University of Qom Civil Infrastructure Researches 2783-140X 11 2 2025 12 22 Optimal Design of Space Structures Considering Connection Stiffness Using a Machine Learning Based Surrogate Model Optimal Design of Space Structures Considering Connection Stiffness Using a Machine Learning Based Surrogate Model 67 82 3695 10.22091/cer.2025.13271.1635 FA Majid Ilchi Ghazaan School of Civil Engineering, Iran University of Science and Technology, Tehran, Iran 0000-0001-6689-9301 Mostafa Sharifi School of Civil Engineering, Iran University of Science and Technology, Tehran, Iran. 0009-0003-8992-4321 Sevim Rahimbaksh Khiabani School of Civil Engineering, Iran University of Science and Technology, Tehran, Iran. 0009-0000-4252-9234 Journal Article 2025 07 02 In this research, the optimal design of space structures considering the actual stiffness of connections is performed using a surrogate model based on a machine learning algorithm. Space structures, as one of the most important types of lightweight and robust structural systems, often have semi-rigid connections whose behavior is conventionally idealized as either rigid or pinned. This unrealistic assumption can lead to an increase in structural weight or construction costs. Therefore, incorporating the actual stiffness of connections into the optimal design process can result in reduced overall structural weight and improved efficiency. Since accurately calculating the total structural weight including the weight and costs associated with connections, which account for approximately 15 to 45 percent of the total weight is essential, connection weight has also been included in the objective function. To reduce computational costs, a surrogate model based on a machine learning algorithm is employed. Moreover, to enhance the accuracy and efficiency of the surrogate model, an active learning method is used for the intelligent selection of training data. The results indicate that the proposed method is capable of finding optimal solutions with fewer analyses compared to metaheuristic algorithms. According to the results, 800-member and 1016-member space structures with semi-rigid connections have 4.25% and 14.48% less weight, respectively, compared to structures with pinned and rigid connections. In this research, the optimal design of space structures considering the actual stiffness of connections is performed using a surrogate model based on a machine learning algorithm. Space structures, as one of the most important types of lightweight and robust structural systems, often have semi-rigid connections whose behavior is conventionally idealized as either rigid or pinned. This unrealistic assumption can lead to an increase in structural weight or construction costs. Therefore, incorporating the actual stiffness of connections into the optimal design process can result in reduced overall structural weight and improved efficiency. Since accurately calculating the total structural weight including the weight and costs associated with connections, which account for approximately 15 to 45 percent of the total weight is essential, connection weight has also been included in the objective function. To reduce computational costs, a surrogate model based on a machine learning algorithm is employed. Moreover, to enhance the accuracy and efficiency of the surrogate model, an active learning method is used for the intelligent selection of training data. The results indicate that the proposed method is capable of finding optimal solutions with fewer analyses compared to metaheuristic algorithms. According to the results, 800-member and 1016-member space structures with semi-rigid connections have 4.25% and 14.48% less weight, respectively, compared to structures with pinned and rigid connections. https://cer.qom.ac.ir/article_3695_536cdaf2e07c43f71f27fee68ce0c213.pdf