Use this URL to cite or link to this record in EThOS:
Title: Application of data mining techniques to predict the performance of matured vertical-flow constructed wetlands systems treating urban wastewater
Author: Iliyasu, U.
ISNI:       0000 0004 8504 4781
Awarding Body: University of Salford
Current Institution: University of Salford
Date of Award: 2019
Availability of Full Text:
Access from EThOS:
Access from Institution:
The rapid urbanisation and industrialisation, due to technological advancement, led to severe environmental pollution. The environmental pollution in the last few decades resulted in an adverse impact on the environment causing massive accumulation of wastewater. Wastewater is one of the closest sources of environmental problems, at the same time water scarcity is becoming alarming due to its high demand as the global population is increasing. Hence, the application for managing available water resources becomes crucial. The ever-increasing demand for water brings the need for wastewater treatment as an alternative source of water. Constructed Wetlands (CW) have gained broader research attention due to their environmental and safety benefits for wastewater treatment. In this study, over three years of monitoring performance data from 3rd December 2014 to 28th March 2018 (thirty-nine months) of the vertical flow vertical wetlands system, receiving and treating domestic wastewater, were collected and utilised to assess and investigate the treatment performance efficiency of the Vertical Flow Constructed Wetland Systems (VFCWs) for removing pollutants from wastewater. Different laboratory-scale vertical-flow constructed wetlands filters filled with gravel and planted with common reed were built to remove removal from wastewater. The overall evaluation of the system treatment performance was calculated using percentage removal efficiency. The results were recorded it was observed that all vertical flow constructed wetland filters had recorded high removal performance for the water quality parameters, irrespective of filter set-up and operation. The system was discovered to be very useful in pollutants removal (water quality parameters) with significant efficiency. However, the high cost of analysis laboratory tests, time-consuming parameters couple with uncertainties associated with an analysis of water quality variables, lead to the development of two data mining technique models Multiple Linear Regressions (MLR) and Multilayer Perceptron (MLP). To predict the wastewater treatment performance of CW by predicting selected output water quality parameters these include Chemical Oxygen Demand (COD), Biological Oxygen Demand (BOD), orthophosphate phosphorous (PO4-P), ammonium nitrogen (NH4-N) and suspended solids (SS) with respect to other known input parameters that will provide comfortable, reliable and cost-effective methods. Correlation analysis was conducted to select the most highly correlated input parameters to be used for the model development (prediction of output parameter). The monitoring dataset of all the parameters used was divided into training dataset to build prediction models (MLR and MLP) and testing dataset to validate the models constructed. In this current work, 70% of the whole data was used as a training dataset while the remaining 30% of the data set was used as a testing dataset. The prediction models built were evaluated and compared using two model evaluation criteria: graphical model evaluation (scatter plot and hydrograph) and numerical model error evaluation criteria using five model evaluation criteria, these include: Root Mean Square Error (RMSE), regression coefficient (r), Relative Absolute Error (RAE), mean absolute error (MAE) and root relative squared error (RRSE). The results obtained indicated that the predicted values of output parameters were in good agreement and relationship with their respective measured parameters. Thus, this showed that the two models built yielded satisfactory predictions and both models had performed reasonably well in predicting output variables concentrations accurately given the value of input dependent variable. Furthermore, the comparison between the model's outcomes showed that MLP model prediction performance was discovered to be better than the MLR model in a majority of water quality parameters. Both models built could be effectively used as a tool for predicting removal of water quality parameters efficiency of vertical flow constructed wetlands treating domestic wastewater and in predicting constructed wetland performance in wastewater treatment process in term of pollutants removal. The results demonstrated the potentiality of vertical flow constructed wetlands to treat domestic wastewater and remove pollutants for future reuse.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available