Optimisation of castellated beams with metaheuristic algorithms

This thesis aims to develop a scientific approach utilizing advanced and intelligent optimization algorithms to minimize the total cost in the design of castellated beams. A complex and multivariable design environment was established by considering different opening types circular, hexagonal, and sinusoidal as well as various beam lengths ranging from 300 cm to 800 cm. Five optimization algorithms were employed within the proposed framework: CPA, IRO, TWO, VPS, and ICA. Each algorithm was executed independently, and their performances were analyzed in detail. One of the most original contributions of this study is the integration and simultaneous application of these five algorithms, which has not been previously addressed in the literature. This comprehensive integration not only increases the diversity of solution methods but also enables a comparative evaluation in terms of computational time, stability, statistical reliability, and the ability to reach optimal solutions. Additionally, castellated beams designed with sinusoidal openings seldom addressed in existing research were modeled and analyzed in detail for the first time in this study. Results related to sinusoidal openings were evaluated comparatively with other opening types. Since unrealistic results were obtained for beams with a length of 800 cm, this case was excluded from the analysis. This approach reflects not only algorithmic assessment but also an original evaluation process supported by engineering rationale. Furthermore, for the first time, a trend analysis was conducted for each algorithm's performance, examining the effects of design variables such as opening type and beam length in depth. The consistency and distribution of final solutions for the algorithms were assessed. A sensitivity analysis was also performed, observing the objective function behavior of each algorithm up to 10,000 iterations. This allowed the determination of how quickly and consistently each algorithm approached the optimal solution. The study also evaluated statistical metrics such as standard deviation, processing time, and correlation coefficient, analyzing their relationships with each algorithm, opening type, and beam length in detail. From a rarely examined perspective in the literature, the behavior of the algorithms toward key design variables was assessed. Four main dimensions web height, flange width, flange thickness, and web width were chosen as the primary design variables, since other dimensions depend on these parameters. This approach helped reveal the algorithms' reactions to these variables and allowed a clearer evaluation of the design's stability. Moreover, the accuracy of the cross-sectional dimensions obtained through the optimization algorithms was validated using ANSYS finite element analyses, which reinforces the engineering reliability of the proposed approach. In conclusion, the thesis demonstrates a high level of originality in terms of both methods used and the depth of analysis, making it a potentially valuable scientific resource for the intelligent design of castellated beams in modern engineering applications.