Using Neural Networks to Process a Large Database to Quantify Causes of Reverse Osmosis Fouling

Artificial neural network (ANN) modeling was used to analyze a multi-year process database to extract detailed information about Reverse osmosis membrane fouling in a complex, dynamic Ground Water Recovery System (GWRS) process run by the Orange County Water District (OCWD). OCWD wants to improve the RO?s performance and sought to mine the large database it has compiled to obtain new process knowledge. The analyses groundtruthed cause-effect parameter relationships that were generally known, but had not been previously quantified. This suggests a possible role for predictive models in extending membrane run times. The analysis also produced previously unknown information about RO stage interactions and the relative effectiveness of different membrane cleaning procedures