潘杰,陈凡,杨炜,金闻达.基于SPSO–BP神经网络的自适应抛光工艺参数匹配[J].表面技术,2022,51(8):387-399. PAN Jie,CHEN Fan,YANG Wei,JIN Wen-da.Adaptive Polishing Process Parameter Matching Based on SPSO-BP Neural Network[J].Surface Technology,2022,51(8):387-399 |
基于SPSO–BP神经网络的自适应抛光工艺参数匹配 |
Adaptive Polishing Process Parameter Matching Based on SPSO-BP Neural Network |
|
DOI:10.16490/j.cnki.issn.1001-3660.2022.08.035 |
中文关键词: 抛光 材料去除率 表面粗糙度 SPSO 神经网络 预测模型 工艺参数 自适应 |
英文关键词:polish material removal rate surface roughness SPSO neural networks predictive model process parameters self-adaptive |
基金项目:国家重点研发计划(2018YFB1308700) |
|
Author | Institution |
PAN Jie | HUST-Wuxi Research Institute, Jiangsu Wuxi 214174, China |
CHEN Fan | HUST-Wuxi Research Institute, Jiangsu Wuxi 214174, China;School of Mechanical Science & Engineering, Huazhong University of Science and Technology, Wuhan 430074, China |
YANG Wei | Jiangsu Jitri-Hust Intelligent Equipment Technology Co., Ltd., Jiangsu Wuxi 214174, China |
JIN Wen-da | Jiangsu Jitri-Hust Intelligent Equipment Technology Co., Ltd., Jiangsu Wuxi 214174, China |
|
摘要点击次数: |
全文下载次数: |
中文摘要: |
目的 在湿性物理抛光作业中,根据不同工件的表面抛光质量和效率要求,实现抛光工艺参数的自适应匹配,达到理想的抛光效果。方法 基于工件表面材料去除原理,建立工艺参数与材料去除率(MRR)和表面粗糙度的数学关系模型,明确影响抛光效果的工艺参数。针对工艺参数与抛光质量和效率之间的复杂且交互影响的关系,以及理论计算的抛光效果与实际结果存在差异的问题,提出SPSO–BP预测模型,分别以20组不同的抛光工艺参数与对应抛光结果为训练样本,训练SPSO–BP模型,并与传统PSO–BP模型进行对比。基于训练好的预测模型,根据不同的基础条件与抛光质量和抛光效率的要求,通过模型自适应匹配抛光工艺参数。针对SUS304板材,设定表面粗糙度目标Ra1—Ra5和材料去除率目标Rm1—Rm5,分别通过SPSO–BP和PSO–BP模型预测获得的工艺参数进行抛光试验,将获得的真实粗糙度Raz1—Raz5和材料去除率Rmz1—Rmz5与目标值进行对比验证。结果 SPSO–BP预测模型比PSO–BP预测模型具有更高的收敛精度,SPSO–BP和PSO–BP预测模型的收敛精度分别为1.26×10−6、0.180,并且SPSO–BP模型对样本具有较好的跟踪能力和泛化能力。以SPSO–BP模型预测的工艺参数进行抛光,获得的真实粗糙度Raz和真实材料去除率Rmz,相较于PSO–BP预测模型与目标值更接近。通过SPSO–BP和PSO–BP预测模型获得的真实粗糙度值Raz与目标值Ra的最大误差比分别为8.00%和20.00%,平均误差比分别为5.77%和14.07%,最小误差比分别为2.50%和10.00%;真实材料去除率Rmz与目标值Rm的最大误差比分别为3.00%和8.57%,平均误差比分别为2.14%和7.46%,最小误差比分别为1.11%和4.38%。结论 根据不同的基础条件及抛光质量和抛光效率要求,可以通过SPSO–BP预测模型自适应匹配抛光工艺参数,与传统PSO–BP预测模型相比具有更高的收敛精度,可以获得与抛光目标更接近的真实抛光效果。 |
英文摘要: |
With the development of science and technology, the requirements for the surface roughness value and precision polishing efficiency of key parts in the fields of aviation, aerospace, national defense, and medical treatment were getting stricter. The wet physical polishing method can reduce the deformation of the material during the polishing process and obtain a lower surface roughness value. When testing the polishing process parameters, it was necessary to manually select the polishing process parameters, observe the polishing results, and repeatedly adjust the process parameters based on experience to achieve the desired polishing effect. The test process required a lot of time and energy, relying on people's subjective experience to adjust the parameters, the accumulated knowledge and experience were difficult to transfer among different operators. The surface roughness and material removal rate are usually measured after the parts are polished, when the test does not meet the requirements, it often leads to scrapped parts. This paper aims to achieve the self-adaptive matching of polishing parameters according to the requirements of different workpiece surface polishing quality and efficiency, and endeavors to achieve the ideal polishing effect. Based on the principle of material removal on the surface of the workpiece, this paper established a mathematical model of the relationship between process parameters, material removal rate and surface roughness value, and the process parameters that affected the polishing effect was clarified. Aimed at the complex and interactive relationship between process parameters and polishing quality and efficiency, as well as the difference between the theoretically calculated polishing effect and the actual result, the SPSO-BP prediction model was proposed. 20 sets of different polishing process parameters and corresponding polishing results were taken as training samples. The SPSO-BP model was trained with the samples and compared with the traditional PSO-BP model. Based on the trained prediction model, the polishing process parameters are adaptively matched through the model according to different basic conditions, polishing quality and polishing efficiency requirements. For SUS304 plates, the surface roughness value targets Ra1-Ra5 and the material removal rate targets Rm1-Rm5 were set. Moreover, the process parameters in the SPSO-BP and PSO-BP models were predicted, then polishing test was performed. The actual roughness values Raz1-Raz5 and the material removal rate Rmz1-Rmz5 were achieved, compared and verified with the target values. Compared with the PSO-BP prediction model, the SPSO-BP prediction model had higher convergence accuracy. The convergence accuracy of the SPSO-BP and PSO-BP were 1.26×10−6 and 0.180 respectively, and the SPSO-BP model has good tracking ability and generalization ability for samples. The real roughness value Raz and the real material removal rate Rmz obtained by the SPSO-BP prediction model were closer to the target value than the PSO-BP prediction model. The maximum error ratios of the real roughness value Raz and the target value Ra obtained by the SPSO-BP and PSO-BP prediction models were 8.00% and 20.00%, the average error ratios were 5.77% and 14.07%, and the minimum error ratios were 2.50% and 10.00%; the maximum error ratios of the true material removal rate Rmz and the target value Rm are 3.00% and 8.57%, the average error ratios were 2.14% and 7.46%, and the minimum error ratios were:1.11% and 4.38%. According to different basic conditions, polishing quality and polishing efficiency requirements, the SPSO-BP prediction model can be used to adaptively match the polishing process parameters. In comparison with the traditional PSO-BP prediction model, it had higher convergence accuracy, which can achieve a more closer real polishing result to the target requirement. |
查看全文 查看/发表评论 下载PDF阅读器 |
关闭 |
|
|
|