New fault zero sample identification of equipment based on feature attribute description
-
摘要:
自动化升级背景下,设备间耦合性不断提高,故障表现形式繁杂多样. 单一故障不及时处理极易造成影响范围的扩大,使得事故进一步升级. 为保证设备的正常运转,对传统的基于案例分析生成的故障诊断方法提出了新的要求,具体包括：低成本、长期监测、少样本或零样本故障识别. 针对这些新需求,本文提出将图像处理领域中使用的零样本分类识别思想引入故障诊断领域. 通过研究现有故障样本的特征参量,对其进行寻优确定用于状态监测的特征,采用模糊神经网络构成特征属性描述器,将特征描述为设备属性,再由ART网络以属性描述为基础,对设备进行长期监测的同时增量学习. 即以少量设备样本或相似样本的分析为基础构建监测与学习机制,识别原有故障的同时学习并记录新类故障. 为说明本方法的可行性与有效性,文章利用电机故障数据集以少量样本为先验知识构建系统,混合未知故障样本进行系统测试. 实验结果表明,零样本分类思想的应用有望解决当前技术背景下设备故障诊断的新挑战.
Abstract:Under the background of automation upgrading, the coupling between equipments is constantly improving, and the fault manifestations are complex and diverse. Failure to deal with a single fault in a timely manner is likely to expand the scope of influence, which further escalates the accident. In order to ensure the normal operation of the equipment, new requirements are put forward for the traditional fault diagnosis methods based on case analysis, including low cost, long-term monitoring, small sample or zero sample fault identification. To meet these new requirements, this paper proposes to introduce the zero sample classification recognition idea used in the image processing field into the fault diagnosis field. By studying the characteristic parameters of the existing fault samples, the characteristics used for condition monitoring are determined through optimization. The fuzzy neural network is used to form a feature attribute descriptor, which describes the characteristics as equipment attributes. Based on the attribute description, the ART network conducts incremental learning for long-term monitoring of equipment at the same time. That is to build a monitoring and learning mechanism based on the analysis of a small number of equipment samples or similar samples, identify the original faults and learn and record new faults at the same time. In order to illustrate the feasibility and effectiveness of this method, this paper uses a small number of motor fault data sets as prior knowledge to build a system, and mixed unknown fault samples for system testing. The experimental results show that the application of zero sample classification is expected to solve the new challenges of equipment fault diagnosis under the current technical background.
-
Key words:
- Characteristic attribute /
- New fault /
- Zero sample /
- Fault diagnosis /
- Incremental learning.
-
图 3 感应电机10种不同状态下的振动信号波形
Figure 3. Vibration signal waveform of induction motor under 10 different states
图 5 不同寻优方法下的特征重要性评分
Figure 5. Feature importance score under different optimization methods
表 1 振动信号常用特征参数
Table 1. Common Characteristic Parameters of Vibration Signal
振动信号常用特征 $ {p_1} = \dfrac{{\displaystyle\sum\nolimits_{n = 1}^N {x(n)} }}{N} $ $ {p_{13}} = \sqrt {\dfrac{{\displaystyle\sum\nolimits_{k = 1}^K {{f_k}^2s(k)} }}{{\displaystyle\sum\nolimits_{k = 1}^K {s(k)} }}} $ $ {p_2} = {(\dfrac{{\displaystyle\sum\nolimits_{n = 1}^N {\sqrt {\left| {x(n)} \right|} } }}{N})^2} $ $ {p_{14}} = \sqrt {\dfrac{{\displaystyle\sum\nolimits_{k = 1}^K {{f_k}^4s(k)} }}{{\displaystyle\sum\nolimits_{k = 1}^K {{f_k}^2s(k)} }}} $ ${p_3} = \sqrt {\dfrac{{\displaystyle\sum\nolimits_{n = 1}^N {{{(x(n))}^2}} }}{N}} $ $ {p_{15}} = \dfrac{{\displaystyle\sum\nolimits_{k = 1}^K {{f_k}^2s(k)} }}{{\sqrt {\displaystyle\sum\nolimits_{k = 1}^K {s(k)} \displaystyle\sum\nolimits_{k = 1}^K {{f_k}^4s(k)} } }} $ ${p_4} = \max \left| {x(n)} \right|$ $ {p}_{16}=\dfrac{{\displaystyle {\displaystyle\sum }_{k=1}^{K}{f}_{k}\cdot s(k)}}{{\displaystyle {\displaystyle\sum }_{k=1}^{K}s(k)}} $ $ {p_5} = \sqrt {\dfrac{{\displaystyle\sum\nolimits_{n = 1}^N {{{(x(n) - {p_1})}^2}} }}{{N - 1}}} $ $ {p_{17}} = \dfrac{{\displaystyle\sum\nolimits_{k = 1}^K {{{(s(k) - {p_{12}})}^2}} }}{{K - 1}} $ $ {p}_{6}=\dfrac{{\displaystyle {\displaystyle\sum }_{n=1}^{N}{(x(n)-{p}_{1})}^{3}}}{(N-1)\cdot {p}_{5}{}^{3}} $ $ {p}_{18}=\dfrac{{\displaystyle {\displaystyle\sum }_{k=1}^{K}{(s(k)-{p}_{12})}^{3}}}{K\cdot {(\sqrt{{p}_{17}})}^{3}} $ $ {p}_{7}=\dfrac{{\displaystyle {\displaystyle\sum }_{n=1}^{N}{(x(n)-{p}_{1})}^{4}}}{(N-1)\cdot {p}_{5}{}^{4}} $ $ {p}_{19}=\dfrac{{\displaystyle {\displaystyle\sum }_{k=1}^{K}{(s(k)-{p}_{12})}^{4}}}{K\cdot {p}_{17}{}^{2}} $ $ {p_8} = \dfrac{{{p_4}}}{{{p_3}}} $ $ {p_{20}} = \sqrt {\dfrac{{\displaystyle\sum\nolimits_{k = 1}^K {{{(f(k) - {p_{16}})}^2}s(k)} }}{K}} $ $ {p_9} = \dfrac{{{p_4}}}{{{p_2}}} $ $ {p_{21}} = \dfrac{{{p_{20}}}}{{{p_{16}}}} $ ${p_{10} } = \dfrac{ { {p_3} } }{ { {1 /N }\displaystyle\sum\nolimits_{n = 1}^N {\left| {x(n)} \right|} } }$ $ {p}_{22}=\dfrac{{\displaystyle {\displaystyle\sum }_{k=1}^{K}{({f}_{k}-{p}_{16})}^{3}s(k)}}{K\cdot {p}_{20}{}^{3}} $ ${p_{11} } = \dfrac{ { {p_4} } }{ { {1/ N }\displaystyle\sum\nolimits_{n = 1}^N {\left| {x(n)} \right|} } }$ $ {p}_{23}=\dfrac{{\displaystyle {\displaystyle\sum }_{k=1}^{K}{({f}_{k}-{p}_{16})}^{4}s(k)}}{K\cdot {p}_{20}{}^{4}} $ $ {p_{12}} = \dfrac{{\displaystyle\sum\nolimits_{k = 1}^K {s(k)} }}{K} $ ${p}_{24}=\dfrac{ {\displaystyle {\displaystyle\sum }_{k=1}^{K}{({f}_{k}-{p}_{16})}^{ 1/ 2 }s(k)} }{K\cdot \sqrt{ {p}_{20} } }$ 
下载: 导出CSV
表 2 特征寻优前后分类识别准确度比较
Table 2. Comparison of classification and recognition accuracy before and after feature optimization
特征
维数KNN 逻辑
回归决策树 朴素贝
叶斯LDA 平均 寻优前 24 1 0.92 1 1 1 0.984 寻优后 8 0.99 0.82 0.99 0.97 0.98 0.95
下载: 导出CSV
表 3 新故障零样本识别性能
Table 3. New Fault Zero Sample Identification Performance
已知故障数 未知故障数 平均准确率 10 - 99.91% 9 1 91.8% 5 4 79.94% 4 5 74.95%
下载: 导出CSV
-
[1] 石颉, 袁晨翔, 孔维相, 等. 基于电-热加速老化的LED寿命评估检验方法研究[J]. 电子元件与材料,2020,39(8):89-95. DOI: 10.14106/j.cnki.1001-2028.2020.0279.SHI J, YUAN C X, KONG W X, et al. Research on the test method of LED life evaluation based on electro-thermal accelerated aging[J]. Electronic Components and Materials,2020,39(8):89-95. DOI: 10.14106/j.cnki.1001-2028.2020.0279. [2] 罗东亮, 蔡雨萱, 杨子豪, 等. 工业缺陷检测深度学习方法综述[J]. 中国科学:信息科学,2022,52(6):1002-1039. DOI: 10.1360/SSI-2021-0336.LUO D L, CAI Y X, YANG Z H, et al. Survey on industrial defect detection with deep learning[J]. Science in China:Information Sciences,2022,52(6):1002-1039. DOI: 10.1360/SSI-2021-0336. [3] 武红鑫, 韩萌, 陈志强, 等. 监督和半监督学习下的多标签分类综述[J]. 计算机科学,2022,49(8):12-25. DOI: 10.11896/jsjkx.210700111.WU H X, HAN M, CHEN Z Q, et al. Survey of multi-label classification based on supervised and semi-supervised learning[J]. Computer Science,2022,49(8):12-25. DOI: 10.11896/jsjkx.210700111. [4] CHEN D M, YANG S, ZHOU F N. Transfer learning based Fault diagnosis with missing data due to multi-rate sampling[J]. Sensors,2019,19(8):1826. DOI: 10.3390/s19081826. [5] 王凯, 李元辉. 迁移学习在机械设备预测性维护领域的应用综述[J]. 中国仪器仪表,2019(12):64-68. DOI: 10.3969/j.issn.1005-2852.2019.12.018.WANG K, LI Y H. Summary of application of transfer learning in predictive maintenance of machinery and equipment[J]. China Instrumentation,2019(12):64-68. DOI: 10.3969/j.issn.1005-2852.2019.12.018. [6] LAMPERT C H, NICKISCH H, HARMELING S. Learning to detect unseen object classes by between-class attribute transfer[C]//2009 IEEE Conference on Computer Vision and Pattern Recognition. Miami: IEEE, 2009: 951-958. [7] 贾霄, 郭顺心, 赵红. 