Unique Responses in Graphene Sensor to Predict Temperature and Humidity using ANN Model

This article introduces a method to determine the values of temperature (T) and relative humidity (RH) in a mixed environment. Graphene with defective sp ² lattice gives a non-linear reading of resistance (R) when deployed into the test environments with varied T = 30 – 60 °C, RH = 10 – 60%, and biases of 1 – 100 mV. Perturbing the RH during a very low bias of 1 – 5 mV makes the carrier transportation less sensitive to the interaction between the dipole water droplets with graphene lattice, thus creating unique responses of R values. The parameter of T/RH and four unique responses of Δ1 – Δ4 generated by the biases of 5, 10, 50, and 100 mV are fed into an artificial neural network (ANN) model. A total of 326 datasets with a fine combination of T/RH results in training and testing performance with regression of 0.916 and 0.865, respectively. The best prediction of T and RH by the ANN model has an accuracy of at least 97% and 68%, respectively.