多芯片LED光源的可靠性分析涉及到光、电、热多个物理场, 高精度的多场分析结果会导致计算资源过多、计算时间过长、计算难度大等问题。为解决上述问题, 本文分别利用传统的有限元算法(FEM)和高效的人工神经网络方法(ANN)进行LED光源温度分析, 并讨论两种方法的优劣性。最后, 通过将FEM分析单一传热物理场的优势与ANN计算时间短、计算资源需求低的优势相结合, 归纳出一种更为高效的方法来进行多芯片LED光源的散热分析。利用该方法, ANN的预测数据与训练数据之间的相关系数达到了0.997 79, 预测结果与实际热分布图有良好的匹配, 计算资源相比传统的FEM方法节约了59％。该方法的应用能够在满足精度的前提下耗费更少的计算资源和时间, 同时提高了分析的灵活性。除此之外, 该方法对求解大功率LED光源寿命等可靠性问题也具有一定的参考价值。

The reliability analysis of multi-chip LED light sources involves multiple physical fields of light, electricity and heat. The high-precision analysis results will lead to too many calculation resources, too long calculation time and difficult calculation. To solve the above problems, the traditional finite element method(FEM) and efficient artificial neural network(ANN) method are used to analyze the temperature of LED light source, and the advantages and disadvantages of both are discussed. Finally, by combining the advantages of FEM analysis in a single heat transfer physics field with the advantages of ANN in little calculation time and low computational resource requirements, a more efficient method for heat dissipation analysis of multi-chip LED light sources is summarized. Using this method, the correlation coefficient between the prediction data and the training data of ANN reaches 0.997 79, and the prediction result has a good match with the actual heat distribution. The computational resource saves 59% compared with the traditional FEM method. The application of this method can consume fewer computing resources and time based on satisfying the accuracy, while improving the flexibility of analysis. In addition, this method has certain reference value for solving the reliability problems such as the lifetime of high-power LED light source.