基于声发射信号稀疏表示的铝合金板材拉伸过程识别方法
Sparse Representation of Acoustic Emission Signals for Identifying Tensile Process of Aluminum Alloy Sheets
焦敬品 1孙延东 1李光海 2赵鹏经 1吴斌 1何存富1
作者信息
- 1. 北京工业大学材料与制造学部 北京 100124
- 2. 中国特种设备检测研究院 北京 100029
- 折叠
摘要
针对钣金成形过程质量监测的需要,提出一种基于声发射信号稀疏表示的铝合金板材拉伸过程自动识别方法.该方法对钣金拉伸过程监测的声发射信号进行非负矩阵分解,提取其在低维子空间映射的特征系数,用于构造训练字典和测试样本,并利用l1范数进行稀疏分解和信号重构,进而实现对拉伸过程中弹性、塑性、屈服、强化和颈缩5个不同应力-应变状态的自动识别.同时,研究了声发射信号的数据类型对不同应力-应变状态识别准确率的影响.结果表明,提出的基于声发射信号稀疏表示方法可以很好实现铝合金拉伸过程中5个不同应力-应变状态的识别,相较于快速傅里叶变换类型数据,利用短时傅里叶变换数据的识别准确率更高.研究工作为钣金成型质量监控提供了可行的解决方案.
Abstract
Aiming at the need of quality monitoring in sheet metal forming process,an automatic identification method of aluminum alloy sheet drawing process based on sparse representation of acoustic emission signals was proposed.The method to monitor the process of sheet metal drawing of the acoustic emission signal is non-negative matrix factorization,and extract the mapping on the low-dimensional subspace characteristic coefficient of dictionary is used to construct the training and testing samples,and the use of l1 norm for sparse solution and the signal reconstruction,thus realize the tensile elasticity,plasticity,yield,hardening,and in the process of necking five different stress-strain state of automatic identification.At the same time,the influence of data types of acoustic emission signals on the recognition accuracy of different stress-strain states is studied.The results show that the proposed sparse representation method based on acoustic emission signals can well realize the identification of five different stress-strain states in the tensile process of aluminum alloy.Compared with Fast Fourier Transform data,the identification accuracy of Short-Time Fourier Transform data is higher.The research provides a feasible solution for the quality control of sheet metal forming.
关键词
拉伸过程/稀疏表示/声发射/应力-应变状态/非负矩阵分解Key words
tensile process/sparserepresentation/acoustic emission/stress-strain state/non-negative matrix factorization引用本文复制引用
基金项目
国家自然科学基金资助项目(11972053)
国家自然科学基金资助项目(12274012)
出版年
2024
NETL
NSTL
维普
万方数据