Machine Learning-Driven Framework for Predicting Aging Induced Delay in CMOS Circuits

View at publisher Summarise with VIRAT
Authors: K. Yogeshwaran, C. Ramesh, V. Govindaraj, R. Sri Bharath, V. Suruthi, G. Raj
Cited by: 0
References: 16

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

As CMOS technology advances into deep nanometer regimes, age-related reliability degradation has become a major design concern for modern integrated circuits. Aging mechanisms such as Bias Temperature Instability (BTI) and Hot Carrier Injection (HCI) gradually shift transistor threshold voltages, increasing propagation delay and potentially causing timing violations in critical logic paths. While SPICE- based simulations like Cadence Spectre provide high accuracy in aging analysis, their computational cost limits large-scale circuit evaluation. To address this, the proposed work introduces a machine learning-driven predictive framework for estimating aging-induced delay degradation (△Delay) in CMOS benchmark circuits. A comprehensive dataset of over 7200 samples was generated through Spectre simulations under varying voltages (0.7-1.2 V), temperatures (25-125°C), and stress times$(10^{5}-10^{8} ~\mathrm{s}$), capturing features such as$\Delta$Vth, W/L ratio, switching activity, and initial delay. Using permutation-based feature selection and grid search optimization, Random Forest (RF) and Extreme Gradient Boosting (XGBoost) models were trained with 5-fold crossvalidation to ensure robustness. Experimental evaluation showed that XGBoost achieved MAE = 0.18 ps, RMSE = 0.25 ps, and$\mathbf{R}^{\mathbf{2}}=0.973$, closely matching SPICE results while being over$100 \times$faster. The framework effectively scales from gatelevel to circuit-level prediction across ISCAS-85 and ITC-99 benchmark circuits, demonstrating its capability for fast, accurate, and scalable aging-aware delay prediction and providing a strong foundation for reliability-driven VLSI design optimization.

Fields of Study & Subjects

Low-power high-performance VLSI design Radiation Effects in Electronics VLSI and Analog Circuit Testing

Publication details

Journal / source
International Conference on Communication, Computing and Electronics Systems (ICCCES)
Pages
79–84
Publication date
21 January 2026
Language
EN