Thermal Analysis And Structural Optimization Of COB-LED Fluorescent Lens

LED is widely used in lighting,decoration and display fields due to their high luminous efficiency,long life,green energy saving,fast response and compact structure.Among various packaging forms of LEDs,COB package has achieved rapid market development in recent years due to its simple packaging process and short heat dissipation path.However,due to the large density distribution of the chip in the COB package,the power density is high,the thermal coupling between the chips is serious,and the self-heating of the phosphor is obvious,which seriously affects the service life and reliability of the LED light source.Therefore,systematic analysis of the heat dissipation of the COB-LED is critical and the optimization of the package structure is required.First,this dissertation tests and analyzes the performance of COB-LED samples.Then based on the actual size and measurement results of the model,we built a three-dimensional thermal model in the Icepak software,which is used to analyze the thermal performance of the COB-LED.The results show that the white LED has a 46.38 ? higher temperature than the blue LED which with the same package structure.This indicating that the phosphor has a self-heating effect and causes the junction temperature to rise by 10 ?,and the highest temperature of the fluorescent lens is slightly lower at the center of the lens surface.The optical simulation model was established by TracePro software to obtain the heating power of the fluorescent lens structure under natural convection,and then combined with the thermal model to analyze the influence of fluorescent lens parameters include phosphor concentration,thickness,and the coating thickness of the silicone under the remote phosphor and the phosphor sedimentation structure on the maximum temperature of the lens,average junction temperature and color temperature of COB-LED based on single factor test.Secondly,according to the single factor test results,the orthogonal test is designed to determine the sample data.Aiming at the TPmax,TJ and CCT,we established BP neural network prediction models and support vector machine prediction models respectively.The results show that the correlation coefficients between the prediction of the BP neural network model and the original data are 0.98712,0.98879,and 0.97796,respectively,and the mean square errors are 0.00037,0.00637,and 0.00378,respectively.The correlation coefficients between the prediction of the support vector machine model and the original data are 0.9915,0.99165,0.98157,respectively,and the mean square errors are 0.00034,0.00294,and 0.00094,respectively.Regardless of the decision coefficient or the mean square error,the prediction accuracy of the support vector machine is higher than that of the BP neural network.Therefore,the prediction model established by the support vector machine is used for subsequent optimization.Finally,the support vector machine prediction model is called in Matlab by the non-dominated sorted genetic algorithm-? to optimize the performance of COB-LED.Results show that,while ensuring the color quality,the optimized chip junction temperature is reduced by 0.4 ?,but the maximum lens temperature is reduced by 5.04 ?.In addition,the paper also provides the optimal structure at a color temperature of 3500 K to 5000 K.In practical applications,different solutions can be selected according to different needs.