Geração de dados com algoritmo genético e aplicação de machine learning na otimização de estruturas metálicas

Abstract

This work presents an integrated approach to optimizing metal structure design by combining genetic algorithms, data collection, and Machine Learning. Genetic Algorithms were employed to generate a robust dataset from various structural configurations, providing parameters such as metallic profile characteristics and structural positioning. These data were statistically analyzed and used in predictive models, including Random Forest, Gradient Boosting, and XGBoost. The results indicate that boosting-based models (Gradient Boosting and XGBoost) outperformed Random Forest across all evaluated metrics. Gradient Boosting achieved an MAE of 4.7674, MSE of 53.6243, and RMSE of 7.3229, while XGBoost reported an MAE of 4.7255, MSE of 54.0651, and RMSE of 7.3529. Both models reached similar R² scores (0.5844 for Gradient Boosting and 0.5809 for XGBoost), demonstrating a strong ability to explain data variance. This study highlights the potential of integrating genetic algorithms, statistical analysis, and Machine Learning, showing that boosting-based models can better capture the structural patterns in the data. The findings suggest that this approach can contribute to more efficient, cost-effective, and sustainable metal structure engineering projects.