基于偏最小二乘回归的荧光粉涂覆型白光LED的光谱预测研究

为了更便捷高效地对荧光粉涂覆型白光LED的发光光谱进行预测，利用GaN蓝光LED芯片与杭州萤鹤光电材料有限公司的YH-S525M绿色荧光粉和YH-C640E红色荧光粉进行实验样品的制备。分别测量其单色荧光光谱，测得蓝光芯片的发射峰波长为453 nm，选用的红色和绿色荧光粉的发射峰波长分别为631和526 nm。制备红色和绿色荧光粉通过AB胶混合并涂覆于蓝光芯片上的LED实验样品，红粉/绿粉质量比设置为1∶3，1.2∶3，1.4∶3，1.6∶3，1.8∶3，2∶3，红粉混胶后的浓度为7%，9%，11%，13%，15%，17%。每组质量比和混胶浓度条件下的样品制备3~5份，利用杭州远方色谱有限公司的HAAS-2000高精度快速光谱辐射计测试样品的发光光谱，并进行蓝峰归一化处理得到共36组光谱数据。将白光光谱视为蓝色，绿色和红色三种单色荧光光谱的线性叠加，蓝色和红色峰项直接选用对应的发射谱，而绿色峰项选用两个高斯线型方程拟合，系数均由强度决定。通过循环搜索算法，分别计算36组实验条件下的模型参数最优值，对拟合结果进行优度检验，R2的范围为99.33%~99.88%。然后运用偏最小二乘回归方法建立荧光粉质量比和浓度与模型参数间的回归方程，最终得到一种能够精确预测两种荧光粉混合涂覆的白光LED发光光谱的新方法。用一组新制备的样品测得的光谱功率分布进行预测效果检验，得到的预测光谱相对于实测光谱的拟合优度为99.62%，证明该方法的预测效果良好。该研究建立的预测模型从该类型的白光LED的发光机理出发，分析发光时两种荧光粉之间的相互作用，并引入绿色荧光谱线的展宽效应，更加简单有效地建立起两种荧光粉的质量比和混胶浓度与白光光谱间的数学关系。该方法具有更好的普适性，为荧光粉涂覆型LED的光源参数优化提供了一种新的思路。

Study on the Spectral Prediction of Phosphor-Coated White LED Based on Partial Least Squares Regression

To predict the luminescence spectrum of phosphor-coated white LEDs more conveniently and efficiently, GaN Blue LED chip and YH-S525M green phosphor and YH-C640E red phosphor from Hangzhou Yinghe Optoelectronic Materials Co., Ltd. were selected for preparing experimental samples. The monochromatic fluorescence spectra were measured respectively. The emission peak wavelength of the blue-chip is 453 nm, the emission peak wavelength of red and green phosphor is 631 and 526 nm respectively. The red and green phosphors were mixed with AB glue and coated on the blue-chip. The mass ratio of red and green phosphors was set as 1∶3, 1.2∶3, 1.4∶3, 1.6∶3, 1.8∶3 and 2∶3. The concentration of red phosphors was set as 7%, 9%, 11%, 13%, 15% and 17%. 3~5 samples were prepared under each proportion and concentration, and the luminescence spectrum of each sample was measured by HAAS-2000 high-precision fast spectral radiometer of Hangzhou Yuanyuan chromatography Co., Ltd. A total of 36 groups of SPD (spectral power distribution) data were obtained by normalizing the relevant data. The white light spectrum was regarded as the linear superposition of blue, green and red monochromatic fluorescence spectra. The corresponding emission spectrum was directly selected for blue and red peak terms, while two Gauss linear equations were used for fitting the green peak term, and the intensity determined the coefficient. Therefore, a prediction model of the white light spectrum was established. The circular search algorithm calculated the optimal values of the model parameters under 36 groups of experimental conditions, and the model’s goodness of fit was tested. R2 ranged from 99.33% to 99.88%. Then, the partial least squares regression method was used to establish the regression equation between the mass ratio, concentration of phosphors and the model parameters. Finally, a new method that can accurately predict white LEDs’ emission spectrum coated with red and green phosphors was obtained. The SPD of a group of newly prepared samples was used to test the prediction effect. The goodness of fit of the predicted spectrum is 99.62%, which proves that the prediction effect of this method is good. Based on the physical mechanism of phosphor-coated LEDs, the mathematical relationship between the mass ratio, concentration of phosphors and the white light spectrum is established more simply and effectively. Meanwhile, the interaction between the two phosphors was analyzed, and the broadening effect of the green phosphor spectrum was introduced to the prediction model. There is good universality, and this method provides a new idea for optimising the light source parameters of the phosphor-coated LEDs.