Optimal Design of Welded Structure Using SVM
Engineers and researchers can develop more efficient, load-resistant, reliable, and cost-effective structures using optimization techniques, Sensitivity Analysis (SA), and support vector machine (SVM) applications. This study evaluated the SA of welding design parameters and the optimal cost design...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Pouyan Press
2024-07-01
|
| Series: | Computational Engineering and Physical Modeling |
| Subjects: | |
| Online Access: | https://www.jcepm.com/article_209701_3a271699999f453800a450e7dedc6c0a.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Engineers and researchers can develop more efficient, load-resistant, reliable, and cost-effective structures using optimization techniques, Sensitivity Analysis (SA), and support vector machine (SVM) applications. This study evaluated the SA of welding design parameters and the optimal cost design using Geometric Programming (GP) and Lingo Program (LP). By combining GP and LP techniques, the minimum cost for the welded beam was obtained as 2.3806, demonstrating that it is the most optimal value compared to other optimization methods. This comparison indicates that the proposed hybrid method is an effective, robust, and efficient approach for optimizing linear and nonlinear engineering problems. SA is used to find the optimal values and assess the sensitivity of the optimal solution, and the cost of welded beam 2.3876 was obtained using this method. Moreover, Support Vector Machine with Radial Basis Function Kernel (SVM-RBF) is trained using the optimal design condition data. The values of the root mean square error of SVM-RBF in the prediction of the cost function are 0.00137 (correlation coefficient of R2=0.9999) and 0.217 (R2=0.9656) for training and testing, respectively. After successful training and testing, the model accurately predicted the cost function and optimal design parameters and satisfied all the design constraints, indicating that GP, LP, and SA are among the most effective techniques for the optimal design of engineering problems. SVM-RBF is a powerful tool that performs excellently in predicting optimal design parameters values and cost function in optimization problems. |
|---|---|
| ISSN: | 2588-6959 |