Automated Anomaly Detection in Power Integrity Simulations via HyperScore-Driven Reinforcement Learning

**Abstract:** This paper presents a novel methodology for automated anomaly detection within power integrity (PI) simulations, a critical step in modern electronic design. Leveraging a modified Reinforcement Learning (RL) agent assessed by a HyperScore-driven evaluation pipeline, our system automatically identifies and flags unexpected simulation behaviors indicative of design flaws or solver instability, surpassing traditional manual inspection in speed and accuracy. We demonstrate this approach's effectiveness on a complex multi-layer PCB design, showcasing improvements in detection rate and a significant reduction in debugging time compared to conventional methods. This research provides a scalable solution for enhancing the efficiency and reliability of PI verification, directly impacting time-to-market for high-speed digital systems.