Abstract: In view of the weakness of existing bridge evaluation methods‚a neural network method was first intro-duced to evaluate the damage conditions of a bridge．The evaluation effects of several common artificial neural network （ANN） models were then compared．The ANN models were finally trained and tested based on the maintenance data of1018bridges on the nationa- l grade roads in Guangdong．It is found that the neural network method is effective in e-valuating the bridge conditions‚more than60％ of the bridge grade being correctly evaluated and the average evaluation error of each bridge being0．25grades．

Key words: artificial neural network, bridge evaluation, LVQ network, RBF network, Elman network