Poster: Deep Learning-Assisted Fingerprint-Inspired Flexible Pressure Sensor for Tension Monitoring in Carbon Fiber Production
Date:
We present a novel fingerprint-inspired flexible pressure sensor for real-time tension monitoring in wide carbon fiber bundle arrays. Carbon fiber production relies on continuous pre-oxidation lines, where thermal contraction induces tension fluctuations that can compromise fiber quality. Conventional tension sensors are bulky and difficult to integrate, making multi-bundle monitoring challenging.
Key features of our approach include:
- A high-performance piezoresistive sensing layer inspired by fingerprint ridge depth and spacing, optimized for sensitivity and linearity across a wide pressure range.
- End-to-end modeling enabling high-accuracy detection of anomalous tension events during pre-oxidation.
- Reliable, real-time monitoring suitable for high-temperature, continuous, and complex industrial environments.
Future work will integrate advanced deep learning models, such as Transformers, combined with transfer learning and anomaly detection algorithms, to address challenges of nonlinearity, temporal variation, and data imbalance, enhancing adaptability and recognition capability under limited samples and rare anomalies.
This system provides a scalable solution for industrial carbon fiber production while expanding the application potential of flexible sensors in advanced manufacturing processes.