| This dissertation presents comparisons of several theories on multiaxial fatigue, a new design formula for out-of-phase torsion and bending, and data on the fatigue behavior of two pressure vessel steels under out-of-phase bending and torsion.;Discriminating specimens were used to compare the predictive powers of three formulas. The design of these specimens is presented.;Three sets of tests were performed. Two of them were performed under synchronous practically sinusoidal, fully reversed, loading with 90 degree phase lag between torsion and bending. The first set is the exploratory test with stepwise loading. The second set is the main tests. Results for the main tests were presented in two groups: one is with annealed stainless steel type 304, the other, A533B steel. Failures occurred at the location which experienced "equivalent" stress ranges lower than other locations if equivalence is calculated by conventional theories. Failure was defined as a 10% loss in the specimen stiffness, which corresponds to visible cracking.;For multiaxial, out-of-phase loadings, generated by combined bending and torsion in these tests, the fatigue danger of ductile steels could be underestimated by the commonly used theories. The underestimation is most pronounced at 90 degree phase difference between bending and torsion and ratios of torsion to bending stress of approximately 0.5. For these conditions, the difference was about 10% in terms of stress.;The third set was performed with A533B pressure vessel steel to find the mean tensile stress effect on the fatigue strength under multiaxial cyclic stresses. Mean tensile stress superimposed on alternating torsion and bending produced no significant, additional fatigue damage as long as ratchetting or creep were avoided. This implies that a Goodman stress correction, such as used in Section III of the ASME Boiler and Pressure Vessel Code, could be quite conservative.;Compared to more commonly used methods, discriminating specimens used here saved the number of test specimens required to arrive at conclusions and the need for precise control of tests. They require more effort in the design of the specimen and they are more expensive to make. The present investigation with discriminating specimens confirmed that the advantage are well worth the cost of the difficulties. |
