packageUsersGuide

Information

Welcome to the SMArtInt Library

The SMArtInt (Simple Modelica Artificial Intelligence Interface) library provides components for integrating external neural network models—such as TensorFlow Lite or ONNX—into Modelica-based simulations. It is designed to work with Dymola and OpenModelica and enables hybrid modeling approaches that combine physics-based models with machine-learned surrogate models.

Purpose

SMArtInt facilitates the connection between dynamic Modelica models and external AI-based computations by providing reusable interface blocks and examples. Typical use cases include:

• Embedding neural networks as surrogate models
• Applying hybrid modeling and control strategies, combining physics-based and data-driven components

Library Structure

The SMArtInt library consists of the following main Modelica packages:

• UsersGuide: Introduction and overview of the library, usage patterns, and documentation structure.
• Tester: Ready-to-run example models demonstrating how to use SMArtInt components in simulations.
• BaseClasses: Abstract base models and partial classes providing reusable structure and logic for custom blocks.
• Blocks: Usable interface blocks for loading and running AI models (.onnx or .tflite) inside Modelica models.
• Internal: Utility components and internal implementations not intended for direct use.

Resources/ExampleNeuralNets: Python scripts and neural network files (e.g., .tflite, .onnx) used in the tester examples. They show how to prepare compatible models from common ML frameworks.

Getting Started

1. Open the SMArtInt library in your preferred Modelica environment (Dymola or OpenModelica).
2. Explore the SMArtInt.Tester package to run ready-to-use examples.
3. Review the documentation included with each block for configuration and usage instructions.
4. The Overview section also provides a general introduction to the library.
5. If needed, use the provided Python scripts to generate your own AI models for integration.

Documentation Details

Detailed usage instructions, parameter descriptions, and model behavior are documented directly within the Modelica components/packages. To understand how a specific block or model works, simply navigate to it in your Modelica tool and view its embedded documentation.

The tester models also include additional explanations and typical application patterns. They are intended as both verification models and starting templates for your own projects.

System Requirements

The library relies on external C-libraries and works on 64-bit Windows systems. Please ensure the following for successful use:

• Use 64-bit Dymola (tested) or recent versions of OpenModelica
• Tested versions:
• Dymola 2026x Refresh 1
• OpenModelica v1.26.2
• Modelon Impact 4.142.3
• MWorks Sysplorer 2026b(x64)
• Set Advanced.CompileWith64 = true in Dymola if required

Further Resources

• GitHub Repository: https://github.com/xrg-simulation/SMArtInt
• SMArtInt+ (Commercial Version): Additional features such as extrapolation warnings, dedimensionalization, and automatic block generation are available via XRG Simulation’s SMArtInt+ Library.

Contact

For feedback, questions, or contributions, please visit the GitHub repository or contact XRG Simulation GmbH via their website.

Contents

Name	Description
Overview
LicenseConditions
Contact