Seoultech Researchers Use Machine Learning to Predict Strength of Steel Columns
In a groundbreaking development for the construction industry, researchers from Seoul National University of Science and Technology (Seoultech) have harnessed the power of machine learning to improve the structural design of steel columns. The team’s innovative hybrid model is designed to predict the ultimate axial strength of carbon fiber-reinforced polymer (CFRP)-strengthened concrete-filled steel tube (CFST) columns—a crucial parameter for ensuring the safety and performance of modern construction.
The innovative CFRP-strengthened CFST columns combine the load-bearing capabilities of traditional steel with the corrosion-resistant and lightweight properties of CFRP, offering a durable solution for skyscrapers, high-rise buildings, and even offshore structures. These columns not only enhance structural strength but also reduce maintenance needs, making them a promising material for sustainable construction. However, the scarcity of data on CFRP-strengthened CFST columns has hindered the development of accurate predictive models for their design. To overcome this challenge, Dr. Jin-Kook Kim, Associate Professor at Seoultech, and his team utilized machine learning, alongside generative AI, to create a synthetic database that mimics the characteristics of real-world data. This database was then used to train a hybrid machine learning model combining the Extra Trees (ET) technique and the Moth-Flame Optimization (MFO) algorithm.
In their study, published in Expert Systems with Applications, the team demonstrated that their new model outperforms existing empirical models, offering higher accuracy and lower error rates across key performance metrics. The hybrid model’s ability to consistently provide accurate predictions under various conditions further solidified its reliability. The hybrid model promises to be a game-changer for engineers, providing them with a tool to design stronger and safer structures. Its application can also be extended to retrofitting older buildings and bridges with CFRP materials, boosting their durability and resilience against natural elements. With the increasing frequency of extreme weather events and climate change, the corrosion-resistant properties of CFRP-strengthened CFST columns make them an even more vital solution for modern infrastructure.
To make the model widely accessible, the research team has also developed a web-based tool that allows engineers to predict the ultimate axial strength of CFRP-strengthened CFST columns for free, directly from any device without the need to install software. This machine learning-powered innovation is set to improve the safety and efficiency of both new and existing structures, helping to create safer and more sustainable buildings at a lower cost.
