Long-gage fiber optic sensors for dynamic strain measurement and structural identification

B Xu,ZS Wu

2003-01-01

Abstract:A structural parametric identification strategy with neural networks using dynamic macro-strain measurements in time domain from the experimental measurement of a series of long-gage Fiber Bragg Grating (FBG) sensors is testified. A beam model with known mass distribution is considered as an object structure. An array of long-gage FBG sensors with different gage length is bounded on the structure to measure reliably and accurately free vibration macro-strains. By the proposed methodology, without any eigenvalue analysis, the structural stiffness in the form of Young's Modulus can be identified. First, a reference structure with the same degree of freedoms and topology as the object structure is assumed. An emulator neural network is presented to identify the reference structure. Free vibration macro-strain responses of the reference structure in time domain are directly used to train the emulator neural network. The trained emulator neural network can be used to forecast the free vibration macro-strain responses of the reference structure with enough precision and decide the difference between the free vibration macro-strain responses of other assumed structures with different structural parameters and those of the reference beam structure. The root mean square (RMS) error vector is presented to evaluate the difference. Subsequently, corresponding to each assumed structure with different structural parameters, the RMS error vector can be calculated. By using the training data sets composed of the structural parameters and the corresponding RMS error vectors, a parametric evaluation neural network can be constructed and trained. Based on the trained parametric evaluation neural network, the stiffness of the object beam structure can be identified by the direct use of the macro-strain responses from long-gage FBG sensors. It is shown that the parametric identification strategy using macro-strain measurement from long-gage sensors has the potential of being a practical tool for a health monitoring methodology applied to existing damaged structures or newly constructed structures.