Graduation Year


Document Type




Degree Name

Doctor of Philosophy (Ph.D.)


Civil Engineering

Degree Granting Department

Civil and Environmental Engineering

Major Professor

Mahmood Nachabe, Ph.D.

Committee Member

Bo Zeng, Ph.D.

Committee Member

Qiong Zhang, Ph.D.

Committee Member

Maya Trotz, Ph.D.

Committee Member

Ambe Njoh, Ph.D.


Calibration, Contamination, Sampling Design, Security, Water Saving


A potable water distribution system (WDS) consists of pipes, pumps, valves, storage tanks, control and supporting components. Traditionally, it has two basic functions. First, provides end users with potable water at sufficient pressures and good water quality. Second, provides sufficient pressure and flow for fire fighting. Currently, potable water is still the least expensive material for fire fighting. To accomplish these two goals, water utilities have to consider the integrity and security of the water network. As a result, this research selected three research topics that are closely related to the daily operation of water utilities and water quality simulation.

The first study is on optimal sampling design for chlorine decay model calibration. Three questions are investigated: (1) What is the minimum number of chlorine sample locations a water network needs? (2) How many combinations of sampling locations are available? (3) What is the optimal location combination? To answer the first two questions, the mathematical expressions of the chlorine concentrations between any two sampling locations are developed and sampling point relationship matrices are generated, then a mixed integer programming (MIP) algorithm is developed. Once obtained, the solutions to the first two questions are used to calculate the chlorine decay wall reaction coefficients and sensitivity matrix of chlorine concentration wall reaction coefficients; then, sampling location combinations achieved in the second question are sorted using a D-optimality algorithm. The model frame is demonstrated in a case study. The advantage of this method, compared to the traditional iterative sensitivity matrix method, is that a prior knowledge or estimation of wall reaction coefficients is not necessary.

The second study is on optimizing the operation scheduling of automatic flushing device (AFD) in water distribution system. Discharging stagnant water from the pipeline through AFD is a feasible method to maintain water quality. This study presents a simulation-based optimization method to minimize total AFD discharge volume during a 24-hour horizon. EPANET 2.0 is used as hydraulics and water quality simulator. This is formulated as a single objective optimization problem. The decision variables are the AFD operation patterns. The methodology has three phases. In the first phase, AFD discharge capacities are calculated, whether existing AFDs are able to maintain chlorine residuals in the water network is also evaluated. In the second phase, the decision variables are converted to AFD discharge rates. A reduced gradient algorithm is used to quickly explore and narrow down the solution space. At the end of this phase, decision variables are switched back to the AFD operation patterns. In the third phase, simulated annealing is used to search intensively to exploit the global minimum. The method is demonstrated on the water system located at the south end of Pinellas County, Florida where AFD optimal operation patterns are achieved.

The third study is on simulating contaminant intrusion in water distribution system. When contaminant matrix is introduced into water distribution system, it reacts with chlorine in bulk water rapidly and causes fast disinfectant depletion. Due to the difficulties in identifying contaminant types and chemical and biological properties, it is a challenging task to use EPANET-MSX to simulate chlorine decay under contaminant attack. EPANET 2.0 is used in the study to accomplish this goal. However, EPANET 2.0 cannot directly simulate chlorine depletion in the event of contamination attack because it assigns one time-independent bulk reaction coefficient to one specific pipe during the simulation. While under contaminant intrusion, chlorine decay bulk coefficient is not a constant. Instead, it is a temporal and spatial variable. This study presents an innovative approach for simulating contaminant intrusion in water distribution systems using EPANET multiple times. The methodology has six general steps. First, test bulk reaction coefficients of contaminant matrix in chemical lab. The uniqueness of this study is that the contaminant matrix is studied as a whole. The investigations of chemical, biological properties of individual aqueous constituents are not needed. Second, assume the contaminants as nonreactive, using EPANET 2.0 to identify where, when and at what concentrations of the inert contaminants will pass by in the water network. Third, determine the number of chlorine residual simulations based on the results in step two. Fourth, use EPANET to simulate the chlorine residual in the water network without the occurrence of contamination. Fifth, assign contaminated bulk coefficients to contaminated pipes; use EPANET to simulate the chlorine residual in the pipe network. Lastly, the chlorine concentrations of the impacted moments of impacted junctions are replaced with the results calculated in step five. This methodology is demonstrated in the south Pinellas County water distribution system.