Design And Prediction Of Temperature-Strain Relationships In Fiber Bragg Grating Sensors Using Machine Learning

A wide range of research has been conducted in the field of optical fiber sensors on Fiber Bragg Grating sensors for the measurement of various physical properties such as strain, temperature, pressure, motion, etc. The most significant advantages of fiber-optic sensors are their low weight, compact size, passive nature, and immunity to electromagnetic interference (EMI). In addition, they have a broad bandwidth, great sensitivity, and environmental ruggedness. They also use less power and cause little attenuation. In this paper, the FBG sensor is designed to measure strain and temperature for a given interval of time. The calibration factor of 10.04 pm/degrees c for temperature and 1.22 pm/mu epsilon for strain, is achieved in this work, which is on par with standard FBG calibration factor values. Both the transverse effect and the strain transfer error are intrinsically prevented by the designed FBG sensor. The respective experimental results are verified using simulation carried out in Matrix Laboratory. In addition to this Machine Learning algorithms were employed to predict the strain, which will directly measure the strain reducing the need of the extensive physical set up. This study explores the application of two machine learning algorithms, Linear Regression (LR) and Random Forest (RF), to predict strain in a material based on temperature data. Using a dataset collected over a time series of strain and temperature readings, both models were evaluated for accuracy, and Random Forest demonstrated superior performance with R2 value of 0.94. This study contributes to the field of sensing technology by introducing a unique method.