and for the function S where £ = kiu + .An eddy viscosity model of the form e+ = kiS for the wall region has been proposed by Spalding. 4 Clearly, Spalding's formula as well as the model proposed by Kleinstein 2 coincide with the present model in the region where T + = 1.It is apparent that once the asymptotic condition lim (s/y + ) -* 1 has been realized in 2, as y + increases this condition continues to hold. An increase in y + however implies a farther penetration into the core region, but on the other hand as was shown in the core analysis the model e + = kiy + r + is valid for 2 + C. Thus we are led to the conclusion that for sufficiently large Reynolds numbers the eddy viscosity model e + = kiSr + holds through the entire pipe. It can be shown (Ref. 5) that "sufficiently large" means 0(S r 1/4 In5 r ) « 1.
III. Comparison with Experimental ResultsThe turbulent pipe flow problem as obtained by the aforementioned analysis is denned by the differential equation 0 (22) subject to the boundary conditions, u + = 0 at y+ = 0 (22a) where r+ = 1 -y + d T , e+ = /bi>Sr + , S = fe~1[exp(J) -(1 + £ + i^2)], £ = kiu + provided \im(S/y + ) -> 1 as y + -+• y^+.This theory is complete with the exception of the two arbitary constants ki and & 2 which must be determined experimentally. From Laufer's 6 experiment at # max = (u ma *D/v) •= 500,000, the value of d T as determined from pressure drop and ^m ax measurements was found to be 0.000114. Taking ki -the von Karman constant-as 0.4, the value of & 2 is determined in such a way that upon integrating Eq. (22) with d T = 0.000114 the resulting Reynolds number # max = (u maK D/v) = 2w max + 5 r~1 is equal to 500,000. A calculation with k 2 = 11.0 gave w max + = 28.69 and # max = 5.033 X 10 5 . With k% = 11.0 as an accepted value a large number of calculations has been carried out for different values of the parameter d T . In Fig. (1) viscous dissipation w^ = v^Ur~^(du/dyY = (du + /dy + ) 2 and turbulent energy production P r = e + w M as computed by the present analysis are compared with Laufer's 6 results at the 50,000 and 500,000 Reynolds numbers. In Fig. (2) , the mean velocity distributions for ten different Reynolds numbers are compared with Nikuradse's data as presented in Goldstein. 7 As can be seen from the above figures the agreement with experimental data is good. Additional comparisons for the friction law and turbulent shear distributions could be found in Ref. (5).