Year of Publication

2007

Document Type

Dissertation

College

Engineering

Department

Civil Engineering

First Advisor

Srinivasa Lingireddy

Abstract

The need for optimal air vessel sizing tools, in protecting large pipe networks from undue transient pressures is well known. Graphical and other heuristic methods reported in literature are limited to sizing the air vessels for simple rising mains. Although attempts have been made to utilize optimization techniques, they have been largely unsuccessful due to their impractical computational requirements. This research work proposes a robust framework for developing surge protection design tools and demonstrates the usefulness of the framework through an example air vessel sizing tool. Efficiency and robustness of the proposed framework are demonstrated by developing a design aid for air vessel sizing for protecting large pipe network systems against excessive high pressures generated by rapid valve closures. The essence of the proposed framework is in identification of key transient response parameters influencing air vessel parameters from seemingly unmanageable transient response data. This parameterization helps in exploiting the similarity between transient responses of small pipe networks and sub-sections of large pipe networks. The framework employs an extensive knowledgebase of transient pressure and flow scenarios defined from several small network models and corresponding optimal air vessel sizes obtained from a genetic algorithm optimizer. A regression model based on an artificial neural network was used on this knowledgebase to identify key parameters influencing air vessel sizes. These key parameters were used as input variables and the corresponding air vessel parameters as output variables to train the neural network model. The trained neural network model was successfully applied for large complex pipe networks to obtain optimal air vessel sizes for transient protection. The neural network model predictions were compared with optimal air vessel parameters to assess the efficacy of the proposed framework. The validity and limitation of the design aid developed and areas in the framework that need further research are also presented. The proposed frame work requires generation of hundreds of optimization data for small and simple network systems which is a daunting task since genetic algorithm-based optimization is computationally expensive. Selection of a numerically efficient and sufficiently accurate transient analysis method for use inside a genetic algorithm based optimization scheme is crucial as any reduction in transient analysis time for a network system would tremendously reduce the computational costs of bi-level genetic algorithm optimization scheme. This research work also demonstrate that the Wave Plan Method is computationally more efficient than the Method of Characteristics for similar accuracies and the resulting savings in computational costs in the transient analysis of pipe networks and subsequently in the genetic algorithm based optimization schemes are significant.

Share

COinS