AI-ROMs: From CAE Data to Real-Time Predictive Models

Course category

Vertical

Level

Advanced

Duration

8 hours [4 hours a day, for 2 days]

Time

09:00 - 13:00 | 14.00 - 18.00

Language

English

Tutor

Stefano Carrino

Description

EnginSoft, within its extensive portfolio of training courses, offers a theoretical and practical course on the creation of low computational impact mathematical-statistical models based on Reduced Order Modeling (ROM) methodologies.

The course aims to provide both a theoretical and practical perspective on the application of reduced order models in the field of CAE, specifically focusing on the generation of predictive models from data obtained through numerical analyses across multiple disciplines (thermo-structural analysis, electromagnetism, CFD, etc.). These predictive models will be developed by applying ROM algorithms.

Reduced Order Models (ROMs) offer numerous advantages, including a significant reduction in computation time and computational impact, making them easily integrable into existing design workflows. The use of Artificial Intelligence (AI) algorithms represents an advanced and innovative approach that is crucial for improving accuracy and reducing computational effort. The integration of AI and ROMs not only enhances computational efficiency and effectiveness but also transforms how CAE models are used, opening new frontiers for complex and real-time applications. The use of ROMs in the CAE context makes simulations more accessible (contributing to the democratization of software and enterprise tools) and accelerates the entire design process, from the pre-design phase to final validation, while ensuring accuracy and robustness of the analyses. ROMs in the CAE field find applications in the following scenarios:

• Preliminary analysis and design optimization
• Parametric analyses
• Virtual prototyping and reduction of costs associated with experimental testing
• Integration into multi-disciplinary optimization (MDO) processes
• Digital Twin for monitoring, diagnostics, and predictive maintenance
• Decision-making support

Target Audience

The course is aimed at users (all engineers, designers, researchers) who need to:

• reduce the computational effort of complex detailed models
• predict output in real-time
• use reduced models in system modeling

Pre-requisites

Agenda

Theory: Detailed description of Static and Dynamic ROMs:

• Data preparation (formatting)
• Data dimensionality reduction (Singular Value Decomposition - SVD - & Machine Learning)
• ROM construction (Static ROMs, Dynamic ROMs)
• ROM validation
• Export (FMU creation) & Run

Practical: Hands-on exercises on application test cases with the generation of the following types of ROMs:

• LTI ROM (State-Space Model of a cantilever beam)
• Static ROM
• Scalar Dynamic ROM

For the Reduced Order Modeling (ROMs) methodologies, a commercial code will be used.