一种基于深度学习的荧光粉型 LED色温预测新方法
CSTR:
作者:
作者单位:

(1.东华理工大学江西省核地学数据科学与系统工程技术研究中心,南昌 330013;2.东华理工大学人工智能与信息工程学院,南昌 330013;3.南昌科瑞普光电科技有限公司,南昌 330117)

作者简介:

郭兰辉(2000-),男,江西省赣州市人,硕士研究生,主要从事 LED光谱特征的研究;何剑锋(1977-),男,江西省婺源市人,教授,博士,主要从事核数据采集与信号处理、嵌入式系统应用开发的研究。

通讯作者:

中图分类号:

TN312. 8

基金项目:


Study on Deep-Learning-Based Color Temperature Prediction Methods for LED Phosphors
Author:
Affiliation:

(1. Jiangxi Engineering Technology Research Center of Nuclear Geoscience Data Science and System,East China University of Technology,Nanchang 330013,CHN;2. College of Artificial Intelligence and Information Engineering, East China University of Technology,Nanchang 330013,CHN;3. NanChang CREP Optoelectronics & Technology Co. Ltd.,Nanchang 330117,CHN)

Fund Project:

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
  • |
  • 文章评论
    摘要:

    针对传统荧光粉型发光二极管(LED)封装方式在实现目标色温调控过程中需频繁实验而导致效率低下、生产成本高的问题,提出一种基于多层感知机(MLP)的荧光粉型 LED色温预测方法,旨在实现荧光粉配比的智能化和色温的精准预测。该方法利用 Light Tools光学仿真软件建立 LED封装模型,系统研究三色荧光粉配比、荧光粉粒子总数及粒径大小与色温的映射关系,构建了相应的训练数据集。在此基础上建立了 MLP神经网络预测模型,通过独立数据集验证模型的性能。结果表明,模型预测值与真实值具有良好的一致性,平均决定系数 R2达到 0.915 8,表明所构建的模型具有优异的泛化能力和预测精度,可为 LED封装工艺优化提供技术支持。

    Abstract:

    Achieving target color temperatures in LED packaging using traditional red,green,and blue phosphors combined with ultraviolet LED chips is restricted by three factors—requirement of frequent experimentation,low efficiency,and high production costs. Therefore,in this study,a color temperature prediction method,based on a multilayer perceptron(MLP),is proposed for LED phosphors. This method enables intelligent optimization of phosphor ratios, supporting accurate prediction of color temperatures. An LED packaging model was established,using the Light Tools optical simulation software,to systematically investigate the relationship mapping among the ratios of three color phosphors, total number of phosphor particles,particle size,and color temperature. A corresponding training dataset was constructed,and an MLP neural network prediction model was developed,whose performance was then validated using an independent dataset. The values predicted by the proposed model are in good agreement with the actual values,with an average coefficient of determination(R2)of 0. 915 8. This result confirms the excellent generalization ability and prediction accuracy of the proposed model,which can thus provide technical support for optimizing LED packaging processes.

    参考文献
    相似文献
    引证文献
引用本文

郭兰辉,何剑锋,徐杰,余鑫,崔剑,李卫东,曾民晨.一种基于深度学习的荧光粉型 LED色温预测新方法[J].半导体光电,2025,46(5):892-898. GUO Lanhui, HE Jianfeng, XU Jie, YU Xin, CUI Jian, LI Weidong, ZENG Minchen. Study on Deep-Learning-Based Color Temperature Prediction Methods for LED Phosphors[J].,2025,46(5):892-898.

复制
分享
相关视频

文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:2025-07-09
  • 最后修改日期:
  • 录用日期:
  • 在线发布日期: 2025-10-31
  • 出版日期:
文章二维码