AMS-Anomaly-Dataset

🚀 Overview

The AMS-Anomaly-Dataset is an open-source collection of well-structured data curated for anomaly detection in analog and mixed-signal (AMS) circuits. This dataset aims to empower researchers and engineers to develop, benchmark, and evaluate models for anomaly detection, especially in automotive System-on-Chip (SoC) applications. It provides a comprehensive set of signals covering various types of faults, parametric variations, and environmental conditions across multiple AMS circuit types.

📜 Background

Analog and Mixed-Signal (AMS) circuits are integral to automotive SoCs, acting as bridges between continuous analog signals and digital logic. However, these circuits are susceptible to issues like parametric faults, open/short circuit faults, and operating region anomalies, which can compromise functional safety (FuSa) and reliability. This dataset provides a structured repository of AMS circuit data to help researchers standardize and improve machine learning (ML) models for fault detection and anomaly analysis.

🗂️ Directory Structure

The dataset is organized into the following directory structure, capturing both RAW and Feature-Extracted data from key AMS components:

|-ADC
|---Converted
|-----NON-Anomalous
|-----shorttrans
|-------1stbit
|-------1stbitmux
|-------3rdbitand
|-------4thbitand
|-------4thbitmux
|-----tempsweep
|-------Anomalous
|-------Non_Anomalous
|-----trans
|-------1stbit
|-------1stbitmux
|-------3rdbitand
|-------4thbitand
|-------4thbitmux
|---RAW
|-----NON-Anomalous
|-----shorttrans
|-------1stbit
|-------1stbitmux
|-------3rdbitand
|-------4thbitand
|-------4thbitmux
|-----tempsweep
|-------Anomalous
|-------Non_Anomalous
|-----trans
|-------1stbit
|-------1stbitmux
|-------3rdbitand
|-------4thbitand
|-------4thbitmux
|-DAC
|---Converted
|-----bulk
|-------N1
|-------P1
|-----DAC Open and Short Circuit Data
|-------100K OHM
|-------100M OHM
|-------100 OHM
|-------10K OHM
|-------10M OHM
|-------1G OHM
|-------1K OHM
|-------1M OHM
|-----DAC Temperature Data
|-------Anomalous
|-------Non_anomalous
|-----Non_anomalous
|---RAW
|-----bulk
|-------N1
|-------P1
|-----DAC Open and Short Circuit Data
|-------100K OHM
|-------100M OHM
|-------100 OHM
|-------10K OHM
|-------10M OHM
|-------1G OHM
|-------1K OHM
|-------1M OHM
|-----DAC Temperature Data
|-------Anomalous
|-------Non_anomalous
|-----Non_anomalous
|-OPAMP
|---FeatureExtractedData
|-----DualStage
|-------AC_Parametric_Features
|-------DC_Parametric_Features
|-----SingleStage
|-------AC_Parametric_Features
|-------DC_Parametric_Features
|-----TrippleStage
|-------AC_Parametric_Features
|-------DC_Parametric_Features
|---RawParametricTempGain
|-----DualStage
|-----SingleStage
|-----TrippleStage
|-VRef
|---LinearRegion
|---OpenFaults
|---ParametricFault_Temp
|---ShortFaults

📊 Dataset Components and Fault Scenarios

The dataset consists of the following key AMS components, each characterized by various fault injection scenarios:

1. Analog-to-Digital Converters (ADCs):

   • Faults Simulated: Short circuits, open circuits, temperature sweeps, and bit-level anomalies during conversion.
   • Applications: Vital for translating real-world analog signals into digital values within SoCs.

2. Digital-to-Analog Converters (DACs):

   • Faults Simulated: Bulk parametric faults, temperature-induced variations, open and short circuits at various resistance values.
   • Applications: Essential for providing precise analog outputs based on digital inputs in automotive electronics.

3. Operational Amplifiers (OpAmps):

   • Data Types: AC Parametric Features, DC Parametric Features, and Raw Temperature Gain Data.
   • Faults Simulated: Gain errors, non-linearity, and saturation-region anomalies.
   • Applications: Core components for signal amplification in various control and signal processing systems.

4. Voltage Reference (VRef) Circuits:

   • Faults Simulated: Linear region faults, open faults, short faults, and temperature-related parametric shifts.
   • Applications: Serve as stable reference points for other analog circuits within the SoC.

📈 Anomaly Injection and Simulation Methodology

The dataset includes comprehensive data reflecting:

• Field-effect Transistors (FETs) Anomalous Behavior: Simulated operation in triode/linear regions instead of the expected saturation region.
• Parametric Faults: Variations induced by temperature shifts outside the ideal range (-40°C to 125°C).
• Open and Short Faults: Faults modeled using varying resistances, representing different open or short circuit conditions across terminals.

🤖 GAN-based Data Augmentation

To address data imbalance issues and enhance the quality of training datasets, Generative Adversarial Networks (GANs) were employed to generate synthetic anomaly data. The GAN-based augmentation strategy achieves up to a Correlation Similarity Score of 0.96, indicating the high fidelity of the generated data with respect to real-world conditions.

License

This dataset is provided under the MIT License.

Contact

For any questions or issues, please contact the dataset maintainers at sanjay.das@utdallas.edu