Reproduce “Reliable Piezoresistive Strain Sensing Through Physical Limits and Uncertainty Monitoring” on OpenResearch

About this paper

Carmen Ballester, Víctor Muñoz, Dorin Copaci, Dolores Blanco

A new framework for piezoresistive soft strain sensors enables self-monitoring by combining physics-informed modeling with probabilistic uncertainty and physical operating limits. It detects 95% of out-of-range conditions and 100% of abnormal events by reporting a reliability state alongside strain estimates, eliminating the need for labeled fault data.

Key insights

• Integrating physics-guided features, such as relative resistance rate and a Standard Linear Solid memory state, significantly improves the predictive accuracy and uncertainty calibration of the inverse model.
• A multi-component risk factor, which fuses epistemic uncertainty with the probability of violating physical operating limits, provides a more robust and comprehensive reliability assessment than either component alone.
• Reliability monitoring for soft sensors can be effectively achieved without requiring labeled fault data, by defining physical limits and leveraging uncertainty-aware modeling from nominal data.

Summary by alphaXiv.