Scholarlink

<p>The Reaction Wheel (RW) actuator provides three-axis attitude control and ensures the pointing accuracy of spacecraft. It consists of a Brushless DC motor (BLDC), with the rotor connected to a flywheel through precision ball bearings, using different lubricants to meet the spacecraft’s mission life requirements. The slow acceleration and deceleration of the reaction wheel, driven by motor torque and high moment of inertia, can be impacted by non-linear bearing friction, potentially compromising the spacecraft’s attitude control accuracy and stability during docking. Typically, friction torque and health assessment of RW bearings are evaluated through factors like motor current and bearing temperature. Despite existing literature on friction torque prediction for statically loaded radial journal bearings that considers temperature, load, and rotational speed, there remains a significant gap in the systematic analysis of friction torque and health assessment for precision ball bearings. Specifically, the influence of lubricant distribution at varying speeds has not been thoroughly investigated, highlighting the need for further research in this area. This paper presents an approach for measuring friction torque through free deceleration by employing machine learning (ML) techniques—namely Regression Tree (RT), Artificial Neural Network (ANN), and Ensemble Boosted Tree (EBT)—to predict friction torque in precision ball bearings used in reaction wheels. The performance of the ML models has been compared using computational complexity analysis and an ablation study of the ANN. Additionally, the research investigates friction torque under various conditions, using different lubricants—namely Kluber and Nye Torr—across a range of temperatures and rotational speeds. The study also delves into the prediction of run-down time and lubricant type, employing ANN techniques. By utilizing machine learning models, this approach allows for swift data processing, the detection of patterns, anomaly identification, and the assessment of complex interrelationships.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Friction torque prediction of precision ball bearing unit for reaction wheel actuators for spacecraft applications

  • 期刊论文

  • 发表日期：

    2025年2月1日

查看全文
  • P. Muthuganapathy,
  • Sanjay K. Chaturvedi,
  • Heeralal Gargama,
  • P. Sasikumar

摘要

The Reaction Wheel (RW) actuator provides three-axis attitude control and ensures the pointing accuracy of spacecraft. It consists of a Brushless DC motor (BLDC), with the rotor connected to a flywheel through precision ball bearings, using different lubricants to meet the spacecraft’s mission life requirements. The slow acceleration and deceleration of the reaction wheel, driven by motor torque and high moment of inertia, can be impacted by non-linear bearing friction, potentially compromising the spacecraft’s attitude control accuracy and stability during docking. Typically, friction torque and health assessment of RW bearings are evaluated through factors like motor current and bearing temperature. Despite existing literature on friction torque prediction for statically loaded radial journal bearings that considers temperature, load, and rotational speed, there remains a significant gap in the systematic analysis of friction torque and health assessment for precision ball bearings. Specifically, the influence of lubricant distribution at varying speeds has not been thoroughly investigated, highlighting the need for further research in this area. This paper presents an approach for measuring friction torque through free deceleration by employing machine learning (ML) techniques—namely Regression Tree (RT), Artificial Neural Network (ANN), and Ensemble Boosted Tree (EBT)—to predict friction torque in precision ball bearings used in reaction wheels. The performance of the ML models has been compared using computational complexity analysis and an ablation study of the ANN. Additionally, the research investigates friction torque under various conditions, using different lubricants—namely Kluber and Nye Torr—across a range of temperatures and rotational speeds. The study also delves into the prediction of run-down time and lubricant type, employing ANN techniques. By utilizing machine learning models, this approach allows for swift data processing, the detection of patterns, anomaly identification, and the assessment of complex interrelationships.