基于神经网络粉末冶金摆线转子成形模具磨损分析与优化  

Wear Analysis and Optimization of Powder Metallurgy Cycloidal Rotor Molding Die Based on Neural Network

作　　者：王兴 林育阳 贺利乐[1] 贠军辉 吕哲 WANG Xing;LIN Yuyang;HE Lile;YUN Junhui;LYU Zhe(College of Mechanical and Electrical Engineering,Xi'an University of Architecture&Technology,Xi'an 710055,China;Shaanxi Provincial Machinery Research Institute,Xianyang 712000,China;Xian XD Switchgear Electric Co.,Ltd.,Xi'an 710077,China)

机构地区：[1]西安建筑科技大学机电工程学院,陕西西安710055 [2]陕西省机械研究院,陕西咸阳712000 [3]西安西电开关电气有限公司,陕西西安710077

出　　处：《热加工工艺》2022年第16期44-48,共5页Hot Working Technology

基　　金：陕西研发计划——一般项目(2020GY-245)。

摘　　要：在粉末压制成形过程中,模具磨损分析对保证产品质量与精度有着不可忽视的作用。以铁基材料摆线转子为例,基于Archard磨损模型和BP神经网络,结合有限元模拟,研究了成形过程中工艺参数对磨损量的影响规律。首先采用粉末压制试验获得铁基材料的弹性模量、泊松比与密度之间的映射关系;选择摩擦系数、初始硬度、下压速度、压制方式作为工艺参数,构建3因素4水平以及1因素2水平的正交试验。其次基于Deform-3D有限元软件实现压制过程的数值模拟,将4种不同工艺参数作为输入,模具的磨损量作为输出,构建4×13×1的3层BP神经网络模型。最终结合遗传算法在各参数取值范围中完成迭代优化获取最佳工艺参数组合。结论表明,神经网络的预测值与实际值相对误差RE的平均值仅为4.45%,最佳工艺参数组合为:摩擦系数0.125、初始硬度63 HRC、下压速度1.558 mm/s、压制方式为双向压制,获得最小磨损量为1.5521×10^(-5)mm,提升了模具的使用寿命。In the process of powder compaction, the die wear analysis plays an important role in ensuring the quality and precision of products. Based on the Archard wear model, BP neural network and finite element simulation, the influences of process parameters on the wear of iron based cycloidal rotor were studied. Firstly, the mapping relationship among Poisson’s ratio, elastic modulus and density of iron base material was obtained by powder compaction test. The friction coefficient,initial hardness, lowering speed and pressing mode were selected as technological parameters to construct the orthogonal tests with three factors and four levels and one factor and two levels. Secondly, the numerical simulation of the pressing process was realized based on the finite element software DEFORM-3D. Four different process parameters were taken as the input and the wear of the die as the output, and a three-layer BP neural network model of 4 ×13 ×1 was built. Finally, the genetic algorithm was combined to complete the iterative optimization in the range of parameter values to obtain the best combination of technological parameters. The results show that the average relative error between the predicted value and the actual value of the neural network is only 4.45%. The optimal process parameters are as follows: friction coefficient 0.125, initial hardness63 HRC, press speed 1.558 mm/s. The pressing method is two-way pressing, and the minimum wear quantity is 1.5521 ×10-5mm, which can improve the service life of the die.

关 键 词：铁基材料 正交试验 神经网络 磨损 数值模拟 

分 类 号：TG141[一般工业技术—材料科学与工程] TG305[金属学及工艺—金属材料]