Smart Strain Approximation Surface-Mounted Optical Fiber Strain Sensor

Abstract

This study is aimed to develop a smart neural network perceptron model for strain prediction using fiber optic sensor signals. Optical parameters corresponding to surface-mounted optical fiber are obtained experimentally under static loading conditions. Four variations are used by creating external damages to study the strain variations on healthy, single damage and multiple damage beam structures. The strain values are obtained by using phase difference and change in intensities as input data for the feed-forward backpropagation neural network model. A comparative study of pre-existing analytical solutions, conventional strain gauge measurement, and finite element analysis is performed. The neural network model proposed in this work provides more close results to the results obtained by strain gauge and FEA analysis as compared to analytical analysis carried out by Haslach. © 2020, Springer Nature Singapore Pte Ltd.