以梁桥节点最大位移改变率作为损伤程度伤识别指标,分别采用广义回归神经网络(GRNN)算法和ε-支持向量回归机(ε-SVR)算法,进行损伤程度识别研究。通过对一座铁路双线简支钢桁梁桥某杆件的损伤程度识别研究发现:(1)GRNN损伤程度识别模型具有一定的抗噪能力,不具有泛化性。(2)SVR损伤程度识别模型具有很强的抗噪能力和很好的泛化性。(3)以桥梁节点最大位移改变率作为损伤程度识别指标时,数据回归算法不能采用GRNN算法,应采用ε-SVR算法。 Taking the rate of change of maximum displacement of the bridge as the identification index of injury degree, the generalized regression neural network (GRNN) algorithm and ε-SVR (ε-SVR) algorithm were used respectively to study the damage degree identification. Through the identification of the degree of damage to a bar member of a railway double-tracked simply supported steel truss girder bridge, it is found that: (1) The GRNN damage degree identification model has certain anti-noise ability and does not have generalization. (2) SVR damage identification model has strong anti-noise ability and good generalization. (3) When the change rate of maximum displacement of bridge nodes is used as the identification index of damage degree, the data regression algorithm can not adopt GRNN algorithm, ε-SVR algorithm should be adopted.