Development of an artificial neural network system for sulphate-reducing bacteria detection by using model-based design technique

Abstract

The paper demonstrates an approach for the implementation of artificial neural network (ANN) as a pattern recognition system in the development of electronic nose devices for sulphate-reducing bacteria (SRB) detection. The neural network algorithm was modeled using Xilinx System Generator and can be realized in Xilinx FPGA. The neural network was trained using the Levenberg-Marquardt back propagation algorithm in Matlab based on data collected from experiments, in which samples that contained either the broth medium with SRB or only the broth medium were used. In Matlab, sigmoid function is used as the activation function for the neurons in hidden and output layers. However, the block sets provided by Xilinx System Generator do not contain the sigmoid function. Thus, the sigmoid function was modeled through the expansion of the Maclaurin power series. Adders and multipliers blocks are used to calculate the series equation. This paper discusses on the similarity of sigmoid function generated using Matlab and approximation of sigmoid function generated using Maclaurin power series. The simulated results show that the absolute error between the values of sigmoid function based on Matlab and Xilinx System Generator using Maclaurin power series are not more than 1.38% when the inputs fall in the range between -0.5 to 0.5. Thus, the modeling of sigmoid function in Xilinx System Generator has been successfully modeled with the accuracy of the ANN to detect SRB is 100%.