Third public release (v1.0.2) of the official eFMI® Test Suite of the Modelica Association Project Functional Mock-up Interface for embedded systems (MAP eFMI®). The release comprises:
eFMIModelica library: Support library to ease preparing Modelica models for eFMI® tooling, particularly for embedded table interpolation.eFMI_TestCasesModelica library: Official MAP eFMI® library with test-scenarios for evaluating and crosschecking eFMI® tooling.- GALEC-examples directory: Set of eFMI® GALEC programs intended as GALEC compiler test suite, e.g., to test production code generators.
The release is for the upcoming eFMI® 1.0.0 Standard. The Modelica libraries are based on the Modelica Standard Library (MSL) 4.1.0.
The main differences compared to the previous release are:
- Migration to MSL 4.1.0.
- New
eFMI.NeuralNetworkspackage, enabling the modeling of artificial neural networks (NN) in Modelica. All NN components are completely equation-based, leveraging symbolic manipulation and optimization of Modelica tooling and enabling preservation of equation-based tensor-flows as multi-dimensional GALEC expressions, such that the explosion of production code sizes due to scalarized tensor-flows can be avoided. - New
eFMI_TestCases.M11_NeuralQVMexample, demonstrating the use of NNs combined with physics-equations (hybrid Modelica model, PeN-ODE: physics enhanced neural ordinary differential equations) to incorporate the unknown non-linearities of the suspension (spring and damper) of a vertical dynamics quarter vehicle model (QVM) used as predictor/virtual sensor.
Screenshot of new eFMI.NeuralNetworks package:
Screenshot of new eFMI_TestCases.M11_NeuralQVM example:
