| Title |
Damage Estimation Technique of a Structure Using Committee of Neural Networks |
| Keywords |
Committee of Neural Networks ; Damage Estimation ; Modal Parameter ; Distributed Fiber Optic Sensor |
| Abstract |
In this study, committee technique for neural networks is applied to damage estimation of structure to enhance estimation result of single neural network. Local minima sensitive to initial synaptic weights and non-uniqueness of solution due to noise and limited number of measurements may be resolved by employing the committee technique, which is a statistical approach averaging damage indices in functional space. First, a numerical study for a simple beam was carried out. Damage locations and severities of the simple beam were reasonably estimated using change of modal parameters due to damage and committee of neural networks, and the estimation results of the single neural network could be effectively enhanced by applying the committee technique. The precedent damage estimation results for a cantilever beam using static and dynamic strain data from the distributed fiber optic sensor and the single neural network could also be considerably improved by applying the committee technique for neural networks. |