Long-Term Guided Wave Dataset Under Real-World Conditions Dataset
Summary
What happens when we take structural health monitoring (SHM) out of the lab and into the real world?
The Dataset on Guided Waves from Long-Term Structural Health Monitoring under Uncontrolled and Dynamic Conditions, created by Kang Yang and the SmartDATA Lab, offers one of the first comprehensive public resources for doing just that. This large-scale, real-world dataset captures guided ultrasonic waves over two years of continuous monitoring, encompassing millions of waveforms, changing environmental conditions, and true operational variability.
It was introduced in the open-access journal Scientific Data by Yang et al. (2025):
📄 Yang, K., et al. (2025). Dataset on guided waves from long-term structural health monitoring under uncontrolled and dynamic conditions.
Nature Scientific Data, https://doi.org/10.1038/s41597-025-05300-5
📦 What’s in the Dataset?
This dataset includes:
- 📈 Over 7.6 million ultrasonic guided wave signals, collected automatically
- 📅 4.5 years of measurements, recorded every 30 seconds
- 🌡️ Paired temperature and humidity data, enabling environmental correlation studies
- 🧭 Multiple distinct propagation paths across an aluminum panel instrumented with 14 piezoelectric sensors
- 📍 Real-world uncertainties: signal drift, sensor degradation, seasonal temperature swings, and unmonitored activity
Data is stored in efficient HDF5 format, and is accompanied by:
- ✅ Python code for data loading
- ✅ Scripts for basic preprocessing and exploratory analysis
🔍 Why This Dataset Matters
Most SHM and AI-guided wave studies use simulated or lab-controlled data. These are valuable—but they don’t reflect the uncertainty, noise, and environmental variation found in real infrastructure.
This dataset helps researchers:
- 🧠 Train and evaluate AI algorithms under real, imperfect conditions
- 🔄 Develop unsupervised learning, drift compensation, and anomaly detection techniques
- 🌦️ Understand the influence of temperature, humidity, and long-term aging on signal quality
- 🧪 Benchmark signal processing, sensor fusion, and domain adaptation methods in SHM
With this data, AI and SHM researchers can finally move beyond idealized scenarios and begin tackling the complexities of real-world monitoring.
💡 Ideal For:
- Machine learning & signal processing research in SHM
- Benchmarking algorithms under environmental drift
- Simulating long-term sensor networks
- Educating students with real sensor data
- Developing temperature-robust damage detection algorithms
🚀 Get Started
- 📂 Access the dataset:
▶ View on Figshare - 💻 Load and explore the data:
▶ GitHub code repository
📚 Citation
If you use this dataset in your research, please cite:
Yang, K., et al. (2025). Dataset on guided waves from long-term structural health monitoring under uncontrolled and dynamic conditions. Scientific Data. https://doi.org/10.1038/s41597-025-05300-5
🌐 In Summary
This is not just another SHM dataset—it’s a step toward real-world AI in structural monitoring. Whether you’re building smarter algorithms, modeling sensor behavior over time, or exploring environmental robustness, this dataset provides a rich, messy, and meaningful foundation for innovation.
Long-term. Real-world. Open-access. This is SHM data for the future.
Lamb Wave Long-Term Machine Learning Real Data Structural Health Monitoring Variations