Drift Compensation for pH IsFET Sensor Using NARX Neural Networks

Abstract

This paper introduces a Nonlinear Autoregressive Neural Network (NARX) to predict the sensor error of IsFET pH drift with accuracy over the long period. The Bayesian Regularization (BR) backpropagation was used as network training function for this problem and combined with different delay and hidden layer. The results were compared to predict the sensor error in buffer solution pH 4, pH 7 and pH 10 over the time. The NARX performance will be measure based on the value of Mean Squared Error (MSE) and coefficient of determination (R2). The results proved by using Bayesian Regularization with 10 hidden nodes and 50 delays produced the accurate sensor error prediction. This research will provide the significant contributions to the implementation of IsFET pH sensor drift compensation over the time.