Triboelectric linear bearing sensor for self-powered condition monitoring using wavelet transform and lightweight CNN

The working mode of a linear bearing is reciprocating linear motion. Therefore, it is difficult to form a lubricating oil film while most of the surface of the shaft is unprotected, resulting in serious wear and a high failure rate. In this study, a triboelectric linear bearing sensor (TLBS) is developed and integrated into the linear bearing's noload outer circulation track area while maintaining its structural and functional integrity. First, the output characteristics of the TLBS and variation characteristics with external load resistance are measured, and the influences of the structure design on its output characteristics are discussed. Subsequently, the effects of load carrying, ambient humidity, and lubrication on the output characteristics under its actual operating conditions are investigated. Finally, the sensing characteristics experiments show that TLBS can detect the displacement and velocity of linear bearings (25-125 mm/s) while the error rate of velocity detection is less than 0.5 %. The continuous wavelet transforms the time-frequency diagram and lightweight convolutional neural network fault diagnosis method is used to achieve online and fast fault monitoring of bearing with classification accuracy of up to 100 %. This work presents a novel methodology for monitoring linear bearing operating condition, which has important practical implications.

Automated summary

This study develops a triboelectric linear bearing sensor (TLBS) and integrates it into the linear bearing's noload outer circulation track area while maintaining its structural and functional integrity. The output characteristics of the TLBS and its variation with external load resistance are measured, and the influences of the structure design on its output characteristics are discussed. The sensing characteristics experiments show that TLBS can detect the displacement and velocity of linear bearings while achieving high classification accuracy in fault monitoring.