首页|基于多源信息融合与神经网络的汽车塑件无损检测方法研究

基于多源信息融合与神经网络的汽车塑件无损检测方法研究

扫码查看
原文链接
  • NETL
  • NSTL
  • 万方数据
针对汽车塑件内部缺陷检测精度不高、检测方法单一等亟需解决的问题,基于多源信息融合与深度学习神经网络等技术,探究高鲁棒性、稳定性、精确性的检测模型,旨在为汽车塑件批量无损检测提供理论基础和技术支持.通过超声波探伤、激光光斑测量、紫外线探伤、智能电参数测量和红外热成像等无损检测方法,试验样本经异质信息采集、预处理以及多特征变量提取,根据检测要求,用于定量/定性神经网络模型构建.再以样本训练、规则判别、深度学习和模型类比,选取最优模型用于试验样本内部缺陷预测和品级分析,以求得所设计的"深度学习"神经网络模型与检测方法具有通用性,可进一步用于汽车铸件、冲压件或焊接件等无损检测,为匹配智能装备研发提供研究基础.
Research on nondestructive testing method of automobile injection parts based on multi-source information fusion and neural network
In view of the problems of low accuracy and single detection method of internal defect detection of automobile injection moulded parts,based on multi-source information fusion and deep learning neural network and other technical means,the detection model with high robust-ness,stability and accuracy was explored,aiming to provide theoretical basis and technical sup-port for batch nondestructive testing of injection moulded parts in the automobile industry.Through nondestructive testing methods such as ultrasonic flaw detection,lasers spot testing,ul-traviolet flaw detection,intelligent electrical parameter measurement and infrared thermal imag-ing.The test samples were collected by heterogeneous information acquisition,pretreatment and multi-characteristic variable extraction,and were used to construct quantitative/qualitative neural network models according to the detection requirements.Then,through sample training,rule dis-crimination,deep learning and model analogy,the optimal model was selected for internal defect prediction and grade analysis of test samples,so as to obtain the generality of the designed"deep learning"neural network model and detection method.It could be further used for nondestructive testing of automobile castings,stampings or elements,and provide a research basis for the develop-ment of matching intelligent equipment.

nondestructive testingmulti-source informationneural networkprincipal compo-nent analysisautomobile injection parts

李吉生、孙潇鹏、张胜宾

展开 >

广西水利电力职业技术学院 交通工程学院,广西 南宁 530023

广东交通职业技术学院 汽车与工程机械学院,广东 广州 510650

无损检测 多源信息融合 神经网络 主成分分析 汽车塑件

广东省职业教育教学改革研究与实践委托项目广西高校中青年教师科研基础能力提升项目(2023)广东省高等职业院校交通运输类专业教指委教育教学改革项目(2022)清远市产教融合社会科学专项(2023)广东交通职业技术学院校级教科研项目(2023)

JG-WT2021X032023KY1136JTYS-JZW2022B02ZJCYJY202314GDCP-ZX-2023-013-N3

2024

10.16787/j.cnki.1001-2168.dmi.2024.06.003

模具工业
桂林电器科学研究所

模具工业

影响因子：0.637
ISSN：1001-2168
年,卷(期)：2024.50(6)