A hybrid NCA-CNN deep learning approach for high-accuracy parametric fault diagnosis in MOSFET-C analog circuits

ERKAL K., TURHAL Ü. Ç., ÖNAL Y., Yesil A.

AEU-INTERNATIONAL JOURNAL OF ELECTRONICS AND COMMUNICATIONS, cilt.212, 2026 (SCI-Expanded, Scopus) identifier identifier

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 212
  • Basım Tarihi: 2026
  • Doi Numarası: 10.1016/j.aeue.2026.156306
  • Dergi Adı: AEU-INTERNATIONAL JOURNAL OF ELECTRONICS AND COMMUNICATIONS
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Compendex, INSPEC
  • Anahtar Kelimeler: Deep learning, Convolutional neural network, Neighborhood components analysis, Parametric fault diagnosis, MOSFET-C filter circuits, Frequency response analysis
  • Bilecik Şeyh Edebali Üniversitesi Adresli: Evet

Özet

Diagnosing faults in large-scale integrated electronic systems is challenging due to their complex operational mechanisms. Fault identification is crucial, as faults significantly impact the performance of filter circuits. This paper proposes a hybrid methodology for diagnosing parametric faults caused by component variations in MOSFET-C-based designs. The method integrates neighborhood components analysis for feature extraction with a convolutional neural network for fault diagnosis, leveraging the strengths of both traditional machine learning and deep learning techniques. This hybrid approach enhances feature extraction, improves performance, increases robustness against noise, and improves computational efficiency through dimensionality reduction. The study used data generated through Monte Carlo simulations conducted in the LTspice environment using TSMC 0.18 μm CMOS technology parameters, focusing on a resistor-free circuit in which faults were simulated by introducing a 10% variation in the capacitance values. Fault detection relied on frequency response data obtained from these simulations. Comparative analysis against various hybrid machine learning and deep learning combinations with a success rate of 99% showed that the proposed methodology outperformed all alternatives, demonstrating superior accuracy, generalization, and diagnostic reliability.