Our products for cement and concrete production are all based on the same technology; we use machine learning for predictive quality control.
During the production of cement and concrete, large amounts of data are generated which currently remain unused. Our software structures and connects data from different systems, e.g. LIMS and production data acquisition. We then clean and filter the data so that measurement errors or mixed-up samples do not affect the models and thus falsify the subsequent processes.
With the collected data we train machine learning models that can predict different quality properties of cement and concrete. The models are not trained once, but regularly. This allows them to take into account the changing conditions in the plant. In addition, our models are continuously evaluated and compared to ensure that the most suitable models are used at all times.
Based on our prediction models, we determine optimal target values for production parameters in order to achieve the desired quality targets for concrete or cement. These target values are transmitted to the control stations of the cement and concrete plants.
Our machine learning pipeline and web application run in a modern cloud-based infrastructure. We use "Infrastructure as Code" with Terraform to implement changes to our production system securely and quickly. Terraform also enables us to create test environments that support fast and flexible software development.
Concrete is not a dull grey mass. Despite changing raw materials, weather and traffic conditions, achieving that the concrete on the building site has a consistency of honey to flow without segregation, solidifies at the given time, develops the compressive strength required for the structural design, and finally withstands 50-100 years of attack by de-icing salt or liquid manure with the help of a specific porosity, is similar to the work of an alchemist in earlier centuries, described in a unique way in the book of the same title by Paolo Coelho.
Responsible for the development of these properties are various crystals  in cement, the central ingredient of concrete. As soon as they come into contact with water (usually in a concrete plant), the crystals  start to grow and thus determine the properties of the concrete. The study of this growth, in all its facets, provides information on how the development of the properties of concrete can be better understood or even deliberately controlled. It is therefore highly relevant for our company.
This video was produced by Dr. Alexander Herb, who in his dissertation  succeeded for the first time to observe exactly these crystals during their growth and their partial transformation into other crystals under a special scanning electron microscope and to record this in videography. We have also used photographs of Dr. Herb on other parts of our website, for which we would like to express our gratitude.
1] Called clinker phases, e.g. tricalcium aluminate (3 CaO * Al2O3, or C3A in cement notation)
2] Then called hydrate phases, mainly calcium silicate hydrates (m CaO * SiO2 * n H2O)
3] Dr.-Ing. Alexander Herb (2003). Indirekte Beobachtung des Erstarrens und Erhärtens von Zementleim, Mörtel und Beton mittels Schallwellenausbreitung. Dissertation, University of Stuttgart, Department of Civil and Environmental Engineering.