This　study　presents　a　novel　simulation-based　optimization　approach　to　enhance　drive　shaft　fatigue　strength　using　Bayesian-Kriging　surrogate　model.　The　methodology　incorporates　three　key　innovations:　(1)　a　Python-based　parametric　model　development　using　ABAQUS,　(2)　an　entropy　weight　TOPSIS　method　for　dimensional　contribution　analysis,　and　(3)　a　Bayesian　Expected　Improvement　(BYEI)　strategy　for　enhanced　Kriging　model　performance.　The　proposed　method　significantly　reduces　computational　cost　while　maintaining　accuracy　through　intelligent　sampling　and　model　updating　strategies.　Validation　using　three　test　functions　demonstrates　superior　convergence　speed,　stability,　and　accuracy　compared　to　traditional　methods.　Application　to　drive　shaft　optimization　achieved　a　56.11　%　improvement　in　fatigue　life　while　maintaining　structural　constraints.　Compared　with　the　other　two　models,　Bayesian-Kriging　model　has　obvious　advantages　in　prediction　accuracy.　The　results　demonstrate　the　method＇s　effectiveness　for　complex　mechanical　component　optimization,　particularly　in　scenarios　requiring　balance　between　computational　efficiency　and　accuracy.