This　study　comprehensively　investigates　the　mechanical　behavior　and　failure　mechanisms　of　aluminum　alloy　bending-torsion　gusset　joints　under　static　loading　conditions.　Initially,　the　stress　response　of　the　gusset　joint　during　the　elastic　phase　was　confirmed　through　static　experiments,　followed　by　an　in-depth　analysis　of　the　ultimate　bearing　capacity　and　failure　mechanisms　using　FEA.　Furthermore,　a　rigid　joint　model　was　constructed,　revealing　a　bearing　capacity　6.07　%　greater　than　that　of　the　bending-torsion　gusset　joint,　thereby　characterizing　it　as　a　typical　semi-rigid　joint.　The　study　further　investigates　the　effects　of　key　parameters—including　torsion　angle,　initial　curvature,　flange　thickness,　and　web　thickness—on　the　ultimate　bearing　capacity　of　the　joint　through　parametric　analysis.　To　enhance　computational　efficiency,　the　study　develops　models　based　on　the　BP　neural　network　and　RF　algorithms.　The　results　indicate　that　the　dominant　failure　mode　of　aluminum　alloy　bending-torsion　gusset　joints　is　characterized　by　compressive　plastic　deformation　near　the　upper　flange　joint　and　localized　web　buckling,　exhibiting　notable　symmetry.　While　the　torsion　angle　minimally　impacts　the　joint＇s　bearing　capacity,　other　parameters　have　a　significant　influence　within　specific　ranges.　Comparisons　with　finite　element　simulation　results　demonstrate　that　the　Back　Propagation　(BP)　neural　network　excels　in　addressing　complex　nonlinear　problems,　effectively　capturing　the　nonlinear　relationships　governing　the　joint＇s　mechanical　properties.　Although　the　BP　neural　network　offers　superior　predictive　accuracy　compared　to　the　Random　Forest　(RF)　model,　it　requires　extensive　training　time　and　complex　parameter　tuning.　Conversely,　the　RF　model　is　advantageous　for　its　rapid　training　capabilities　and　strong　interpretability.　This　research　not　only　establishes　a　theoretical　foundation　for　designing　gusset　joints　in　aluminum　alloy　bending-torsion　members　but　also　offers　critical　insights　for　optimizing　joint　design　in　practical　engineering　applications.　©　2024　Institution　of　Structural　Engineers