The project research computational methods for handling granular material in digital twins of transportation and processing systems. A particular focus is on enabling development of machine learning models for state determination, tracking and prediction where no real sensors can measure. The results will contribute to a virtual testbed of solutions for fully automated mining operations, to be used both for engineering and academic research.

A digital twin is a virtual replica of a real cyberphysical system, like manufacturing or processing plant, or a vehicle. It is based on a large collection of physics-based and geometric models in conjunction with controllers and sensor data in an IoT infrastructure. The digital twin functions as an interface to the real system, making it more easily accessible for monitoring, analysis, machine learning and, ultimately, creating software applications for smart control, predictive maintenance and early detection of risk for catastrophic failure. The major benefit of digital twins is the possibility of experimentation using full-system simulation to investigate radical changes in control strategies and system design. Physics based simulation enables also determination of the internal state in processes and measurement of quantities that are hidden in the real system for practical or fundamental reasons.