Theoretical and experimental investigation of the differential planetary roller screw mechanism condition monitoring and public dataset

Abstract

Abstract This study introduces an innovative condition monitoring test rig for the differential planetary roller screws (DPRS), focusing on their performance in extreme conditions. In practical applications such as aerospace and humanoid robotics, DPRS operate under complex conditions involving varying temperatures and multi-directional loads. Accurately monitoring the operational state under these challenging environments is difficult, which can hinder the assessment of the system’s stability and reliability. This model was developed to address these challenges, it underscores the importance of anti-jamming and real-time monitoring for the DPRS reliability. The research presents a dynamic friction model using the Lagrange method, enhancing understanding of DPRS operations. Advanced signal processing techniques, including discrete wavelet transforms and a convolutional neural network, are implemented for effective feature extraction. A DWTC-BiGRU network is utilized to capture temporal dependencies, vital for monitoring DPRS under varying conditions. Experimental validation is conducted under thermal stress and load variations, demonstrating the system’s durability and reliability. The study compares its method with existing algorithms, showing superior accuracy and robustness by combining mechanical modeling with computational techniques for real-time industrial monitoring. The dataset is publicly available at GitHub—haomjc/HealthMoni.