Roughness prediction of Al2O3-based ceramic insulation coating on bearing surface
-
摘要: 为了提升轴承表面Al2O3基陶瓷绝缘涂层的粗糙度预测精度,提出基于光谱共焦原理的砂轮表面测量及磨粒特征参数量化方法,以砂轮表面的磨粒特征参数K,砂轮线速度vs,工件进给速度f,切削深度ap及法向磨削力F为输入参数,建立能够直接反映砂轮表面时变状态的工件表面粗糙度BP神经网络预测模型,并通过已知磨削样本及砂轮磨损后的4组未知样本对网络预测模型性能进行验证。结果表明:已知样本的BP网络模型粗糙度预测结果与实际结果的规律及数值较为一致,其网络输出误差均 < ± 0.04 μm;4组未知样本的网络预测精度下降,但其相对误差最大值的绝对值不超过20.00%。建立的包含砂轮表面磨粒特征参数的神经网络预测模型,可以适应砂轮磨粒磨损时变状态下的轴承表面Al2O3基陶瓷绝缘涂层的粗糙度预测,且其对未知样本具有一定的泛化能力。Abstract: To improve the roughness prediction accuracy of Al2O3-based ceramic insulation coating on bearing surfaces, a method based on the spectral confocal principle was proposed for measuring the surface of grinding wheels and quantifying the characteristic parameters of abrasive particles. The abrasive characteristic parameter K of the grinding wheel surface, the grinding wheel line speed vs, the workpiece feed speed f, the cutting depth ap, and the normal grinding force F were taken as input parameters. A BP neural network prediction model of workpiece surface roughness, which directly reflects the time-varying state of the grinding wheel surface, was established. The prediction performance of the network model was verified using known grinding samples and four groups of unknown samples after grinding wheel wear. The results show that the predicted roughness results of the BP network model with known samples are consistent with the actual roughness results in terms of regularity and numerical values, with network output errors are all less than ± 0.04 μm. The network prediction accuracy for the four unknown samples decreases, but the absolute value of the maximum relative error does not exceed 20.00%. The neural network prediction model, which includes the characteristic parameters of abrasive particles on the grinding wheel surface , can be used to predict the roughness of Al2O3-based ceramic insulation coating on the bearing surface under the time-varying state of abrasive wear on the grinding wheel. It also demonstrates a certain generalization ability for unknown samples.
-
Key words:
- Al2O3-based ceramics /
- insulating coating /
- roughness prediction /
- BP neural network /
- abrasive wear
-
表 1 正交试验因素及水平
Table 1. Orthogonal test factors and levels
水平 因素 砂轮线速度
vs / (m·s−1)
A切深
ap / μm
B工件进给速度
f / (mm·min−1)
C1 15 5 300 2 25 15 900 3 35 25 1 500 4 45 35 2 100 表 2 正交试验表
Table 2. Orthogonal test table
试验序号 A B C 1 15 5 2 100 2 15 15 1 500 3 15 25 900 4 15 35 300 5 25 5 1 500 6 25 15 900 7 25 25 300 8 25 35 2 100 9 35 5 900 10 35 15 300 11 35 25 2 100 12 35 35 1 500 13 45 5 300 14 45 15 2 100 15 45 25 1 500 16 45 35 900 表 3 1#砂轮表面磨粒的特征参数与磨削试验结果
Table 3. Characteristic parameters of abrasive particles on the surface of grinding wheel 1# and grinding resuilts
序号 特征参数 K 法向磨削力 F / N 表面粗糙度 Sa / μm 1 30.3 26.87 0.319 2 30.4 29.73 0.373 3 30.3 66.73 0.531 4 30.1 87.53 0.599 5 30.2 25.94 0.239 6 29.9 21.88 0.361 7 30.4 73.87 0.528 8 29.8 88.79 0.593 9 30.3 13.96 0.225 10 29.8 16.20 0.312 11 30.0 88.40 0.455 12 30.3 90.41 0.500 13 30.4 13.38 0.202 14 29.7 84.48 0.332 15 30.6 72.02 0.365 16 30.4 88.51 0.545 表 4 2#砂轮表面磨粒的特征参数与磨削试验结果
Table 4. Characteristic parameters of abrasive particles on the surface of grinding wheel 2# and grinding resuilts
序号 特征参数 K 法向磨削力 F / N 表面粗糙度 Sa / μm 1 15.8 40.73 0.254 2 15.8 34.05 0.312 3 15.4 25.55 0.453 4 15.5 73.24 0.507 5 15.7 33.69 0.204 6 15.4 39.63 0.309 7 16.1 90.16 0.430 8 15.4 86.58 0.486 9 15.4 27.80 0.178 10 15.4 32.49 0.268 11 15.4 89.57 0.387 12 15.6 103.64 0.423 13 15.5 38.41 0.151 14 16.1 38.37 0.287 15 15.6 103.69 0.315 16 15.4 104.42 0.467 表 5 磨削试验参数
Table 5. Grinding test parameters
参数 规格或取值 砂轮编号 1# 砂轮线速度vs / (m·s−1) 20,32,37,50 工件进给速度 f / (mm·min−1) 1400,900,600,200 切削深度 ap / μm 24,27,11,5 表 6 磨削试验网络输入向量的试验值与预测值对比
