#8 Patterns for smart manufacturing for SMEs

Manufacturing industries need to evolve to the next technology level and to improve their competitiveness
in terms of production performance, component quality, and maintenance cost. This testbed is devoted to
the definition of a general and replicable methodology for SMEs based on the convergence of data
analytics, AI, IoT and physics-based simulation. The first phase of the testbed consists in the
development, adaptation and testing of the methodology in a Lab context at CETIM, where the needed
research/development competences and the infrastructure facility are already available (and already
operating on existing projects). In the second phase, the methodology will be replicated and validated on
two real industrial use cases, including turning and milling machines.
The testbed foresees the development of:
1. additional steps of optimal data processing (for removing correlations and determining goal-
oriented data only);
2. process learning and knowledge extraction, by combining synthetic big-data produced via efficient
real-time simulations and AI techniques to emulate rich sets of possible defaults;
3. real-time decision making techniques to control production quality (when machines allow online
parameter tuning), including monitoring of surface defects (short-time loop), monitoring of small
deviations (tool wear, corruption, tool breakage, ….) for medium-time loop, and performing
predictive maintenance for long-time loop.
The distributed digital twins developed within this test-bed will be specific for each machine and will evolve
in time.