Hybrid Model of a Wastewater-Treatment Electrolytic Process

This work is focussed on the development of models which can be implemented in efficient control devices to properly schedule batch electrolysis processes for wastewater treatment. The main goal has been to develop a hybrid approach based on a phenomenological mathematical model and a neural network methodology, in order to improve the quality of the model designed for electrolysis process of wastewaters. A well-known phenomenological model for the electrolyses of wastewaters and different neural network architectures were used, along with homogenous and heterogenous transfer functions. The models were validated with the results of a series of experiments consisting of the electrochemical oxidation of phenol wastes in different process conditions. The best results were obtained designing a feed-forward neural network with tangent hyperbolic axon as homogeneous transfer function (relative error of 8 %).