Feasibility of a Specklegram-Based Quasi-Distributed Temperature Sensor With Principal Component Analysis and Variational Autoencoder

Specklegram-based optical fiber sensors have gained attention for their advantages of sensitivity, low cost, and intelligent sensing ability in fields such as force, small deflection, single-point temperature sensing, and so on. To enhance specklegram-based distributive temperature sensing through a multimode optical fiber (MMF), this work proposes a quasi-distributed approach involving a triple-color illumination with light emitting diodes (LEDs) and a hybrid model that combines principal component analysis and variational autoencoder (PCAVAE). The proposed MMF optical sensor demonstrates strong performance in sensing the distributive temperature configurations of four heaters, showcasing its capability in detecting quasi-distributed temperature information in the range of 30 °C-60 °C, achieving a spatial resolution of 3 cm and a temperature resolution of 1 °C with a mean absolute error (MAE) of 1.85~^{\circ } C and a root mean square error (RMSE) of 2.63~^{\circ } C, respectively. The proposed sensor opens a gateway for exploiting multiple sites sensing capability with a single optical fiber specklegram and exhibits application potentials for low-cost quasi-distributed optical temperature detection.