Research on PSO-SSTCA Algorithm of Rolling Bearings Under Different Loads
-
摘要: 针对不同负载下滚动轴承故障诊断准确率不高和样本稀缺的问题，本文提出了一种基于粒子群优化的半监督迁移学习（PSO-SSTCA）算法。在迁移学习算法的基础上，引入希尔伯特-施密特独立性系数（HSIC）增强迁移学习过程中不同数据标签的依赖性，加入粒子群优化算法自适应寻找多核函数的最优系数，缩小数据集的类内间距，并利用K-近邻算法进行不同负载间滚动轴承的故障诊断。对4种不同负载工况下的滚动轴承振动信号进行分析，结果表明：在单-单、多-单负载工况下，PSO-SSTCA算法的平均准确率分别为85.92%与88%，与重构信号相比分别提高了10.75%与19.42%。该方法有效地为机械设备的状态监测与故障诊断提供了技术支撑。Abstract: Aiming at the problems of low accuracy of fault diagnosis of rolling bearing under different loads and scarcity of samples, this paper proposes a semi-supervised transfer learning (PSO-SSTCA) algorithm based on particle swarm optimization. On the basis of the transfer learning algorithm, the Hilbert-Schmidt independence coefficient (HSIC) is introduced to enhance the dependence of different data labels in the transfer learning process, and the particle swarm optimization algorithm is added to adaptively find the optimal coefficients of the multi-core function. The intra-class spacing of the data set is reduced, and the K-nearest neighbour algorithm is used for fault diagnosis of rolling bearings between different loads. The analysis of rolling bearing vibration signals under four different load conditions shows that the average accuracy of the PSO-SSTCA algorithm is 85.92% and 88% respectively under single-single and multiple-single load conditions, which are lower than the original weight. Compared with the reconstructed signal, it increased by 10.75% and 19.42% respectively. This method effectively provides technical support for the condition monitoring and fault diagnosis of mechanical equipment.
-
Key words:
- rolling bearing /
- particle swarm algorithm /
- transfer learning /
- feature extraction /
- fault diagnosis
-
图 3 D、C数据集PCA映射后的四维特征
Figure 3. Four-dimensional features after PCA mapping for D and C datasets
图 4 D、C数据集迁移变换后的四维特征
Figure 4. Four-dimensional features of D and C datasets after migration transformation
表 1 待测故障滚动轴承参数
Table 1. Parameters of faulty rolling bearings to be tested
型号 内径/mm 外径/mm 节径/mm 滚动体个数 测量位置 宽度/mm 采样频率/kHz 6205-2RS JEM SKF 25 52 52 9 驱动端 15 12 
下载: 导出CSV
表 2 滚动轴承重构信号时域计算结果
Table 2. Rolling bearing reconstruction signals' time domain calculation results
类别 时频域特征 IMF0 IMF1 IMF2 IMF3 无故障 奇异熵 0.37 0.27 0.19 0.16 排列熵 0.76 0.55 0.40 0.31 内圈故障 奇异熵 0.72 0.16 0.07 0.05 排列熵 0.85 0.75 0.64 0.54 外圈故障 奇异熵 0.55 0.18 0.16 0.11 排列熵 0.66 0.66 0.56 0.44
下载: 导出CSV
表 3 滚动轴承重构信号频域计算结果
Table 3. Rolling bearing reconstruction signal frequency domain calculation results
类别 重心频率 均方频率 均方根频率 频率方差 无故障 254.04 966520.86 983.12 901983.96 内圈故障 2269.56 8663659.59 2943.41 3512749.75 外圈故障 2277.24 8698418.07 2949.31 3512593.66
下载: 导出CSV
表 4 滚动轴承重构信号时频域计算结果
