For　improving　the　serving　life　of　the　threaded　connection,　an　elastic-plastic　contact　finite　element　method　was　used　for　analyzing　the　stress　distribution　of　the　connecting　threads.　After　that,　the　parametric　modeling　technique　was　used　for　establishing　the　multi-objective　optimization　model　where　the　load　distribution　uniform　and　the　stress　peak　values　were　considered.　This　model　was　then　solved　by　the　radial　basis　function　(RBF)　neural　network　method　and　the　elitist　non-dominated　sorting　GA　(NSGA-II).The　numerical　results　indicate　that　compare　with　the　original　design,　the　non-uniform　rates　of　load　distribution　between　the　teeth　in　the　optimized　design　are　reduced　by　13.6%,　and　the　maximum　of　equivalent　stresses　is　dropped　by　6.9%,　and　the　peak　values　of　the　first　principal　stress　are　cut　down　by　11.5%.　©　2017,　China　Mechanical　Engineering　Magazine　Office.　All　right　reserved.