| 摘 要: 针对包装机故障信号受噪声影响且样本稀少导致传统的诊断方法不能满足实际场景应用要求的问题,提出一种新的轴承故障诊断方法。首先,利用连续小波变换(Continue Wavelet Transform,CWT)将振动信号转换为二维图像。其次,将其输入深度网络模型进行训练。再次,利用极限学习机(Extreme Learning Machine,ELM)进行故障分类。最后,通过麻雀搜索算法(Sparrow Search Algorithm,SSA)对ELM进行优化。试验结果显示,在强噪声干扰且少样本训练的情况下,所提方法的准确率仍能够达到98.91%,并且模型在不同的轴承数据集中的准确率均达到98.92%,证明所提方法具有一定的实用价值。 |
| 关键词: 故障诊断;深度学习;特征提取;极限学习机 |
|
中图分类号: TP307
文献标识码: A
|
| 基金项目: 浙江省科技计划项目(2022C01065) |
|
| Fault Diagnosis of Packaging Machine Bearings Based on Deep Learning and Enhanced Extreme Learning Machine |
|
RU Xin1, MENG Jinxin1, LI Jianqiang2, PENG Laihu1,2
|
(1.School o f Mechanical Engineering, Zhejiang Sci-Tech University, Hangzhou 310018, China;
2.Research Institute of Zhejiang Sci-Tech University in Longgang, Wenzhou 325000, China)
ruxin@zstu.edu.cn; 934412020@qq.com; wzcnljq@126.com; laihup@zstu.edu.cn
|
| Abstract: Aiming at the problem of packaging machine fault signal affected by noise and sparse samples, making traditional diagnostic methods unsuitable for practical applications, a new bearing fault diagnosis method is proposed. Firstly, the vibration signals are transformed into two-dimensional images by using Continuous Wavelet Transform (CWT). Subsequently, these images are input into a deep neural network for training. Following that, Extreme Learning Machine (ELM) is used for fault classification. Finally, the ELM is optimized by the Sparrow Search Algorithm (SSA). Experimental results demonstrate that the proposed method achieves an accuracy of 98.91% even in scenarios with strong noise interference and limited training samples. Additionally, the model achieves an accuracy of 98.92% across various bearing datasets, proving the practical value of the proposed method. |
| Keywords: fault diagnosis; deep learning; feature extraction; extreme learning machine |
