R. von Mises - Mathematical theory of compressible fluid flow (798534), страница 49
Текст из файла (страница 49)
Then from (44)(45)Αθ =6 =*({,)m-θ&ηο) =(ηι-f)-(ηο -ξ)=ηι -ηο .Thus this angle depends only upon the two fixed C~-lines, and not on thevariable ξ. Also, interchanging the role of the two families, we obtain(45')Δ'β =e' =*(&,„)-θ(ξ ,η)0=|o -ξιSimilarly for the angle Q(q), which is a function of q, M, p, or p:(45")Δ<? = Q F C 1,0 -Q(l;, ηο) =IH -η„ =-€,Thus, it is seen that the angle between the flow directions atcorrespondingpoints of two fixed characteristics remains the same all along these characteristics) and an analogous constancy relation holds for the differences in Q.sIn Sec. 4 we indicated how the net of M a c h lines is determined in a certain domain by specific types of boundary conditions. T o construct thisnet we used the compatibility relations (24) in the form of finite differenceqqF I G .
96. Constancy of ΔΘ along characteristics.16.7259N E T S OF C H A R A C T E R I S T I C Sequations (25) and the direction conditions (23). This most direct procedure, essentially due to J. Massau, becomes a practical approximationmethod if the compatibility relations are used in the form (43) together with(23) and with a tabulation of Θ (or Q) as a function of q along a M a c h line(see end of preceding section). A first approximation thus found can beimproved b y iteration. I n the following we describe a few other usefulprocedures, both numerical and graphical, for the determination of theMach lines, if the net of Γ-characteristics is considered known.T o fix ideas, consider the characteristic boundary-value problem where weknow that two arcs OA, OB in the #,?/-plane are parts of a £-line and an17-line, respectively. T h e case where the whole or part of one of these curvesis a straight line will be considered in A r t .
18 where we shall study so-calledsimple waves; here they are assumed to be curved. Consider a subdivision£0, £ 1 , · * · , £m , * · · on OA and η ,· · · , η , · · · on OB and a corresηι,0ηponding lattice where a general nodal point has the coordinates (ξπι,ηη)or briefly (m,n).Then from Eqs. (44), 0(46)—6mnθτηΟ —m n= 77 — £ , we havenθθη +m000 =0.Similarly(460Qmn~QmO~Qon+QoO= 0.Hence knowing 0 and Q at the subdivision points on OA, OB, we can determine them at all nodal points.W e still need, however, the coordinates x ymnin the physical plane cormThresponding to ( £ , η ) to which qmηm nbelongs. T o find these coordinates, i.e.,our original independent variables, we use a step-by-step procedure.
T h esimplest is the recurrence procedureχ„t/mny-m—l ,n— t»anVmnVm,n—1—Φτη-Ι,η+Φ mn+Φ mn~/\%mn\Xm—1 ,n)j(47)xwhere the x o ,y ,o,m>mpoints along OAΦτη,η-l—tan/\\XmnXm,n—l)jΧο,η , 2/o.n are the known coordinates of the subdivisionand OB,respectively; the φ^η, Φ^η are known at allnodal points since 0 and q are known there. Thus (47) are two simultaneouslinear equations which determine x ,mny -i,nmym noncex ,n-i,my n-\mtand x _ i , ,mnhave been found. According to the quasi-linear character of ourproblem the original unknowns 0, q are found first through the linear compatibility relations, or in other words through our knowledge of the net inthe hodograph, while for the independent variables χ and y we need a stepwise procedure based on the direction conditions.
T h e indicated procedure,200IV.PLANESTEADYPOTENTIALFLOWwhich is, of bourse, essentially a rearrangement of that in Sec. 4, can likewise be made more accurate by means of iterations.T h e best known graphical procedure for constructing the Mach net isbased on the property that the C , C~ through Ρ are perpendicular to theΓ~, Γ through P' in the hodograph. Consider a C~ in the physical planewith points P i , P , P3, · · · on it and known velocity vectors q i , q ,q , · · · .
If we plot these in the hodograph, their endpoints P[,P ,P 3 , · · · lie on the corresponding Γ". Then the tangent at P[ to this Γis perpendicular to the characteristic C t through P (Fig. 97). This is usedsystematically in a procedure due to A . Busemann (see Fig. 98). Denotelattice points in the hodograph by P'. Then, starting from the givenboundary values, the Mach net of points Pis constructed successivelyaccording to the rule++2232-tmnmnΡmnPm,n-\-\ -LPmnPm—1 ,n >Ρ m n P m - f l ,n -LΡmnPm, π—1 ·In this way the r~-curves correspond to the C~, and the Γ to the C ;dotted line segments in the hodograph which have r e d i r e c t i o n are perpendicular to dotted segments in the flow plane which have C -direction,+++F I G .
