Identification of damage in dome-like structures using hybrid sensor measurements and artificial neural networks

A damage detection scheme using multi-type sensor-based hybrid sensing and artificial-neural-network- (ANN-) based information processing was developed for dome-like structures used in civil infrastructure. Accelerometers and strain sensors were used to provide a hybrid measurement with the purpose of acquiring rich information associated with structural damage. The optimal placement of multiple sensors was explored so as to capture the most appropriate and sensitive signal features (damage parameter vectors) for damage characterization. A back-propagation ANN was constructed with the inputs extracted from the hybrid measurement. To validate the capacity of the proposed damage identification scheme, finite element analysis was conducted to identify damage in a Schwedler dome structure as an example. The performance of ANNs, trained by three kinds of damage parameter vector extracted from signals captured by (i) a sole accelerometer, (ii) a sole strain sensor, and (iii) both kinds of sensor was compared, to observe that the one trained by hybrid sensor measurement outperformed the others. Error analysis for a series of parametric studies, in which noise at different levels was included in the training input, was further carried out, and robustness of the proposed damage identification scheme under noisy measurement was demonstrated.