Hybrid fuzzy neural network for diagnosis - application to the anaerobic treatment of wine distillery wastewater in a fluidized bed reactor

In this paper, we present a hybrid approach that uses both fuzzy logic and artificial neural networks for on-line detection and analysis of problems occurring in a 120 liter anaerobic digestion fluidized bed reactor for the treatment of wine distillery wastewater. The raw data available on the process (i.e., pH, temperature, recirculation flow rate, input flow rate and gas flow rate) are preprocessed using fuzzy logic to build a vector of features (i.e., a pattern vector). This feature vector is classified into a prespecified category (i.e., a class) which is a state of the system, according to discrimination fuzzy rules. An artificial neural network is then used to classify the process states and to identify the faulty or dangerous ones. This approach was developed to handle in real time problems such as, for example, foam forming, sudden changes in the effluent to be treated (due to a change in concentration), pipe clogging (due to struvite formation) or bad temperature regulation (due to improper setting of the control parameters).