DiffDA-Net Benchmark Datasets for Analog Circuit Fault Diagnosis

Four datasets provide time-domain fault responses for two benchmark analog circuits, simulated with PSpice Monte Carlo analysis. The data includes 50% and 25% parametric deviations for a Sallen-Key bandpass filter and a four-opamp biquad high-pass filter, with 2000 Monte Carlo runs per fault mode. The datasets accompany a 2026 paper submitted to Scientific Reports and are released for academic use.

Use Cases

• Training fault diagnosis models based on time-domain response data.
• Benchmarking algorithms for class-imbalanced learning based on the provided 9-class and 13-class structures.
• Developing domain adaptation methods for cross-severity settings based on the 50% and 25% deviation levels.
• Evaluating model performance on early-stage (incipient) fault detection.

Strengths

• Includes 18,000 rows per file for the Sallen-Key circuit and 26,000 rows per file for the four-opamp circuit.
• Simulates two distinct fault severity levels (50% and 25% deviation) for each circuit.
• Specifies component tolerances (5% for resistors, 10% for capacitors) and excitation pulse parameters.
• Data is structured with clearly defined, sequentially stored fault classes.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• The dataset's 1.2 KB size indicates only metadata files are present; the actual data files must be located separately.
• License restricts use to academic purposes only; commercial use requires contacting the author.

Provenance

Source

figshare, accompanying the paper 'DiffDA-Net: Diffusion-Augmented Domain Adaptive Network for Analog Circuit Fault Diagnosis...'

Collection Method

Generated via PSpice Monte Carlo simulation with specified component tolerances and excitation pulse.

Freshness

Last updated 2026-05-13 06:54:18