Abstract: The compressive strength of damaged concrete reinforced with fiber reinforced polymer composite (FRP) has an important guiding significance in repairing of concrete columns. However, the existing model cannot capture the compressive strength of FRP hardening and softening confined damaged concrete with circular and square cross section. In order to fill this gap, an experimental database of 46 FRP hardening confined square damaged concrete, 210 FRP hardening confined circular damaged concrete and 35 FRP softening confined circular damaged concrete was established. Based on the characteristics of generalized regression neural network (GRNN) and database, the GRNN compressive strength model of FRP confined damaged concrete was developed. The GRNN model was compared with the existing model. The results show that the GRNN model can accurately predict the strength of FRP confined damaged concrete columns.