LED Photoelectric Thermal Coupling Analysis Based On Artificial Neural Network

It is necessary and meaningful to analyze and study the multi-physical fields of LED,as the application of LED in the lighting field is expanding constantly.And many scholars have put forward different models in LED multi-physics research field,and they have given corresponding calculation and verification results.Among many models,the finite element method(FEM)is a widely used model.It can been seen from the calculation results that the accuracy of the finite element method is very high,however the high accuracy is at the expense of huge computing resources and computing time.Especially when the number of physical field is increases,the computing resources and computing time will exponential growth,which greatly limits the generalization and application of finite element method in multi-physical fields.In order to solve the contradiction between the calculation accuracy and computation time of the traditional multi-physical field method,the artificial neural network(ANN)combined with finite element method is mentioned.The thermal distribution of LED lamp is calculated in this article based on ANN combined with FEM and the lifetime of LED lamp is predicted,too.In order to calculate and analyze the multi-physical field model of LED,the photo-electro-thermal characteristics of a single LED is measured and analyzed and the main testing parameters include LED electric power,optical power and luminous flux.What's more,the LED photo-electro-thermal(PET)BP ANN is established and trained based on the experimental data.Meanwhile,the LED PET characteristics is predicted based on the ANN,and the accuracy of LED PET BP ANN is evaluated by compared the ANN prediction results as well as the measurement results.The research results show that the trained LED PET BP ANN can be used to predict the PET characteristics of LED,and the error between prediction results and experimental data can be controlled within 5%.Secondly,a LED optical-electro-thermal analysis method based on artificial neural network is raised,and the thermal distribution of a LED and its LED lamp is calculated and analyzed.The calculation results show that the thermal distribution results of LED and LED lamp made by this method are very close to the measured results,while the thermal distribution of LED lamp is in the range of 0.3K.It is found that the traditional multi-physical field model consumes at least 6.7 times computational resources than this new LED optical-electro-thermal analysis method.And it can get satisfying and accuracy results consume less computing resources and computing time in solving the LED optical-electro-thermal problems.Finally,the lifetime of LED lamp is predicted and analyzed based on the LED artificial neural network and the LED lamps' heat distribution results obtained by the above analysis method.And the lifetime of LED lamp is predicted to be 73518 hours,while the normalized luminous flux is reduced to 0.7.