摘要详情

ID / 提交时间

279 / 2025-06-15 23:16:14

标题

Sliding-Based Surface Material Recognition Using Flexible Magnetic Cilia Sensor for Intelligent Maintenance

关键字

Bio-inspired Sensor, Tactile Texture Recognition, Flexible Magnetic Sensing, Material Classification, Real-Time Tactile Perception

主题及专题

5、Sensors Technology and Internet of Things Technology for EOM

状态

全文待审

作者

Yangqianhui Zhang / Zhejiang University；School of Mechanical Engineering

Huayong Yang / Zhejiang University；School of Mechanical Engineering

Dong Han / Zhejiang University；School of Mechanical Engineering

Qingyuan Zhang / Beihang University；Hangzhou International Innovation Institute

摘要

Intelligent maintenance of equipment requires timely detection of surface condition changes, even when visual inspection is impossible. This paper presents a flexible, bioinspired magnetic tactile sensor system for non-visual surface condition monitoring. The sensor consists of vertically magnetized 3D micro-cilia arrays embedded in an elastomer and underlying Hall effect magnetic sensors. As the cilia slide over a surface, microscopic texture features bend the cilia, producing distinctive magnetic field fluctuations. The low-power sensor is highly flexible and durable, capable of enduring repeated contact. A sliding scan experiment was performed on eight different material surfaces, with multiple trials simulating human-like tactile exploration. The resulting magnetic field time-series were analyzed using a machine learning pipeline. After preprocessing and feature extraction, a supervised classification model was trained to recognize each material. The classifier achieved over 99% accuracy in identifying the surface textures, with a nearly diagonal confusion matrix and per-class ROC AUC ≈ 1.0. We further applied explainability techniques: feature importance analysis and SHAP values revealed which signal features differentiate textures. These results demonstrate that the magnetic cilia-based sensor can reliably “feel” and identify surface textures, enabling non-visual condition monitoring. The compact form factor and high sensitivity of this tactile sensor make it suitable for integration into manufacturing lines, robotic end-effectors, and wearable diagnostic devices for real-time surface monitoring. This study highlights the potential of bioinspired magnetic tactile sensing for intelligent maintenance applications.