基于全连接卷积神经网络的铸造缩松缩孔缺陷快速预测

张建明; 廖敦明; 孙飞

doi:10.15980/j.tzzz.2020.08.007

计算机应用技术 | 浏览量 : 0 下载量: 107 CSCD: 0

导出
分享
收藏
专辑

基于全连接卷积神经网络的铸造缩松缩孔缺陷快速预测
Fast Prediction Method of Casting Shrinkage Defects Based on Fully Convolutional Networks
2020年40卷第8期页码：841-845
纸质出版日期： 2020 ，
DOI： 10.15980/j.tzzz.2020.08.007

扫描看全文

[1]张建明,廖敦明,孙飞.基于全连接卷积神经网络的铸造缩松缩孔缺陷快速预测[J].特种铸造及有色合金,2020,40(08):841-845.

Zhang Jianming, Liao Dunming, Sun Fei. Fast Prediction Method of Casting Shrinkage Defects Based on Fully Convolutional Networks[J]. Special Casting & Nonferrous Alloys, 2020,40(8):841-845.
[1]张建明,廖敦明,孙飞.基于全连接卷积神经网络的铸造缩松缩孔缺陷快速预测[J].特种铸造及有色合金,2020,40(08):841-845. DOI： 10.15980/j.tzzz.2020.08.007.

Zhang Jianming, Liao Dunming, Sun Fei. Fast Prediction Method of Casting Shrinkage Defects Based on Fully Convolutional Networks[J]. Special Casting & Nonferrous Alloys, 2020,40(8):841-845. DOI： 10.15980/j.tzzz.2020.08.007.

摘要

在预测铸造缩孔、缩松缺陷时引入机器学习

采用具有针对性的数据预处理方式处理铸造工艺数据

同时采取多模态的数据输入方式增加数据维度

并使用全连接卷积神经网络对铸造缩孔、缩松缺陷进行快速计算。通过对某轴承座铸件进行分析

验证了该方法的有效性。相比于数值模拟方法

该方法具备较高的精度

同时可大幅缩短铸造缩孔、缩松缺陷的计算时间。

Abstract

The machine learning method was applied to predict shrinkage defects in casting process. The targeted pre-processing method was employed to process casting data

and the multi-modal data input method was adopted to increase the data dimension. Fully convolutional networks were used to fast calculate casting shrinkage porosity defects. Based on the fast prediction method

a typical casting was selected for analysis

which verified the effectiveness of the method. Compared with the numerical simulation method

this method has high accuracy

and can greatly reduce the calculation time of casting shrinkage defeets.