Abstract: Loading condition of a three-span truss structure was monitored by using electromechanical impedance method and BP neural network was used for the localization and quantification of structural loading. First,the effect of excitation frequency on the impedance of PZT element was studied and the most sensitive frequency band of 190- 200 k Hz was selected. Second,the independence of the two monitoring nodes was tested so well and this is helpful for loading positioning. Finally,when the truss structure was loaded,the imaginary part of impedance data that were compressed reasonably was collected and parts of these were regarded as the input samples of the neural network. The comparison of the data before and after compression illustrates the reliability of this data compression method. The rest parts after the same process were served as the training samples to test the trained network.Resultsshow that the trained neural network has excellent abilities of recognition and classification of information that can also accurately locate and quantify the load in the truss structure.