03.06: A study on the behaviour of steel beam-column connections to deep wide flange columns for seismic applications

Nonlinear time-history analysis of steel braced frames, especially bucklingrestrained braced frames (BRBF), subjected to earthquake ground motions has revealed that columns in the bottom stories are often subjected to combined high axial load and inelastic rotation demand resulting from story drift. The reliability of column members under this reversing high axial loads and drift demand has not been previously experimentally verified. To provide a basis for performance evaluation of columns under high axial load and drift demand, nine W14 section column specimens have been subjected to laboratory and analytical investigation.

Since a loading protocol for column testing did not exist, the first phase of this project consisted of development of a statistically based loading sequence for braced frame column testing. 3-story and 7-story BRBF prototype building models were designed and analyzed. Nonlinear time-history analysis of frame models, subjected to a suite of 20 properly scaled earthquake ground motions, was conducted. Time histories of first-story drift ratio for the 20 records were processed using a rainflow cycle counting procedure. Statistical analysis was used to quantify maximum and cumulative story drift and maximum column axial load demand for development of a reasonable loading sequence for experimental testing. The first step in the developed loading protocol consisted of imposing a compressive axial load offset of 0.15Pn to simulate gravity load.

Then in-phase, increasing amplitude cyclic axial load and story drift were applied. The experimental program consisted of cyclic testing nine full-scale fixed-base columns. ASTM A992 steel wide-flange sections typical of braced frame columns, representing a practical range of flange and web width-to-thickness ratios, and 15 ft story height were subjected to different levels of axial force demand (35%, 55%, and 75% of column yield strength) combined with story drift demand of up to 10% for these simulated first-story columns. Column specimens were tested at the University of California, San Diego (UCSD) Seismic Response Modification Device (SRMD) Test Facility.

The test data agreed well with the established P-M interaction surface. Significant overstrength was observed between the test data and the P-M interaction surface based on nominal material properties. Specimens achieved drift capacities of 7% to 9%, corresponding to a ductility of approximately 10. These drift capacities were calculated assuming a 10% reduction from peak moment resistance was used to define the drift capacity. These drift capacities were more than four times the expected BRBF story drift from earthquake nonlinear time-history analysis. For all specimens only minor yielding was observed at a story drift of 1.5%, the maximum expected drift from nonlinear time-history analysis. Flange local buckling was observed for all specimens with the exception of Specimen W14×370-35. No web local buckling was observed.

The relatively small amplitude of flange local buckling observed at 6% drift (more than three times the maximum expected drift) provided an indication that strength degradation due to flange local buckling is not expected to present a problem for the seismic design of the tested W14 column sections.

The finite element program ABAQUS was used to model the steel column specimens. Models predicted global behavior, yielding, and strength degradation resulting from local buckling at large drifts. Models were subjected to both monotonic and cyclic loading sequences. Analysis results were observed to be well correlated with experimental results. It was determined that a typical initial residual stress distribution did not significantly effect the P-M interaction or moment versus drift response. The behavior of deep columns with higher width/thickness ratios than the tested W14 sections was also investigated. Models of W27 deep column sections showed significant strength degradation due to flange and web local buckling.