An international team of scientists from Saudi Arabia, China, Pakistan, India and Australia have collaborated on a new paper in Cement and concrete composites using machine learning to test aerated concrete.
Study: Compressive Strength Estimation of Lightweight Aerated Concrete Using a Neural, Genetic, and Ensemble Machine Learning Approach. Image Credit: olpo/Shutterstock.com
What is Foam Concrete?
Concrete has excellent mechanical and physical properties, especially when reinforced with steel rebar, making it an ideal building material. However, concrete production comes at a huge environmental cost due to its carbon footprint and is a key source of climate change-inducing emissions. In addition, concrete waste is a major source of environmental pollution.
Several sustainable types of concrete and cementitious additives have been studied over the past decades to partially or completely replace traditional ordinary Portland cement and improve the environmental friendliness of the construction industry, as well as to improve the properties mechanical, physical and physicochemical aspects of concrete. Foam concrete is currently being explored by researchers.
This innovative material has a light cell structure with variable density. The weight reduction in the bonding mortar is caused by the inclusion of random air voids. Although not a new material, aerated concrete has become the subject of research due to its environmental and economic advantages over conventional concrete.
Other names for aerated concrete include low density aerated concrete and lightweight aerated concrete. This material has been applied in fire and earthquake resistant structures due to its excellent properties and is composed of cement, aggregates, foaming agents and water. Several researchers have also investigated the inclusion of waste binders to improve the material.
The properties of aerated concrete can be influenced by the characteristics of the material, such as the mineralogy of the cement, the type of foaming agent and the grain size of the aggregates, as well as the proportions of the mixture, the uniformity and nature of the pores and the water quality. Additionally, curing methods can affect the final properties and performance of the material.
Determining the optimal mix for aerated concrete performance
Determining the optimal mix of aerated concrete material components is crucial to improving the performance of the final product. Getting the right mix of materials gives aerated concrete an optimal high strength-to-weight ratio, which is vital for structural applications. In addition, determining the optimal ratio of materials improves fire resistance, energy consumption and thermal conductivity.
The benefits of this material have led to its adoption in several countries like Korea, UK and Canada. Currently, determining the optimal mix of materials in aerated concrete is usually done using empirical models and experiments, which can be inefficient.
Empirical models based on fundamental models such as those of Feret and Balshin have been developed over the years, but these methods present challenges. For example, the relationship between material properties and compressive strength is complex, and various constants must be used that are not easily determined.
Due to the shortcomings of current empirical models in terms of complexity, time and cost, researchers have turned to advanced AI-powered models based on machine learning. These techniques offer several unique advantages over conventional methods due to built-in capabilities to overcome the complexity of critical problems and provide superior predictive power.
Emphasis has been placed on applying machine learning to predict the compressive qualities of concretes prior to fabrication, saving time, cost and waste. Additionally, structural behavior can be predicted using machine learning algorithms. Machine learning-based techniques are quickly becoming an integral part of research in the construction industry.
Research in Cement and concrete composites used three powerful machine learning algorithms (GEP, GBT, and ANN) due to their nonlinear capabilities. Specifically, the algorithms are used to predict the compressive strength of aerated concrete.
The water/cement and sand/cement material ratios were optimized using parametric analysis by the authors. Model performance, parametric analysis, and variable sensitivity analysis are presented in the paper, and it has been proposed that a machine learning-based approach can be used to select the optimal foam concrete composition. .
The study revealed a strong correlation between the density of aerated concrete and the compressive strength. The optimal algorithm parameters for the three machine learning-based models have also been revealed in the author’s work. All the optimized AI models produced a strong R correlation, which reflects a strong agreement between the predicted and experimental results.
The GBT model achieved the best performance, with the highest validation data among the three machine learning-based methods. In terms of accuracy, it outperformed the other two models, with ANN second and GEP being the least accurate.
Future opportunities exist to further improve modeling and prediction using machine learning algorithms, and the scientists involved in the paper proposed that future research investigate variable foaming agent dosages. Overall, the new paper demonstrated the advantages of machine learning methods over conventional empirical and experimental models.
More from AZoM: How does laser cleaning technology work?
Salami, BA et al. (2022) Estimating Compressive Strength of Lightweight Aerated Concrete Using a Neural, Genetic, and Ensemble Machine Learning Approach Cement and concrete composites 104721 [online, pre-proof] sciencedirect.com. Available at: https://www.sciencedirect.com/science/article/abs/pii/S0958946522003146