Experiments
Intelligent prediction of woodworking machines performance (IPER)
OPEN CALL 1
ITALY, Lombardia, Veneto
Fondazione Speedhub
Experiment objective
Aims of this experiment is to replace high-fidelity simulation (e.g. FEA) of machine tools, and in particular of woodworking ones, with a resource-saving data-driven (AI-based) surrogate model within the early stages of the virtual product development process. Based on a main design parameters/data (size, strokes, working volume, materials, etc.) the surrogate model is supposed to give shortcut evaluations of some key figures, such as static stiffness and firsts modal vibration frequencies, that quickly permit to assess if an “early concept design” of a machine tool is appropriate in order to enable engineers to modify/optimise the concerned key design parameters.
Challenges
FEA simulation of machine tools often requires very large amount of time for a proper product modelling; this in turn will slow down the design loop before converging to an optimal early design solution (to be then finalised in subsequent stages). The main challenges of IPER are:
- Challenge 1: Development of a AI based data-driven model (surrogate model) of a machine tool
- Challenge 2: Prediction of structural performance of a new designed machine tool by a data-driven approach that ‘learns’ from historical results obtained by physical simulations /tests
- Challenge 3: Quick identification of optimal design choice in early design stage
Implementation Solution
The experiment includes three main technical workpackages, that are: WP1 – Data structuring and workflow specifications, WP2- Data analytics (where the Data-driven surrogate model will be developed and a trained), and finally the WP3 – Testing and validation (where the model will be validated and assessed against KPIs). Finally, WP4 has been included to implement exploitation and dissemination actions for promoting the results of the experiment towards the relevant stakeholders, DIH-WORLD members and scientific community.