¾ÅÖÝÓ°Ôº

Automating mix design for 3D concrete printing using optimization methods

Vasileios Sergis and Claudiane M. Ouellet-Plamondon

This study attempts to automate the development process of mortar mixtures for 3D concrete printing applications. 3D printing brings automation to construction processes, and the customization potential of this technology is a notable advantage.

However, customizability adds complexity to mix design. This complexity arises from the intricate compositions required for 3D printing, which differ significantly from traditional concrete. As more materials are integrated, the workload during the development phase surges exponentially.

To address these complexities, the study employs optimization methods to automate the development of mortar mixtures. The core objectives are enhancing workability, buildability, and compressive strength. This investigation spans eight factors, including cement, sand, superplasticizer types, water-to-binder ratio, sand-to-binder ratio, and admixture dosages.

We designed an initial D-optimal set of 18 mixtures to reduce resource requirements and experimentation efforts. Incorporating advanced computational tools, the study employs feedforward neural networks to simulate part of the mix design by predicting the mixtures’ properties. A genetic algorithm fine-tunes the models’ architectures to improve their performance. The Pareto-optimization algorithm then harmonizes material choices and dosages to optimize all three conflicting objectives simultaneously.

Remarkably, after only five iterations, the study successfully formulated mixtures meeting all criteria. This methodology reduces labour demands and generates optimized mixture compositions, adapting to the application needs. The goal is to develop the algorithms further to add other design functions, such as low CO2 impact, durability and others. The algorithms are easily adaptable to changing the components of the material composition.