Minimization of sensing elements for full-range optical pH device formulation

Abstract

The goal of this work is to find the minimum number of sensing elements that can be used to build a sensor array suitable for pH prediction for a full-range (0–14) optical pH sensor array based on hue (H) data from the HSV color space. The hue of each element coming from the pH sensor array imaging is used as input data for a neural network that provides the pH prediction approximation as the output response. This problem may be considered a multicriteria optimization task with the dual objectives of error minimization between the network pH prediction and the reference pH in the calibration data, and of minimizing the network complexity and the number of network inputs. To that end, this work proposes a multi-objective optimization method applied to a collection of 11 sensing elements that returns a set of optimal networks considering the Pareto optimality criterion. A solution from the Pareto front was selected to achieve a minimum pH prediction error with the minimum number of sensing elements. After this analysis, it can be concluded that the use of a sensor array made up of 4 sensing elements offers a good pH prediction (Mean Square Error of 0.052) over the full-range working with 4 hidden neurons.