High entropy alloy reinforced composite strain sensor with long-cycle durability for machine learning assisted human motion recognition

Flexible piezoresistive sensors have attracted considerable attention in wearable healthcare, motion monitoring, and human-machine interaction. However, their relatively slow response/recovery behavior and limited cyclic reliability still restrict their practical applications. In this work, a flexible piezoresistive strain sensor (PHM-FPSS) based on a polydimethylsiloxane (PDMS) composite incorporating high-entropy alloy (HEA) powders and multi-walled carbon nanotubes (MWCNTs) was developed. The introduction of HEA fillers improved interfacial bonding and mechanical reinforcement within the PDMS matrix, providing additional conductive and structural pathways. In synergy with MWCNTs, these effects led to the formation of a continuous and stable three-dimensional conductive network, ensuring efficient electron transport and long-term durability. Thus, the PHM-FPSS exhibited a rapid response time (88 ms), short recovery time (115 ms), moderate electrical resistivity (0.851 Ω·m), and excellent durability over 11,000 loading-unloading cycles with negligible degradation. Furthermore, a multilayer perceptron (MLP) machine learning model was implemented to classify human motion signals acquired from the sensor, enabling accurate recognition of joint bending angles including the fingers, wrist, elbow, and knee. This study provides an effective materials design approach for achieving balanced sensitivity and stability in flexible piezoresistive composites and establishes an integrated sensing-intelligence platform for potential applications in sports science, rehabilitation monitoring, and human-machine interfaces.