Publication: Study on adaptive neuro-fuzzy inference system for sulphate-reducing bacteria detection
Loading...
Date
2012-06-01
Authors
Lim, Khai Sian
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Bacteria are one type of microorganism which contributes majorly in microbial contamination. Sulphate-reducing bacteria (SRB) are anaerobic and may lead to corrosion of iron material. The bacteria have also contributed for pitting corrosion in copper containing alloys. The prediction of existence sulphate-reducing bacteria in a water system is very crucial to cope with corrosion issue in the water system. In this regard, a method of using an adaptive neuro-fuzzy inference system (ANFIS) is studied for the modeling and predicting the existence of SRB in the medium. The general structure and criteria of the system are studied. Experimental data collected from two different medium are used for training the ANFIS system. Three parameters (voltage, temperature and humidity) of the medium are selected as major factors to indicate the existence of the bacteria. The parameter which gives most significant impact for the prediction model is identified. Three membership functions (trapezoidal, bell-shaped, triangular) are used for training the data for comparison purpose. Results showed that the ANFIS with trapezoidal membership function is the best with average error, 1.66E-07 at epoch 250. The comparison analysis is further validated with 10 experiments, by using the input-output pairs (parameters of medium) which are chosen randomly from the total data set.