Leveraging AI and MBSE for ECU Building Blocks
Speaker: A Chief Systems Engineer from Honeywell (presenting on behalf of the project team)
Event: INCOSE IW 2026
Overview of Key Themes
The presentation detailed Honeywell’s significant digital transformation, focusing on a pilot project to reinvent their digital engineering ecosystem. The core theme was the development of a modular, Model-Based Systems Engineering (MBSE) approach for creating Engine Control Unit (ECU) building blocks. The speaker emphasized the shift from rigid, custom-developed systems to a more flexible, efficient architecture leveraging SysML v2, AI, and a suite of integrated tools. This new method aims to accelerate development cycles, improve design validation through early co-simulation, and streamline the management of component obsolescence, ultimately delivering greater business value through product line engineering.
Key Points
- Project Background: Honeywell is undergoing a massive transformation, moving towards a more robust digital engineering and MBSE framework. This project is one of over thirty pilots, aiming to build on decades of modeling expertise. The previous approach to creating ECU building blocks was hampered by heavy custom development and a rigid structure, making updates challenging.
- The New Architecture: The new solution utilizes an ecosystem of modern tools to facilitate rapid development.
- It pairs SysML v2 for system modeling with MATLAB/Simulink for simulation and analysis.
- This is integrated with OpenMBEE (including Starcat, Starforge, and Flexo) which provides a queryable model of models.
- This allows the team to focus on developing the core deterministic libraries of electronic components rather than building the entire support structure from scratch.
- Use Case & Business Value: The primary use case is the Engine Turbofan Digital Integrated Control Unit. The main business goal is to improve cycle time. This project is a foundational component of Honeywell’s broader Product Line Engineering (PLE) strategy. A key benefit is the ability to perform “pre-verification” earlier in the design cycle using co-simulation.
- AI-Assisted Design: A central feature of the new process is a conversational AI interface (built on LLMs). Engineers use natural language in a Jupyter Notebook to instruct the AI to:
- Pull pre-defined, validated building blocks from a library.
- Connect these blocks according to specified rules.
- Generate the corresponding SysML v2 textual code and visual diagrams.
- The AI is not creating new components from scratch but is orchestrating the assembly of known, trusted parts, ensuring a deterministic outcome.
- Managing Obsolescence: This model-based library is key to managing component lifecycle. When a subcomponent becomes obsolete, the system can be queried to find a replacement with matching or similar performance specifications from the library. This drastically limits the redesign effort required for sustainment engineering.
- Project Status and Next Steps:
- The prototype has successfully proven the concept from requirements through architecture and co-simulation.
- The next steps involve completing the digital thread integration into hardware and software groups, expanding the component libraries, and building out the entire technology stack (including OpenMBEE and LLMs) internally to handle proprietary and export-controlled data.
- The goal is to have a production-quality system by 2026, with plans to apply the framework to other system-level components like fans, compressors, and integrated modular avionics across Honeywell by 2027.
Key Quotes
- “Honeywell’s going through a massive transformation. We’re splitting away from the other Honeywell companies, and because of that, it’s an opportunity for us to reinvent and improve the digital engineering ecosystem that we have in place.”
- “It’s not magic; they’re not magically created… These building blocks at the core are already made; the AI is making the links between blocks. And there are rules to making those links.”
- “By being able to push these analysis results into the test management platform, we can actually plot that over time and see the trajectory of our design, iterations and our progressive elaboration of the design and the model towards compliance to the requirement.”
- “In order to use our proprietary data… We can’t be running on an external LLM, so we’re going to have to move that. So we’re going to have to build the entire Open MBE stack internal.”
Action Points for the Audience
- Embrace Modular, Library-Based Design: Focus on creating libraries of trusted, reusable “building blocks” for your systems. This standardizes components and accelerates the assembly of new design variants.
- Integrate AI as an Orchestrator: Explore using conversational AI and Large Language Models (LLMs) not to invent novel engineering, but to orchestrate the assembly of pre-validated components. This can serve as a powerful user interface, lowering the barrier to entry for complex modeling tasks.
- Prioritize the Digital Thread: Establish a clear digital thread from requirements through design, analysis, and testing. Connecting tools like requirements managers (e.g., DOORS, Jama), SysML model repositories, and test platforms enables traceability and data-driven validation throughout the lifecycle.
- Link Engineering Efforts to Business Value: Ensure that all digital engineering and MBSE initiatives are directly tied to tangible business outcomes, such as reduced cycle time, improved quality, or streamlined sustainment.
- Plan for In-House Infrastructure: For organizations dealing with proprietary or sensitive data (e.g., CUI, ITAR), develop a long-term strategy to host and manage your digital engineering stack, including LLMs, internally to maintain data security and control.
