engleski

DATA ANALYSIS FROM EXPERIMENT DESIGN TO ARTIFICIAL NEURAL NETWORKS – APPLICATION IN WATER TREATMENT AND QUALITY CONTROL

Over the last 20 years, many scientists have had certain theories about the number of experiments and the amount of data needed to successfully test and analyse a certain processes. Design of experiments is a branch of applied statistics which can provide an answer to that question. The importance of a well-designed experiment is significant because the data obtained in that way can be successfully analysed by different statistical methods, optimized or further used in development of artificial neural networks (ANNs) which can help in process monitoring and prediction. The primary aim of this paper was to test the applicability of experiment design on processes which involve different water treatments (flocculation, ozonation). The secondary aim was to determine whether the data obtained from designed experiments can further be used in ANN training, testing and validation. Examples shown in this paper confirm the benefits of experiment design and confirm that ANNs can be successfully used for water treatment process monitoring and control, as well as for prediction of specific responses (targets) defined in the treatment process with high accuracy (the achieved values of R2 for train, test and validation were higher than 0.9 in all presented cases).

water treatment ; quality control ; ANN ; experiment design

nije evidentirano