3.1 Research Design
H2S generation can lead to corrosion of concrete and metals in wastewater collection and treatment tanks. Identification of the parameters influencing H2S generation can help in limiting the corrosion and thus reduce the investment associated with rehabilitation works in a wastewater collection and treatment plant. This can be established by measuring parameters from the wastewater treatment facility. These parameters were correlated with corresponding H2S concentration using Minitab.
Regression analysis was performed to identify significant factors influencing H2S generation. Later, ANN model was developed using these significant factors.
3.2 Study Area - Sewer Field Description
The study was conducted in Al Saad wastewater treatment plant (Figure 3).
The plant treats raw domestic sewage coming from neighboring houses and wastewater collected in tankers. The research was performed at the sewage collection unit before the screening unit. The samples were also collected from the same unit.
The width of the tank was 6 meters and the depth of wastewater varied with the wastewater flow.
Figure 3: Arial view of Al Saad wastewater treatment plant from google earth
3.3 Field Investigation
Different parameters were measured at the inlet of the wastewater treatment plant using a mobile collection unit. Samples were collected coincidentally and stored in a refrigerator. Sampling events were performed on the 8th, 9th, and 10th of October 2020 and repeated on the 16th, 17th, and 18th of June 2021 at similar climatic conditions.
The air quality parameters that were measured during this project are H2S concentration in the headspace, the temperature of headspace air, moisture content, headspace airflow. The hydraulic parameters measured were flowrate of wastewater, wastewater depth, and velocity. The water quality parameters measured were temperature of wastewater, dissolved oxygen (DO), chemical oxygen demand (COD), sulfates, sulfides, total suspended solids (TSS), total dissolved solids (TDS), pH, total organic carbon (TOC) and electrical conductivity. Significant parameters were identified and interrelated using Minitab. An overall model correlating these
parameters and the corresponding H2S concentration for wastewater treatment plant applications was developed.
3.4 On Site Data Collection
The mobile data collection unit, installed inside the wastewater storage tank, was used to collect relevant data at Al-Saad Wastewater Treatment Plant (ASWWTP), Al Ain, UAE (Appendix 1). It consists of temperature and humidity probes, AcruLog© H2S analyzer, and ACURITE© anemometer. A laser gun was used to measure the temperature of the wastewater onsite. The mobile unit was carried to the wastewater treatment plant for measurements. The unit was placed at the headworks of the ASWWTP right before the mechanical screens. The data was collected for 48 hours in 2-hour intervals from 12:00 PM to 12:00 PM. Measurements of H2S concentrations, moisture content, headspace airflow, the temperature of wastewater and headspace air, flowrate, and wastewater depth were recorded manually by personnel at the wastewater treatment plant in a data entry sheet. Wastewater samples of 50 ml were collected at each sampling event coinciding with the 2-hour recording intervals.
Wastewater samples were collected in fluoropolymer containers and stored in the refrigerator at 4°C.
3.5 Laboratory Analysis of the Collected Samples
The data recorded in the AcruLog© analyzer were downloaded using a computer after it is moved back to the laboratory. Parameters such as DO, COD, TSS, TDS, TOC, chloride, sulfates, sulfides, and electrical conductivity in the wastewater samples were measured in the laboratory. DO was measured using EXTECH® DO probe, while COD, sulfates, and sulfides were measured using HACH kits (LCK514,
LCK353, 2244500) respectively using a spectrophotometer (HACH DR 3900).
Procedures are described in Appendix B, C, and D. HORIBA EC probes were used to measure conductivity and TDS. Sulfates and chloride were measured using ion chromatography (Thermo Scientific©) at 1000 times dilution (Appendix E). TSS was measured using a spectrophotometer (Hach DR 3900). TOC was measured using a TOC analyzer (Appendix F).
3.6 Statistical and Correlation Analysis of Each Parameter:
The parameters resulting from on-site observations and laboratory testing were correlated using MinitabTM. The relationship of parameters with H2S and their interrelation was studied using scatterplots and contour plots. These plots examine the interrelation of parameters and patterns. Correlation analysis was performed for parameters with H2S concentration to identify the significant parameters. Significant parameters were observed at a 95% significance level. In MinitabTM correlation can be performed through the following steps, Stat > Basic statistics > Correlation.
3.7 H2S Prediction Modeling
The significant parameters identified in Task 1.4 were correlated with the H2S concentration. This was performed by utilizing the multi-parameter regression analysis function in Minitab. Regression analysis is used to examine the relationship between a dependent variable, (H2S concentration) and independent variables (significant parameters). Multi-parameter regression analysis helps to predict an outcome using multiple explanatory variables. The model with the highest R-sq value was chosen as the model for the collected data. In MinitabTM regression analysis can be performed through the following steps, Stat > Regression > Regression > Fit regression model.
3.8 Artificial Neural Network Modeling
MATLAB© 2018a software was used to quantitate the amount of H2S generated at the headworks of the wastewater treatment plant. It was developed using a Neural net fitting app. The artificial neural networks consist of three layers, input, hidden, and target layer as shown in Figure 4. The modeling process is comprised of three steps, (i) training, (ii) validation, and (iii) testing. The collected data of the significant parameters were divided into 70%, 15%, and 15% for training, validation, and testing, respectively. The model was trained using the training data set. The validation data set was combined with the training data set to decide when the training process should be stopped for the model to have good generalization properties. The test data set allows to evaluate the prediction abilities of the model. Steps to be followed for the modeling in MATLAB are briefly explained in Appendix G.
The number of significant parameters corresponds to the number of input layers. In this case, there were 4 input layers based on the analysis. The desired output from the model is H2S concentration, and hence the number of target layers is 1. Based on the number of input and target layers, the number of hidden layers was chosen to be 2.
Figure 4: Artificial neural network model