A Ball Mill Load State Identification Method in Combination with CEEMDAN and Sample Entropy
-
摘要: 针对磨矿作业过程中球磨机的筒体内部情况复杂难以仅靠经验估计负荷状态的问题,提出一种自适应白噪声的完整经验模态分解(CEEMDAN)结合样本熵和极限学习机(ELM)的球磨机负荷状态识别方法。首先,通过CEEMDAN算法对不同负荷状态下原始信号进行分解,利用相关系数法筛选有效的IMF分量;然后,通过分析3种负荷状态下振动信号的有效IMF分量的样本熵在不同数据长度、嵌入维度和相似容限下的值,来确定计算样本熵的最佳参数。结果表明,3种负荷状态下振动信号的有效IMF分量样本熵有明显区别,可以有效识别出磨机不同负荷状态。将各组信号有效IMF分量样本熵作为ELM的输入,球磨机负荷状态为输出,建立了磨机负荷状态识别模型。利用磨矿实验进行验证,表明此种方法应用在球磨机负荷识别上的有效性,整体识别率高达96.81%,且对比于EMD-样本熵和MEEMD-样本熵,总体识别率分别提高了12.41%和9.01%。Abstract: Owing to the complex internal condition of a ball mill in its grinding process, it is difficult to estimate its load state with experience alone. For the complete empirical mode decomposition (CEEMDAN) of adaptive white noise combined with sample entropy and extreme learning machine (ELM), the paper proposes a ball mill load state identification method. Firstly, the eigenmode decomposition of the original signal under different load conditions is carried out withthe CEEMDAN algorithm, and the effective IMF component is filtered with the correlation coefficient method. Then, the sample entropy of the effective IMF component of the vibration signal under three load states is analyzed for different data. The values for length, embedded dimension and similar tolerance determine the optimal parameters for calculating sample entropy. The results show that there are significant differences in the effective IMF component sample entropy of vibration signals under three load states, which can be used to effectively identify the different load states of the ball mill. The effective IMF component sample entropy of each group of signals is taken as the input of ELM and the load states of the ball mill as output, and the load state recognition model is established. The grinding test is used to verify the effectiveness of this method in theball mill load state recognition. The overall recognition rate is as high as 96.81% and increases by 12.41% compared with that of EMD-sample entropy and MEEMD-sample entropy, namely 9.01%.
-
表 1 不同负荷信号去噪效果比较
负荷状态 原始信号SNR/dB 重构信号SNR/dB 欠负荷 9.12 24.52 正常负荷 10.36 25.53 过负荷 8.25 27.41 表 2 不同算法处理后的信噪比
算法 SNR/dB EEMD 11.23 MEEMD 24.35 CEEMDAN 27.41 表 3 不同负荷状态五组信号的有效IMF分量样本熵平均值
磨机负荷状态 IMF分量 IMF1 IMF2 IMF3 IMF4 欠负荷 1.2542 0.2564 0.5895 0.4562 正常负荷 0.9523 0.1245 0.3526 0.2564 过负荷 0.5426 0.0254 0.1257 0.0859 表 4 不同负荷状态的部分信号IMF分量样本熵值
负荷状态 IMF分量 IMF1 IMF2 IMF3 IMF4 欠负荷 1.1379 0.2133 0.6275 0.4689 1.0596 0.2461 0.4853 0.4437 1.2255 0.2831 0.3955 0.4026 1.2630 0.0273 0.3978 0.4020 正常负荷 0.7150 0.1562 0.3413 0.2199 0.6109 0.1228 0.3238 0.4099 1.0493 0.1829 0.3561 0.2984 0.7637 0.1433 0.300 0.2491 过负荷 0.5687 0.0771 0.1406 0.1130 0.5003 0.1565 0.2836 0.2606 0.5732 0.0873 0.1605 0.0893 0.2677 0.0450 0.0725 0.0339 表 5 不同特征提取算法磨机负荷识别结果
特征提取
算法球磨机不同负荷状态识别率 总体识别率/% 欠负荷识别率/% 正常负荷识别率/% 过负荷识别率/% EEMD-样本熵 86.7 83.3 83.3 84.4 MEEMD-样本熵 90 86.7 86.7 87.8 CEEMDAN-样本熵 96.46 95.14 98.85 96.81 -
[1] 蔡改贫, 宗路, 罗小燕, 等. 基于CEEMDAN-云模型特征熵和LSSVM的磨机负荷预测研究[J]. 振动与冲击, 2019, 38(7): 128-133CAI G P, ZONG L, LUO X Y, et al. Prediction of ball mill's load based on IEDA-cloud model feature entropy and LSSVM[J]. Journal of Vibration and Shock, 2019, 38(7): 128-133 (in Chinese) [2] 刘卓, 柴天佑, 汤健. 一种多尺度球磨机筒体振动频谱分析与建模方法[J]. 东北大学学报, 2015, 36(3): 305-308LIU Z, CHAI T Y, TANG J. Multi-scale shell vibration frequency spectrum analysis and modeling approach of ball mill[J]. Journal of Northeastern University, 2015, 36(3): 305-308 (in Chinese) [3] 李腾飞, 林蜀勇, 张博, 等. 不同转速率下球磨机内钢球的碰撞研究[J]. 中南大学学报, 2019, 50(2): 251-256 doi: 10.11817/j.issn.1672-7207.2019.02.001LI T F, LIN S Y, ZHANG B, et al. Study on collisions of steel balls in grinding mill at different rotation speeds[J]. Journal of Central South University, 2019, 50(2): 251-256 (in Chinese) doi: 10.11817/j.issn.1672-7207.2019.02.001 [4] 张建宇, 张随征, 管磊, 等. 