From Data to Decisions: Using Digital Twins for Smarter Maintenance

How do you turn raw sensor data into reliable maintenance decisions when systems operate under constantly changing conditions?

In this talk, Priyanka Schnell will share a real-world case study from the aerospace industry, showing how data and engineering models can be used to move from traditional, schedule-based maintenance to smarter, condition-based decisions.

The example comes from aircraft wing manufacturing, where robotic drills operate under extremely strict quality requirements. "Replacing tools too early wastes resources, while replacing them too late can cause costly damage." Priyanka will explain how combining sensor data (such as power, vibration, and torque) with physical knowledge of the process makes it possible to detect early warning signs of failure.

Using a concept called a digital twin—a live data model that mirrors the real process—the team was able to identify problems hours before failure occurred. This approach improved efficiency, reduced waste, and prevented expensive damage.

To make the ideas concrete, parts of the session will include live exploration of data and models in Google Colab, focusing on intuition, workflow, and decision-making rather than heavy mathematics. The talk is designed to be accessible and will emphasize practical lessons, interpretability, and why combining data science with domain knowledge is key when working in production systems—no prior knowledge of manufacturing or physics required.