Table 4. Rolling bearing reconstruction signal time-frequency domain calculation results
类别 均方根 偏斜度 峰值 峭度 波形 脉冲指标 无故障 0.23 0.006 1.31 3.78 4.94 85.5258 内圈故障 0.08 −0.056 1.69 6.59 11.13 150058.50 外圈故障 0.10 −0.012 2.36 4.99 11.81 74125.97
下载: 导出CSV
表 5 不同负载工况滚动轴承振动信号数据集构造
Table 5. Construction of rolling bearing vibration signal dataset for different loading conditions
项目功率/W 无故障 故障为0.177 8 mm 故障为0.533 4 mm 样本名称 内圈 滚动体 外圈 内圈 滚动体 外圈 0 30 30 10 10 10 10 10 A/90 735.49 30 30 10 10 10 10 10 B/90 1470.98 40 40 10 10 10 10 10 C/100 2206.47 40 40 10 10 10 10 10 D/100
下载: 导出CSV
表 6 单-单负载情况下各算法诊断准确率
Table 6. Diagnostic accuracy of each algorithm in single - single load case
类别 A-B A-C A-D B-A B-C B-D C-A C-B C-D D-A D-B D-C 原始信号 0.55 0.58 0.52 0.45 0.50 0.51 0.52 0.51 0.48 0.52 0.53 0.54 重构信号 0.77 0.78 0.65 0.80 0.74 0.77 0.74 0.69 0.84 0.76 0.68 0.82 重构 + TCA 0.80 0.81 0.63 0.83 0.86 0.80 0.77 0.74 0.9 0.84 0.76 0.94 重构 + POS -SSTCA 0.82 0.89 0.76 0.88 0.90 0.86 0.79 0.82 0.9 0.89 0.80 1.00
下载: 导出CSV
表 7 多-单负载情况下各算法诊断准确率
Table 7. Diagnostic accuracy of each algorithm in the case of multiple-single load
类别 AB-C AB-D AC-B AC-D AD-B AD-C BC-A BC-D BD-A BD-C CD-A CD-B 原始信号 0.58 0.62 0.63 0.62 0.61 0.68 0.59 0.62 0.64 0.66 0.56 0.59 重构信号 0.74 0.78 0.66 0.69 0.70 0.79 0.63 0.66 0.62 0.71 0.60 0.65 重构 + TCA 0.9 0.72 0.78 0.81 0.81 0.92 0.81 0.8 0.8 0.87 0.76 0.77 重构 + POS -SSTCA 0.9 0.84 0.89 0.88 0.88 0.96 0.88 0.88 0.88 0.93 0.79 0.85
下载: 导出CSV
-
[1] 陈伟. 深度学习在滚动轴承故障诊断中的应用研究[D]. 成都: 西南交通大学, 2018CHEN W. Application of deep learning in rolling bearing fault diagnosis[D]. Chengdu: Southwest Jiaotong University, 2018. (in Chinese) [2] 张习习, 顾幸生. 基于集成学习概率神经网络的电机轴承故障诊断[J]. 华东理工大学学报(自然科学版), 2020, 46(1): 68-76. doi: 10.14135/j.cnki.1006-3080.20181206001ZHANG X X, GU X S. Motor bearing fault diagnosis method based on integrated learning probabilistic neural network[J]. Journal of East China University of Science and Technology (Natural Science Edition), 2020, 46(1): 68-76. (in Chinese) doi: 10.14135/j.cnki.1006-3080.20181206001 [3] QIN Y, JIN L, ZHANG A B, et al. Rolling bearing fault diagnosis with adaptive harmonic kurtosis and improved bat algorithm[J]. IEEE Transactions on Instrumentation and Measurement, 2021, 70: 3508112. [4] 赵志宏, 赵敬娇, 魏子洋. 基于BiLSTM的滚动轴承故障诊断研究[J]. 振动与冲击, 2021, 40(1): 95-101. doi: 10.13465/j.cnki.jvs.2021.01.013ZHAO Z H, ZHAO J J, WEI Z Y. Rolling bearing fault diagnosis based on BiLSM network[J]. Journal of Vibration and Shock, 2021, 40(1): 95-101. (in Chinese) doi: 10.13465/j.cnki.jvs.2021.01.013 [5] 赵庆恩, 黄宏伟, 冯坤, 等. 基于小波包能量-决策树的滚动轴承混合故障诊断[J]. 