Interfacial Bond Strength of Concrete-Filled Galvanized Steel Tubes: Push-Out Experiments, Geometry-Grouped Machine Learning, and Multi-Criteria Optimization

In the skeleton of modern construction, from the soaring columns of skyscrapers to the deep foundations of bridges, a quiet partnership often holds the weight. Engineers frequently rely on concrete-filled steel tubes, where a hollow steel pipe is packed tight with wet concrete that hardens into a solid core. The steel shell provides flexibility and tensile strength, while the concrete inside offers crushing power and stability. For this hybrid structure to work as a single, unified unit, the two materials must grip each other tightly. If the concrete slips inside the steel tube, the structure loses its strength, and the carefully calculated safety margins vanish. This grip, known as the interfacial bond, is not a simple glue; it is a complex interaction of chemical adhesion, surface friction, and the physical pressure the concrete exerts against the steel walls as it shrinks and settles.

The challenge lies in predicting exactly how strong this grip will be. Too often, designers rely on assumptions about how the size of the tube or the strength of the concrete affects this bond, but these factors interact in ways that are difficult to guess without testing. A new study by researchers at Jakson Green Limited and several engineering colleges in India has taken a close look at this problem, specifically focusing on tubes made of galvanized steel. Galvanizing is a process where steel is coated in a layer of zinc to prevent rust, a common choice for exposed structures. However, this zinc coating changes the texture of the steel surface, making it rougher and harder than bare steel, which could alter how well the concrete sticks. The researchers set out to map the relationship between the geometry of the tube and the strength of the concrete to find the most reliable configuration for these composite columns.

To solve this puzzle, the team did not rely on a single massive experiment, but rather a carefully organized series of twenty-seven tests. They used a method called a Taguchi design, which is like a highly efficient sampling strategy that allows scientists to test many different combinations of variables without having to build every single possible version. They varied three main things: the diameter of the tube, the thickness of its steel wall, and the ratio of the tube's length to its diameter. These were tested against three different grades of concrete, ranging from a standard mix to a very high-strength version. In each test, a steel tube was held in place while a heavy load pushed the concrete core out from the bottom, measuring exactly how much force was required to break the bond between the two materials.

The results revealed a clear hierarchy of importance among the variables. The most powerful factor influencing the bond strength was not the size of the tube or the strength of the concrete, but the length-to-diameter ratio. Specifically, shorter tubes relative to their width held the concrete much better than longer ones. The thickness of the steel wall was the second most important factor; thicker walls provided significantly more grip. The strength of the concrete mattered, but only up to a point, with the highest-strength concrete showing a notable jump in performance. Surprisingly, the diameter of the tube itself had a much smaller and less predictable effect than many might expect. The study found that the strongest bond occurred in a specific configuration: a tube with a diameter of 42.2 millimeters, a wall thickness of 4.0 millimeters, and a length-to-diameter ratio of 12. This specific geometry consistently produced the highest bond strength across all three types of concrete, with values reaching up to 2.062 megapascals.

To ensure these findings were not just a lucky match for the specific data they had, the researchers turned to machine learning, a field where computers learn patterns from data. They tested several different computer models to see which one could best predict the bond strength for a tube geometry it had never seen before. Many complex models, such as those based on decision trees, failed to generalize well, essentially memorizing the test data rather than learning the underlying rules. The most successful model was a type of support vector regression, which managed to predict new outcomes with a high degree of accuracy. This success highlighted a crucial lesson for small-scale engineering studies: when data is limited, the way you test your computer models is just as important as the models themselves. If you do not test the model on completely new geometry groups, you risk getting a false sense of confidence.

The researchers also used a multi-criteria decision-making approach to confirm their findings. They weighed the performance of every geometry against the others using different mathematical methods to see if they all pointed to the same winner. Every method agreed: the short, thick-walled tube with a specific diameter was the most robust choice. This configuration not only offered the highest strength but also the most stable performance, meaning it was less likely to fail unexpectedly. The study explicitly notes that these results are specific to the range of sizes and materials tested and cannot be blindly applied to much larger or smaller structures. Furthermore, because the study only used galvanized tubes, it cannot separate the effect of the zinc coating from the geometry itself.

Ultimately, this work provides a clear, data-driven path for engineers designing with galvanized steel tubes. It suggests that to maximize the strength of the bond between steel and concrete, one should prioritize a shorter, thicker-walled design over simply increasing the concrete strength or the tube diameter. While the study does not offer a universal formula for every possible construction scenario, it establishes a reliable baseline for the specific conditions it examined. By combining rigorous physical testing with careful statistical validation, the researchers have shown that even with a limited number of experiments, it is possible to uncover the most effective geometric relationships, ensuring that the hidden connections in our buildings remain as strong as the materials that compose them.