The current experimental base for double-tee joints in tension was screened for reliable data for evaluating various prediction formulas. The existing formulas were found to be very unreliable, especially when the chord and branch have the same diameter. Eleven tests were conducted on DT joints to provide additional data needed for the development of more reliable formulas for predicting the first crack load and ultimate load. New formulas are processed that reduce the coefficient of variation in the statistical analysis almost 50% compared to current recommendations.
The API-RP 2A [2] ultimate strength recommendation for a double-tee (DT) connection with the branches in tension is (Mathematical equation available in full paper) where ? = chord diameter to branch diameter ratio, Fy = yield stress of the chord, and T = chord thickness. Equation 1 was based on only three tests available in 1978 in which the load at first crack was reported in a screened data base [18]. The first crack was used to define the joint strength because of its potential adverse effect on fatigue strength. However, most other published recommendations are based on ultimate strength. It is unlikely that tension loading cracks will be present in the connection at the load levels which cause fatigue damage, since a factor of safety or load factor is used with the maximum design storm loading, unless the connection strength is more than 30% higher than the first crack load. The relative merits of first crack or ultimate as the proper limit will not be discussed. New recommendations are developed for both conditions based on a critical examination of current experimental data and eleven tests reported herein.
Ochi et al [10] has summarized the results of 57 tests on DT joints with branch tension loading. These tests and a few others were examined with the same criteria that were used to develop the data base [18] for the current API static strength recommendations for tubular joints. Small test specimens with chord diameters less than 5.5 in. were eliminated. In addition, the load deformation response of the tests was examined for excessive deflection prior to the attainment of maximum load. The deflection of the chord wall at each branch was limited to 2Fy (30d)/E where d = branch diameter and E = modulus of elasticity. This limit is discussed in detail elsewhere [18].
The screened data base of 31 tests is given in Table 1. Only 6 tests with loads at first crack, Pfc, lower than the maximum load, Pu, were reported and the strength was controlled by the deformation limit in 9 tests. The non dimensionalized static strength vs ? is shown in Fig. 1. The solid data points indicate the strength was controlled by the deformation limit. All but one of these cases occurs at ? < 0.5. Only one data point represents a replicate. For 0.24 < ? < 0.8, the data show little scatter and the strength varies linearly with ?. At ? = 1.0, where there are no replicates, the data are widely dispersed indicating experimental scatter, an inability of the variables T2Fy to fully represent strength, or both.
