A simulation-optimization methodology was used to assess monitoring strategies for a drinking water distribution network. Multiple simulation trials of contamination events were used to create input data for an integer optimization problem. A network model, based on the Blacksburg, VA water distribution system, was used as the basis for a case study of contaminant transport under conditions of uncertainty. The model was not calibrated due to the lack of reliable field data. Optimization of monitoring plans was performed within the context event based simulation trials. This precluded the design of monitoring plans that were directly compatible with requirements of water quality regulations. However, the results of the optimization did provide information that may be of use to the broader problem of compliance monitoring. Optimal plans were assessed in comparison with several alternative plans using a separate set of simulation trials. Optimization of monitoring plans derived from simulated source node contamination events was generally effective at choosing points that provided better detection of source node contamination than alternative plans based on random sampling or judgement sampling. Optimal monitoring plans derived from simulated random node contamination events were ineffective at detecting random node contamination. The results of optimization and the separate analysis of monitoring plan performance indicated that the number of simulation trials may have been inadequate to completely describe the stochastic behavior of the system. Additionally, comparison of these results with those obtained from a previous simulation study indicate that the results of any simulation of distribution system contamination may be very sensitive to the level of contaminant loading and the size and layout of the system. / Master of Science

Drinking water systems aim to remove, reduce, and prevent microbial contamination in water by usingmultiple barriers from catchments to consumers. Water distribution networks are vulnerable tocontamination from external sources if they lose their physical or hydraulic integrity. The leading causeof intrusion is losing hydraulic integrity due to low pressure in the water distribution networks. Eventsthat lead to low pressure in the water distribution networks can result in transient or sustained lowpressure lasting from milliseconds in a transient to hours and days in sustained low-pressure events.This study studied two sustained low-pressure events with durations of one to five hours, leading tointrusion in the water distribution network. The first event was the pump shut down, and the secondwas the pipe repair. Different durations, start times, and locations were simulated for the pumpshutdown and pipe repair events. Hydraulic and water quality modelling using EPANET 2.2 was usedto simulate low-pressure events and intrusion of microbial contamination in the drinking waterdistribution networks. Quantitative microbial risk assessment (QMRA) was used to estimate potentialpublic health risks using the Swedish QMRA tool. Campylobacter, Norovirus, and Cryptosporidiumwere selected as reference pathogens for simulating intrusion transport within the drinking waternetwork based on their health problem severity, persistence in water supplies, and resistance to chlorinecompound disinfectants. The study area was taken from the virtual network files generated usingHydroGen. This study showed that the volume of intrusion depended on the magnitude but mainly onthe duration of pressure drop. Also, the length of the pipes experiencing pressure drop and the numberof intrusion nodes affected the volume of intrusion. The location and magnitude of maximum nodalpathogen concentration changed significantly by changing the pump shutdown's start time and locationof pipe repair. Generally, the pump shutdown event affected extended areas with low pressure in thewater distribution network than the pipe repair. The QMRA results showed a considerable infection riskin all studied pump shutdown scenarios. The pipe repair duration was crucial in increasing or decreasingthe infection probability. The findings of hydraulic modelling and QMRA could benefit the watermanagers in deciding mitigation strategies.