Simcenter MBSE 2404 – new ways to master complexity
Complexity is a fact of modern life and a continuing challenge for companies looking to engineer new products, processes and systems which operate under the complex laws of physics and in harmony with the natural environment, but which deliver us the results we require to the high standards we demand.
Model-based systems engineering (MBSE) helps to master this complexity by creating a link between the requirements of a product (what we would like it to do) with real-world behavior (what it will really do). The resulting virtual test bench unleashes the design to evolve much more rapidly, and with a greater visibility of future roadblocks, incompatibilities or unforeseen complications.
Simcenter MBSE 2404
The recent release of Simcenter Studio and Simcenter System Architect includes several new capabilities which empower our users to do more with MBSE and to do it faster and more effectively. This blog explores some headline news – starting with a powerful weapon to defeat complexity….
Enter Reinforcement Learning
One challenge that can crop up with an MBSE-accelerated design process is system development outpacing controls development. In some ways a good problem to have – but to test a system, you do need to be able to control it. And that control system also needs to be carefully designed, developed and tested first. Or does it?
We hear with ever increasing frequency about the latest achievements of AI – according to a recent article in Nature – now rivalling human beings at many basic tasks. A few articles further on and you might find yourself reading about how AI is also rather good at very specialized non-basic tasks too. A great example is this inspirational paper describing how reinforcement learning was used to effectively control a complex plasma by manipulating magnetic fields, avoiding “substantial engineering effort” in controller design and development. The best part is that reinforcement learning technology may be in your hands already in the form of Simcenter Studio, and it just got better.
Reinforcement Learning workflow in Simcenter Studio
Reinforcement Learning is a powerful tool for tackling challenging, highly tuned, control (as required for a Tokamak plasma), as well as for simpler applications where a control algorithm is required for immediate validation and testing of a system design.
The recent 2404 release of Simcenter Studio brings this to life with an end-to-end workflow which delivers a deployable controller in the FMU format, ready for immediate use. Here’s the recipe:
- Take an existing plant model in need of a controller
- Connect the model to Simcenter Studio 2404
- Choose from a set of state-of-the-art RL agents to control your model, learning through experience and arriving rapidly at an efficient and effective solution
- Export to FMU for validation and deployment of the trained controller
Further connecting the worlds of simulation and system architecture
A key role of simulation in an MBSE workflow is to enable performance targeting and verification by executing simulation models in the context of system architectures. This requires a seamless flow of information between these two worlds, in both directions.
That bridge is Simcenter System Architect – a powerful co-simulation platform, able to consume architectures from the systems engineering world, and to drive them using models from the simulation world.
In the 2404 release, Simcenter System Architect deepens this connection with an enhanced ability to import multi-level system architectures from Cameo, Rhapsody and System Modeling Workbench, and to maintain that hierarchy in the simulation architecture.
Two-way traffic is important too
On the subject of a seamless flow of information between simulation and system architecture, the 2404 release of Simcenter System Architect is also able to detect the presence of a Simulation Scenario attached to an imported Internal Block Diagram (IBD). It can automatically import associated parameters, sharing simulation results back to the architecture authoring environment, facilitating collaboration and data flow.
Simplifying simulation architectures with senders and receivers
Complexity can also be visually challenging when it comes to implementing simulation architectures. Simcenter System Architect 2404 has a solution – rather than displaying every connection between model blocks, you can choose which ones are the important ones to see. The others can be replaced with colored sender and receiver blocks which are easy to inspect and toggle on/off for visualization.
Now supporting Windows 11
In the 2404 release you will also find more user experience enhancements – both Simcenter System Architect and Simcenter Studio now support Windows 11, with a simplified installation process for Simcenter Studio and upgraded, interactive documentation on Support Center.
As always, those are only the headlines – find out more!
The new capabilities I have described in this blog deliver some exciting new possibilities, but there are many other enhancements to user experience, improved client and solver performance and extended SysML v2.0 support to go along with them.