基于多小波包样本熵的轴承损伤程度识别方法[J]. 振动、测试与诊断, 2015, 35(1): 128-132ZHANG J Y, ZHANG S Z, GUAN L, et al. A method for identifying bearing damage degree based on multi-wavelet packet sample entropy[J]. Journal of Vibration, Measurement & Diagnosis, 2015, 35(1): 128-132 (in Chinese) [5] 刘晓东, 刘朦月, 陈寅生, 等. EEMD-PE与M-RVM相结合的轴承故障诊断方法[J]. 哈尔滨工业大学学报, 2017, 49(9): 122-128 doi: 10.11918/j.issn.0367-6234.201604066LIU X D, LIU M Y, CHEN Y S, et al. Rolling bearing fault diagnosis based on EEMD-PE coupled with M-RVM[J]. Journal of Harbin Institute of Technology, 2017, 49(9): 122-128 (in Chinese) doi: 10.11918/j.issn.0367-6234.201604066 [6] ZHANG W, JIA M P, ZHU L, et al. Comprehensive overview on computational intelligence techniques for machinery condition monitoring and fault diagnosis[J]. Chinese Journal of Mechanical Engineering, 2017, 30(4): 782-795 doi: 10.1007/s10033-017-0150-0 [7] 耿读艳, 王晨旭, 赵杰, 等. 基于CEEMDAN-PE的心冲击信号降噪方法研究[J]. 仪器仪表学报, 2019, 40(6): 155-161GENG D Y, WANG C X, ZHAO J, et al. Research on BCG signal de-noising method based on CEEMDAN and PE[J]. Chinese Journal of Scientific Instrument, 2019, 40(6): 155-161 (in Chinese) [8] 李军, 李青. 基于CEEMDAN-排列熵和泄漏积分ESN的中期电力负荷预测研究[J]. 电机与控制学报, 2015, 19(8): 70-80LI J, LI Q. Medium term electricity load forecasting based on CEEMDAN-permutation entropy and ESN with leaky integrator neurons[J]. Electric Machines and Control, 2015, 19(8): 70-80 (in Chinese) [9] 王文哲, 吴华, 王经商, 等. 基于CEEMDAN的雷达信号脉内细微特征提取法[J]. 北京航空航天大学学报, 2016, 42(11): 2532-2539WANG W Z, WU H, WANG J S, et al. Subtle intrapulse feature extraction based on CEEMDAN for radar signals[J]. Journal of Beijing University of Aeronautics and Astronautics, 2016, 42(11): 2532-2539 (in Chinese) [10] 张宁, 刘友文. 基于CEEMDAN改进阈值滤波的微机电陀螺信号去噪模型[J]. 中国惯性技术学报, 2018, 26(5): 665-669, 674ZHANG N, LIU Y W. Signal de-noising model for MEMS gyro based on CEEMDAN improved threshold filtering[J]. Journal of Chinese Inertial Technology, 2018, 26(5): 665-669, 674 (in Chinese) [11] 刘霞, 宋启航. CEEMDAN自适应阈值去噪算法在地震方向的应用[J]. 重庆大学学报, 2019, 42(7): 95-104LIU X, SONG Q H. CEEMDAN adaptive threshold denoising algorithm in application to seismic direction[J]. Journal of Chongqing University, 2019, 42(7): 95-104 (in Chinese) [12] 张建伟, 侯鸽, 暴振磊, 等. 基于CEEMDAN与SVD的泄流结构振动信号降噪方法[J]. 振动与冲击, 2017, 36(22): 138-143ZHANG J W, HOU G, BAO Z L, et al. A signal de-noising method for vibration signals from flood discharge structures based on CEEMDAN and SVD[J]. Journal of Vibration and Shock, 2017, 36(22): 138-143 (in Chinese) [13] WU Q L, LIN H X. Daily urban air quality index forecasting based on variational mode decomposition, sample entropy and LSTM neural network[J]. Sustainable Cities and Society, 2019, 50: 101657 doi: 10.1016/j.scs.2019.101657 [14] 向北平, 周建, 倪磊, 等. 基于样本熵的改进小波包阈值去噪算法[J]. 振动、测试与诊断, 2019, 39(2): 410-415XIANG B P, ZHOU J, NI L, et al. Research on improved wavelet packet threshold denoising algorithm based on sample entropy[J]. Journal of Vibration, Measurement & Diagnosis, 2019, 39(2): 410-415 (in Chinese) [15] 谢国民, 黄睿灵, 丁会巧. 基于VMD样本熵和KELM的输电线路故障诊断[J]. 电子测量与仪器学报, 2019, 33(5): 73-79XIE G M, HUANG R L, DING H Q. Fault diagnosis of transmission lines based on VMD sample entropy and KELM[J]. Journal of Electronic Measurement and Instrumentation, 2019, 33(5): 73-79 (in Chinese) [16] 施莹, 林建辉, 庄哲, 等. 基于振动信号时频分解-样本熵的受电弓裂纹故障诊断[J]. 振动与冲击, 2019, 38(8): 180-187SHI Y, LIN J H, ZHUANG Z, et al. Fault diagnosis for pantograph cracks based on time-frequency decomposition and sample entropy of vibration signals[J]. Journal of Vibration and Shock, 2019, 38(8): 180-187 (in Chinese)