Optimization of the graph model of the water conduit network, based on the approach of search space reducing

2017 5th International Conference on Mechanical Engineering, Materials Science and Civil Engineering

机械制造;材料科学;土木建筑工程

Korovin, Iakov S.^1 ; Tkachenko, Maxim G.^1

Southern Federal University, Taganrog, Russia^1

关键词: Conduit networks;    Efficient architecture;    Flow restriction;    Heuristic approach;    Quadratic function;    Search space reduction;    Uniform distribution;    Water distribution networks;   

Others  :  https://iopscience.iop.org/article/10.1088/1757-899X/324/1/012019/pdf
DOI  :  10.1088/1757-899X/324/1/012019

来源: IOP

【 摘 要 】

In this paper we present a heuristic approach, improving the efficiency of methods, used for creation of efficient architecture of water distribution networks. The essence of the approach is a procedure of search space reduction the by limiting the range of available pipe diameters that can be used for each edge of the network graph. In order to proceed the reduction, two opposite boundary scenarios for the distribution of flows are analysed, after which the resulting range is further narrowed by applying a flow rate limitation for each edge of the network. The first boundary scenario provides the most uniform distribution of the flow in the network, the opposite scenario created the net with the highest possible flow level. The parameters of both distributions are calculated by optimizing systems of quadratic functions in a confined space, which can be effectively performed with small time costs. This approach was used to modify the genetic algorithm (GA). The proposed GA provides a variable number of variants of each gene, according to the number of diameters in list, taking into account flow restrictions. The proposed approach was implemented to the evaluation of a well-known test network - the Hanoi water distribution network [1], the results of research were compared with a classical GA with an unlimited search space. On the test data, the proposed trip significantly reduced the search space and provided faster and more obvious convergence in comparison with the classical version of GA.