Discussion and conclusions.

An analytical procedure is presented to establish the effective friction factors and Nusselt numbers for a heat exchanger core consisting of cylindrical passage channels having different cross sections. The illustrative numerical results demonstrate the validity of using a standard deviation parameter &c, defined by equations (32) and (37), to characterize the nonuniformities in channel cross sections for the geometries described in Figs. 2 and 8. Moreover, Mondt's experimental results [3] suggest that a bulginess standard deviation parameter may be combined with fin spacing standard deviation factor into a single total deviation factor for deepfold surfaces. A comparison of the results in Figs. 3, 6 and 7 clearly show that the deterioration in performance is less severe for a given channel deviation (1 ci/c r), if the nonuniformity is considered as having n passages (n > 2) instead of two passages. For example, (1 Ci/cr) of 0.25 for a two-passage model results in <5C = 0.25. For the same lowest and highest aspect ratios, an eleven-passage model with Gaussian distribution results in <5t. = 0.12. Thus the two-passage model will provide the greatest deterioration in performance if the passageto-passage nonuniformity is considered to be symmetrical around the nominal passages. Figs. 7 and 10 illustrate that a channel standard deviation as small as 5 percent result in heat transfer penalties of the order of 10 percent for the © boundary condition and a smaller amount for the (ffl) boundary condition. The corresponding gain of reduced pressure drop is negligible. It is recommended, therefore, that manufacturing tolerance goals should be set for Sc < 5 percent. The foregoing comments apply to passage nonuniformities that are distributed in a symmetrical manner about the nominal or reference passage dimension. A computer test of some skewed distributions demonstrated that &c as defined by equations (32) and (37) is inadequate as a correlating parameter.