Optimum Coagulant Dose Prediction for Water Treatment using Artificial Neural Network (Case of Legedadi Water Treatment Plant)
| dc.contributor.advisor | Agizew, Nigussie (PhD) | |
| dc.contributor.author | Eyerusalem, Alemayehu | |
| dc.date.accessioned | 2022-05-13T09:14:05Z | |
| dc.date.accessioned | 2023-11-28T14:01:09Z | |
| dc.date.available | 2022-05-13T09:14:05Z | |
| dc.date.available | 2023-11-28T14:01:09Z | |
| dc.date.issued | 2022-04 | |
| dc.description.abstract | Water obtained from surface or subsurface sources is crucial for life. But direct use of raw water has serious health risks. Different water treatment processes can be used to make the raw water safe for domestic purposes. Coagulation and Flocculation is one of the treatment stages used to remove colloidal particles through the use of chemicals (coagulants) that enable the formation of larger flocs that can be easily removed by sedimentation and filtration. Determination of the optimum coagulant dosage is important to meet the required water quality. For instance, a high coagulant dosage may lead to high residual chemical in the treated water, high sludge volume, and increased load on the filter units. All these can result in poor water treatment performances, high operational costs, and process complications. Jar test has been used widely to determine the optimal coagulant dosage for a given water quality. However, this practice has some drawbacks such as it is time-consuming, the probability of making errors is high and it is impractical for highly variable raw water turbidity. Consequently, the need to develop new tools and techniques to determine optimum coagulant dosage becomes important. However, quantifying the relationship between the process inputs and output in the water treatment unit process is very difficult with the existing process model. Thus, Artificial Neural Network was used to develop the models in this research as it is a robust technique that allows the development of a multi-variable and complex non-linear relationship. ANN Multi-Linear Perceptron type with one hidden layer was used to simulate the jar test for the optimum coagulant dosage forecasting. Two models were developed and their performances were evaluated based on Root mean squared error (RMSE) and coefficient of determination (R 2 ). The first model which enables prediction of turbidity for a given coagulant dosage and other factors was the process model. The second model which allows determination of optimum coagulant dosage to attain the desired turbidity level is the Inverse Process Model. The RMSE and R 2 values were found to be 0.0748 NTU and 0.6121, respectively for the Process model. For the Inverse Process model the RMSE and R 2 values were found to be 0.225mg/l and, 0.9823, respectively. These indicate that a good performance of the ANN-based model to simulate the jar test could be obtained. The models can be used to optimize the coagulation process within the available raw water turbidity range of Legedadi Water Treatment Plant. | en_US |
| dc.identifier.uri | http://etd.aau.edu.et/handle/12345678/31652 | |
| dc.language.iso | en | en_US |
| dc.publisher | Addis Ababa University | en_US |
| dc.subject | Artificial Neural Network | en_US |
| dc.subject | Optimum coagulant | en_US |
| dc.subject | prediction, process | en_US |
| dc.subject | inverse process model | en_US |
| dc.title | Optimum Coagulant Dose Prediction for Water Treatment using Artificial Neural Network (Case of Legedadi Water Treatment Plant) | en_US |
| dc.type | Thesis | en_US |