<p> A compression member end-supported rotationally about its non-principal axis deflects in both directions of its principal axes. The elastic eigenvalue of this issue was theoretically solved by Rasmussen and Trahair, while neither elastic instability nor inelastic collapse of such struts with initial crookedness has been investigated.</p><p> Thus, experiment was first performed as shown in Fig. 1. The strut is a steel flat bar, whose centroid and shear center coincide and then the flexural behavior is uncoupled with torsion. The flat bar is welded at both ends to a circular plate bolt-connected to a knife-edge block in order to alter the oblique angle between the knife-edge direction and the weak axis. Eight specimens in Table 1 were tested. Compressive stress-strain curves of the flat bar stubs are shown in Fig. 2, from which Young’s modulus and yield strength of the material are determined. Each specimen’s initial crookedness was measured as in Fig. 3, whose shapes are compared with cosine curves utilized in theoretical analysis.</p><p> The experimental results are shown in the left-hand sides of Figs. 4 and 5 for elastic loading, and in Fig. 6 for collapse loading. From Fig. 4 showing the relationships between compressive load and mid-deflection due to weak-axis bending, the curves shift upward with an increase of the oblique angle. From Fig. 5 showing the relationships between mid-deflections about weak-axis and strong-axis, the deflection due to weak axis bending always follows the same direction of the initial crookedness, while the deflection due to strong axis bending does not.</p><p> The load vs. lateral deflection curves of the flat bar struts are analytically derived by solving Eq.(1) so as to satisfy the boundary conditions of Eq.(6) considering the rigid-end offsets as well as the cosine crookedness and referring to the section forces and lateral displacement in Fig. 7. The solution is detailed in the appendix. The analytical curves are shown in the right-hand sides of Figs. 4 and 5 showing an acceptable level of resemblance with the experimental curves in the left hand sides. The major cause of the deviation is attributed to the difference of actual crookedness and assumed one. Buckling load as an eigenvalue can be obtained either by finding the load which makes the lateral deflection diverge or by solving the eigenequation of Eq.(a28) in the appendix. The elastic buckling loads P_cr of the tested specimens are calculated as in Table 2.</p><p> The strength curve is influenced by the oblique angle as observed in Fig. 4, and also by the bending stiffness ratio of weak-axis to strong-axis. This is analytically investigated as exemplified in Fig. 8, where r is defined by Eq.(7). For a small r–value, i.e., for strong-axis rigidity much higher than weak-axis rigidity, the oblique angle is very influential for the strength curve. This is accounted for by high end-fixity of the strut as demonstrated in Fig. 9, where the end-fixity is represented by the minus sign ratio of end curvature to mid-length curvature about weak-axis deflection.</p><p> The relation of maximum compressive stress to slenderness ratio is simply called the column curve, in which the slenderness ratio is the effective one regarding the end-fixity. The experimental data plotted in Fig. 10 are found well coincident with the column curves in current design specifications.</p>