Steel Interlocking Connection Nonlinear FEM & Test Correlation

A full nonlinear FEM simulation of a Steel Interlocking Sideplate Connection (SISC) — a beam-to-column joint where machined teeth on the beam flanges engage directly with slotted sideplates, transferring load through bearing contact rather than through conventional welded or simple bolted interfaces. The simulation was built to replicate a physical test campaign at component level, and validated against measured force-strain data through three distinct behavioural phases from elastic loading to failure.

The central challenge was building a model capable of capturing all the nonlinear physics simultaneously: material yielding, large-displacement geometry, frictional contact between interlocking tooth and plate surfaces, and bolt preload — each interacting with the others throughout the load history.

The assembly model comprised seven distinct component types — built-up beam, column, four interlocking sideplates, two web splice plates, bolts, and lateral restraints — each meshed individually to match its local stress field. The interlocking sideplate slots and the beam flange teeth that engage with them drove the most demanding meshing decisions: the tooth-slot interface was discretised with five elements around the slot perimeter and three elements through the plate thickness, sized to resolve the contact pressure and through-thickness stress gradients at the bearing surfaces without numerical smearing. Bolt hole regions used radial mesh patterns emanating from each hole, with the first ring of elements sized at 1 mm to capture bearing stress concentrations accurately. The complete assembly totalled approximately 164,000 elements, predominantly C3D8 linear hexahedral bricks.

Material behaviour was defined separately for each component type rather than using a single steel model throughout. The primary structural steel (S275JR) was represented with a bilinear elastic-plastic formulation including isotropic hardening, so that post-yield stiffness evolution — critical for capturing the transition and plastic phases of the structural response — was physically represented rather than truncated at yield. The Grade 8.8 bolts used their own higher-strength plastic curve, while the wooden lateral restraints in the experimental rig were modelled as isotropic linear elastic to replicate their relatively soft stiffness contribution without absorbing load that should flow through the steel connection.

Three distinct contact interaction types governed load transfer through the assembly. The interlocking tooth-hole interfaces used hard contact in the normal direction (penalty stiffness enforcement) combined with penalty friction in the tangential direction, with a finite sliding formulation to allow relative motion within the manufacturing clearance gap — itself explicitly defined in the model rather than geometrically built in, so that gap sensitivity studies across the range of realistic cutting tolerances could be run without remeshing. Bolted connections between web splice plates and the beam web used a preload step applied as an axial bolt force, with bolt length locked at the start of the main loading step to preserve the clamping pressure throughout; this two-step sequence (preload, then external load) is essential for correctly predicting how friction at the contact interfaces between clamped surfaces evolves as external loads are applied and the bolt stress state changes. Rigid tie constraints handled surfaces where relative motion was physically impossible, such as the stiffener-to-beam interactions. The overall analysis accounted for geometric nonlinearity throughout, allowing the model to correctly track large-displacement effects as the connection loaded into the post-elastic regime.

Validation against physical test data focused on the force-strain relationship at the sideplate, tracked across all three behavioural phases identified in the test: a linear elastic phase up to approximately 40 kN, a nonlinear transition zone where contact gap closure and local yielding at the first tooth interface changed the stiffness response, and a plastic deformation phase at higher loads approaching the experimental failure condition. The model reproduced all three phases with a maximum strain deviation of less than 8% at any load level, including correctly predicting the load at which the transition to nonlinear behaviour initiated. At the system level, the failure mode — upper compression flange buckling at mid-span — matched the physical test both in shape and in the load at which it occurred, confirming that the combined nonlinear model correctly represented not just local connection behaviour but the global structural response all the way to failure.