From Data to Decisions: Digital Twins for Smarter Maintenance
R-Ladies Rome Tutorials
Registered Attendees (65)
On February 16, 2026, R-Ladies Rome hosted From Data to Decisions: Digital Twins for Smarter Maintenance, a practical session focused on how data science moves beyond notebooks and into real production systems.
The talk presented a physics-informed approach to Condition-Based Maintenance (CBM) through a real-world aerospace case study. The central question of the session was clear:
How do we detect failure before it happens — and make decisions that reduce cost and risk?
🧩 Tutorial Overview
The session was delivered by Priyanka Schnell, Chief Technology Officer at Insurfox. With experience in enterprise-scale digital systems and aerospace manufacturing, Priyanka guided us through how sensor data, physics models, and anomaly detection can work together in production.
The context: robotic drilling in A321 wing assembly. A single snapped drill bit can cause burrs or delamination, potentially scrapping a wing skin valued at €150,000. Traditional maintenance policies replaced drill bits after a fixed number of holes — a strategy that often led to premature replacements or unexpected failures.
Instead of relying on static thresholds, the session introduced a digital twin approach.
We explored:
- how high-frequency sensor data (torque, vibration, speed, feed rate) can be transformed into meaningful features
- why Specific Cutting Energy helps detect “wasted energy” before functional failure
- how regime normalization allows comparison across aluminum and titanium cutting
- how Isolation Forest can identify deviations without labeled failure data
- why persistence rules (e.g., “6 of 12”) reduce alert noise
- how early detection within the P–F interval provides actionable lead time
The session included a live walkthrough in Google Colab using simulated data to demonstrate the workflow step-by-step.
🎥 Recording
🎬 Watch the Recording
The recording is suitable both for participants who attended live and for anyone interested in understanding how anomaly detection and digital twins operate in high-stakes environments.
🧠 What You’ll Learn
By watching the recording, you will learn how to:
- understand what a digital twin means in practical data terms
- engineer physics-informed features from raw sensor signals
- apply unsupervised anomaly detection in production settings
- use persistence logic to stabilize alert systems
- think beyond time-based rules toward dynamic maintenance strategies
- connect domain knowledge with statistical modeling
📦 Resources & Materials
🔊 About the Tutorial
This session is part of the R-Ladies Rome Tutorials series, which focuses on practical workflows and real-world data applications.
Our goal is to show how data science connects to operational systems — where decisions have financial and technical consequences. We aim to make advanced topics accessible without removing the rigor behind them.
Keep learning and exploring, subscribe to our YouTube channel, and revisit past events on rladiesrome.org!