초음파 감쇠계수 기반 3D 프린팅용 UHPC의 유변학적 특성 추정을 위한 인공신경망 모델

Abstract

To ensure structural stability and buildability in 3D concrete printing (3DCP), it is essential to measure and evaluate the rheological properties of fresh concrete in real time. However, conventional rheometer-based tests, although highly precise, have limited applicability in practice due to the high cost of equipment, long measurement time, and reliance on operator expertise. In this study, the rheological properties (yield stress and viscosity) and ultrasonic signals of ultra-high-performance concrete (UHPC) for 3D printing were measured, and their correlations were analyzed. The experimental variables were set as the water-to-binder ratio (W/B) and the dosage of superplasticizer (SP). Ultrasonic measurements were performed using a transmitter and receiver operating at 54 kHz, and the attenuation coefficient was calculated from the acquired signals. The results showed that the attenuation coefficient increased as the ultrasonic signal was weakened by particle dispersion and pore structure during the early stage of hydration. In particular, mixtures with lower W/B and lower SP content exhibited a rapid increase in attenuation coefficient over time, whereas those with higher W/B and higher SP content showed a more gradual increase. These variations in attenuation coefficient were consistent with the time-dependent changes in yield stress and viscosity. From these results, an artificial neural network (ANN) model was developed using attenuation coefficient, mixture parameters, and initial rheological properties as input variables to predict the yield stress and viscosity of fresh concrete. The proposed model achieved coefficient of determination (R²) of 0.81 and 0.82 for yield stress and viscosity, respectively, in the test set.