Evaluation of sequential batch reactor wastewater treatment plant using artificial neural network /
Domestic wastewater is one of the main pollution sources in municipal areas. Wastewater treatment plant is obviously the most important component in eliminating the unwanted materials in domestic and industrial effluents before discharging into the water bodies. Wastewater treatment process is in us...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
Kuala Lumpur :
Kulliyyah of Engineering, International Islamic University Malaysia,
2012
|
| Subjects: | |
| Online Access: | http://studentrepo.iium.edu.my/handle/123456789/4792 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Domestic wastewater is one of the main pollution sources in municipal areas. Wastewater treatment plant is obviously the most important component in eliminating the unwanted materials in domestic and industrial effluents before discharging into the water bodies. Wastewater treatment process is in use since 17th century for treating both domestic and industrial effluent, but its performance is not as expected due to lack of adequate knowledge on the mechanisms involved. The deterministic models are unable to fully comprehend the complex nature of the processes that comprises physical, chemical and biological processes of the wastewater treatment plant. The combined application of Principal Component Analysis and Artificial Neural Network tools is a recent development in modelling wastewater treatment plant characterization. Therefore, in this study artificial neural network and principal component analysis were integrated using MATLAB® to train neural network for Sequential Batch Reactor wastewater treatment plant. The study comprises optimization of the network properties; artificial neural network development and hybrid network development for influent biological oxygen demand, influent ammonical nitrogen, effluent biological oxygen demand, and effluent ammonical nitrogen of Bandar Tun Razak sewage treatment plant in Kuala Lumpur. The models were verified using separate wastewater samples collected from the plant. The results showed that hybrid model outperformed its corresponding normal artificial neural network and recorded a higher correlation coefficients for training (0.7362), testing (0.7678) and verification (0.7699) datasets with the respective mean absolute errors of 13.75, 11.29 and 12.76. The hybrid networks for the remaining model performances were lower than the normal corresponding network. The correlation coefficients (and mean absolute errors) for training, testing and verification of the best effluent biochemical oxygen demand network were 0.7990 (1.43), 0.8276 (1.76) and 0.7344 (1.73), respectively. Generally, influent and effluent ammonical nitrogen model predictions ability were weak compared to biological oxygen demand model. The best network for prediction of influent ammonical nitrogen was the normal artificial neural network model with correlation coefficients and (mean absolute errors) of 0.5648 (2.54), 0.6024 (3.48) and 0.6284 (7.63) for training, testing and verification datasets, respectively. The best model for effluent ammonical nitrogen simulation was the normal artificial neural network. It recorded the highest correlation coefficients and least (mean absolute errors) for training, testing and verification as; 0.7984 (4.18), 0.6224 (5.59) and 0.6507 (4.15), respectively. The model performances were satisfactory when compared with the published results as shown in Chapter four. The combination of principal component analysis and artificial neural network techniques to evaluate the strength of training neural networks and method of model generalization are the major contributions of this study. |
|---|---|
| Item Description: | Abstract in Englisah and Arabic. "A dissertation submitted in fulfilment of the requirements for the degree of Master of Science (Biotechnology Engineering)."--On t.p. |
| Physical Description: | xix, 150 leaves : ill. : 30cm. |
| Bibliography: | Includes bibliographical references (leaves 129-134). |