基于图像属性的零样本分类方法综述[J]. 南京大学学报(自然科学),2021,57(4):531-543. DOI: 10.13232/j.cnki.jnju.2021.04.001.JIA X, GUO S X, ZHAO H. A review of zero-shot learning classification methods based on image attributes[J]. Journal of Nanjing University (Natural Sciences),2021,57(4):531-543. DOI: 10.13232/j.cnki.jnju.2021.04.001. [8] WANG W, ZHENG V W, YU H, et al. A survey of zero-shot learning: settings, methods, and applications[J]. ACM Transactions on Intelligent Systems and Technology,2019,10(2):13. DOI: 10.1145/3293318. [9] 丁飞, 石颉, 吴宏杰. 改进YOLOv4的轻量级遥感图像建筑物检测模型[J/OL]. 计算机工程与应用, 1-10[2022-06-30]. http://kns.cnki.net/kcms/detail/11.2127.TP.20220524.1035.007.html.DING F, SHI J, WU H J. Lightweight building detection model based on YOLOv4 optimization for remote sensing images[J/OL]. Computer Engineering and Applications, 1-10[2022-06-30]. http://kns.cnki.net/kcms/detail/11.2127.TP.20220524.1035.007.html. [10] 王干军, 李锦舒, 吴毅江, 等. 基于随机森林的高压电缆局部放电特征寻优[J]. 电网技术,2019,43(4):1329-1335. DOI: 10.13335/j.1000-3673.pst.2018.2652.WANG G J, LI J S, WU Y J, et al. Random forest based feature selection for partial discharge recognition of HV cables[J]. Power System Technology,2019,43(4):1329-1335. DOI: 10.13335/j.1000-3673.pst.2018.2652. [11] 杨戬. 非确定性模糊神经网络方法在正、反分析中的应用[D]. 北京: 清华大学, 2008.YANG J. The application of uncertain fuzzy neural network method in forward and back analysis[D]. Beijing: Tsinghua University, 2008. [12] 江水, 徐启胜, 李军, 等. 基于T-S模糊神经网络的液压设备故障诊断[J]. 锻压装备与制造技术,2022,57(2):67-71. DOI: 10.16316/j.issn.1672-0121.2022.02.019.JIANG S, XU Q S, LI J, et al. Fault diagnosis of hydraulic equipment based on T-S fuzzy neural network[J]. China Metalforming Equipment & Manufacturing Technology,2022,57(2):67-71. DOI: 10.16316/j.issn.1672-0121.2022.02.019. [13] CHEFROUR A. Incremental supervised learning: algorithms and applications in pattern recognition[J]. Evolutionary Intelligence,2019,12(2):97-112. DOI: 10.1007/s12065-019-00203-y. [14] 刘冰瑶, 刘进锋. 增量学习研究综述[J]. 现代计算机,2022,28(13):72-75. DOI: 10.3969/j.issn.1007-1423.2022.13.012.LIU B Y, LIU J F. Literature review of incremental learning[J]. Modern Computer,2022,28(13):72-75. DOI: 10.3969/j.issn.1007-1423.2022.13.012. [15] 王跃龙. 笼型异步电动机多故障智能诊断及分离方法的研究[D]. 太原: 太原理工大学, 2017.WANG Y L. Research on the multi-fault intelligent diagnosis and separation methods for squirrel cage asynchronous motor[D]. Taiyuan: Taiyuan University of Technology, 2017. [16] 申海锋, 石颉. 电机转子振动信号故障特征提取方法[J]. 噪声与振动控制,2022,42(4):138-143. DOI: 10.3969/j.issn.1006-1355.2022.04.023.SHEN H F, SHI J. Fault feature extraction method of motor rotor vibration signals[J]. Noise and Vibration Control,2022,42(4):138-143. DOI: 10.3969/j.issn.1006-1355.2022.04.023. [17] WANG P P, LU J J, SHI L P, et al. Method for extracting current envelope for broken rotor bar fault detection of induction motors at time-varying loads[J]. IET Electric Power Applications,2020,14(6):1067-1077. DOI: 10.1049/iet-epa.2019.0779. [18] LIU D, XIAO Z H, HU X, et al. Feature extraction of rotor fault based on EEMD and curve code[J]. Measurement,2019,135:712-724. DOI: 10.1016/j.measurement.2018.12.009. [19] SHI J, SHEN H F, DING Z K. Quantitative analysis of broken rotor bars in cage motor based on energy characteristics of vibration signals[J]. Computational Intelligence and Neuroscience,2022,2022:9312876. DOI: 10.1155/2022/9312876. -
下载:
点击查看大图
计量
- 文章访问数: 3
- HTML全文浏览量: 3
- PDF下载量: 0
- 被引次数: 0
