In　recent　years,　single-layer　spherical　lattice　shell　structure　has　garnered　extensive　attention　in　the　fields　of　spatial　grid　structures　and　prefabricated　construction　techniques.　The　design　of　joints　constitutes　an　indispensable　component　of　single-layer　spherical　lattice　shell　structures.　In　order　to　ensure　that　the　joint　design　aligns　with　both　mechanical　performance　and　ease　of　construction,　a　comprehensive　approach　to　joint　design　is　imperative.　This　study　introduces　a　novel　intelligent　design　framework　for　rigid-jointed　joints　in　single-layer　spherical　lattice　shell　structures.　The　architecture　and　loss　function　of　this　framework　are　specifically　tailored　to　evaluate　the　performance　of　various　joint　design　schemes.　By　integrating　a　multi-head　attention　mechanism,　the　Graph　Attention　Neural　Network　can　extract　global　structural　design　parameters.　In　contrast　to　conventional　design　methods,　the　latter　lacks　consideration　for　the　overall　structural　performance　and　optimization　of　joints.　By　incorporating　the　research　group＇s　enveloping　tubular　staggered　flange　spatial　joint,　this　method　not　only　addresses　the　stability　of　the　overall　structure　but　also　accounts　for　the　rigidity　requirements　of　the　joints.　Numerical　experiments　demonstrate　that　the　model　loss　value,　when　employing　both　multi-head　attention　mechanism　and　graph　attention　mechanism,　decreases　faster　compared　to　the　model　loss　value　with　only　the　graph　attention　mechanism,　and　it　tends　to　plateau　after　approximately　70　iterations.　The　proposed　intelligent　design　method　for　rigid　joints　in　prefabricated　single-layer　spherical　lattice　shell　structures　offers　a　fresh　perspective　on　the　design　of　such　structures　and　advances　the　development　of　spatial　grid　structures　and　prefabricated　construction　techniques.　©　2024