A Prediction of the Printability of Concrete Through Artificial Neural Networks (2023-08)¶

Digital Fabrication with Concrete (DFC) is acquiring potential because of the need of a technological advancement in the construction industry. The rheology of concrete finds evident meaning, because of the necessity of “handling” concrete in its early ages, when still fluid, and allowing it to immediately start developing its mechanical performance for the building process to advance. In order to make the material suitable for the printing process, it must comply with the printability requirements that are governed by parameters including yield stress and tensile and shear strength. Nowadays, the study behind the mix of a concrete material that fits all the printing requirements is still something that goes through an iterative trial and error process, due to the difficulty and the novelty that 3D Concrete Printing (3DCP) represents. For this reason, it can be helpful to develop a model, based on experience and the literature, that is able to predict if, for a given concrete mix design and printing process, all the requirements tied to 3DCP are satisfied. The employment of Artificial Intelligence (AI) can represent a solution to it. Applications of AI are very wide: referring to Civil Engineering, AI can be applied to concrete science and technology to handle several different topics including mix proportions, workability and overall mechanical performance, being instrumental to establish and assess multi-parameter correlations among the different involved material and process input parameters. AI tries to mimic the human brain, that is made up of many nerve cells driven by neurons which are in control of the external stimuli. The purpose of the paper is to analyse the printability through the implementation of AI techniques, designing a neural network between the parameters that control the mix composition and rheological properties of printable concrete mixes in the printability window.