A method is proposed in this paper to automatically estimate the workability of six different strength grades concrete by recording the mixing process based on deep learning. The concrete mixing videos were collected in a specially designed set up fixed on a mixer located in the mixing station. These videos were transformed into a series of image sequences to fit the deep learning model to predict the slump and slump flow values of concrete, with six groups in total and more than twenty thousand image sequence samples. The workability of six groups concrete with different strength grades learned by the DL model, was estimated. The results indicate that the trained deep learning model with CNN and LSTM can estimate concrete workability effectively. Our goal to estimate concrete workability in different strength grades is achieved.

本文提出了一种基于深度学习通过记录搅拌过程自动估计六种不同强度等级混凝土和易性的方法。混凝土搅拌视频是在一个专门设计的装置中收集的，该装置固定在位于搅拌站的搅拌机上。这些视频被转化为一系列图像序列，以拟合深度学习模型来预测混凝土的坍落度和坍落度流动值，总共六组，两万多个图像序列样本。对通过 DL 模型学习的具有不同强度等级的六组混凝土的和易性进行了估计。结果表明，经过训练的带有 CNN 和 LSTM 的深度学习模型可以有效地估计混凝土的可加工性。我们估计不同强度等级的混凝土和易性的目标已经实现。