Achieving Consistency of Flexible Surface Acoustic Wave Sensors with Artificial Intelligence

Flexible surface acoustic wave technology has garnered significant attention for wearable electronics and sensing applications. However, the mechanical strains induced by random deformation of these flexible SAWs during sensing often significantly alter the specific sensing signals, causing critical issues such as inconsistency of the sensing results on a curved/flexible surface. To address this challenge, we first developed high-performance AlScN piezoelectric film-based flexible SAW sensors, investigated their response characteristics both theoretically and experimentally under various bending strains and UV illumination conditions, and achieved a high UV sensitivity of 1.71 KHz/(mW/cm²). To ensure reliable and consistent UV detection and eliminate the interference of bending strain on SAW sensors, we proposed using key features within the response signals of a single flexible SAW device to establish a regression model based on machine learning algorithms for precise UV detection under dynamic strain disturbances, successfully decoupling the interference of bending strain from target UV detection. The results indicate that under strain interferences from 0 to 1160 με the model based on the extreme gradient boosting algorithm exhibits optimal UV prediction performance. As a demonstration for practical applications, flexible SAW sensors were adhered to four different locations on spacecraft model surfaces, including flat and three curved surfaces with radii of curvature of 14.5, 11.5, and 5.8 cm. These flexible SAW sensors demonstrated high reliability and consistency in terms of UV sensing performance under random bending conditions, with results consistent with those on a flat surface.