Scholarlink

<p>Ultrasonic motors (USMs) are widely used in precision transmission for power systems, but their friction material wears down due to coupled working conditions. Most existing technologies focus on single functions, suffering from slow dynamic response, weak anti-interference capability, and a lack of closed-loop control, which limits their adaptability to complex operational requirements. This study constructs a coupling system that integrates a spiking neural network (SNN) with a biologically grounded compensation strategy. The SNN employs leaky integrate-and-fire (LIF) neurons and surrogate gradient learning for temporal spike encoding. The compensation strategy maps muscle damping dynamics—specifically the length-tension relationship and calcium-mediated activation—to a state-dependent feedback law. Through a working condition adaptation module, the parameter calibration logic is optimized. Subsequently, a hierarchical control trigger and smooth correction model are designed to achieve closed-loop control across the entire “perception-recognition-control” process of wear state. The innovations include the construction of a comprehensive quantification factor for wear characteristics and the optimization of the synergistic mechanism between the SNN dynamic learning algorithm and the bionic compensation, thereby overcoming the bottleneck of single-function traditional technologies. Simulation results show that the comprehensive performance index of this technology is 0.91 ± 0.01, which is 16.7%–40.0% higher than those of support vector regression (SVR) and LSTM-SPH. The pre-judgment error is controlled at 2.1% ± 0.3%, and the regulation delay is shortened to 0.4 ± 0.1&#xa0;s. Furthermore, the performance attenuation rate is only 6.3% under 20% disturbance, with both robustness and dynamic adaptability indexes exceeding 0.9. Clearly, this technology provides core support for the long-term stable operation of USMs in power systems, reducing unplanned downtime losses and promoting intelligent upgrades in power equipment operation and maintenance.</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

Wear state of friction material in ultrasonic motor based on dynamic learning of spiking neural network and state-dependent proportional compensation

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月1日

查看全文
  • Honglin Wu,
  • Zhe Wang,
  • Junyan Wang

摘要

Ultrasonic motors (USMs) are widely used in precision transmission for power systems, but their friction material wears down due to coupled working conditions. Most existing technologies focus on single functions, suffering from slow dynamic response, weak anti-interference capability, and a lack of closed-loop control, which limits their adaptability to complex operational requirements. This study constructs a coupling system that integrates a spiking neural network (SNN) with a biologically grounded compensation strategy. The SNN employs leaky integrate-and-fire (LIF) neurons and surrogate gradient learning for temporal spike encoding. The compensation strategy maps muscle damping dynamics—specifically the length-tension relationship and calcium-mediated activation—to a state-dependent feedback law. Through a working condition adaptation module, the parameter calibration logic is optimized. Subsequently, a hierarchical control trigger and smooth correction model are designed to achieve closed-loop control across the entire “perception-recognition-control” process of wear state. The innovations include the construction of a comprehensive quantification factor for wear characteristics and the optimization of the synergistic mechanism between the SNN dynamic learning algorithm and the bionic compensation, thereby overcoming the bottleneck of single-function traditional technologies. Simulation results show that the comprehensive performance index of this technology is 0.91 ± 0.01, which is 16.7%–40.0% higher than those of support vector regression (SVR) and LSTM-SPH. The pre-judgment error is controlled at 2.1% ± 0.3%, and the regulation delay is shortened to 0.4 ± 0.1 s. Furthermore, the performance attenuation rate is only 6.3% under 20% disturbance, with both robustness and dynamic adaptability indexes exceeding 0.9. Clearly, this technology provides core support for the long-term stable operation of USMs in power systems, reducing unplanned downtime losses and promoting intelligent upgrades in power equipment operation and maintenance.