Deep Learning for Predicting the Strength of 3D-Printable Engineered Cementitious Composites

The construction industry, grappling with challenges like high costs, labor shortages, and environmental issues, is undergoing a transformative phase with the advent of 3D concrete printing (3DCP). This innovative approach, characterized by automation and sustainability, has introduced 3D-printable engineered cementitious composites (3DP-ECC), enabling the creation of complex structures without the need for traditional formwork or reinforcement. However, predicting the mechanical properties of 3DP-ECC, especially compressive strength, is challenging due to the influence of numerous variables such as printing parameters, material composition, and curing conditions. This paper presents a novel approach using a deep learning methodology, specifically a Recurrent Neural Network (RNN) with Long Short-Term Memory (LSTM) architecture, to predict the compressive strength of 3DP-ECC. The study involves meticulous feature engineering, including the selection and preprocessing of relevant input features like printing speed, nozzle diameter, fiber content, and water-cement ratio. The performance of the LSTM model is evaluated in terms of accuracy, precision, and robustness against traditional empirical models. Our results demonstrate that deep learning, particularly RNN with LSTM, offers a reliable and efficient tool for predicting the compressive strength of 3DP-ECC, outperforming conventional models. This research not only highlights the potential of deep learning in advancing construction methodologies but also provides a foundation for future innovations in the field.