|
| 面向壁厚均匀性的拉深成形TRB板形设计及神经网络预测模型 |
| Neural Network Prediction Model of Deep Drawing TRB Sheet for Wall Thickness Uniformity |
| |
| DOI：10.3969/j.issn.1674-6457.2023.06.005 |
| 中文关键词: 圆筒件 壁厚 TRB 正交试验 神经网络 |
| 英文关键词: cylinder wall thickness TRB orthogonal test neural network |
| 基金项目:重庆市自然科学基金面上项目(CSTB2022NSCQ–MSX1444) |
|
| Hits: 1409 |
| Download times: 794 |
| 中文摘要: |
| 目的 研究连续变截面板拉深成形过程中圆筒件的壁厚均匀性问题。方法 基于正交试验设计和极差分析方法，研究连续变截面板结构参数对圆筒件拉深成形壁厚均匀性的影响，并结合SSA–BP神经网络预测并验证连续变截面板成形的壁厚均匀性。结果 通过正交试验及极差分析，得到连续变截面板各结构参数按对最大壁厚影响由大到小的顺序依次为薄区厚度＞厚区厚度＞左侧过渡区长度＞右侧过渡区长度;按对最小壁厚影响由大到小的顺序依次为薄区厚度＞厚区厚度＞右侧过渡区长度＞左侧过渡区长度;按对最大壁厚差影响由大到小的顺序依次为薄区厚度＞左侧过渡区长度＞厚区厚度＞右侧过渡区长度。综合考虑最大壁厚、最小壁厚及最大壁厚差，得到最优参数组合如下:厚区厚度为1.1 mm，薄区厚度为0.8 mm，左侧过渡区长度为2.5 mm，右侧过渡区长度为29.5 mm。基于正交试验分析结果建立的SSA–BP神经网络模型具有良好的预测能力，正交试验外5组数据的预测值与真实仿真值的最大误差均在11%以下。结论 基于正交试验分析结果建立的SSA–BP神经网络模型能够实现对TRB板圆筒件拉深成形壁厚的准确推测，该方法具有良好的工程应用价值。 |
| 英文摘要: |
| The work aims to study the wall thickness uniformity of cylinders in deep drawing of continuous variable cross-section sheets. Based on orthogonal test design and range analysis, the effect of structural parameters of continuous variable cross-section sheets on the wall thickness uniformity of cylinder deep drawing was studied, and SSA-BP neural network was used to predict and verify the wall thickness uniformity of continuous variable cross-section sheets. Through orthogonal test and range analysis, it was found that the structural parameters of the continuously variable cross-section sheet, in descending order of their effect on the maximum wall thickness, were as follows:thin zone thickness>thick zone thickness>left transition zone length>right transition zone length; in descending order of their effect on the minimum wall thickness, were as follows:thin zone thickness>thick zone thickness>right transition zone length>left transition zone length; in descending order of their effect on the maximum wall thickness difference, were as follows:thin zone thickness>left transition zone length>thick zone thickness>right transition zone length. Considering the maximum wall thickness, minimum wall thickness, and maximum wall thickness difference, the optimal parameter combination was as follows:the thickness of thick area was 1.1 mm, the thickness of thin area was 0.8 mm, the length of left transition area was 2.5 mm, and the length of right transition area was 29.5 mm. The SSA-BP neural network model established based on the analysis results of orthogonal test had good prediction performance, and the maximum error between the predicted values and the real simulation values of five groups of data outside the orthogonal test was less than 11%. The SSA-BP neural network model established based on the results of orthogonal test analysis can accurately predict the wall thickness of TRB sheet cylinder deep drawing, so this method has high value in engineering application. |
|
View Full Text
View/Add Comment Download reader |
| Close |
渝公网安备 50010702501719号