Mathematical Modeling of Ion Sensitive Field Effect Transistor and Metamodel Based Optimization Simulation for Detection of Aflatoxin B1

Mathematical modeling is a powerful tool for prediction of Ion Sensitive Field Effect Trnsistor (ISFET) response and optimization of its functional parameters. In this study the optimal values of drain current, drain voltage and initial concentrations of substrate and enzyme parameters were determined to achieve maximum of ISFET response for detection of Aflatoxin B1 (AFB1). Optimization was performed by using Genetic Algorithm (GA) and based on numerical solution of ISFET governing differential equations by means of Finite Element Method (FEM) and COMSOL Multiphysics software. The objective function of GA was defined through substituting simulated model by Artificial Neural Network (ANN) metamodel. The results showed that ISFET simulated FEM model has a MAPE equal to 1.06 % in prediction of ISFET response compared with experimental results. With FEM model, 1296 virtual experiments were simulated to achieve necessary data base for train ANN metamodel. By evaluation of different ANN structures, trained ANN with 4-45-1 structure was selected which has MAPE equal to 0.04 %, 0.07% and 0.05% at train, validation and test phase respectively. ISFET optimization results states that by using of GA determined optimal values of drain current, drain voltage and initial concentrations of substrate and enzyme parameters, extremum response of ISFET equal to 44.44 % was achieved.