Study of Members Damage Early-Warning System for Frame Structures Based on Neural Network

Two damage anomalous filters which were set up by BP neural network have been used to alarm the damage in structural members. After dealing with eigenparameter extracted from damaged and intact structure, different input data is considered for setting up different damage warning anomalous filters. Filter □: the first eight natural frequencies are chosen as input data of network. Filter □: one mode damage index DSI₁ is chosen as input data of network. Six damaged work conditions have been discussed in the paper, and the result of analysis shows that using the composite indicator for structural damage detection is accurate, efficient and convenient in engineering.