QUASIM

QC-Enhanced Service Ecosystem for Simulation in Manufacturing

© Projekt QUASIM

Project description
The QUASIM project investigates how quantum computing (QC) can be used to develop medium- and long-term benefits for manufacturing industry applications. The manufacturing industry is one of the central German economic sectors and at the same time requires the fulfillment of the highest quality standards in order to be globally competitive. In order to avoid errors in manufacturing and to derive optimized parameterizations of the machines, simulations are used. These are based on physical and material science models and mathematical equation systems, which place considerable demands on the engineering knowledge in modeling and the computational resources for simulations. Particularly SMEs are often overburdened with the use of such approaches. In this context, QUASIM will test different QC approaches that accelerate simulations and embed them in Quantum Services in a practical way. This should also enable manufacturing companies, which only have limited expertise in performing simulations, to benefit from quantum computers.

Challenge and innovation
In manufacturing, simulations are used to eliminate faulty manufacturing configurations and plans. However, the simulation of complex modeling through numerical methods, such as Finite Element Method (FEM), reaches limits in runtime and memory capacity for complex manufacturing. The QUASIM project is therefore investigating the acceleration of these simulations using an approach that combines numerical methods with QC. Given that quantum computers are still very small, their use is particularly interesting in time-critical applications, i.e., applications that are solvable on classical computers but for which unacceptable time is required. In parallel, alternative modeling by means of quantum machine learning (QML) is being investigated. For this purpose, QML will be applied to data from real manufacturing environments in QUASIM. The correlation between ML model types and QC as well as their effects on manufacturing quality will be investigated. By comparison with previous approaches, innovative solutions based on QC will be designed, implemented, integrated into low-threshold quantum services, which can also be made available via GAIA-X environments in distributed environments. In this way, QUASIM sets a milestone of innovation for the competitiveness of the manufacturing industry in Germany.

Solution approach
QUASIM follows a three-phase solution path including the development of a "baseline" with which the two QC-based solution paths (numerical methods + QC and QML) are compared. For the baseline, dynamic models are simulated via numerical methods and used as a basis for the computational derivation of manufacturing parametrizations. The second way accelerates the numerical models by using QC technologies. Here, both obvious methods, i.e. methods that can be implemented with available hardware, and methods that are promising in the long term are selected. For this purpose, focused QC approaches based on prefabricated solutions, if possible, will be adapted to numerical methods, simulated as well as tested on real hardware and evaluated for feasibility. The third approach addresses the addition of ML models in combination with QC technologies (QML) to the numerical approach to investigate whether complex modeling can be replaced by a data-driven approach. To move from QC requirements to ML models, different hybrid QML models are analyzed. The central element here is the derivation of a method for the structured develop-ment of QML systems.
All three solution paths are tested in QUASIM in parallel against each other in order to be able to quantitatively identify promising approaches. This is intended to discover and avoid technological dead ends at an early stage. The results of the numerical and ML-based QC models are compared to the baseline model by simulation runs and expert analyses.

Consortium
Deutsches Forschungszentrum für Künstliche Intelligenz GmbH (DFKI) (project lead), Forschungszentrum Jülich GmbH, Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e. V. - Institut für Produktionstechnologie IPT, ModuleWorks GmbH, TRUMPF Werkzeugmaschinen GmbH + Co. KG

Duration
January 2022 – December 2024

Budget
Total costs: € 5.2 million
Funding volume: € 3.9 million