Analog Circuit Fault Diagnosis and Prognostics

Analog circuits form the backbone of numerous critical systems, from sensor interfaces and signal processors to power converters. Their continuous‐time operation and sensitivity to component tolerances render them susceptible to both abrupt “hard” faults (such as open‐ or short‐circuit failures) and gradual “soft” faults (including parameter drift and ageing). Fault diagnosis seeks to detect, locate and classify such failures by analysing circuit responses—often via time-domain measurements, frequency-domain transforms or hybrid methods. Traditional model-based techniques compare observed signals with analytical or simulated benchmarks, while data-driven approaches leverage statistical learning and deep neural networks to extract distinguishing features from raw waveforms. Prognostics extends diagnosis by predicting remaining useful life, enabling on-condition maintenance and minimising downtime. Recent progress has shifted from handcrafted feature extraction towards automated, hierarchical representations and from standalone classifiers to integrated frameworks coupling deep learning with optimisation heuristics. These advances support real-time monitoring of complex analog functions, enhance resilience in safety-critical applications and offer pathways to self-adaptive electronic systems.

Research from Nature Portfolio

Recent studies have demonstrated the power of combining unsupervised deep feature learning with metaheuristic optimisation for analog fault diagnosis. One approach employs a deep belief network to learn hierarchical representations of circuit output signals and applies a grey wolf optimisation algorithm to fine-tune a support vector machine classifier. Benchmarks on Sallen–Key band-pass and biquad high-pass filters achieved near-perfect diagnostic accuracy while reducing computation time by over 75%. This hybrid methodology exemplifies how deep architectures and nature-inspired search strategies can accelerate and refine fault detection in analogue integrated systems. Further foundational work has delineated frameworks for adaptive feature learning, underscoring the transition from manual signal processing to automated representation discovery.

Analog Circuit Fault Diagnosis and Prognostics publication trend

The graph below shows the total number of articles in analog circuit fault diagnosis and prognostics across all publications each year (not limited to Nature Index journals).

Technical terms

Analog circuit: An electronic network that processes continuous-time signals using passive and active components.

Hard fault: A sudden, discrete failure such as an open or short in a circuit element.

Soft fault: A gradual deviation of component parameters, leading to performance degradation.

Deep belief network (DBN): A layered, unsupervised neural model that captures hierarchical feature representations.

Grey wolf optimisation (GWO): A metaheuristic algorithm inspired by grey wolf social hierarchies, used for parameter tuning.

Empirical wavelet transform (EWT): A data-adaptive method for decomposing signals into meaningful frequency bands.

Residual network (ResNet): A convolutional neural architecture that uses skip connections to facilitate the training of deep models.

Wavelet scattering network: A cascade of predefined or learnable wavelet transforms that extract stable time-frequency features.

Remaining useful life (RUL): The predicted operational lifespan of a system or component before failure.

References

1. Application of DBN and GWO-SVM in analog circuit fault diagnosis. Scientific Reports (2021).
2. A Novel Fault Diagnosis Method for Analog Circuits Based on Multi-Input Deep Residual Networks with an Improved Empirical Wavelet Transform. Applied Sciences (2022).
3. Learnable Wavelet Scattering Networks: Applications to Fault Diagnosis of Analog Circuits and Rotating Machinery. Electronics (2022).