Research on wear state monitoring of ultrasonic vibration drilling bits based on dual signal fusion
-
摘要: 为了监测超声振动钻削过程中钻头的磨损状态, 构建了超声振动钻削钻头的振动信号和AE信号的采集系统, 通过采集不同磨损状态下钻头的振动信号和AE信号, 对其进行小波分解, 得到与钻头磨损相关的特征值, 将二者融合后作为神经网络的输入, 输入至构建的12-10-3的BP神经网络中, 进行钻头磨损状态的识别。试验结果表明, 所建BP神经网络通过振动和AE的融合信号对钻头的有效识别率为91.7%, 可以有效对钻头的磨损状态进行识别。Abstract: In order to monitor the wear status of the drill bit in the process of ultrasonic vibration drilling, an acquisition system for the vibration signal and AE signal of the ultrasonic vibration drilling bit is constructed. By collecting the vibration signal and AE signal of the drill bit under different wear conditions, it is subjected to wavelet decomposition. Obtain the eigenvalues related to the wear of the drill bit, merge the two as the input of the neural network, and input it into the constructed 12-10-3 BP neural network to identify the wear status of the drill bit. The test results show that the built BP neural network has an effective recognition rate of 91.7% for the drill bit through the fusion signal of vibration and AE, which can effectively identify the wear state of the drill bit.
-
Key words:
- AE signal /
- wavelet decomposition /
- BP neural network /
- bit wear condition
-
表 1 YG8与45钢板的材料特性
材料属性 YG8 45#钢板 密度/(g/cm3) 14.8 7.85 抗弯强度/MPa 1 670 400 硬度 89 HRA 48 HRC 表 2 钻头状态识别方式
序号 钻头状态 期望输出向量 1 正常磨损 [1 0 0] 2 严重磨损 [0 1 0] 3 崩刃 [0 0 1] 表 3 神经网络参数设置
神经网络参数设置 参数 训练函数 Traingd 迭代次数 2 000 收敛误差 0.000 1 学习率 0.01 表 4 神经网络输出结果
序号 钻头状态 期望输出 实际输出 1 正常磨损 [1 0 0] [0.995 90.016 30.002 1] 2 正常磨损 [1 0 0] [0.987 90.016 10.002 1] 3 正常磨损 [1 0 0] [0.998 20.002 90.018 3] 4 正常磨损 [1 0 0] [0.979 90.019 00.060 7] 5 严重磨损 [0 1 0] [0.002 70.986 40.448 4] 6 严重磨损 [0 1 0] [0.001 90.956 20.019 0] 7 严重磨损 [0 1 0] [0.008 80.995 70.001 3] 8 严重磨损 [0 1 0] [0.023 20.995 60.005 2] 9 崩刃 [0 0 1] [0.005 50.010 30.990 2] 10 崩刃 [0 0 1] [0.004 40.017 50.981 3] 11 崩刃 [0 0 1] [0.064 40.977 30.002 8] 12 崩刃 [0 0 1] [0.017 30.013 30.969 4] -
[1] 张存鹰, 赵波. 超声振动辅助加工表面微结构及其特性研究进展[J]. 表面技术, 2019, 48(5): 259-274. https://www.cnki.com.cn/Article/CJFDTOTAL-BMJS201905039.htm [2] 库祥臣, 曹贝贝, 郭跃飞, 等. 刀具磨损监测的一种数据处理方法[J]. 机床与液压, 2017, 45(17): 105-109. doi: 10.3969/j.issn.1001-3881.2017.17.024 [3] Laddads, Si-Chaib M, Benkedjouh T et al. Tool wear conditionmonitoring based on wavelet transform and improved extreme learning machine[J]. Proceedings of the Institution of Mechanical Engineers, Part C: Journal of Mechanical Engineering Science, 2020, 234(5): 095440621988854. doi: 10.1177/0954406219888544 [4] 王二化, 颜鹏, 李昕. 基于小波包能量分布特征的齿轮故障诊断方法. 研究[J]. 机床与液压, 2020, 48(1): 188-192. doi: 10.3969/j.issn.1001-3881.2020.01.039 [5] 李涛, 黄新宇, 罗明. 基于小波包分解的刀具磨损特征分析[J]. 组合机床与自动化加工技术, 2020(7): 10-15. https://www.cnki.com.cn/Article/CJFDTOTAL-ZHJC202007003.htm [6] 齐添添, 陈尧, 何才厚, 等. 损伤声发射信号小波包神经网络的特征识别方法[J]. 北京邮电大学学报, 2021, 44(1): 124-130. https://www.cnki.com.cn/Article/CJFDTOTAL-BJYD202101020.htm [7] 李帆, 闫献国, 陈峙, 等. 基于遗传算法优化BP神经网络的YG8硬质合金耐磨性预测模型[J]. 金属热处理, 2019, 44(12): 244-248. https://www.cnki.com.cn/Article/CJFDTOTAL-JSRC201912058.htm [8] Bi J X, Fan W Ze, Wang Y, et al. A fault diagnosis algorithm for wind turbine blades based on BP neural network[J]. IOP Conference Series: Materials Science and Engineering, 2021, 1043(2): 022032. doi: 10.1088/1757-899X/1043/2/022032 [9] 史尧臣, 刘红岩, 张学忱, 等. 基于声发射和振动信号的振动钻削钻头故障诊断试验研究[J]. 机床与液压, 2020, 48(10): 197-200. doi: 10.3969/j.issn.1001-3881.2020.10.041 [10] 赵亭. 难加工材料钻削性能及刀具磨损特性研究[D]. 西安: 西安理工大学, 2020. [11] 金铃子, 曹越操, 亓元浩, 等. 基于声发射与D-S证据理论的截齿磨损状态识别[J]. 煤炭科学技术, 2020, 84(5): 120-128. https://www.cnki.com.cn/Article/CJFDTOTAL-MTKJ202005017.htm [12] 周芸梦. 基于声发射技术的刀具磨损监测研究[D]. 洛阳: 河南科技大学, 2014. [13] 陈洪涛. 基于多参量信息融合的刀具磨损状态识别及预测技术研究[D]. 成都: 西南交通大学, 2013.