DOI: 10.1109/TMTT.2026.3731590   IEEEXplore: https://ieeexplore.ieee.org/document/11701406    

A dual-hypermutation immune algorithm (DHIA) for multi-chip power distribution network decoupling.

Not relevant to develop the algorithm

The proposed method and algorithm address challenging optimization applications with multidimensional constraints in high-dimensional decision spaces that often yield low value feasible solutions with poor optimization efficiency due to the severe constraint coupling. The proposed algorithm incorporates a physics-aware violation-mapping mechanism based on impedance morphology, together with morphology-guided directed hyper-mutation, probability-modulated stochastic hypermutation, and concentration-controlled competitive selection. The algorithm is demonstrated for multi-chip power distribution network decoupling for space-constrained high-density system-in-package (SiP) designs to simultaneously minimize capacitor count and occupied area.

Power distribution networks for single package SiP designs and other highly compact form factors with high density heterogeneous integration of chiplets from a variety of process nodes.

9.9% to 14.1% reduction in occupied area with the minimum capacitor count with weak-reward weighting with the lowest final fitness compared to five conventional optimization approaches. A further 9% reduction in occupied area was achieved with strong-reward weighting without excessive reward-driven component accumulation.