| 珩磨缸套表面粗糙度预测及多目标优化研究 |
| |
| 引用本文： | 吕延军,李杰,强程,李鹏洲,张永芳,常欢. 珩磨缸套表面粗糙度预测及多目标优化研究[J]. 摩擦学学报, 2022, 42(4): 728-741. DOI: 10.16078/j.tribology.2021131 |
| |
| 作者姓名： | 吕延军 李杰 强程 李鹏洲 张永芳 常欢 |
| |
| 作者单位： | 1.西安理工大学 机械与精密仪器工程学院, 陕西 西安 710048 |
| |
| 基金项目： | 国家自然科学基金(51775428, 52075438)、陕西省重点研发计划(2020GY-106)和机械制造系统工程国家重点实验室开放课题(sklms2020010)资助. |
| |
| 摘 要： | 为了对粗珩阶段缸套内孔表面粗糙度Rk粗糙度集中的Rk、Rpk和Rvk进行预测，进而对粗珩加工参数进行优化，以珩磨压力(P)、珩磨头旋转速度(VR)和往复速度(VRe)为决定因素，Rk粗糙度集为目标响应，进行多目标优化.建立基于广义回归神经网络(Generalized regression neural network, GRNN)与响应曲面法(Response surface methodology,RSM)的粗糙度预测模型，并采用三因素三水平的全因子珩磨试验进行验证，结果表明所建立模型的预测结果与试验结果具有很好的一致性. GRNN预测模型决定系数R2的均值为0.959,RSM多元回归预测模型决定系数R2的均值为0.963，与RSM所建立的多元回归预测模型相比，GRNN预测模型在预测Rk和Rpk时，预测精度更高，预测误差更小，R2分别提高了0.025和0.020，在预测Rvk时RSM多元回归模型更优，R2提高了0.057.进一步结合响应曲面法分析了3个决定因素对粗糙度的影响显著性并进行了排序，对于...
|
| 关 键 词： | 珩磨加工 粗糙度预测 神经网络 响应曲面法 多目标优化 |
| 收稿时间： | 2021-06-23 |
Surface Roughness Prediction and Multi-Objective Optimization of Honing Cylinder Liner |
| |
| Affiliation: | 1.School of Mechanical and Precision Instrument Engineering, Xi’an University of Technology, Shaanxi Xi’an 710048, China2.State Key Laboratory for Manufacturing Systems Engineering, Xi’an Jiaotong University, Shaanxi Xi’an 710049, China3.School of Printing, Packaging Engineering and Digital Media Technology, Xi’an University of Technology, Shaanxi Xi’an 710054, China |
| |
| Abstract: | The surface machining quality of cylinder liner is usually evaluated by surface roughness Rk set, it is of great significance to establish a prediction model between roughness Rk set and machining parameters, and analyze the influence of machining parameters on honing quality of cylinder liner. Based on the rough honing experiments on cylinder liner under different honing pressure (P), different honing head rotation speed (VR) and different reciprocating speed (VRe), the values of the surface roughness Rk set were obtained. In order to predict the Rk, Rpk and Rvk of the surface roughness Rk set of the inner bore of cylinder liner during the rough honing process, and to optimize the honing parameters, the P, VR and VRe were taken as decision variables, and Rk roughness set was taken as the object response, multi-objective optimization was implemented to obtain the optimal machining parameters of P, VR and VRe. The roughness prediction models based on generalized regression neural network (GRNN) and response surface methodology (RSM) were established respectively. For the GRNN method, it was necessary to use the K-fold cross validation method to optimize the smoothing factor (i.e., spread), the value of the smoothing factor can greatly affect the prediction performance of the network. The minimum MSE can be obtained in the process of five-fold cross validation, the smoothing factor was 0.01 when the MSE had the minimum value. Twenty-one groups of experiment data were trained to achieve the prediction model by GRNN, and other 6 groups were regarded as test samples to verify the accuracy of the proposed model. The Box-Behnken Design (BBD) method of response surface method was used to design the experiment, and 17 groups of experiments were conducted to establish the response relationship between the honing parameters and the surface roughness Rk set. The surface roughness Rk set can be predicted by the optimized regression models based on RSM. And comparing with a full factorial honing test with three factors and three levels, the result showed that the prediction results obtained by the proposed models were in good agreement with the experimental results. The mean value of R2 of the coefficient of determination of GRNN prediction model was 0.959, the mean value of R2 of the coefficient of determination of the multiple regression prediction model established by RSM was 0.963. Comparing with the multiple regression prediction model established by RSM, the GRNN prediction model had higher prediction accuracy and smaller prediction error for prediction of Rk and Rpk, and R2 of Rk and Rpk increased by 0.025 and 0.020 respectively. The multiple regression model established by RSM was better than the GRNN prediction model when predicting Rvk, and R2 increased by 0.057. The influence significance of three decision variables on the roughness was analyzed by the RSM, and the ranking result for Rk was VRe > P > VR, for Rpk was P > VRe > VR, for Rvk was P > VRe > VR.In order to select the appropriate values of P, VR and VRe to make the Rk roughness set relatively small, based on the regression model established by the RSM, a multi-objective optimization was implemented to obtain the Pareto front of the Pareto optimal solutions by combining the multiple regression models with NSGA-II (non-dominated Sorting Genetic Algorithm II) optimization algorithm. More machining parameters combinations for rough honing of cylinder liner were provided to get high honing quality. Eight groups of optimum roughness values and the corresponding rough honing parameters of cylinder liner were given. |
| |
| Keywords: | |
|
| 点击此处可从《摩擦学学报》浏览原始摘要信息 |
|
点击此处可从《摩擦学学报》下载全文 |
|
