Evaluation of the Effectiveness of Eccentric Column Stiffeners in Strengthening Beam-to-Column Moment Connections Using Response Surface Methodology

Xiang Lu, Purdue University


After the 1994 Northridge Earthquake, analysis and studies were conducted and proved that transverse stiffeners with full-depth of beam have a better effectiveness on the strength of connection than other stiffeners. Meanwhile, engineers have questions on the allowable maximum eccentricity of transverse stiffeners of a beam-to-column connection. Hence, in this project, the effectiveness of column transverse stiffeners for beam-to-column moment connections were investigated under gravity load, and the method used to assess the strength of a beam-to-column connection under gravity load, response surface methodology, was introduced. In this research, a set of finite element analysis was constructed on the experimental test models in the past study by Graham et al. using ANSYS. The finite element analysis results matched the test results, which proved the validity of the finite element models in this research. A thorough literature review on stiffener design and the limit states of beam-to-column moment connections, including flange local bending, web local yielding, web crippling, web compression buckling, and web panel-zone shear, was finished. The JMP software were utilized to do design of experiments analysis on the 5 limit state strengths on beam-to-column moment connections, and the significant parameters to connection limit state strengths, d, tw, and e, were identified. 16 models with a typical size column, 14WF257, and different d, tw, and e were designed and simulated in ANSYS. The strengths of the 16 connection models were studied by analyzing the beam force ratios (P/Py) vs. deflection ratios (d/dy). Then, the relationship between connection strength and the three significant parameters was analyzed using response surface methodology in JMP software, and a linear equation which could predict connection strength was generated. Additionally, finite element analysis was finished on a random model, R-1, to set the “Target” P/ Py, which was 1.182. The equation was successful to predict the allowable maximum eccentricity of a random model, Run 11, but the equation showed an error to predict the allowable maximum eccentricity of R-1. Hence, a factor to make the equation more conservative was provided using “3 Sigma Limits” method, which was 0.932. Also, a table indicating the controlling limit state of a connection, either web yielding or web buckling, was created, and the allowable maximum eccentricities for connections were listed. From the table, connections controlled by web buckling need more eccentricity than connections controlled by web yielding, and it was proved by two random models, as Run 11, which was controlled by web buckling, had a lower allowable eccentricity in simulation than R-1, which was controlled by web yielding.




Chen, Purdue University.

Subject Area

Engineering|Civil engineering

Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server