Optimal Design and Sensitivity Analysis of welded structure and Support Vector Machine application Prediction

Abstract

This study aimed to evaluate the sensitivity of welding design parameters and optimum cost design via Geometric Programming (GP) and the Lingo Program (LP). The first section begins with basic definitions and steps for transforming problems into the GP format, followed by comparing the GP and LP with several other optimization methods. This comparison demonstrates that GP is an effective, robust, and reliable method for optimizing linear and nonlinear problems, whereas Lingo is a simple and suitable program for locating optimal values and analyzing the solutions to engineering problems. The next part of the study concerned sensitivity analysis of the problem and determining the sensitivity of the optimal solution to certain changes in the original model. To perform sensitivity analysis, the optimal response and design parameters of the welded beam for a certain percentage of changes in P, L, б_d, and τ_d values were evaluated and plotted as diagrams. Sensitivity analysis is a crucial component of optimization, as it determines the percentage of change at which the optimal response remains constant. To this end, the support vector machine (SVM) is trained using data from optimal design conditions. In this study, the SVM was utilized to predict the weld thickness variable (h), weld length variable (l), bar thickness (t), bar width (b), and objective function (F_X). After successful testing, the model accurately predicted the parameters above, adhering to all design constraints.This study aimed to evaluate the sensitivity of welding design parameters and optimum cost design via Geometric Programming (GP) and the Lingo Program (LP). The first section begins with basic definitions and steps for transforming problems into the GP format, followed by comparing the GP and LP with several other optimization methods. This comparison demonstrates that GP is an effective, robust, and reliable method for optimizing linear and nonlinear problems, whereas Lingo is a simple and suitable program for locating optimal values and analyzing the solutions to engineering problems. The next part of the study concerned sensitivity analysis of the problem and determining the sensitivity of the optimal solution to certain changes in the original model. To perform sensitivity analysis, the optimal response and design parameters of the welded beam for a certain percentage of changes in P, L, б_d, and τ_d values were evaluated and plotted as diagrams. Sensitivity analysis is a crucial component of optimization, as it determines the percentage of change at which the optimal response remains constant. To this end, the support vector machine (SVM) is trained using data from optimal design conditions. In this study, the SVM was utilized to predict the weld thickness variable (h), weld length variable (l), bar thickness (t), bar width (b), and objective function (F_X). After successful testing, the model accurately predicted the parameters above, adhering to all design constraints.