Elemental strain mode difference combined with RBF neural network is proposed to detect the damage of space truss structures. Finite element models of square pyramid space truss and honeycomb shaped triangular pyramid space truss were established. With elemental strain mode difference as criterion, the damage locations of single pole or multi-poles were determined. The damage degrees of the structures were investigated by using RBF neural network. The results show that this method can be used to detect the damage locations and damage degrees of space truss structures。