Short name: COGNIWARE
Long name: Cognition-driven Maintenance supported by Computing Continuum in Industry 4.0: Cognition4.0
Call: F4Fp-09 (see call details)
Proposal number: F4Fp-09-M33
SUMMARY REMARKS & TESTBEDS
This proposal is about experimentation how the cognition-driven maintenance, a novel approach for predictive maintenance driven by the cognition process, can be efficiently realized using powerful computing infrastructure, by focusing on the trade-off between the price of the infrastructure and the performances of the system. We argue that the modern concept of Computing Continuum can be applied for the realization of the cognition-driven maintenance, since it seamlessly combines resources and services at the center (e.g., in Cloud datacenters), at the Edge, and in-transit, along the data path. Typically, data is first generated and preprocessed (e.g., filtering, basic inference) on Edge devices, while Fog nodes further process partially aggregated data. Then, if required, data is transferred to HPC-enabled Clouds for Big Data analytics, Artificial Intelligence (AI) model training, and global simulations.
This federation of computing resources requires a federation platform like FED4FIRE+. As a main enabler, we have selected Grid5000 testbed. The main outcome of the experiment is performance measurement and scalability test of new cognition-driven maintenance that is able to detect unusualities in the operation as soon as they appear (and before a problem arises), but without any explicit knowledge about the problem, like the human-cognition is able to resolve unknown situations. From the innovation point of view, this proposal is about testing a novel architecture for predictive maintenance in manufacturing (cognition-based maintenance) on the novel computing infrastructure that will be provided by FED4FIRE federation of testbeds. We argue that this work can influence many other application domains for implementing cognition-based processing through computing continuum infrastructure Main argument for FED4FIRE+: we cannot find a suitable commercial environment where this complex computation process can be tested in such a convenient way.