model geometries fall within the following range of best fit.
These powers, the minimizing set of con-
parameters: stants, and the resulting standard deviations are:
0.523 < di/DT < 6.32 P1 P, p3 P4 P. P.
Eq.(l) 1.25 -.25 -.75 1.00 -.25 -.25 Eq. (2) 0.00 0.00 -1.50 1.25 -.50 0.50 19 < DIT < 999 Eq. (3) 0.25 0.00 -.50 0.00 -.25 0.50 Eq. (4) -0.50 0.00 -.75 -.50 0 -.25 4 < dlt < 999 a0 a1 a, Std. Dev. 0.039 < diD < 0.515 Eq. (1) 0.5315 -.06342 0.4372 4.7 Eq. (2) 1.0048 -.01427 0.8605 9.1 0.388 < Sis < 6.26 Eq. (3) 0.2728 -.04706 0.9551 5.2 Eq. (4) 0.3377 -.5272 1.4229 7.5 0.1 < tiT < 3.0 Eq. (1) provides the maximum membrane stress The above range covers most of the practical cases intensity in the vessel of cylinder intersections. Some of the models have Eq. (2) provides the maximum surface stress inten- diameter to thickness ratios which violate the thin sity in the vessel shell assumptions of the program; however, the calcu- Eq. (3) provides the maximum membrane stress lated stresses are still fairly accurate and do not intensity in the nozzle distort the data. The models have been made long Eq. (4) provides the maximum surface stress inten- enough so that the effects of the boundary conditions sity in the nozzle are negligible. 4.0 Results Most codes which have rules for fatigue analysis, have adopted the concept of stress intensity. Stress In order to illustrate the results, several graphs intensity is defined as the maximum stress difference have been made showing the difference in percent between any pair of principal stresses. Since some of versus lambda (A). The "% difference" is defined as the maximum stresses occur off the cardinal axes and the difference between the maximum F AST2 stress and the stress from the equation being considered have shear stress associated with them, the principal divided by the maximum FAST2 stress, multiplied by stress is a more meaningful criteria to use than radial 100. Each geometry is represented by a symbol drawn and tangential shell stresses. Values reported in this at the proper coordinates. The type of symbol used is Bulletin are all based on stress intensity concept and determined by the tIT ratio. That is, the range 0.1 < the proposed design formulas provide the maximum tiT < 3.0 is divided into 8 equal regions with a value of stress intensity. (The words "stress" and symbol for each region. For example, points for "stress intensity" are used synonymously, from geometries with 0.1 < tiT < 0.463 are indicated by hereon, in this Bulletin). symbol # 1. The use of lambda (A) as the abscissa is to Since the nozzle behaves differently than the ves- spread the data out evenly and is arbitrary. Likewise, sel, and the membrane solutions are different from the choice of tiT for the symbol is arbitrary and any the bending solutions, four separate equations have other geometric parameter could have been used. A been developed. Each equation has the following number of parameters were tried and no definite form: patterns were detected. vessel Figs. la-1d show the results of using Eqs. 1 thru 4, respectively. As can be seen, the differences are rrv d PD] = [a] [/DtT scattered rather randomly. 2 Fig. 1e indicates the difference between the maxi- mum values of Eqs. 1 and 3 and the maximum nozzle membrane stress of applicable F AST2 results. Fig. 1f indicates the difference between the maximum value rr. D Pd] = [a] [/dT 2t of Eqs. 2 and 4 and the maximum surface stress of applicable F AST2 results. As indicated by Figs. 1a-1d, the percentage errors where between the proposed equations and the F AST2 data are mostly limited to about 20%, except for a few _ = ao CJ (D)P' (!!_)~'• +a, d T (!_)P' T +a d (!!_)P' (!!_)P' 2 T (!___)~'• T isolated points. This degree of accuracy is normally acceptable for analysis of cylinder intersections. If The powers P;(i = 1, 2, ... 6) are allowed to take on Eqs. 1 thru 4 were reduced to only two equations, values of 0.25 either side of the minimum combina- providing the value of maximum membrane and tion while the coefficient!:! a/J = 0, 1, 2) are calculated surface stress, regardless of locations, the percent by minimizing the square of the error for each power differences will be those shown on Figs. 1e and lf. combination. The combination of powers that pro- These differences are somewhat higher than those of duces the lowest standard deviation is taken to be the Figs. 1a-1d.