基于Beer-Lambert理论与人工神经网络的混合荧光粉发射光谱预测

Predicting of emission spectrum for mixed phosphors using Beer-Lambert theory and artificial neural network

摘要:

本文针对由钇铝石榴石基、硅基氮氧化物和氮化物混合制备的三元高显色荧光粉体系，采用修正Lambert-Beer模型结合BP（Back Propagation, BP）人工神经网络的方法实现了任意比例混合荧光粉体系的发射光谱预测。首先，通过修正Lambert-Beer模型计算出三种荧光粉以任意比例混合后的发射光谱比例系数；其次，采用BP人工神经网络对比例系数进行训练和预测；最后，实现三种荧光粉混合后发射光谱的预测并通过实验测量验证。研究结果表明：（1）比例系数预测值与理论计算值之间的误差可以控制在5%以内；（2）通过BP神经网络预测的发射光谱图谱与试验测量图谱符合程度较好，两者的均方根误差RMSE和色度差异Δxy较小，平均值分别为0.019和0.0016。

Abstract:

In this paper, the emission spectra of a high color rendering phosphors, mixed with the Yttrium Aluminium Garnet, Silicon based Oxynitride and Nitride based phosphors, were predicted by using the Lambert-Beer theory and Back Propagation Neural Network (BP NN). Firstly, the modified Lambert-Beer model was used to calculate the proportional coefficient of the emission spectra of the mixed phosphors in ratios. Next, the BP NN was implemented to train and predict the proportional coefficients. Finally, the prediction of the emission spectra of the mixed phosphors were estimated and verified by the experimental measurements. The results show that: (1) The prediction error percentage of the scale factor can be controlled within 5%; (2) The predicted emission spectra by BP NN keep high agreement with the experimental measurements with lower RMSE andΔxy as 0.019 and 0.0016, respectively.