Prediction of the Yield Stress of Printing Mortar Ink

Abstract

The development of printable cement-based material is a high priority in the field of 3D printing for construction. There are many admixtures available for the design of the printing mortar ink which can influence the wet and final properties of the mortar. In this work, artificial intelligence has been utilized to predict those properties and guide the dosage of each admixture. The algorithms were developed from a factorial experimental plan. The mortar investigated consists of cement blended with silica fume to reduce the embodied carbon of the mixture. The selected admixtures were a superplasticizer, a viscosity modifying agent, nano-clay, C-S-H seeds and an accelerator with a water-reducing effect. A rotary rheometer was used to measure the viscosity and the dynamic yield stress of both mortar and cement-paste mixtures. Additional tests were conducted such as the small Abrams cone and the ASTM C1437 flow test. Several predictive algorithms were developed and compared, in which artificial neural networks were used. Furthermore, to enhance the performance of the neural network, a genetic algorithm was used to optimize the network parameters. To evaluate the performance of the models, the normalized root mean square error (NRMSE), and coefficient of determination (R 2 ) were calculated. This approach is a single-objective prediction which yields promising capability to predict the wet properties of both mortar and cement pastes, which can be later expanded into a multi-objective approach.

Automating mix design for 3D concrete printing using optimization methods

“…The methodology is similar to that made in a previous study presented in 57 . Here, the data is divided into training, validation, and testing sets.…”

“…Please do not adjust margins the absence of water, sand, or superplasticizer. Additionally, the data from the study presented in 57 was added to improve the accuracy of the predictions.…”

“…The methodology is similar to that made in a previous study presented in. 57 Here, the data is divided into training, validation, and testing sets. The proportions of splits are 70% for training, 15% for validation and 15% for testing with the unseen data.…”

“…Apart from the data collected from the 18 formed mixtures, the extra data was the same combinations of materials, but with the absence of water, sand, or superplasticizer. Additionally, the data from the study presented in 57 was added to improve the accuracy of the predictions.…”