Deep Learning-Based Fast DC IR Drop Prediction for Power Distribution Networks

Efficiently predicting the DC resistances of power distribution networks (PDNs) is important for the optimization of printed circuit board (PCB) design. This paper proposes a methodology based on boundary integration and algebraic graph theory to calculate the DC resistance of PDNs efficiently, which is much more quickly than traditional full-wave simulation. Then, for PCB design with higher requirements in speed, a deep neural network (DNN) is trained to predict the PDN resistance for different stack-up structures, which can be accomplished within fractions of a second.