Blog
How TechSim Helps Companies Accelerate Simulations Using Neural Networks
Reduced Order Modeling in Simcenter Amesim in Practice
Modern engineering systems are becoming increasingly complex, combining mechanics, hydraulics, thermodynamics, electrification, and control software. While it is now possible to create highly accurate simulation models, many teams face the same challenge: high‑fidelity models are computationally expensive and difficult to use in real‑time applications.
At TechSim, we focus on making simulation models not only accurate, but also efficient and practically usable throughout the development process. One of the key technologies we use to achieve this is the Neural Network Builder in Simcenter Amesim—Siemens’ solution for creating Reduced Order Models (ROMs) based on neural networks.
When High‑Fidelity Models Slow Development Down
Based on our experience in automotive, energy, and industrial projects, engineers typically face these challenges:
- simulations taking minutes or even hours,
- optimization studies requiring thousands of runs,
- models unsuitable for HIL / SIL testing,
- inability to run in real time with a fixed time step,
- the need to share models with customers while protecting sensitive know‑how.
This is exactly where Reduced Order Modeling using neural networks delivers the greatest value.
Reduced Order Models: The Same Behavior, Faster Execution
A Reduced Order Model is a simplified representation of a detailed system that:
- preserves the system’s essential behavior,
- runs orders of magnitude faster than the original model,
- is suitable for real‑time, optimization, and control development.
Simcenter Amesim offers multiple ROM approaches, with neural networks being particularly suitable for:
- highly nonlinear systems,
- systems with high‑frequency dynamics,
- applications where speed and robustness are critical.
Thanks to the Neural Network Builder, engineers can create ROMs without any machine‑learning expertise, which is a major advantage in day‑to‑day engineering practice.
Neural Network Builder in Simcenter Amesim: Practical AI for Engineers
The Neural Network Builder, integrated into Simcenter Amesim since version 2019.2, allows users to:
- generate training data from simulations or measurements,
- easily select model inputs and outputs,
- use predefined neural‑network templates,
- automatically scale data and tune the learning process,
- validate accuracy and deploy the model with a single click.
At TechSim, we use this workflow in both customer projects and the development of digital prototypes and virtual sensors.
From Static Mapping to Complex Dynamics
The Neural Network Builder supports multiple network types:
- Static (Dense) networks for steady‑state mappings,
- Dynamic (RNN) networks for time‑dependent systems,
- GRU / LSTM architectures for advanced dynamic behavior with memory.
This enables TechSim to address a wide range of applications, including:
- air‑conditioning and cooling loops,
- vehicle and chassis dynamics,
- pressure regulators and control valves,
- thermo‑electrical behavior of power electronics.
Proven Benefits in Real‑World Applications
Experience with Simcenter Amesim shows that neural‑network‑based ROMs can deliver:
- simulation speed‑ups of tens to thousands of times,
- typical accuracy levels above 98–99%,
- execution on real‑time targets,
- easy export as FMU 2.0 or native Amesim submodels.
This significantly simplifies:
- controller design and tuning,
- HIL testing,
- digital twin development,
- secure model sharing without exposing detailed physics.
Why Use This Technology with TechSim
TechSim is not just a software reseller. We help customers choose the right model fidelity for the right task.
With strong expertise in:
- Simcenter Amesim,
- Reduced Order Modeling,
- FMU integration and real‑time solutions,
we can:
- assess where ROMs provide real value,
- design appropriate training simulations,
- verify accuracy and operating limits,
- support deployment in development, testing, and operation.
Summary
Neural networks in Simcenter Amesim are not experimental technology.
They are proven engineering tools that TechSim uses to:
- accelerate development cycles,
- reduce simulation costs,
- enable real‑time applications,
- bring simulation models closer to control and operation.
If you are looking to get more value from your simulation models, contact TechSim. We will be happy to discuss your use case and propose the optimal solution.
???? www.techsim.cz
