Identifying Bingham parameters from the video of mini-slump flow or channel flow

The performance of modern concrete, such as self-compacting concrete and 3D printing concrete, depends on precise control of its rheology in addition to the mix proportioning and quality production. It would have defects such as segregation and early-age cracking when its rheological properties are not qualified for a placing and casting process. A way for in-situ measurement possibly linking with the control requires vision technology, and this study proposes an attention-based long short-term memory (LSTM) framework to estimate rheological properties with a video of flow in two conventional flow tests, mini-slump and channel flow test. The proposed framework provides Bingham parameters showing close agreement with the concrete flow, which can be achieved by training the model with volume-of-fluid simulation data. Analysis of the attention weights furthermore reveals the initial flow dominantly determines plastic viscosity, while the flow stage to get stopped is related with yield stress as expected. The proposed framework was applied to different temporal segments of the flow video. The initial flow alone provides sufficient information to estimate both Bingham parameters, eliminating the need to observe the entire flow process. This finding highlights its potential for real-time monitoring during concrete casting.