押海龙,贺利乐,林育阳,等.基于神经网络的粉末冶金摆线外转子整形模具磨损和应力分析[J].精密成形工程,2023,15(6):86-93.
YA Hai-long,HE Li-le,LIN Yu-yang,et al.Wear and Stress Analysis of Powder Metallurgy Cycloid Outer Rotor Shaping Die Based on Neural Network[J].Journal of Netshape Forming Engineering,2023,15(6):86-93. |
| 基于神经网络的粉末冶金摆线外转子整形模具磨损和应力分析 |
| Wear and Stress Analysis of Powder Metallurgy Cycloid Outer Rotor Shaping Die Based on Neural Network |
| |
| DOI：10.3969/j.issn.1674-6457.2023.06.011 |
| 中文关键词: 粉末冶金摆线外转子 模具磨损 模具应力 神经网络 多目标粒子群算法 |
| 英文关键词: powder metallurgy cycloid outer rotor die wear die stress neural network multi-objective particle swarm optimization algorithm |
| 基金项目:陕西研发计划(2020GY–245) |
|
| 摘要点击次数: 1602 |
| 全文下载次数: 847 |
| 中文摘要: |
| 目的 解决粉末冶金制品生产过程中模具磨损和应力集中对产品质量影响较大的问题。方法 以粉末冶金摆线外转子整形模具为研究对象，基于Archard磨损模型和有限元模拟方法，分析整形过程中阴模的磨损和所受应力情况，确定模具下压速度、摩擦因数、模具初始硬度和阴模圆角半径4个影响因素并设计正交实验，以仿真数据为样本，采用BP神经网络建立模具磨损量和应力的预测模型，结合多目标粒子群算法进行参数优化。结果 基于神经网络的模具磨损深度和应力预测值与模拟值之间的平均相对误差仅为3.01%和3.34%、最大相对误差为4.00%和6.16%，均小于10%，说明所构建的BP神经网络模型预测效果较好。获得最佳参数组合如下:模具下压速度为1.95 mm/s，摩擦因数为0.122，模具初始硬度为59.63HRC，阴模圆角半径为4.85 mm。此时模具的磨损深度为1.713´10−5 mm，模具应力为2 036 MPa，与数值模拟结果的相对误差仅为3.19%和4.13%，均小于5%，说明多目标粒子群算法参数优化的准确性较高。结论 通过实验验证发现模具实际磨损量和粒子群算法优化结果的误差仅为2.95%，表明优化后的参数可以有效改善模具磨损和应力集中情况，并提高模具使用寿命。 |
| 英文摘要: |
| The work aims to solve the problem that die wear and stress concentration have a great effect on product quality in the production process of powder metallurgy products. With the powder metallurgy cycloid outer rotor shaping die as the research object, the wear and stress of the female die during the shaping process were analyzed based on the Archard wear model and finite element simulation method. Then, the four affecting factors, such as die pressing speed, friction coefficient, initial hardness of die and fillet radius of female die, were determined and orthogonal experiments were designed. With the simulation data as the sample, the BP neural network was used to establish the prediction model of die wear and stress, and the multi-objective particle swarm optimization algorithm was used to optimize the parameters. The average relative errors between the predicted values and the simulated values of the die wear depth and stress based on the neural network were only 3.01% and 3.34%, and the maximum relative errors were 4.00% and 6.16%, which were less than 10%, indicating that the BP neural network model had a good prediction effect. The optimal parameter combination was obtained:the die pressing speed was 1.95 mm/s, the friction coefficient was 0.122, the initial hardness of the die was 59.63HRC, and the radius of the die fillet was 4.85 mm. At this time, the wear depth of the die was 1.713´10−5 mm, and the die stress was 2 036 MPa, having relative errors of only 3.19% and 4.13% with numerical simulation results, which were both less than 5%, indicating that the multi-objective particle swarm optimization algorithm had high accuracy of parameter optimization. Through experimental verification, it is found that the error between the actual wear amount of the die and the optimization result of the particle swarm optimization algorithm is only 2.95%, indicating that the optimized parameters can effectively reduce the wear of the die and improve the stress concentration of the die. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|
渝公网安备 50010702501719号