Scholarlink

<p>As a critical load-bearing component of mechanical parking system, track was one of the primary factors resulting in failure since it served under long-term working cases of high-frequency and variable loads. Based on Timoshenko beam model with double-layer elastic discrete support of the sleeper, dynamic parameters were analyzed during several running cycles especially referred to the influence of track unevenness. According to this theoretical analysis, online monitoring method by vibration sensors was applied to obtain vibration characteristic samples of tracks under different obstacles in thickness. Based on feature extraction in time and frequency domain, a PHM model was proposed to predict three types of working states. Wavelet package transform was adopted firstly to generate a simulation dataset, and then BiLSTM-CNN algorithm was integrated into train and learn procedure to predict normal and fault states. Compared with CNN, LSTM, BiLSTM, and BiLSTM-Attention, it showed that the proposed model was more precise which approached 98.8%, and supplied a theoretical basis to realize intelligent operation and maintenance of mechanical parking system.</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

Prognostics Health Management of Mechanical Parking Systems Based on Vibration Dynamics of Wheel–Track

  • 期刊论文

  • 发表日期：

    2025年8月7日

查看全文
  • Bin Xu,
  • Teng Wang,
  • Weiping Ouyang,
  • Changyu Li,
  • Jianyong Zheng,
  • Zengjiang Chen

摘要

As a critical load-bearing component of mechanical parking system, track was one of the primary factors resulting in failure since it served under long-term working cases of high-frequency and variable loads. Based on Timoshenko beam model with double-layer elastic discrete support of the sleeper, dynamic parameters were analyzed during several running cycles especially referred to the influence of track unevenness. According to this theoretical analysis, online monitoring method by vibration sensors was applied to obtain vibration characteristic samples of tracks under different obstacles in thickness. Based on feature extraction in time and frequency domain, a PHM model was proposed to predict three types of working states. Wavelet package transform was adopted firstly to generate a simulation dataset, and then BiLSTM-CNN algorithm was integrated into train and learn procedure to predict normal and fault states. Compared with CNN, LSTM, BiLSTM, and BiLSTM-Attention, it showed that the proposed model was more precise which approached 98.8%, and supplied a theoretical basis to realize intelligent operation and maintenance of mechanical parking system.