Asynchronous Hierarchical Parallel Evolutionary Algorithm-Based Framework for Water Distribution Systems Analysis

Water distribution systems (WDSs) are vulnerable to accidental and intentional contamination that can have serious effect on the public health. Accurate characterization of the contaminant source is usually the first step in designing a response strategy for control and containment of the contaminant during an event. Contaminants can spread quickly in the network due to complex hydraulic conditions requiring a procedure for near real-time identification of the source characteristics and design of response strategy. Limitations associated with the quality and quantity of network monitoring data, unknown nature of the contaminant and its interaction with other chemicals in the water, complexity of the network, and inherent uncertainties make contaminant source characterization a difficult and challenging problem to solve. Solution of source characterization problems requires iterative simulations of flow and transport, which are highly computing intensive posing a computational challenge for near real-time solution of the problem.;This research focuses on the development and demonstration of a computational framework for real-time contamination threat management in water distribution systems. Methodologies for contaminant source identification and characterization are developed. An array of contamination scenarios is explored and simulation-optimization-based methodologies are developed to address the source characterization problems. Effects of quality and quantity of available continuous as well as filtered water quality sensor data are studied. The methodologies are tested and evaluated under various conditions of noise and uncertainties in the data. Contamination scenarios involving reactive contaminants are also studied to characterize the source using routinely monitored chlorine levels as surrogate for detection and identification. The uncertainties in the problem due to unknown reaction kinetics of the contaminant in the system are investigated and methodologies for source characterization under the conditions of such uncertainties and noise in the system are examined. These methods are applied and tested for an array of contamination scenarios in two different example networks. To enable a near real-time solution, a massively parallel computational framework is developed for modern shared/distributed memory architecture-based parallel computers. This includes the development of a new Asynchronous Hierarchical Parallel Evolutionary Algorithm (AHPEA) to solve complex large-scale global optimization problems. Performance and robustness of AHPEA algorithm is tested using a suite of benchmark function optimization problems and its application to address water distribution contaminant threat management problems is demonstrated. A WDS simulation model and the optimization methodologies are fully integrated into the computational framework for real-time analysis of the system. The computational framework is tested for parallel performance and scaling on different state-of-the-art parallel computers.;The simulation-optimization framework developed in this research is successfully able to address the WDS contamination threat management problem by narrowing down the potential source location to a small set of concentrated nodes in the network. This is shown for different conditions of data availability, noise and uncertainties, as well as for reactive as well as non-reactive contaminants. Issue of non-uniqueness in the solution is addressed by identifying all possible solutions. A near real-time solution of the problem is enabled by this framework through efficient use of computational resources. The efficient and scalable framework developed in this dissertation research provides a robust tool that can be applied to solve large-scale complex engineering design problems in general.