With　the　development　of　science　and　technology,　the　accuracy　requirements　for　solving　engineering　problems　are　getting　stricter　than　before.　Most　structural　design　optimization　problems　in　civil　and　mechanical　engineering　have　proven　to　be　the　non-deterministic　polynomial　hard　problems.　The　artificial　bee　colony　(ABC)　algorithm　has　been　proven　to　be　an　effective　method　of　design　optimization　problems.　This　paper　proposes　an　improved　ABC　algorithm　(DSM-ABC)　combined　with　dual-search　mechanism　containing　Lévy　flight　and　differential　self-perturbation　and　applies　it　to　three　classical　structural　design　problems,　including　cantilever　beam　design,　gear　train　design,　and　three-bar　truss　design.　The　experimental　results　of　benchmark　functions　from　CEC2005　reveal　that　the　proposed　DSM-ABC　algorithm　accelerates　the　convergence　and　improves　the　performance.　Eventually,　the　obtained　results　of　optimization　structural　design　problems　prove　that　the　DSM-ABC　algorithm　has　a　strong　superiority　compared　with　the　state-of-the-art　algorithms　in　solving　optimization　engineering　design　problems.　©　2013　IEEE.