Table 6. Comparison between experimental and predicted values of input vectors in grinding test networks
试验组 网络输入向量矩阵
q表面粗糙度
Sa / μm相对误差
δ / %实际 预测 1 [20,24,1400,20.1,73.64] 0.406 0.379 6.65 2 [32,27,900,20.5,59.19] 0.436 0.418 4.13 3 [37,11,600,20.3,17.66] 0.203 0.241 −18.72 4 [50,5,200,20.6,15.76] 0.170 0.198 −16.47 -
[1] 卜珍宇, 赵晓琴, 郭向东, 等. 电机轴承防护措施及Al2O3陶瓷绝缘涂层研究现状 [J]. 表面技术,2021,50(5):51-59.BU Zhenyu, ZHAO Xiaoqin, GUO Xiangdong, et al. Electromotor bearing protection measures and research status of Al2O3 ceramic coating [J]. Surface Technology,2021,50(5):51-59. [2] 余剑武, 胡其丰, 文丞, 等. 基于支持向量机的电火花加工8418钢表面粗糙度预测模型 [J]. 中国机械工程,2018,29(7):771-774.YU Jianwu, HU Qifeng, WEN Chen, et al. Prediction model of surface roughness of 8418 steel by EDM based on SVM [J]. China Mechanical Engineering,2018,29(7):771-774. [3] PAN Y, WANG Y, ZhOU P, et al. Activation functions selection for BP neural network model of ground surface roughness [J]. Journal of Intelligent Manufacturing,2020(31):1825-1836. doi: 10.1007/s10845-020-01538-5 [4] 任敬心, 华定安. 磨削原理[M]. 西安: 西北工业大学出版社, 1988: 206-612.REN Jingxin, HUA Anding. Grinding principle [M]. Xi'an: Northwestern Polytechnical University Press, 1988: 206-216. [5] NAMETALA C A L, SOUZA A M, JÚNIOR B R P, et al. A simulator based on artificial neural networks and NSGA-II for prediction and optimization of the grinding process of superalloys with high performance grinding wheels. [J]. CIRP Journal of Manufacturing Science and Technology,2020(30):157-173. doi: 10.1016/j.cirpj.2020.05.004 [6] YANG J, ZhANG L, LIU G, et al. Sintered silicon carbide grinding surface roughness prediction based on deep learning and neural network [J]. Journal of the Brazilian Society of Mechanical Sciences and Engineering,2022,44(7):1-13. [7] XIAO G. Prediction of surface roughness of abrasive belt grinding of superalloy material based on RLSOM-RBF [J]. Materials,2021,14(19):5701. [8] GUO W, WU C, DING Z, et al. Prediction of surface roughness based on a hybrid feature selection method and long short-term memory network in grinding [J]. The International Journal of Advanced Manufacturing Technology,2021(112):2853-2871. doi: 10.1007/s00170-020-06523-z [9] LAI J Y, LIN P C. Grinded surface roughness prediction using data-driven models with contact force information: 2022 IEEE/ASME international conference on advanced intelligent mechatronics (AIM) [C]. Sapporo: IEEE, 2022: 983-989. [10] MIRIFAR S, KADIVAR M, AZARHOUSHANG B. First steps through intelligent grinding using machine learning via integrated acoustic emission sensors [J]. Journal of Manufacturing and Materials Processing, 2020, 4(2): 35. [11] 董志刚, 马槐遥, 康仁科, 等. SiCf/SiC复合材料超声辅助干式侧磨砂轮磨损研究 [J]. 机械工程学报,2022,58(15):134-143. doi: 10.3901/JME.2022.15.134DONG Zhigang, MA Huaiyao, KANG Renke, et al. Study on wear of grinding wheel in ultrasonic assisted dry side grinding of SiCf/SiC composites [J]. Journal of Mechanical Engineering,2022,58(15):134-143. doi: 10.3901/JME.2022.15.134 [12] 胡天荣, 郑红伟, 戴士杰. 基于异类信息融合的砂轮磨损状态监测 [J]. 工具技术,2023,57(2):126-131.HU Tianrong, ZHENG Hongwei, DAI Shijie. Grinding wheel wear condition monitoring based on heterogeneous information fusion [J]. Tool Engineering,2023,57(2):126-131. [13] 王津楠, 陈凤东, 刘炳国, 等. 基于白光LED的光谱共焦位移传感器 [J]. 中国测试,2017,43(1):69-73.WANG Jinnan, CHEN Fengdong, LIU Bingguo, et al. White LED-based spectrum confocal displacement sensor [J]. China Measurment & Test,2017,43(1):69-73. [14] 王嵘冰, 徐红艳, 李波, 等. BP神经网络隐含层节点数确定方法研究 [J]. 计算机技术与发展,2018,28(4):31-35.WANG Rongbing, XU Hongyan, LI Bo, et al. Research on method of determining hidden layer nodes in BP neural network [J]. Computer Technology and Development,2018,28(4):31-35. -