Résumé
Akselos provides industrial digital twins through component-based reduced order modeling. This approach bridges the gap between traditional Finite Element Analysis (FEA) and real-time operational needs. Unlike standard FEA, which can be computationally prohibitive for entire assets, Akselos’ technology allows for models with over 100 million degrees of freedom to be solved in under 10 seconds. This is achieved by decomposing large structures into components and interfaces, and by training a low-dimensional representation of the solution manifold for individual components and interfaces. The approach includes a posteriori error estimators to ensure accuracy. During operations, these estimators monitor every solve; if a new scenario causes the error to exceed a set tolerance, the system triggers automated enrichment to "learn" the new state. This integration of high-fidelity physics with rapid evaluation allows operators to incorporate sensor feeds and inspection data into a live digital twin. Consequently, companies can monitor closely their assets, ensuring safe operations with maximum output for the longest time possible.
