Image-based Deep Learning for the time-dependent prediction of fresh concrete properties

  • 2024-02-09 18:42:30
  • Max Meyer, Amadeus Langer, Max Mehltretter, Dries Beyer, Max Coenen, Tobias Schack, Michael Haist, Christian Heipke
  • 0


Increasing the degree of digitisation and automation in the concreteproduction process can play a crucial role in reducing the CO$_2$ emissionsthat are associated with the production of concrete. In this paper, a method ispresented that makes it possible to predict the properties of fresh concreteduring the mixing process based on stereoscopic image sequences of theconcretes flow behaviour. A Convolutional Neural Network (CNN) is used for theprediction, which receives the images supported by information on the mixdesign as input. In addition, the network receives temporal information in theform of the time difference between the time at which the images are taken andthe time at which the reference values of the concretes are carried out. Withthis temporal information, the network implicitly learns the time-dependentbehaviour of the concretes properties. The network predicts the slump flowdiameter, the yield stress and the plastic viscosity. The time-dependentprediction potentially opens up the pathway to determine the temporaldevelopment of the fresh concrete properties already during mixing. Thisprovides a huge advantage for the concrete industry. As a result,countermeasures can be taken in a timely manner. It is shown that an approachbased on depth and optical flow images, supported by information of the mixdesign, achieves the best results.


Quick Read (beta)

loading the full paper ...