张翔,王应刚,陈泓谕,杭伟,曹霖霖,邓辉,袁巨龙.基于BP神经网络与遗传算法的固结磨具制作工艺参数优化[J].表面技术,2022,51(2):358-366.
ZHANG Xiang,WANG Ying-gang,CHEN Hong-yu,HANG Wei,CAO Lin-lin,DENG Hui,YUAN Ju-long.#$NP Optimization of Fixed-abrasive Tool Development Parameters Based on BP Neural Network and Genetic Algorithm[J].Surface Technology,2022,51(2):358-366
基于BP神经网络与遗传算法的固结磨具制作工艺参数优化
#$NP Optimization of Fixed-abrasive Tool Development Parameters Based on BP Neural Network and Genetic Algorithm
投稿时间:2021-11-02  修订日期:2022-01-04
DOI:10.16490/j.cnki.issn.1001-3660.2022.02.036
中文关键词:  固结磨具  蓝宝石  正交试验  BP神经网络  遗传算法
英文关键词:fixed-abrasive tool  sapphire  orthogonal experiment  BP neural network  genetic algorithm
基金项目:
作者单位
张翔 浙江工业大学 特种装备制造与先进加工技术教育部重点试验室,杭州 310014 
王应刚 浙江工业大学 特种装备制造与先进加工技术教育部重点试验室,杭州 310014 
陈泓谕 浙江工业大学 特种装备制造与先进加工技术教育部重点试验室,杭州 310014 
杭伟 浙江工业大学 特种装备制造与先进加工技术教育部重点试验室,杭州 310014 
曹霖霖 北华大学 机械工程学院,吉林 吉林 132013 
邓辉 南方科技大学 工学院,广东 深圳 518055 
袁巨龙 浙江工业大学 特种装备制造与先进加工技术教育部重点试验室,杭州 310014 
AuthorInstitution
ZHANG Xiang Key Laboratory of Special Equipment Manufacturing and Advanced Processing Technology of Ministry of Education, Zhejiang University of Technology, Hangzhou 310014, China 
WANG Ying-gang Key Laboratory of Special Equipment Manufacturing and Advanced Processing Technology of Ministry of Education, Zhejiang University of Technology, Hangzhou 310014, China 
CHEN Hong-yu Key Laboratory of Special Equipment Manufacturing and Advanced Processing Technology of Ministry of Education, Zhejiang University of Technology, Hangzhou 310014, China 
HANG Wei Key Laboratory of Special Equipment Manufacturing and Advanced Processing Technology of Ministry of Education, Zhejiang University of Technology, Hangzhou 310014, China 
CAO Lin-lin Mechanical Engineering, Beihua University, Jilin 132013, China 
DENG Hui College of Engineering, Southern University of Science and Technology, Shenzhen 518055, China 
YUAN Ju-long Key Laboratory of Special Equipment Manufacturing and Advanced Processing Technology of Ministry of Education, Zhejiang University of Technology, Hangzhou 310014, China 
摘要点击次数:
全文下载次数:
中文摘要:
      目的 利用BP神经网络技术与遗传算法寻找固结磨具制作最优工艺参数组合,实现固结磨具制作工艺参数的快速寻优。方法 设计磨粒粒径、磨粒质量分数、成型压力、烧结温度的正交工艺参数表,按正交表工艺参数制作蓝宝石晶片加工用的Cr2O3固结磨具,并且设计不同固化温度下制作的固结磨具的硬度与抗压强度测试试验,验证自制的固结磨具加工的有效性以及固化温度选择的合理性。使用自制的Cr2O3固结磨具在抛光机上进行加工试验,测量蓝宝石晶片的去除率和Cr2O3丸片的磨削比。综合考虑丸片的磨削效率与使用经济性,将去除率与磨削比采用min-max方法归一化后,乘对应的权重值并相加,得到丸片综合评分Y,作为丸片的评价标准。以磨粒粒径、磨粒质量分数、成型压力、烧结温度为输入,综合评分Y为输出,建立基于神经网络的丸片制作工艺参数与丸片综合评分Y之间的BP神经网络预测模型,并使用决定系数R2评价BP神经网络的训练结果。最后,设计初始化种群个体N、交叉概率Pc、变异概率Pm的正交试验表,基于构建的神经网络,根据正交试验表,使用遗传算法进行制作工艺参数的全局寻优。依据寻优结果制作丸片并进行试验,计算综合评分Y,与预测评分对比。结果 构建了4个输入层神经元、12个隐含层神经元、1个输出层神经元的三层BP神经网络。构建的BP神经网络的决定系数R2为0.9313,丸片综合评分Y的预测值与实际值的误差在4%以下,满足工程的实际应用。在给定的工艺参数范围内,在参数组合为初始化种群个体N为80、交叉概率Pc为0.6、变异概率Pm为0.06的条件下,使用遗传算法寻优得到的蓝宝石加工用Cr2O3固结磨具最优的制作工艺参数组合为:磨粒粒径10 μm,磨粒质量分数88%,成型压力80 MPa,成型温度174 ℃。丸片综合评分Y的寻优值为94.09,试验得到的丸片平均综合评分Y为89.87,与寻优值的误差为5%。结论 BP神经网络可以有效建立固结磨具制作工艺参数与丸片综合性能的预测模型。神经网络结合遗传算法寻优,可以为固结模具制作工艺参数组合的优化选择提供指导意义。
英文摘要:
