Development of an artificial neural network model for predicting the performance of a reverse osmosis (RO) unit

Abstract

Desalination is one of the most widely used techniques to produce pure water from seawater, groundwater, wastewater or brackish water. This technique has gained wide spread acceptance throughout the world especially in arid and dry regions like the Middle East which possesses the largest capacity desalination plants in the world. On the other hand, Australia which is characterised by its arid regions does not utilise desalination as a source of providing pure water as compared to the Middle Eastern regions. The increasing population in the capital cities and the inhabitants of the isolated mining towns and smaller remote communities would benefit from using desalination. Reverse Osmosis (RO) is the one the widely used desalination technique in the world. It offers the distinct advantage over the other desalination techniques because it consumes low energy, provides a high quality final product, easy installation and flexible design. RO works on the principle of osmosis where the transfer of the solvent is done through a semi permeable membrane under the influence of a concentration gradient. The quality of the pure water that passes through the membrane during the RO process is a function of the difference between the applied pressure and the osmotic pressure of the solution.From the results obtained the simulated results for solute rejection and permeate flux are close to the analytical i.e. experimental obtained results. Traditionally membrane performance has been predicted by polynomial correlations but the neural network model offers the advantage allowing the user to visualise the entire operation, capability of learning from the experimental results and obtaining highly accurate findings. The model generated in this study will provide the solid foundation for extending the ANN model applicability to cover several feedwater sources over a range of different pressures and concentrations.The thesis describes the development of an Artificial Neural Network Model for predicting the two important parameters of Reverse Osmosis i.e. salt rejection and permeate flux. The thesis comprises of six sections including the conclusions and recommendations for future work.Chapter details the general background of the current state of water supplies in Australia, looks at the existing RO plants that have been set up or being planned for the future and establishes the various uses of RO practices.Chapter 2 contains a detailed literature review on desalination and its various processes, understanding the way RO works and the factors that affect the RO operation and performance.Chapter 3 presents the modelling approach used during this study and introduces the reader to artificial neural networks and the manner in which they function.Chapter 4 contains a brief description of the experimental procedures conducted by Nasir (2005) and this experimental data forms the basis for the model development.Chapter 5 deals with the development of the artificial neural network model for predicting the performance of a RO system handling different feedwater sources and validation of the developed ANN model.Chapter 6 presents the conclusions obtained from this study and the recommendations for future work to be conducted in order to expand the developed ANN code to cover different feedwater samples.