An Adaptive Online Diagnosis and Prognosis Approach for Digital Circuits

Failures of digital circuits can cause significant system downtime in industry sectors including manufacturing, telecommunications, transportation, and healthcare. While testing methods are valuable to understand how failures can occur, they do not provide real-time early warnings needed to prevent catastrophic failures or forecast maintenance. This paper develops an adaptive online Prognostics and Health Management (PHM) solution for digital circuits that uses SPICE simulations and laboratory experiments to obtain circuit fault behaviors, while addressing the domain discrepancies between simulations and experimental data. The developed PHM solution achieves early fault detection by analyzing the transition behavior—the rising and falling edges—of circuit output voltage signals. Temporal, spectral, and temporal-spectral features are extracted from these transitions. Diagnosis includes fault type classification that localizes the degraded circuit components, and fault value prediction. The fault value is used in conjunction with component degradation models to make a prognosis of the circuit’s remaining useful life. The effectiveness of the approach is demonstrated through implementation on a four transistor NAND gate. The developed adaptive PHM achieves 98.6% fault-type classification accuracy and 0.12 V RMSE in fault value prediction.