AI Training Data Generation, Management, and Reuse with SPDM, PIDO, and HPC Episode 1 

Overview

This session focuses on the upstream phase of AI model development: generating high-quality, physics-informed synthetic data using simulation workflows. We’ll demonstrate how Ansys optiSLang enables efficient Design of Experiments (DOE) and sensitivity analysis to structure simulation runs, while leveraging HPC resources to accelerate execution. All simulation artifacts, models, workflows, and results are managed in Ansys Minerva for traceability, revision control, and cross-team collaboration.

The engineering team will set up a multiphysics simulation workflow in Ansys optiSLang, execute it on HPC (including parallelization strategies), and store the resulting design data in Ansys Minerva. The AI/ML team will then reuse this data to train and validate models using Ansys optiSLang’s no-code AI capabilities. We’ll highlight how Ansys Minerva ensures traceability from the original physics model to the trained AI model, enabling a robust simulation digital thread.

What Attendees Will Learn

• Structuring simulation workflows for DOE and sensitivity analysis using Ansys optiSLang
• Accelerating simulation execution with HPC and parallelization (design-level and solver-level)
• Managing models, workflows, and data in Ansys Minerva with revision control and access governance
• Reusing simulation data for AI model training and validation with Ansys optiSLang’s no-code interface
• Ensuring traceability across teams and models using Ansys Minerva’s digital thread capabilities

Who Should Attend

• Audience Simulation/Optimization Expert and Data Engineers

Speaker

• David Schneider 
• Sak Arumugam