Civil　infrastructure　O&M　requires　intelligent　monitoring　techniques　and　control　methods　to　ensure　safety.　Unfortunately,　tedious　modeling　efforts　and　the　rigorous　computing　requirements　of　large-scale　civil　infrastructure　have　hindered　the　development　of　structural　research.　This　study　proposes　a　method　for　impact　response　prediction　of　prestressed　steel　structures　driven　by　digital　twins　(DTs)　and　machine　learning　(ML).　The　high-fidelity　DTs　of　a　prestressed　steel　structure　were　constructed　from　the　perspective　of　both　a　physical　entity　and　virtual　entity.　A　prediction　of　the　impact　response　of　prestressed　steel　structure＇s　key　parts　was　established　based　on　ML,　and　a　structure　response　prediction　of　the　parts　driven　by　data　was　realized.　To　validate　the　effectiveness　of　the　proposed　prediction　method,　the　authors　carried　out　a　case　study　in　an　experiment　of　a　prestressed　steel　structure.　This　study　provides　a　reference　for　fusion　applications　with　DTs　and　ML　in　impact　response　prediction　and　analysis　of　prestressed　steel　structures.