0 7 . Graphical construction of Mach lines.F I G . 98. Reciprocal lattices in flow plane and hodograph.17.1 D I F F E R E N T I A L E Q U A T I O N S F O R L E G E N D R E T R A N S F O R M S261and similarly for Γ and C~. If in the physical plane each mesh is denotedby the two numbers equal to the smallest pair of subscripts of the fourcorners P ,it is seen that the mesh (i,k) corresponds to the point P\ in thesense that the four sides around the mesh (i,k) are respectively normal tothe four sides through P' (reciprocal lattices).+ikkikArticle 17Further Discussion of the H o d o g r a p hMethod1. Differential equations for the Legendre transformsIn the preceding article we started with the two nonlinear partial differential equations of first order, (16.1) and (16.2), which express continuityand irrotationality. Transforming these to natural or intrinsic coordinates,we obtained the basic nonlinear equations (16.7) for q and 0, which together with Eq.
(16.10) defined the problem completely. In Sec. 16.2 thepotential φ and stream function ψ were introduced by means of (16.15),(16.17), and (16.18). T h e equations (16.18) with ψ and ψ as dependent, χand y as independent variables, or their natural form (16.18') again constitute a pair of nonlinear partial differential equations of first order. These,although simple in appearance, are quite complicated since ρ has to be expressed in terms of q, and q in terms of φ. N e x t we derived a nonlinearsecond-order equation for φ alone and one for ψ alone, Eqs. (16.14) and(16.21), and the corresponding intrinsic forms (16.16') and (16.20'). InSec. 16.5 the hodograph transformation was introduced and it led to variouslinear equations.
Using now q and 0 (the previous dependent variables) asindependent variables, we obtained the basic linear equations of first order(16.31), and also the equations of second order (16.32) or (16.32'), (16.33)or (16.33') for ψ alone or φ alone, respectively.In this and the following section we shall give still other useful equationsof the problem. I t is left to the reader to consider and determine variousanalogies with the equations of Chapter I I I .One pair of nonlinear equations in the physical plane for the velocitycomponents q = u, q = ν consists of Eqs. (16.2) and (16.14):xyIV.262PLANESTEADYPOTENTIALFLOWThese equations can be linearized by the direct interchange of variablesintroduced in Sec. 10.6 and used in Chapter I I I .
If the Jacobian. _ du dvJ~ dxdydu dv _d(u,v)dydx~d(x,y)is different from zero we obtain the linear system/ 22s(α -u)dy ,(dxuv[ — +(3)dy\,J + (a -2\dx2v) —Λ= 0,- Ου= οduT o satisfy the second of these equations we introduce a function Φ(μ,ν)such that** = χduthen the first Eq. (3) yields(4)W** = νdv'X )/r\/ 2(5)(α2x 5 Φ .θφ. /2u ) — + 2ui; — — + (α 2-yΟ^Φ2\20ν) —Λ= 0.T h e Φ introduced in this way is the Legendre transformof the potential φ.9From Eqs. (4) and (16.15) we have(4')άΦ = χ du + y dv,a\p = udx+ ν dy,and by integration we obtainΦ = f(x du + y dv) = J d(xu + yv) — J(u dx + ν dy)= xu + yv — φ.Hence, between φ considered as a function of χ and y and its transform Φconsidered as a function of u and v, the relations,.d<p .
dtpΦ = xu+yv — φ = x — + y(6)8,<p = ux +XdφV_dΦ ,dΦvy — Φ = u1- νdudv.Φhold.Using q, Θ as independent variables instead of u, ν and introducing theabbreviations(7)χ cos θ + y sin θ = X,y cos 0 — χ sin 0 =F17.1 D I F F E R E N T I A LEQUATIONS FOR LEGENDRE TRANSFORMS263for the components of the radius vector r in the direction of the flow andnormal to it, we obtain from ( 6 ) : Φ = q (x cos Θ + y sin θ) — φ = qX —φ, so that<«·> M'^and(6')= q^L dqφφ,Φ = Χ ^ - φ.dsT h e following relations, derived from Eqs.
(16.17) and ( 4 ) , serve t o determine the stream function ψ, if Φ(η, V) is known:, v0dxj/(d% ,d% \θψq θθ)'dd ~~θΦ(,2d%\ordq"\dqd6PPV dqdd )'2Likewise, a Legendre transform Ψ of the stream function ψ may be defined.If w e interchange the variables x, y for the variables pu, pv, the continuityequation becomesdxd(pw)4-yd(pv)d=πThis equation may be satisfied by introducing a function Ψ of pw and pi>such that(9)a(pw)= y,ΓΤ—τ =θ(ρν)- ζalso,-y—d(p?)=F.B y integration of ( 9 ) , the following relations, analogous to (6) and ( 6 ' ) ,obtain:ψ = puy -pvx -ψ = x -+ y--ψ = Υ - - φ ,(10),dVdYΨ = pu - — - + py -j—- Ψ = pq —— d(pw)a(py)a(pg)W e now derive α pair o/ first-order equations for Φ and Ψ analogous to Eqs.(16.31).