| Author | Kim, Chan Moon |
| Note | A dissertation submitted in partial fulfillment of the requirements for the
degree of Doctor of Engineering in
Mechatronics |
| Publisher | Asian Institute of Technology |
| Abstract | Coagulation process is known to be crucial in drinking water treatment so as to attain acceptable treated water quality. However, operators still feel that the determination of coagulant dosage is challenging since coagulation is inherently nonlinear and complicated process. In this dissertation, ANNs, ANFIS, and K-means clustering are applied to determine coagulant dosage. For coagulation dosage prediction model by raw water quality, MLP, ANFIS, GRNN are applied. MLP shows the best result in high turbidity zones, meanwhile, ANFIS provides consistent result and the best performance in low turbidity zones and high disorder zone of coagulant dosage. GRNN shows high accuracy in the highest turbidity zone. A hybrid of K-means clustering and ANFIS is designed to estimate ideal coagulant dosage and settled water turbidity. The K-means-ANFIS model proves it is more consistent and precise than a single ANFIS and ANN model. The hybrid model can be used for adjusting both of treated water quality and production expenses. To improve coagulation process model, NARX-ANFIS model is proposed. The model accuracy of NARX-ANFIS is slightly better than K-means-ANFIS by a recurrent input. The proposed NARX-ANFIS can be another good choice for real-time coagulant dosing control using past trend information of settled water turbidity. |
| Year | 2017 |
| Type | Dissertation |
| School | School of Engineering and Technology (SET) |
| Department | Department of Industrial Systems Engineering (DISE) |
| Academic Program/FoS | Microelectronics (ME) |
| Chairperson(s) | Manukid Parnichkun; |
| Examination Committee(s) | Shrestha, Sangam;Mongkol Ekpanyapong; |
| Scholarship Donor(s) | Korea Water Resources Corporation (K-Water)
Republic of Korea; |
| Degree | Thesis (M. Eng.) -- Asian Institute of Technology, 2017 |