Every system, from aircraft to satellites, depends on tiny components that can ensure or compromise safety. A single resistor or capacitor might seem insignificant. Yet, in reliability engineering, it can determine whether a system succeeds or fails. Although fundamental, reliability prediction can become a tedious and time-consuming task. That’s why at Anzen we’ve explored the use of an agentic reliability prediction analysis that helps the engineer classify components according to their type.
At Anzen, in collaboration with Tecnalia during the CORSARIO project, we explored how artificial intelligence could help classify electronic components following the MIL-HDBK-217 handbook. Our goal was simple: use AI to make reliability prediction a bit faster, automating the easiest cases and leaving the most difficult ones to the engineer. This project combines our expertise in reliability engineering with Tecnalia’s applied research in AI carry out an experiment of how an agentic reliability prediction would look like, and where are today’s limits of AI in engineering tasks.
It’s important to mention that this post is the summary of a pilot an academic project, as well as the experience from previous experiments, such as this and this. At Anzen , we believe that AI provides a lot of power to systems engineering processes; however, it must be used with care and expertise, that’s why we don’t use AI in actual certification projects. But this is a fun project anyway, so enjoy the post!
Why Classification Is Essential for Reliability Prediction
Before engineers can predict failures, they must know exactly what each component is. The MIL-HDBK-217 handbook defines how to calculate electronic component failure rates. However, it all starts with accurate classification. Is the part a fixed resistor, a tantalum capacitor, or a MOSFET transistor? The answer changes everything.
When building Fault Trees and performing System Safety Assessments under ARP4761, each classification feeds into the larger reliability model. A single mistake can distort system-level safety predictions. Consequently, reliable classification is not a small detail—it’s the foundation of a reliability prediction for aerospace, automotive, and other high-stakes sectors.
Traditionally, engineers spend hours searching datasheets, cross-referencing catalogs, and manually assigning categories. It’s slow and prone to human error. Therefore, we asked: could an intelligent agent do this automatically, using up-to-date data and consistent logic? Afterwards, the engineer would need only to validate that the allocation is correct, especially in some difficult cases.
Turning Data Into Insight with Agentic Reliability Prediction
To test the idea, Anzen and Tecnalia built a prototype AI agent capable of classifying components automatically. The process starts when the agent receives a part number. It then collects information from trusted data sources such as vendor information or Anzen in-house databases, interprets it, and classifies the component according to the MIL-HDBK-217 model. Finally, it provides an explanation of its reasoning, creating transparency and traceability.
This approach transforms a routine engineering task into an intelligent process. With AI reliability prediction, we can shorten analysis time, and focus in those components that are the main contributors to the system reliability. Moreover, every decision made by the AI must be reviewed and validated by the engineer.
How We Built the Agent
Although we can’t share confidential implementation details, we can explain the general structure. The system uses a GraphQL API to interact with internal databases. GraphQL helps the AI retrieve only the exact data it needs, improving efficiency and security. It’s like giving the AI a structured vocabulary to query knowledge without overexposing information.
The agent itself is powered by large language models (LLMs) orchestrated through modern frameworks with access to the API and the GraphQL Schema. It follows a “reactive” reasoning approach, meaning it decides based only on the current data it sees. This makes the process reproducible and transparent—two must-haves for agentic reliability prediction.
We also developed a simple web interface using Next.js. This interface allows engineers to interact with the AI directly, see classification results, and understand how each conclusion was reached. By combining backend intelligence with a clean frontend, we’re making reliability analysis both powerful and user-friendly.
Building the Future of AI Reliability Prediction
This agent is part of a long-term vision at ANZEN to develop a full-scale reliability prediction tool powered by AI and custom databases. Traditional reliability tools rely on static databases and manual updates. Ours will evolve dynamically, connecting to live data and supporting engineers with real-time reasoning.
Imagine entering a part number and instantly getting a complete classification, a failure rate according to your environment and stress, and a justification for each assumption. No endless spreadsheet updates. No manual data cleanup. Just reliable, traceable insights to be validated and then plug directly into fault trees and safety analyses. That’s the direction agentic reliability prediction is taking us.
With this approach, engineers spend less time searching for information and more time designing safe, efficient systems.
Collaboration That Drives Innovation
This work would not have been possible without the collaboration between ANZEN and Tecnalia during the CORSARIO project. Tecnalia’s experience in applied AI research complements ANZEN’s strong foundation in safety and reliability engineering. Together, we’re proving that advanced AI techniques can deliver tangible results in real engineering contexts.
By combining research and industry, we’re redefining what reliability tools can do. Our collaboration shows that agentic reliability prediction is not a distant goal—it’s already becoming part of daily engineering practice. Each prototype, each test, brings us closer to a fully intelligent reliability ecosystem.
Challenges and Lessons Learned
Deploying AI in safety-critical domains comes with challenges. The agent must be trustworthy, verifiable, and transparent. It must never invent data or make assumptions beyond the available evidence. That’s why we focused heavily on validation and data integrity. Every classification includes traceable reasoning steps and clear data sources.
Another lesson: automation does not replace experts—it empowers them. Engineers remain central to decision-making. The AI helps by removing repetitive work and by ensuring data consistency, but human oversight guarantees that complex or ambiguous cases receive proper analysis. This balance of automation and expertise defines the future of engineering collaboration.
What Comes Next for AI Reliability Prediction
We’re already planning the next phase of development. The focus will be on expanding the agent’s knowledge base, refining its reasoning, and integrating the tool within ANZEN’s broader safety assessment ecosystem. We’re also exploring new ways to visualize reliability data interactively, helping engineers see not only the numbers but also the logic behind them.
This journey is only beginning. AI is rapidly transforming reliability prediction, safety analysis, and system engineering. By combining rigorous data science with decades of safety expertise, ANZEN and Tecnalia are laying the foundation for smarter, safer systems.
Stay tuned for more updates as we continue advancing reliability prediction. We’re committed to sharing insights, testing ideas, and shaping the future of intelligent reliability tools.
Want to learn more? Contact the ANZEN team to explore how our systems engineering solutions can support your next project.
About the author
Daniel Villafañe is an aerospace engineer with expertise in avionics, systems engineering and model-based design and analysis.
At Anzen, Daniel’s work is focused on ATICA, our model-based tool for safety analysis. Daniel is in charge of building system models and applying systems engineering processes while using ATICA to improve results on safety and reliability analyses for aerospace avionics projects.
