S. Bayoumi, K. Ewida, and D. Saad El-Din*
Environmental Engineering Department, Faculty of Engineering, Zagazig University, Egypt.
The existence of strong non-linearity effects, time variant parameters and multivariable
coupling don’t enable the adoption of simple models to predict either the wastewater
treatment processes efficiency or its effluent quality. Nowadays, an Artificial Neural
Networks (ANNs) is recognized as a very promising tool for relating input data to output
data. Especially where processes are complex, the applications of ANNs are widespread and
vary from process optimization for water and wastewater treatment, scenario evaluation,
demand forecasting, on-line steering of wastewater treatment plants and sewage systems,
resource optimization and management up to financial modeling.
Besides delineating the approach and methodology for the development of successful ANNs
model of Waste Water Treatment (WWT) processes, the ability of ANNs to predict the
quality (BODout and CODout) of the raw domestic sewage effluent from Upflow Anaerobic
Sludge Blanket (UASB) model was verified in this paper. Using Matlab (version 6.5), various
network architectures, differing in the number of hidden layers and nodes were tested in order
to find the optimized solution in terms of both precision and learning time. The effectiveness
of each ANN configuration was verified by Mean Absolute Error (MAE) method. Data for
calibrating, testing, and verifying the developed ANN of UASB model was obtained from the
operation of two identical pilot-scale models simulating UASB reactors that designed,
constructed from PVC pipes, 200 mm diameter and 2500 mm height, and built at Kawmeya
sewage pumping station, Zagazig, Egypt.