The　addition　of　fly　ash　(FA)　to　concrete　will　not　only　reduce　the　CO2　content　of　concrete　but　also　affect　the　compressive　strength　of　concrete.　However,　most　studies　on　carbon　emissions　have　underestimated　the　complexity　of　the　relationship　of　emissions　with　the　concrete　mix　design　and　compressive　strength.　In　this　study,　a　database　of　mix　designs　was　established　by　considering　1062　datasets　related　to　the　laboratory　concrete　mix　designs　and　compressive　strength　values,　and　by　using　an　artificial　neural　network　model,　a　multiple　nonlinear　regression　model,　random　forest　model　and　support　vector　machines　to　simulate　and　predict　the　compressive　strength　of　concrete.　Moreover,　the　differences　in　the　life-cycle　CO2　corresponding　to　FA　concrete　among　three　different　allocation　types-no,　mass,　and　economic　value　allocations-based　on　two　functional　units　(FU)-per　unit　volume　and　per　unit　volume　and　strength　were　investigated.　Finally,　sustainable　mix　design　method　of　FA　concrete　was　established　to　find　the　optimal　FAC　mixtures　and　the　correlation　between　CO2　emission　and　concrete　mix　parameters　was　analyzed　by　the　maximal　information　correlation　(MIC).　The　results　indicate　that　the　neural　network　model　has　a　better　regression　effect　than　the　multiple　nonlinear　regression　model　and　other　machine　learning　models　on　the　prediction　of　compressive　strength　and　CO2　emissions.　For　different　functional　units,　the　CO2　emissions　under　FU1　increases　with　the　strength　of　concrete　regardless　of　different　allocation　types;　the　CO2　emissions　under　FU2　varies　considerably　with　changes　in　the　amount　of　added　FA　under　no　allocation　and　economic　value　allocation.　The　CO2　emission　under　FU1　has　a　stronger　correlation　with　various　mix　parameters　than　that　under　FU2.　©　2023　Elsevier　Ltd