Reduced Order Modeling with MATLAB and Simulink

Create AI-based reduced order models

Simulink Add-On for Reduced Order Modeling provides an app for creating reduced order models (ROMs) of subsystems modeled in Simulink, including full-order, high-fidelity third-party simulation models. You can use reduced order models for system-level desktop simulation, hardware-in-the-loop (HIL) testing, control design, and virtual sensor modeling.

With Simulink Add-On for Reduced Order Modeling, you can:

Set up the design of experiments and generate input-output training data from a full-order, high-fidelity subsystem
Train and compare AI-based reduced order models using pre-configured templates
Export AI-based surrogate models to Simulink for system-level simulation, control design, and HIL testing
Export reduced order models as Functional Mockup Units (FMUs) for use outside of MATLAB and Simulink (with Simulink Compiler)

Diagram of reduced order modeling from full-order, high-fidelity third-party simulation models.

Screenshot of the Reduced Order Modeler app designing experiments.

Design Experiments

Select Simulink signals and block parameters to use as ROM inputs, outputs, and parameters. Interactively design simulation experiments by selecting from built-in excitation types to replace or perturb ROM inputs. Visualize coverage of design space.

Reduced Order Model of an Airframe

Screenshop of the Reduced Order Modeler app running experiments and showing experiment results with the app.

Run Experiments

Specify the option to run experiments one at a time or in parallel with Parallel Computing Toolbox and initiate model simulations. Visualize simulation results for signals and parameters of interest using built-in visualization plots.

Reduced Order Model of a Jet Engine Turbine Blade

Screenshot of experiment details for training reduced order models.

Train Reduced Order Models

Train and compare different types of reduced order models. Choose from Neural State Space, LSTM, and Nonlinear ARX models. Optimize hyperparameters sequentially or in parallel with Parallel Computing Toolbox to improve model fit. Compare accuracy metrics for trained models to select the optimal one for your application.

Reduced Order Modeling of Subsystems in Engine Model

Screenshot shows a trained ROM brought into Simulink for control design.

Use Reduced Order Models in Simulink

Bring trained ROMs into Simulink for system level simulation, control design, and HIL testing. Combine ROMs with first principles-based component models.

Reduced Order Model of a Jet Engine Turbine Blade

Diagram of ROM deployment to embedded hardware and export as an FMU.

Deploy and Export Reduced Order Models

Deploy ROMs to embedded systems through automatic code generation. Export ROMs as FMUs (with Simulink Compiler) for use outside of MATLAB and Simulink.

Requirements:

Simulink, Deep Learning Toolbox, Statistics and Machine Learning Toolbox, System Identification Toolbox

MATLAB release compatibility:

Compatible with R2023b and later releases

Download Support Package

Download add-on (beta)

Have Questions?

Talk to a technical expert.

Email us