.png){kind=icon}
.png){kind=icon}
Return
Machine Learning-Based Package-Embedded Inductor Optimization for Integrated Voltage Regulators
.png){kind=icon}
OA
AI
DOI：10.1109/OJPEL.2025.3643511.png)
.png){kind=icon}
.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
This study presents an algorithm based on an artificial neural network (ANN) to optimize the design of inductors used in fully integrated voltage regulators (FIVRs). Because FIVRs are customized for specific chips, even identical inductor structures require adjustments based on the fabrication process and insertion method. The proposed algorithm evaluates the feasibility of inductor structures by considering their actual geometric constraints and then applies additional constraints to ensure that they remain within a user-defined inductance range. Using data from the structure identification process, the algorithm explores and optimizes structures that exceed the user-defined targets using Bayesian optimization (BO) to derive optimal results. After training, the algorithm can successfully identify inductor structures that meet target specifications and comply with relevant package design rules. Results verified through optimization examples of 5 nH and 2 nH inductors showed that in the 5 nH case, the DC resistance per unit volume decreased by 43.2% and the volume decreased by 24.5% under the same volume conditions. The proposed algorithm is versatile and can be applied to a wide range of inductor structures and designs. It efficiently identifies configurations that meet specific inductance targets. Moreover, the user-defined constraints can be adapted to various processes and insertion methods, enabling the algorithm to adjust to new targets effectively.
Keywords:
Inductors
neural network applications
optimization methods
regulators
AI Summary
Key information extracted from the uploaded paper, including a brief overview, abstract, background, key highlights, visual analysis, and future outlook.
Journal
I
IF:
3.9
Papers:
611
Citations:
1.2K