      The work aims to find the optimal combination of fixed-abrasive tool pellet development parameters, and achieve fast optimization of development parameters for fixed-abrasive tool pellets. The Cr2O3 pellet was used to test on the sapphire workpiece in polishing machine. The pellet was developed according to the orthogonal parameter scheme, which was designed according to particle size, mass fraction, molding pressure, and sintering temperature. And designed the experiment to test the hardness and compressive strength of the fixed-abrasive tool pellets which were developed according to the different sintering temperature. The test results verified the validity of the self-made fixed-abrasive tool pellets and the rationality of sintering temperature selection. The removal rate of sapphire wafers and the grinding ratio of Cr2O3 pellets were measured. Considering the grinding efficiency and economy of the pellet, the removal rate and grinding ratio were normalized by min-max method, and the weight values were multiplied by corresponding weight and added together to obtain the comprehensive score Y, which was used as the evaluation standard of the pellet. With particle size, mass fraction, molding pressure, sintering temperature as input values and comprehensive score Y as output values, a prediction model of pellet comprehensive score Y based on neural network was established. The training results of the BP neural network was evaluated by the coefficient of determination R2. Designed the orthogonal parameter scheme of initial population N, crossover probability Pc and mutation probability Pm. According to the orthogonal parameter scheme, genetic algorithm was used to optimize the global process parameters based on the BP neural network. The genetic algorithm was used to optimize the global manufacturing process parameters According to the optimization results, the pellet is developed and tested on the polishing machine. Then calculated comprehensive score and compared such score with the neural network prediction score. A three-layer BP neural network with 4 input layer neurons, 12 hidden layer neurons and 1 output layer neuron was constructed. The determined coefficient R2 of the constructed BP neural network is 0.9313, and the error between the predicted value of the comprehensive score Y and the actual value is less than 4%, which could meet the practical application of the project. Within the given range of development parameters, under the condition that the genetic algorithm parameter combination is the initial population individual N is 80, crossover probability Pc is 0.6, mutation probability Pm is 0.06, the optimal manufacturing development parameter combination of Cr2O3 fixed-abrasive tool for sapphire polishing obtained by genetic algorithm optimization is:abrasive grain size 10 μm, abrasive grain mass fraction 88%, molding pressure 80 MPa, molding temperature 174 ℃, the optimal value of the comprehensive score Y of pellet is 94.09. The average comprehensive score obtained by the experiment is 89.87, and the error is 5% compared with the optimal value. BP neural network can effectively establish a prediction model between the development parameters and processing quality of the abrasive-fixed tool pellets. Neural network combined with genetic algorithm optimization can provide guiding significance for the optimal selection of the development parameter combination of abrasive-fixed tool.
查看全文  查看/发表评论  下载PDF阅读器
关闭

关于我们 | 联系我们 | 投诉建议 | 隐私保护 | 用户协议

您是第19495792位访问者    渝ICP备15012534号-3

版权所有:《表面技术》编辑部 2014 surface-techj.com, All Rights Reserved

邮编:400039 电话:023-68792193传真:023-68792396 Email: bmjs@surface-techj.com

渝公网安备 50010702501715号