Approximately two years ago my former doctoral student, Thai Dam, and I completed an experimental research study of the punching shear strength of typical slab-column connections. The use of slab-column frames is very popular for low to midrise reinforced concrete residential buildings. The modern trend is to use the thinnest possible slabs that often lead to the need for shear reinforcement at slab-column connections. The use of stud rails, which consist of headed steel studs welded to a steel plate, is the most commonly used option by structural designers. Our research studied how different orientations of these stud rails would affect the strength and failure behavior of full-size slab-column connections. Our research was successful in demonstrating that a radial layout of the stud rails around the column was the best option, but the surprising result was that approximately half of our test specimens failed at a load below the strength predicted by the ACI Building Code.
Punching shear strength equations have been part of the ACI Building Code for over 50 years. They are based on the assumption that part of the load being transferred from the slab to the column is carried by concrete and part of the load is transferred by steel reinforcement, similar to the stud rails we were using in our tests. The load being transferred by the concrete depends on friction along “flexural-shear” cracks that form at a load of approximately half the maximum strength of the slab-column connection. Frictional forces transferred across these cracks depend on the cracks remaining quite narrow. Early researchers who experimentally studied the punching shear strength of slab-column connections used a high percentage of flexural reinforcement in the slabs to eliminate the possibility of a bending (flexural) failure as they studied shear failures at slab-column connections. The high percentage of flexural reinforcement tended to keep the flexural-shear cracks very narrow, so frictional load transfer across the cracks was not disturbed.
For our tests we wanted to use flexural reinforcement percentages in the slabs that were typical of modern construction, which commonly range from 0.75 percent to 1.25 percent. Half of our specimens had a reinforcement percentage of approximately 0.8 percent, and they all failed at loads below those predicted by the ACI Building Code. We had carefully instrumented the slab reinforcing bars and were able to determine that most of the bars at the slab-column connection were yielding just before failure of the slab. Yielding of the slab flexural reinforcement allowed the flexural-shear cracks to open more widely, and the frictional forces acting across the crack were lost. The observed failures were sudden, which is typical for a punching shear failure. These were not flexural failures, which would slowly develop over the full width of the slab and would allow a redistribution of loads within the floor system before final failure. We called these “flexurally-triggered punching shear failures” in the title of a paper we published in Engineering Structures by Elsevier.
The important result is that we accidentally found this flaw in the ACI Building Code by conducting full-scale experiments on realistic test specimens.Professor James K. Wight
After our test results were published, other researchers who had observed similar failures, but incorrectly classified them as flexural failures, came forward to verify the results we observed. As a result of these corroborated test results, the ACI Building Code committee decided that this was enough of a public safety issue to justify a change in the next edition of the Code that was recently published. Instead of changing the punching shear strength equations, there will be a requirement for a minimum slab flexural reinforcement percentage of at least 1 percent in the portion of the slab that passes over and around the column.
The important result is that we accidentally found this flaw in the ACI Building Code by conducting full-scale experiments on realistic test specimens. It is unlikely that this issue could have been discovered using analytical models unless those models were calibrated with our test data to look for the type of failures we observed.