Digital fabrication with cementitious materials is a rapidly growing field of research in which the evolution of strength during the various processes, such as 3D printing, is the key controlling parameter. The strength evolves over multiple orders of magnitude during the process, and thus, it is essential to properly characterize the strength evolution in order to guarantee process success. This chapter summarizes the state of the art in these characterization methods for digital fabrication with fresh cementitious materials, reviewing well-known and more recently developed methods.@en

The purpose of this paper is to investigate the feasibility of using artificial neural network programming for the prediction of the fresh properties of self-compacting concrete. The input parameters of the neural network were the mix composition influencing the fresh properties of self-compacting concrete namely, the cement content, the dosages of limestone powder and water, fine aggregate, coarse aggregate, and superplasticizer, and other parameter of time of testing (5, 30 and 60 minutes after addition of water). The model is based on a multilayer feed forward neural network. The details of the proposed ANN with its architecture, training and validation are presented in this paper. Six outputs of the ANN models related to the fresh properties were the slump flow, T₅₀, T₆₀, V-funnel flow time, Orimet flow time, and blocking ratio (L-box). The effectiveness of the trained ANN is evaluated by comparing its responses with the experimental data that were used in the training process. The dosage of water was varied from 188 to 208 L/m³, the dosage of SP from 3.8 to 5.8 kg/m³, and the volume of coarse aggregates from 220 to 360 L/m³ (587 to 961 kg/m³). In total twenty mixes were used to measure the fresh properties with different mix compositions. ANN performed well and provided very good correlation coefficients (R²) above 0.957 for slump flow, T₅₀, V-funnel flow time, Orimet flow time, and L-box blocking ratio. The predicting results for T₆₀ was slightly lower (R²=0.823). With the calculated models these properties of new mixes within the practical range of the input variables used in the training can be predicted with an absolute error for slump flow, T₅₀, T₆₀, V-funnel flow time, Orimet flow time, and L-box ratio of 3.3%, 13%, 16%, 14%, 15%, and 22%, respectively. The results show that the ANN model can predict accurately the fresh properties of SCC.

@en