Increasing Computational Efficiency of Inverse Transient Analysis for Leak Detection using GA-Kriging Surrogate Model
The inverse transient analysis (ITA) method is amongst the successful leak detection methods in water distribution networks. However, determining the unknown leakage parameters such as number, location, and area of leakages is computationally time-consuming and costly due to applying metaheuristic a...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Water and Wastewater Consulting Engineers Research Development
2020-07-01
|
| Series: | آب و فاضلاب |
| Subjects: | |
| Online Access: | http://www.wwjournal.ir/article_96061_cb96874edcce6db04671fe69132251ea.pdf |
| Summary: | The inverse transient analysis (ITA) method is amongst the successful leak detection methods in water distribution networks. However, determining the unknown leakage parameters such as number, location, and area of leakages is computationally time-consuming and costly due to applying metaheuristic algorithms, like the genetic algorithm (GA). This study aimed to present a novel approach to resolve this issue in order to enhance the accuracy and speed of the ITA method while maintaining its computational structure. In this research, surrogate models were incorporated in the optimization process of the ITA method. Mimicking the behavior of the objective function, surrogate models attempt to represent the most similar behavior at a low computational cost. In this regard, a new optimization algorithm based on the Kriging surrogate model, called GA-Kriging was proposed. In this algorithm, according to the structural characteristics of the Kriging surrogate model, an EI index was presented to modify the offspring selection scheme in GA. In order to evaluate the GA-Kriging algorithm and compare its performance with the conventional GA, a reference water distribution network was considered for leak detection. The accuracy and computational efficiency of the results in the GA-Kriging algorithm were 52% and 75% higher than those of the conventional GA, respectively. The present study concluded that appropriate incorporation of surrogate models in the optimization process can make the computations more intelligent, reduce repeated computations and, ultimately, increase computational efficiency. |
|---|---|
| ISSN: | 1024-5936 2383-0905 |