Why Use Network Couplers?

When modeling a complete physical system, the models can become quite large, resulting in longer simulation times. This can become problematic when you need to deploy the physical model to real time for hardware-in-the-loop studies. It can also be an issue for desktop simulation when you need to explore multiple scenarios or design options.

In such cases, the best first step is always to reduce modeling fidelity to the minimum level needed. For example, if modeling an electric vehicle drive cycle, there is no need to model the switching of power electronics because average value or energy-balancing methods should be sufficient. Use the detailed switching power electronics models in separate test harnesses to determine average losses to be used at the system level.

For some systems, however, this is not enough. Even when modeled at a requisite level of detail, system complexity is sometimes too high to run the complete model on a single microprocessor in real time. In this case, you need to split the model into smaller pieces that can be deployed to multiple processors. The blocks in the Network Couplers library help you do that.

A Simscape™ network cannot be directly divided into separate parts for simulation on different processors. This is because the network is parsed, simplified, and presented to the solver as a single entity. Therefore, you need to split your Simscape network into separate smaller networks interfaced to each other through Simulink^® connections. The Simscape Network Couplers library provides the starting point for you to do this. The blocks in the library look like standard Simscape blocks (for example, Inductor), but are actually subsystems that implement the network separation. The design of the coupler blocks enables you to iterate on where to divide your model without initially changing its architecture.

Naming Conventions

Once you use the network coupler blocks to split a Simscape network in your model into multiple coupled networks, each of these networks can have its own solver settings. For example, you can use a variable solver for one of the coupled networks and a fixed-step solver for another, or use two fixed-step solvers with different step sizes. This makes it easier to deploy your system for hardware-in-the-loop simulation.

For clarity, the block interface and the documentation use these naming conventions for coupled networks:

When using the network coupler blocks, connect port 1 of the block to Network 1 and port 2 to Network 2.

If both networks are variable-step, or both networks are fixed-step with the same step size, then the block orientation does not matter.

Best Practices for Splitting Physical Networks

Deciding where to split a network requires some thought. Ultimately, you are looking to balance the computational burden evenly across two or more microprocessors. However, to minimize loss of modeling fidelity and to reduce likelihood of numerical instability, you also want to split the network at places where there are no fast dynamics.

When interfacing the separated networks, it is necessary to introduce some dynamics or delays to remove any algebraic constraints, which may lead to the loss of fidelity or instability issues. One approach to mitigate these issues is to select a dynamic component in the system as a splitting point and use its time constant to remove the algebraic loop. Suitable components include inductors, flexible shafts, and compressible volumes of fluid.

Also, avoid splitting the network at a point where there is high-frequency activity. For example, in a power network, split at a DC link rather than an AC link if you have the choice.

About the Network Couplers Library

The Network Couplers library, available as a sublibrary in the Utilities library, contains network coupler blocks for mechanical, electrical, thermal, and fluid domains:

For the electrical domain, the library provides several blocks for use in a variety of scenarios:

The block icons are similar to the equivalent Foundation library blocks, but have small annotations to indicate whether they act internally as Through or Across sources. These annotations help when deciding which coupler block to use. For example, the Network Coupler (Inductor) block icon shows the standard inductor symbol with two overlapping circles adjacent to each port, to indicate two internal current sources. Therefore, you should never connect the Network Coupler (Inductor) block directly to a current source, or to another inductor. For more information, see Avoiding Numerical Simulation Issues.

Additionally, the Network Couplers library contains a Fundamental Components sublibrary with prediction and smoothing blocks. For more information, see Prediction and Smoothing.

General Principles of Using the Network Couplers Blocks

The blocks in the Network Couplers library look like standard Simscape blocks, but are actually subsystems that implement the network separation.

When you add a Network Couplers library block to your model, the library links are broken. This means that you can customize the coupler blocks and associated subsystem contents, if needed.

When you double-click a coupler block in your model, the underlying subsystem opens. It generally contains two masked subsystems, Port 1 Interface and Port 2 Interface, connected to each other through two rate transition blocks. By default, the rate transition blocks are commented through. Uncomment them if at least one of the coupled networks is running fixed step.

Masking the Port 1 Interface and Port 2 Interface subsystems, rather than the top-level network coupler subsystem, helps with restructuring the model into different task subsystems for real-time deployment.

Typical Workflow

This is the recommended workflow for dividing a network:

For a detailed example of splitting an electrical vehicle model into separate networks and then deploying it to a multirate HIL, see Configuring an EV Simulation for Multirate HIL.