Application of machine learning for real-time evaluation of salinity (or TDS) in drinking water using photonic sensors

Abstract

The world is facing an unprecedented problem in safeguarding 0.4 % of potable water, which is gradually depleting day-by-day. From a literature survey it has been observed that the refractive index (RI) of water changes with a change in salinity or total dissolved solids (TDS). In this paper we have proposed an automatic system that can be used for real-time evaluation of salinity or TDS in drinking water. A photonic crystal (PhC) based ring resonator sensor has been designed and simulated using the MEEP (MIT Electromagnetic Equation Propagation) tool and the finite difference time domain (FDTD) algorithm. The modelled and designed sensor is highly sensitive to the changes in the RI of a water sample. This work includes a real-time-based natural sequence follower, which is a machine learning algorithm of the naive Bayesian type, a sequence of statistical algorithms implemented in MATLAB with reference to training data to analyse the sample water. Further interfacing has been done using the Raspberry Pi device to provide an easy display to show the result of water analysis. The main advantage of the designed sensor with an interface is to check whether the salinity or TDS in drinking water is less than 1000 ppm or not. If it is greater than or equal to 2000 ppm, the display shows “High Salinity/TDS Observed”, and if ppm are less than or equal to 1000 ppm, then the display shows “Low salinity/TDS Observed”. The proposed sensor is highly sensitive and it can detect changes in TDS level because of the influence of any dissolved substance in water.