引用本文:李昌龙,陈松,吴炫炫,赵耀耀,李雨龙,李鑫.基于PSO-ELM的316L不锈钢细长管磁粒研磨内表面粗糙度预测模型[J].中国表面工程,2023,36(2):212~221
LI Changlong,CHEN Song,WU Xuanxuan,ZHAO Yaoyao,LI Yulong,LI Xin.Inner Surface Roughness Prediction Model of 316L Stainless Steel Slender Tube by Magnetic Abrasive Finishing Based on PSO-ELM[J].China Surface Engineering,2023,36(2):212~221
【打印本页】   【HTML】   【下载PDF全文】   查看/发表评论  【EndNote】   【RefMan】   【BibTex】
←前一篇|后一篇→ 过刊浏览    高级检索
本文已被:浏览 712次   下载 505 本文二维码信息
码上扫一扫!
分享到: 微信 更多
基于PSO-ELM的316L不锈钢细长管磁粒研磨内表面粗糙度预测模型
李昌龙, 陈松, 吴炫炫, 赵耀耀, 李雨龙, 李鑫
辽宁科技大学机械工程与自动化学院 鞍山 114051
摘要:
针对 316L 不锈钢细长管磁粒研磨加工过程中,最佳工艺参数难以选择,以及加工后对工件内表面粗糙度(Ra)的预测问题,将影响磁粒研磨 316L 不锈钢细长管内表面粗糙度的四个工艺参数作为输入值,内表面粗糙度作为输出值,构建粒子群(PSO)优化极限学习机(ELM)模型来预测 316L 不锈钢细长管内表面粗糙度,利用 PSO 对工艺参数进行全局寻优, 获得最佳工艺参数组合,最后通过试验与预测结果进行对比。构建的 PSO-ELM 表面粗糙度预测模型拟合优度 R2为 0.984 8, 绝对误差(MAE)为 0.013 4,均方根误差(RMSE)为 0.021 4。得到的最佳工艺参数组合为:主轴转速 2 389.011 r / min, 进给速度 3.167 mm / s,磨料粒径 216.185 μm,加工时间 35.856 min,预测 Ra 为 0.178 μm。对工艺参数进行调整,试验得到的 Ra 为 0.182 μm,与预测值相比误差为 2.24%。基于 PSO-ELM 方法构建 316L 不锈钢细长管内表面粗糙度预测模型,实现对工件内表面粗糙度的精确预测,应用粒子群方法得到最佳工艺参数组合,提高了磁粒研磨 316L 不锈钢细长管的加工效率。 关键词:磁粒研磨;细长管;内表面;粒子群(PSO);极限学习机(ELM);表面粗糙度
关键词:  磁粒研磨  细长管  内表面  粒子群(PSO)  极限学习机(ELM)  表面粗糙度
DOI:10.11933/j.issn.1007?9289.20220513001
分类号:TG176
基金项目:国家自然科学基金(51775258)、辽宁省教育厅(2020FWDF05,2020FWDF07)和辽宁科技大学基金(2018FW05)资助项目
Inner Surface Roughness Prediction Model of 316L Stainless Steel Slender Tube by Magnetic Abrasive Finishing Based on PSO-ELM
LI Changlong, CHEN Song, WU Xuanxuan, ZHAO Yaoyao, LI Yulong, LI Xin
College of Mechanical Engineering and Automation, University of Science and Technology Liaoning,Anshan 114051 , China
Abstract:
In the magnetic abrasive finishing processing of a 316L stainless steel slender tube, it is very important to accurately estimate the surface roughness of the workpiece after finishing under the combination of different process parameters. Several experiments and empirical methods are used to determine the improved surface quality and the processing efficiency of the workpiece to solve for the best process parameter combination, but these ways are inefficient and inaccurate. Considering the rotation speed of the workpiece, feeding speed of the magnetic pole, magnetic abrasive powder size, and processing time as the input values, the inner surface roughness Ra can be obtained using a combination of different process parameters as the output value, and an orthogonal experiment with four factors and levels is designed. The test results are recorded based on the retention method. The weight of the link between the input and hidden layers and the threshold of the hidden layer in the extreme learning machine (ELM) are optimized by the particle swarm optimization (PSO) algorithm to improve the prediction accuracy of the ELM model. Based on the orthogonal test data, a PSO-ELM magnetic abrasive finishing 316L stainless steel slender tube inner-surface roughness prediction model is established. To verify the superiority of the PSO-ELM model, two types of surface roughness prediction models are established using the multivariate nonlinear regression method and the support vector machine (SVM), which are compared with the PSO-ELM surface roughness model. The predictive models are evaluated using machine-learning regression evaluation metrics. Then, the prediction model constructed by PSO-ELM is used as the objective function of the particle swarm optimization algorithm, and again the particle swarm optimization algorithm with the ability of global optimization is used again to optimize the process parameters. Therefore, the best combination of process parameters for the magnetic abrasive finishing of a 316L stainless steel slender tube is obtained. The test is performed using a combination of process parameters obtained after optimization, and the results obtained after the test are compared with the predicted results. The model’s accuracy is evaluated by the evaluation index of machine learning performance, and the constructed PSO-ELM surface roughness prediction model has a high prediction accuracy and small error. The model’s goodness-of-fit R2 is 0.984 8, mean absolute error (MAE) is 0.013 4, and root mean square error (RMSE) is 0.021 4. The optimal combination of process parameters obtained using the particle swarm optimization algorithm is as follows: the speed of the workpiece is 2 389.011 r / min, the feed speed of the magnetic pole is 3.167 mm / s, the abrasive particle size is 216.185 μm, and the processing time is 35.856 min. The surface roughness predicted by the optimal combination of process parameters is 0.178 μm. The optimal combination of process parameters must be rounded-off and converted into a standard form. After the test, according to the fine-tuned process parameters, the surface roughness Ra of the obtained workpiece is 0.182 μm, which error with the predicted value is 2.24 %. Based on the PSO-ELM method, a prediction model of the inner surface roughness of a 316L stainless steel slender pipe is constructed, which realizes the controllability of the accurate prediction of the inner surface roughness of the workpiece and uses the particle swarm optimization ability to obtain the best process parameter combination, which improves the magnetic abrasive finishing efficiency of the 316L stainless steel slender tubes.
Key words:  magnetic abrasive finishing  slender tube  inner surface  particle swarm optimization(PSO)  extreme learning machine (ELM)  surface roughness
手机扫一扫看