Solving the “AI drift” challenge

AI drift is one of the severe challenges faced by AI simulation. When facing the unknown design space, the model will infer inaccurate situation. To address this challenge, the HEEDS AI Simulation Predictor software introduces a verity-conscious AI that proactively self-validates predictions to help engineers develop accurate and reliable simulations for real-world engineering applications.

Behnam Nouri, team leader of Siemens Energy Engineering and Platform Design, said: “With the help of the HEEDS AI Simulation Predictor software, Siemens Energy has achieved significant improvements in gas turbine components and improved thermal fatigue prediction, and we are now able to process approximately 20,000 design components in 24 hours, increasing component life by 20%. This allows us to fully describe the limitations of the existing design space and create more efficient turbine engines. The HEEDS AI Simulation Predictor technology has now saved more than 15,000 hours of computing time.”

Simcenter Reduced Order Modeling, launched by Siemens at the same time, is a solution for training and validating AI/ML models using high-precision simulation and test data. Based on trained and proven models, engineers can execute predictions in a fraction of a second.

5SHY3545L0009

Simcenter Reduced Order Modeling Solution Benefits:
• Speed and accuracy: Simcenter Reduced Order Modeling software uses high-precision data to help engineers make predictions and key decisions quickly, in less time than traditional methods.
Predictive performance: Training AI/ML models based on comprehensive data sets enables engineers to gain robust, reliable, and trusted insights to help eliminate common AI drift problems.

Plastic Omnium simulation engineer Jurgen Dedeurwaerder said: “The Simcenter Reduced Order Modeling solution allows us to increase the speed of our simulation models, allowing detailed fuel cell device models to run faster than in real time, while maintaining the same level of accuracy as full system models. We can now improve the speed of models in loop controller development and testing, reducing the overall development cycle by about 25%. At the same time, the solution provides us with a reliable, IP-protected and cost-effective way to distribute our models to other teams, both internally and at our customers, to enhance their products and processes and deliver better quality products to end users.”

Jean Claude Ercolanelli, Senior Vice President, Digital Industrial Software Simulation and Test Solutions, Siemens, said: “The HEEDS AI Simulation Predictor and Simcenter Reduced Order Modeling deliver a true breakthrough in simulation technology, enabling our customers to get the most out of AI simulation, Improve the speed, accuracy and robustness of its design space exploration. Now, customers can not only apply these innovative technologies to new projects, but also leverage decades of historical simulation data to provide new insights into existing projects.”