Unified AI-Based Predictive Models for the Ultimate Capacity of Multi-Planar Gapped KK Steel Pipe Joints

Multi-Planar KK Joint Ultimate Capacity Unified Model Symmetrical Asymmetrical Artificial Intelligence.

Authors

  • Ahmed Kadry
    ahmed.shaheen@fue.edu.eg
    1) Structural Engineering and Construction Management Department, Faculty of Engineering and Technology, Future University in Egypt, Egypt. 2) Structural Engineering Department, Faculty of Engineering, Ain Shams University, Cairo,, Egypt https://orcid.org/0000-0002-5866-1309
  • Eslam El-Ganzoury Structural Engineering Department, Faculty of Engineering, Ain Shams University, Cairo,, Egypt
  • Abdel Salaam A. Mokhtar Structural Engineering Department, Faculty of Engineering, Ain Shams University, Cairo,, Egypt
  • Said Y. Aboul Haggag Structural Engineering Department, Faculty of Engineering, Ain Shams University, Cairo,, Egypt
  • Ahmed Ebid Structural Engineering and Construction Management Department, Faculty of Engineering and Technology, Future University in Egypt,, Egypt
Vol. 10 (2024): Special Issue "Sustainable Infrastructure and Structural Engineering: Innovations in Construction and Design"
Special Issue "Sustainable Infrastructure and Structural Engineering: Innovations in Construction and Design"

Downloads

The multi-planar steel pipe joints are widely used in communication towers, industrial structures, and offshore platforms. The current design formulas consider this joint as a uniplanar joint and account for the multi-planar effect using empirical correction factors. Recent studies deal with this multi-planar joint as a 3D joint but considering certain loading conditions. Hence, the aim of this research is to develop more general AI-based predictive models for the ultimate capacity of multi-planar gapped KK steel pipe joints, considering both symmetric and asymmetric loading conditions. Three AI techniques were applied to a database of previously published works. These techniques are "Genetic Programming” (GP), "Artificial Neural Network” (ANN), and "Evolutionary Polynomial Regression” (EPR). The prediction accuracies of the developed AI models were compared against two previously published formulas. The results indicated that the developed AI models are much more accurate than the previously published formulas. Also, the results showed that both the ANN and EPR models have almost the same level of accuracy (about 92%), but the EPR model has the advantage of presenting a closed-form equation that could be implemented either manually or using software.

 

Doi: 10.28991/CEJ-SP2024-010-07

Full Text: PDF