深度学习的光学超精密制造设备状态模式识别  

State recognition of optical ultra-precision manufacturing equipment based on deep learning

作　　者：李瑛达[1] 贾宁[1] LI Yingda;JIA Ning(Dalian Neusoft University of Information,Dalian 116023,China)

机构地区：[1]大连东软信息学院,大连116023

出　　处：《激光杂志》2020年第12期161-165,共5页Laser Journal

基　　金：国家自然基金(No.61602075)。

摘　　要：为提升超精密光学元件成品质量以及成品率,提出了深度学习的光学超精密制造设备状态模式识别方法,采集光学超精密制造设备状态数据集后,通过数据处理、关联分析与回归分析获取光学超精密制造设备状态数据集的关联度与拟合函数结果,将其作为训练样本输入卷积神经网络,通过卷积操作、池化操作以及网络结构与设置实施深度学习的模式识别,通过输出深度学习结果模式识别光学超精密制造设备状态,最后实验结果表明,本文方法可有效识别该制造设备的压力、转速、元件摆幅、元件摆速异常状态,总识别准确率高达99%,具有较高实用性能。In order to improve the quality and yield of ultra-precision optical components,a state pattern recognition method of optical ultra-precision manufacturing equipment is proposed.After collecting the state data set of optical ultra-precision manufacturing equipment,the correlation degree and fitting function results of the state data set of optical ultra-precision manufacturing equipment are obtained through data processing,correlation analysis and regression analysis,which are used as training.Through convolution operation,pooling operation,network structure and setting,the input convolution neural network can recognize the state of the optical ultra-precision manufacturing equipment.Finally,the experimental results show that this method can effectively identify the manufacturing equipment abnormal state of the pressure,rotation speed,component swing amplitude and component swing speed.The total accuracy is as high as 99%,and it has high practical performance.

关 键 词：深度学习 光学 超精密 制造 设备状态 模式 

分 类 号：TN297[电子电信—物理电子学]