轴承, 2016(6): 43-46. doi: 10.3969/j.issn.1000-3762.2016.06.011ZHAO Q E, HUANG H W, FENG K, et al. Mixed fault diagnosis for rolling bearings based on wavelet packet energy- decision tree[J]. Bearing, 2016(6): 43-46. (in Chinese) doi: 10.3969/j.issn.1000-3762.2016.06.011 [6] 王化玲, 刘志远, 赵欣洋, 等. 基于故障敏感分量和改进K近邻分类器的故障状态识别[J]. 重庆大学学报, 2020, 43(12): 33-40. doi: 10.11835/j.issn.1000-582X.2020.002WANG H L, LIU Z Y, ZHAO X Y, et al. Fault state identification method based on fault sensitive components and improved KNNC[J]. Journal of Chongqing University, 2020, 43(12): 33-40. (in Chinese) doi: 10.11835/j.issn.1000-582X.2020.002 [7] CHE C C, WANG H W, NI X M, et al. Intelligent fault diagnosis method of rolling bearing based on stacked denoising autoencoder and convolutional neural network[J]. Industrial Lubrication and Tribology, 2020, 72(7): 947-953. doi: 10.1108/ILT-11-2019-0496 [8] SONG X D, ZHU D J, LIANG P, et al. A new bearing fault diagnosis method using elastic net transfer learning and LSTM[J]. Journal of Intelligent & Fuzzy Systems, 2021, 40(6): 12361-12369. [9] 沈飞, 陈超, 严如强. 奇异值分解与迁移学习在电机故障诊断中的应用[J]. 振动工程学报, 2017, 30(1): 118-126. doi: 10.16385/j.cnki.issn.1004-4523.2017.01.016SHEN F, CHEN C, YAN R Q. Application of SVD and transfer learning strategy on motorfault diagnosis[J]. Journal of Vibration Engineering, 2017, 30(1): 118-126. (in Chinese) doi: 10.16385/j.cnki.issn.1004-4523.2017.01.016 [10] DORRI F, GHODSI A. Adapting component analysis[C]//Proceedings of the IEEE 12th International Conference on Data Mining. Brussels, Belgium: IEEE, 2012: 846-851 [11] 张西宁, 余迪, 刘书语. 基于迁移学习的小样本轴承故障诊断方法研究[J]. 西安交通大学学报, 2021, 55(10): 30-37. doi: 10.7652/xjtuxb202110004ZHANG X N, YU D, LIU S Y. Fault diagnosis method for small sample bearing based on transfer learning[J]. Journal of Xi'an Jiaotong University, 2021, 55(10): 30-37. (in Chinese) doi: 10.7652/xjtuxb202110004 [12] PAN S J, TSANG I W, KWOK J T, et al. Domain adaptation via transfer component analysis[J]. IEEE Transactions on Neural Networks, 2011, 22(2): 199-210. doi: 10.1109/TNN.2010.2091281 [13] CWRU. Bearing Date Center, seeded fault test data[EB/OL].[2021-08-10]. http://csegroups.case.edubearingdatacenter/ [14] 张泽宇, 惠记庄, 石泽. 小波包最优基分解树的降噪滤波方法研究[J]. 机械科学与技术, 2020, 39(1): 28-34. doi: 10.13433/j.cnki.1003-8728.20190239ZHANG Z Y, HUI J Z, SHI Z. Research on denoising and filtering method based on wavelet packet optimal base decomposition tree[J]. Mechanical Science and Technology for Aerospace Engineering, 2020, 39(1): 28-34. (in Chinese) doi: 10.13433/j.cnki.1003-8728.20190239 [15] ZHANG Z Y, HUI J Z, SHI Z, et al. Cycle condition identification of loader based on optimized KNN algorithm[J]. IEEE Access, 2020, 8: 69532-69542. doi: 10.1109/ACCESS.2020.2985052 -
点击查看大图
计量
- 文章访问数: 657
- HTML全文浏览量: 22
- PDF下载量: 11
- 被引次数: 0
