An　optimum　mixture　proportion　design　method　of　reactive　powder　concrete　(RPC)　based　on　an　artificial　neural　network　(ANN)　and　harmony　search　(HS)　algorithm　was　developed.　ANNs　were　adopted　to　establish　the　relationship　between　design　parameters　(water-binder　ratio,　silica　fume　content,　sand-binder　ratio,　and　steel　fiber　content)　and　properties　(compressive　strength　under　standard　curing　and　autoclaved　curing,　splitting　tensile　strength　under　autoclaved　curing,　and　slump)　of　RPC,　and　the　HS　algorithm　was　used　to　design　and　optimize　RPC　mixture　proportions　with　the　objective　criterion　of　minimum　cost　while　meeting　all　property　requirements.　The　proposed　method　can　consider　the　influence　of　curing　regimes,　and　its　reliability　was　verified　by　experiment　data.