A Novel Noise-assisted Prognostic Method for Linear Analog Circuits

摘要

Integrated analog circuits play a critical role in modern industrial systems. However, research on prognosis of the remaining useful performance (RUP) of analog circuits is rarely reported. This work presents a novel prognostic scheme for analog circuits. In this scheme first a signal feature is extracted from time-domain output waveforms of the circuit during its initial and faulty states. Then, an auxiliary white-noise estimation methodology based on Kalman filter technique estimates the faulty response waveform. As the embedded planar capacitors have a great influence on integrated circuit (IC) miniaturization, performance and reliability, an empirical model based on gradually decreasing trend of its capacitance is introduced. Due to the influence of capacitor degradation, the change in the circuit output waveform is treated as a fault indicator (FI) for failure prognosis. Next, this FI tendency close to the realistic condition is used to identify the circuit degradation trend. Further, in a data-driven prognostic step, a particle filter is used to predict the RUP of the circuit. Finally, studies on two analog filter circuits verify the proposed method.