Abstract
D-DREAM establish a continuous, data-driven engineering workflow from lab scale to CARE Prefab Plant that enables the seamless transfer of data and information along the entire process chain. The focus is on maintaining consistent component quality across varying materials, environmental influences, and geometric complexities. This interdisciplinary project connects material development (RA1), complex designs automation (RA2), applicable across various manufacturing technologies (RA3).
Aiming to automate a construction workflow, the project integrates simulation environments by designing a digital twin that allows the entire construction pipeline to be replicated at a small scale. It enables efficient testing, optimization, and validation before deploying the workflow at full scale in real-time applications.
ML models will be employed to precisely optimize manufacturing parameters using material data and quality objectives, enabling both process optimization and inverse optimization.
Advanced ML methods, including transformer-based multimodal fusion models, self-supervised learning approaches, foundation models, and lightweight models, will be used for robust sensor data integration, anomaly detection, and efficient on-site inference.
The project creates a digital, data-driven workflow that starts with modeling materials and processes, is tested through lab-scale manufacturing, and is then validated under near-realistic and full production conditions, including in a virtual CARE Prefab Plant.
