Evaluation of AISC Classification of Connections Under Uncertainty with the Interval Finite Element Method
Author(s)
Kim, Terry
Advisor(s)
Editor(s)
Collections
Supplementary to:
Permanent Link
Abstract
Structural steel frames are routinely analyzed under the binary idealization of connections as either perfectly pinned or fully rigid, even though no physical connection satisfies either idealization exactly. The AISC Specification has recognized this limitation since 1986 through its partially restrained (PR) classification, and Section B3.4b(b) of AISC 360-22 provides that the moment–rotation response characteristics of a PR connection shall be included in the structural analysis. The code does not, however, prescribe a standardized procedure for selecting the M–θ curve, leaving the choice to engineering judgment and allowing two engineers to obtain measurably different design forces from the same body of literature. Probabilistic methods have been proposed to close this gap but have seen limited adoption owing to narrow connection-test databases and the computational cost of propagating distributions through nonlinear frame models.
This thesis takes a different direction. The AISC Commentary to Section B3 classifies connections by a nondimensional secant-stiffness parameter and identifies k = 2 and k = 20 as the thresholds separating simple, partially restrained, and fully restrained behavior. If the response of a frame at every admissible PR stiffness lies within the interval bracketed by the responses at these two thresholds, then the classification thresholds themselves constitute an interval model of connection uncertainty, independent of any particular M–θ curve. The Interval Finite Element Method (IFEM) supplies the analytical machinery to propagate that model to displacements, internal forces, and reactions with guaranteed enclosure.
A four-step verification procedure is developed and applied to a fixed–hinge–fixed beam and a single-bay A992 portal frame sized to a W16×67 through a complete AISC 360-22 design process. In both examples the AISC-compatible envelope is enclosed strictly within the physical-extreme envelope at every reported quantity. For the frame, the lateral drift upper bound is 11.82 mm under the AISC envelope against an H/400 limit of 12.50 mm (5.5 % margin), whereas the physical-extreme envelope exceeds the same limit by 28.7 %. The AISC-compatible envelope is approximately 35 % as wide as the physical-extreme envelope uniformly across every output quantity, demonstrating that tightening the assumed range of joint behavior produces a proportionally tighter response envelope without overestimation.
Sponsor
Date
2026-05
Extent
Resource Type
Text
Resource Subtype
Thesis (Masters Degree)