sum of the residues equals W2negative and — Wif θ is positive.* T h e (simple) poles of φ (τ)29(η = 2, 3, · · ·)> with residues p=nwhere C—ηΟ ψ (τ)ηηndepends on theconstants a , b , introduced in E q .