22 Mar. 2001 testing & analysis - 8.5Z073-1E2W

professor ben schafer's thin-walled structures research group - johns hopkins university


The first complete test with fully measured and instrumented specimens was conducted on 22 March 2001. The test was on a pair of 8.5" deep Z's (zeds) with nominal thickness = 0.073 in. The designation of the test is 8.5073-1E2W.


The primary changes from the initial test (19 march 2001) were: (1) modification of the boundary conditions to relieve axial forces in the specimen from the resistance at the ends, (2) the panel was screwed to the purlins on both sides of the raised corrugation in the panel to better engage the panel stiffness, (3) the panel to panel connections were made with typical industry fasteners instead of the panel to purlin fasteners, (4) LVDTs were mounted directly to the specimens and the data acquisition system was updated.

overall test setup for 22 March 2001 testing with loading points and LVDTs shown 
roller support used in 22 March 2001. Roller was resting against square alum block under no load, but translated slightly, as shown, under the application of load (as desired).


The test successfully demonstrated that the new boundary conditions were behaving as desired.  Also the improved data acquisition system worked well. In addition, the test avoided any lateral-torsional bucking instability. 


Failure of the specimen occurred within the constant moment region in the center, and was initiated by buckling of the flange and lip - eventually forming a mechanism through yielding in the flange, lip and web, as shown in the following figures. The specimens failed as a unit (i.e., together).

Initial buckling of the flange and lip, note for scale that the angles fastened to the tension flange are 12" apart.
Failure mechanism for one purlin, note significance of fastener location in relation to the yield lines in the purlin.
Overall view of the purlin, showing the asymmetric nature of the collapse - i.e., if the flange of the purlin on the left buckles down, the flange of the purlin on the right buckles up.
overall view of collapsed specimen (LVDT's removed)
load-displacement response as shown in LabView program written for testing


Work is underway to perform tensile coupon tests on the purlins and post-process the data.



The goal of the testing is to examine the local buckling failure mode of C and Z purlins and accurately determine the role of the web slenderness and web/flange interaction in this failure mode. Despite the addition of the panel the buckling mode is more of a distortional nature than local buckling.


Elastic buckling analysis without the panel (see below) indicates that distortional buckling occurs before local buckling, but analysis modeling the panel as providing a continuous rotational restraint to the flange showed that a fully engaged panel should restrict distortional buckling and allow local buckling to form.

Local buckling mode from finite strip analysis of nominal 8.5Z073, note half wavelength ~ 5"
Distortional buckling mode (with no panel in place*) from finite strip analysis of nominal 8.5Z073, note half wavelength ~ 27"
*Note analysis with a continuous rotational spring at the mid width of the flange indicated that the distortional buckling mode is (1) strongly sensitive to any such torsional restraint, and (2) a fully engaged panel has enough "moment of inertia" to provide the resistance


All indications are that the bending stiffness of the panel is not fully engaged in the current configuration. That is, the discretely fastened panel does not, in this test, behave the same as a continuous restraint.

When the purlins buckle in the test one purlin flange goes up while the other goes down. The result is little bending is engaged in the panel, as for the most part the panel can rotate in a straight line and little bending is engaged - the primary resistance to this rotation is the torsional stiffness of the panel itself - which is quite weak.

A simple idea for engaging a greater portion of the panel bending stiffness is to place two fasteners through the flange. Thus allowing the development of a small moment couple and better engaging the panel's bending stiffness.


In order to better understand these ideas elastic buckling analysis of a finite element model of the test that allows for discrete fastening of the panels was conducted. The dimensions of the finite element model are not identical to the tested specimen, though the depth and thickness are the same, the flange is a bit narrower and the lip a bit shorter. The model indicates that (1) for the current test setup the lowest elastic buckling mode is distortional buckling, and (2) if two fasteners are provided at both the left and right of the raised corrugation of the panel and two fasteners in the center of the pans the local buckling mode is lower than the distortional buckling mode. 

finite element model of test setup for 8.5Z073, red dots indicate location of fasteners in current configuration.
lowest buckling mode predicted by the FE model for current fastener configuration (note center panels removed for visual clarity only, red dots indicate fastener locations.)
lowest buckling mode predicted by the FE model for another fastener configuration (note center panels removed for visual clarity only, red dots indicate fastener locations.)


Based on the performance of the first test and the finite element model we have decided to perform the next test on the 8.5Z073 specimens with the additional fasteners. The elastic buckling analysis indicates that the local mode may occur slightly before the distortional mode, but this may not play out in the post-buckling regime. If the distortional mode still occurs in the next test other options for restricting the distortional mode may be considered .


Additionally, please note that the FE modeling indicates that the additional fasteners do not change the local buckling mode - thus, if the fasteners work it can be rightly assumed that this configuration successfully restrict distortional buckling without overly restricting the local mode.