Scholarlink

<p>The frequency of vibration signals generated by piston pumps under the operating environment of variable speed shows non-stationary and nonlinear characteristics, leading to poor focusing of spectral energy in traditional time–frequency analysis, and the interference of variable speed increases the difficulty of the signal noise reduction and feature extraction process, which in turn enhances the diagnostic complexity of vibration signal processing and health diagnosis. In order to address this problem, this paper proposes a research method based on synchro-extracting transform (SET) and snake optimization (SO) extreme learning machines (ELM) to identify the health status of variable-speed piston pumps. Firstly, the effectiveness of the SET method in processing nonlinear signals and anti-noise is verified by single-signal and multi-signal simulation; then, a test platform for plunger pump health diagnosis is built, and the raw data with different wear levels of the distributor disk are collected, and the data are processed using SET and angular-domain sampling techniques to eliminate the influence of speed changes; and eight angular-domain and order-domain features with significant differentiation are extracted as input features; finally, a method is proposed to identify the health state of variable-speed plunger pumps. as input features; and finally, an SO algorithm is proposed to optimize ELM for health diagnosis, which solves the random assignment problem of ELM by optimizing the initial weights and bias parameters of ELM through the SO algorithm and significantly improves the classification accuracy and computational efficiency and also compares with PSO-ELM and ELM. The results show that SO-ELM can effectively solve the random assignment problem of initial weights and thresholds that exists in ELM and accurately extract fault information; SO-ELM model is better than PSO-ELM and ELM in terms of comprehensive performance such as precision rate, recall rate, and F1-score, and the average diagnostic correctness of SO-ELM model is 96.1%, which is higher than the average diagnostic correctness. The average correct diagnosis rate of SO-ELM model is 96.1%, which is 4.0% and 2.6% higher than that of ELM (92.1%) and PSO-ELM (93.5%), respectively, which can effectively diagnose the health state of the variable-speed axial piston pump, and verifies the effectiveness of the method.</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

A simultaneous extraction transform and snake optimized extreme learning machine for variable speed piston pump health diagnostics

  • 期刊论文

  • 发表日期：

    2025年7月7日

查看全文
  • Rui Guo,
  • Jiaqi Zhang,
  • Yapeng Wen,
  • Dianmin Chen,
  • Xin Li,
  • Hao Zhang,
  • Honglong Liu,
  • Wei Cai,
  • Xiongshuai Luo

摘要

The frequency of vibration signals generated by piston pumps under the operating environment of variable speed shows non-stationary and nonlinear characteristics, leading to poor focusing of spectral energy in traditional time–frequency analysis, and the interference of variable speed increases the difficulty of the signal noise reduction and feature extraction process, which in turn enhances the diagnostic complexity of vibration signal processing and health diagnosis. In order to address this problem, this paper proposes a research method based on synchro-extracting transform (SET) and snake optimization (SO) extreme learning machines (ELM) to identify the health status of variable-speed piston pumps. Firstly, the effectiveness of the SET method in processing nonlinear signals and anti-noise is verified by single-signal and multi-signal simulation; then, a test platform for plunger pump health diagnosis is built, and the raw data with different wear levels of the distributor disk are collected, and the data are processed using SET and angular-domain sampling techniques to eliminate the influence of speed changes; and eight angular-domain and order-domain features with significant differentiation are extracted as input features; finally, a method is proposed to identify the health state of variable-speed plunger pumps. as input features; and finally, an SO algorithm is proposed to optimize ELM for health diagnosis, which solves the random assignment problem of ELM by optimizing the initial weights and bias parameters of ELM through the SO algorithm and significantly improves the classification accuracy and computational efficiency and also compares with PSO-ELM and ELM. The results show that SO-ELM can effectively solve the random assignment problem of initial weights and thresholds that exists in ELM and accurately extract fault information; SO-ELM model is better than PSO-ELM and ELM in terms of comprehensive performance such as precision rate, recall rate, and F1-score, and the average diagnostic correctness of SO-ELM model is 96.1%, which is higher than the average diagnostic correctness. The average correct diagnosis rate of SO-ELM model is 96.1%, which is 4.0% and 2.6% higher than that of ELM (92.1%) and PSO-ELM (93.5%), respectively, which can effectively diagnose the health state of the variable-speed axial piston pump, and verifies the effectiveness of the method.