基于LMBP神经网络的MCM布局电磁场预测模型

Prediction model for the electromagnetic field of MCM with different chips layouts based on LMBP neural network

通信微波多芯片组件(MCM)内不合理的芯片布局将会导致电磁干扰加剧。对具有不同布局的MCM多层布线基板的电磁场进行了仿真分析,并以仿真分析结果作为样本数据,建立了基于LMBP神经网络的MCM布局电磁场预测模型,然后用该模型对两种单面芯片布局下的MCM的坡印廷矢量数值进行了预测。结果显示,MCM电磁场的仿真与模型预测差值分别为0.332e–7W/m2和0.263e–7W/m2,证明了该预测模型的可用性。

Unreasonable chips layouts in the communication-microwave MCM would enhance the electromagnetic interference.The electromagnetic field of MCM with different chips layouts was simulated.With the simulation results as samples,a model based on LMBP neural network was built to predict the electromagnetic field of MCM.The poynting D-vector value of two kinds of MCM with single face chips layout was then predicted using this model.The results show that the differences between the simulation results and prediction results of two MCM are 0.332 e–7W/m2 and 0.263 e–7W/m2,indicating that the model is available to predict the electromagnetic filed of MCM.