Simple Machine Learning Approach for Liquid Concentration Estimation Employing Fiber Optic Tip Sensor

Abstract

Optical fiber sensors have been remarkably demonstrating its capacity for measuring various liquid concentrations. Traditionally, these sensors are characterized with a known concentration value and having its response analyzed in the form of an optical spectrum signal. However, questions arise as to whether there is the possibility to automate the sensing system by detecting an unknown concentration value using this type of sensor especially for real sensing applications. This paper introduced a simple machine learning approach by modelling a single linear regression to predict the unknown liquid concentration value of the sensor by using collected data from experiment work. The result showed that machine learning has the ability to measure the liquid concentration value of the sensor with small error values while accruing an accuracy of 85.67%. This promising result shows the potential of machine learning for integration into optical fiber sensing systems.