High-Fidelity Machine Learning-Based Prediction of the Thermal Behavior of Concrete Mixture Designs and Massive Structures
Presented By: Luna Al Hasani
Affiliation: Kiewit Engineering
Description: The temperature and time-dependent heat of hydration of cementitious pastes is a fundamental concept in understanding the mechanical and thermal behavior of concrete structures. For example, in massive concrete elements, higher curing temperatures increase the rate of heat release and cumulative heat of hydration, therefore increasing the risk of durability issues. Models which currently depict the heat of hydration rely on empirical fits created based on cement sources that are no longer relevant to the industry. Moreover, the empirical fits are not able to explain the behavior of more complex blended cementitious pastes. In recent years, machine learning has emerged as a promising potential to optimize the prediction of properties of different material systems by learning through data and statistical methods. Here, data-driven solutions are proposed for the modeling and evaluation of several aspects of interest in mass concrete: 1) the prediction of time and temperature dependent heat of hydration of cementitious pastes based on their physicochemical properties, 2) the prediction of adiabatic temperature rise of a range of blended mixture designs, and 3) the prediction of maximum temperatures and temperature differentials experienced in concrete elements subjected to different boundary conditions. Results show that advanced data science can predict the output criteria with high- fidelity when compared with experimental data. Moreover, the predicted outputs show good performance when upscaled for use in mass concrete case studies. This creates an opportunity to apply data-driven approaches to perform property predictions of cement pastes and facilitates the selection of mixture designs to satisfy certain performance criteria.
Automatic Detection and Localization of Multiple Cracks Using Distributed Fiber Optic Sensors
Presented By: Yi Bao
Affiliation: Stevens Institute of Technology
Description: Distributed fiber optic sensors (DFOS) offer detailed strain distribution, making them unique tools for monitoring cracks in infrastructure. However, the manual interpretation of DFOS data is error-prone and time-consuming. This paper presents a deep learning approach to achieve real-time monitoring and interpretation of multiple cracks based on DFOS data. The proposed approach incorporates You Only Look Once (YOLO) for accurate detection and localization of multiple cracks. The evaluation of the proposed approach using various experiments with different configurations yielded an F1 score of 0.92 for crack detection. The proposed approach provides an efficient and accurate solution for real-time crack monitoring using DFOS and deep learning, providing benefits for structural health monitoring.
An AI-Framework to Control and Optimize Material and Process in 3D Concrete Printing
Presented By: Giacomo Rizzieri
Affiliation: Politecnico Di Milano
Description: Additive manufacturing (AM) techniques are increasingly being adopted in the construction industry to create optimised and aesthetically pleasing geometries, reduce material waste, and accelerate the construction process. In particular, extrusion-based AM technologies, such as 3D Concrete Printing (3DCP), have been applied to a multitude of small-to-medium scale projects and even more relevant applications are expected in the years to come. However, despite being a promising technology, 3DCP applications are still limited, due to the scarce knowledge about the underlying process and the lack of regulation. In practice, in fact, a lot of experimentation and trial-and-error procedures are often necessary before obtaining reliable outcomes. The reason is that many variables can affect the printing process, such as the material properties (which vary in time), the printing parameters (nozzle shape, printing velocity, extrusion pressure and toolpath), the object design and the environmental conditions (temperature and humidity). There is thus an urge to create efficient and reliable tools to control the complexity of the 3D printing process, making it more standardized, accurate and easy to control. The proposed approach is first of all an attempt to fill this gap. Secondly, there is also the ambition to provide designers with a flexible tool that can be further exploited for the optimization of the printing process, determining the best combination of materials and process parameters for a given objective.
On the Use of Machine Learning and Data-Transformation Methods to Predict Hydration Kinetics and Strength of Alkali-Activated Mine Tailings-Based Binders
Presented By: Sahil Surehali
Affiliation: Arizona State University
Description: The escalating production of mine tailings (MT), a byproduct of the mining industry, constitutes significant environmental and health hazards, thereby requiring a cost-effective and sustainable solution for its disposal or reuse. This study proposes the use of MT as the primary ingredient (³70%mass) in binders for construction applications, thereby ensuring their efficient upcycling as well as drastic reduction of environmental impacts associated with the use of ordinary Portland cement (OPC). The early-age hydration kinetics and compressive strength of MT-based binders are evaluated with an emphasis on elucidating the influence of alkali activation parameters and the amount of slag or cement that are used as minor constituents. This study reveals correlations between cumulative heat release and compressive strengths at different ages; these correlations can be leveraged to estimate the compressive strength based on hydration kinetics. Furthermore, this study presents a random forest (RF) model—in conjunction with fast Fourier and direct cosine transformation techniques to overcome the limitations associated with limited volume and diversity of the database—to enable high-fidelity predictions of time-dependent hydration kinetics and compressive strength of MT-based binders in relation to mixture design. Overall, this study demonstrates a sustainable approach to upcycle mine tailings as the primary component in low-carbon construction binders; and presents both analytical and machine learning-based approaches for accurate a priori predictions of hydration kinetics and compressive strength of these binders.
Artificial Intelligence Method for Low-Carbon, Cost-Effective of Cement Materials
Presented By: Nima Khodadadi
Affiliation: University of Miami
Description: Amid the increasing global concern about carbon emissions, the construction industry emerges as a significant contributor, prompting a re-evaluation of its material and methodological choices. Central to this discussion is the potential of cement replacement materials, which promise enhanced mechanical properties and a considerable reduction in environmental impact. In this research, we introduce a pioneering approach that combines the robust optimization capabilities of metaheuristic algorithms with the deep learning prowess of advanced AI models. Our goal is to provide a comprehensive and precise prediction of cement replacement materials' structural characteristics and environmental implications. Through a systematic fusion of metaheuristics and AI, we aim to achieve a heightened accuracy in predicting compressive strength and embodied carbon values. Our methodology underscores the importance of continuous iterative learning and adaptation, capturing the intricate nuances of cement replacement materials. The expected outcomes of this endeavor include a marked improvement in predictive precision, ushering in substantial economic savings by minimizing exhaustive experimental tests. More profoundly, our research serves as a beacon for a new construction era, one firmly rooted in data-driven insights and an unwavering commitment to sustainable practices. By seamlessly melding the frontiers of AI, metaheuristic algorithms, and green construction, this study stands at the cusp of redefining ecologically conscious construction, setting new benchmarks for innovation and responsibility in the field.