Sizing optimization of exposed steel penstocks supported by ring girders using the combined ANN-GA technique. (April 2020)

• Record Type: Journal Article
• Authors: Haddouch, Manal; Hajjout, Imane; Boudi, El Mostapha
• Is Part Of: Structures. Volume 24(2020)
• Page Start: 705
• Page End: 716
• Publication Date: 2020-04
• Subjects: Exposed penstock -- Sizing optimization -- Ring girders -- ANN-GA; Structural engineering -- Periodicals; 624.1
• DOI: 10.1016/j.istruc.2019.12.011
• Languages: English
• ISSNs: 2352-0124

Abstract

Abstract: The design of penstock is a major concern in the implementation of large hydropower plants, where penstock is a large-scale structure. A design procedure incorporating the combined ANN-GA approach was developed for optimizing the exposed penstock supported by ring girders. The objective function considered is the total mass of the penstock to be minimized under the strength requirements. For this purpose, a numerical finite element analysis (FEA) model of a steel exposed penstock supported by ring girders was used to evaluate the equivalent and local stresses on the penstock shell. Parametric analysis was conducted to examine the influence of penstock thickness, ring girders spacing, and sectional size of the T-shaped ring girders on the weight of the penstock and the maximum equivalent stress. Subsequently, an artificial neural network (ANN) was developed by collecting data from the FEA numerical results, before being combined with a genetic algorithm to provide the optimal design of the penstock. The results of the FEA give a detailed distribution of local stresses in the penstock shell. The performance of the trained ANN was illustrated, and its prediction results show good agreement with the FEA numerical results. The results obtained by the ANN-GA approach proved the efficiency of the proposed process in optimizing the exposed penstock supported by ring girders. This process should be a valuable aid for the design of large penstocks.