Scholarlink

<p>To address the existing challenges in analyzing vibration performance, signal processing, and diagnosing faults in micro-textured rolling bearings, this paper introduces a novel fault diagnosis technique specifically designed for micro-dimple micro-textured rolling bearings. Utilizing the surface texturing method, dimples with a diameter of 200&#xa0;μm, a depth of 10&#xa0;μm, and an area density of 24% were engineered on the outer raceway of a 30,206 tapered roller bearing. A three-stage diagnostic framework was established, incorporating bio-inspired optimization, hybrid denoising, and intelligent recognition. Initially, the Catch Fish Optimization Algorithm (CFOA) was employed to enhance the decomposition efficiency of Feature Mode Decomposition (FMD). Subsequently, a hybrid denoising method integrating mutual information and an improved wavelet thresholding method (MI-IWT) is employed to obtain denoised vibration signals. Subsequently, these denoised signals are transformed into two-dimensional images via the Synchronized Compression the Synchrosqueezed Wavelet Transform (SWT) and then fed into a Multi-Column Convolutional Neural Network (MCNN). The MCNN excels at identifying and extracting significant local features across various regions and scales within the image. Following this, a Bidirectional Gated Recurrent Unit (BiGRU) amalgamates these features derived from different temporal scales. Ultimately, fault classification is performed using the Softmax function. The proposed method was validated using a bearing test rig at Shenyang University of Chemical Technology and the Case Western Reserve University bearing dataset, which included both micro-textured and conventional rolling bearings. The method achieved fault identification accuracies of 98.8% and 98.9%, respectively. These findings indicate strong robustness and practical applicability, offering an effective solution for signal denoising and fault diagnosis in micro-textured rolling bearings and other rotating machinery.</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 new method for micro-textured rolling bearing fault diagnosis based on CFOA-FMD hybrid denoising combined with SWT-MCNN-BiGRU

  • 期刊论文

  • 发表日期：

    2025年9月5日

查看全文
  • Risheng Long,
  • Zusheng Shang

摘要

To address the existing challenges in analyzing vibration performance, signal processing, and diagnosing faults in micro-textured rolling bearings, this paper introduces a novel fault diagnosis technique specifically designed for micro-dimple micro-textured rolling bearings. Utilizing the surface texturing method, dimples with a diameter of 200 μm, a depth of 10 μm, and an area density of 24% were engineered on the outer raceway of a 30,206 tapered roller bearing. A three-stage diagnostic framework was established, incorporating bio-inspired optimization, hybrid denoising, and intelligent recognition. Initially, the Catch Fish Optimization Algorithm (CFOA) was employed to enhance the decomposition efficiency of Feature Mode Decomposition (FMD). Subsequently, a hybrid denoising method integrating mutual information and an improved wavelet thresholding method (MI-IWT) is employed to obtain denoised vibration signals. Subsequently, these denoised signals are transformed into two-dimensional images via the Synchronized Compression the Synchrosqueezed Wavelet Transform (SWT) and then fed into a Multi-Column Convolutional Neural Network (MCNN). The MCNN excels at identifying and extracting significant local features across various regions and scales within the image. Following this, a Bidirectional Gated Recurrent Unit (BiGRU) amalgamates these features derived from different temporal scales. Ultimately, fault classification is performed using the Softmax function. The proposed method was validated using a bearing test rig at Shenyang University of Chemical Technology and the Case Western Reserve University bearing dataset, which included both micro-textured and conventional rolling bearings. The method achieved fault identification accuracies of 98.8% and 98.9%, respectively. These findings indicate strong robustness and practical applicability, offering an effective solution for signal denoising and fault diagnosis in micro-textured rolling bearings and other rotating machinery.