The main objective of this FET Innovation Launchpad project is to evaluate the business feasibility of an open-source digital twin builder software for edge applications and develop and test a demonstrator (alpha) for one or more use cases identified in the market analysis.
The basis of the current project is the capability of accurately modelling highly complex problems using massive computing power (HPC) to extract the essential features of the solution so that similar results can be obtained (within a training range) at very constrained compute environment.
In order to achieve the stated objectives, the project will roughly follow the process outlined in the diagram below, starting with the market analysis, to unveil the requirements for the technical development and demonstration of the reduced order model application in one or more selected use case/s.
Following this methodology the project will be carried-out in 5 Work Packages with a total duration of 18 months:
WP1. Project Management (WP leader: CIM)
Administrative and technical coordination (including innovation).
WP2. Market Analysis & Business Plan (WP leader: EFF)
To analyse potential applications of the EdgeTwins builder, product requirements and market size, and outline a business plan for its commercial exploitation.
WP3. EdgeTwins Tool Simulations and Training (WP leader: BSC)
To define the application problem and set up and optimize the high-performance computing infrastructure.
WP4. Application Downscaling and Demonstration. (WP leader: CIM)
To refine the simulation and training capabilities available within the open source Kratos, and to define a workflow for the construction of reduced order models, suitable for edge or local computations. The workflow will be applied to a testcase/s using a constrained compute device (Raspberry Pi 4 or similar), to showcase the product possibilities.
WP5. Dissemination and Communications. (WP leader: CIM)
To manage the communication of the project activities during its execution and the dissemination of the project results focused on four groups of stakeholders: Institutional, Industry, Developers community, Academia.
Building from ExaQUte.
Integration of solver into a workflow management system:
This provides capability to leverage HPC systems in the training of ROM models. (thus enabling the construction of realistic ROM models in a short time).
ExaQUte: EXAscale Quantification of Uncertainties for Technology and Science Simulation. The ExaQUte project aims at constructing a framework to enable Uncertainty Quantification (UQ) and Optimization Under Uncertainties (OUU) in complex engineering problems using computational simulations on Exascale systems. More information in http://exaqute.eu/.