Overview:

This dataset includes time series data derived from Controller Area Network (CAN) bus log files, representing intermittent faults introduced into an automotive system. This is provided to facilitate reproducibility of results presented in the following paper:

Hespeler, S. C., Moriano, P., Li, M., & Hollifield, S. C. (2025). Temporal cross-validation impacts multivariate time series subsequence anomaly detection evaluation. arXiv preprint arXiv:2506.12183.

The data was collected and converted into time series format to facilitate easy analysis and fault detection without requiring the decoding of raw CAN log files. The dataset contains time series data derived from CAN bus logs collected during experiments designed to introduce intermittent faults in automotive systems. The CAN bus data have been converted into structured time series format to facilitate ease of use in research focused on fault detection, machine learning, and multivariate time-series (MTS) analysis. The data captures various sensor readings and system statuses at 0.1-second intervals, with specific fault conditions annotated in the final column.

Background:

Files:

1. Power_wire.csv: This file contains time series data representing the condition when the common power wire was disconnected, causing intermittent faults.
2. Spark_plug.csv: This file includes time series data recorded when the spark plug power was unplugged, creating a different set of intermittent faults.
3. AST.csv: The two Arbitration ID files include optional information for the time series data from experiments capturing various aspects of the vehicle's operational status under different conditions. The first column contains CAN ID, followed by signal orders within the CAN grouping, and finally the potential physical descrption of the signal.
4. metadata.jason: Provides a brief description of the fault introduced during the data capture, including the nature of the fault and the operational status of the vehicle.

Data Collection Process:

The data was collected in a controlled environment, simulating various fault conditions in a vehicle's electronic control systems. The CAN bus log files were recorded during these simulations and then converted into time series data for easier analysis.

Fault Information:

The faults were introduced intentionally to simulate real-world scenarios where intermittent faults can occur in a vehicle's CAN bus system. The following describes the specific conditions under which the data was collected:

1. 1686085247.018385_IF_candump.log:

   • Description: Disconnected the common power wire on likely #4, closest to the passenger side. The vehicle was stationary at the VSL, with frequent pressing of the gas pedal to activate the combustion engine.
   • Fault Periods:
     • Starts: 1686085247.0177443, 1686085285.0588675, 1686085321.0775447, 1686085355.1128247, 1686085393.1535382, 1686085439.1818333
     • Stops: 1686085269.0425622, 1686085294.0685468, 1686085336.0936263, 1686085376.1355531, 1686085418.160425, 1686085464.208798
     • Durations: 22.0248, 9.0097, 15.0161, 21.0227, 25.0069, 25.0270 (seconds)

2. 1686085676.2174_IF_candump.log:

   • Description: Unplugged spark plug power from cylinder #2 (likely), towards the inside of the passenger side. The vehicle was stationary at the VSL, with frequent pressing of the gas pedal to activate the combustion engine.
   • Fault Periods:
     • Starts: 1686085676.2166712, 1686085703.2368746, 1686085728.2570298, 1686085747.2773526, 1686085786.3179924
     • Stops: 1686085690.2265697, 1686085719.2466102, 1686085732.261957, 1686085770.301622, 1686085807.3200693
     • Durations: 14.0099, 16.0097, 4.0049, 23.0243, 21.0021 (seconds)

File Structure

Each CSV file contains the following columns:

• CAN Signal Columns: These columns represent various CAN signals (MTS features) decoded from the raw CAN messages. Each column corresponds to a specific CAN signal with an ID.
• Timestamp: Time at which the observation was recorded, with a frequency of 0.1 seconds located in the penultimate column 'Time'.
• Fault Status: Last column called 'Fault_status', indicates whether a fault was present at the time of the observation (1 for fault, 0 for no fault).
• CAN Ids: Last row of the csv containing the CAN ID that each signal belongs to.

Description:

Conversion to Time Series:

The raw CAN log files were decoded and converted into time series data. Each time series entry includes a timestamp and the corresponding CAN message details, formatted for straightforward usage in fault detection algorithms.

Usage:

Researchers and practitioners can use this dataset to develop and evaluate fault detection methods for CAN systems. The time series format simplifies integration with machine learning models and other analytical tools, promoting reproducibility and ease of use.

This dataset provides a valuable and open resource for advancing fault detection methodologies in automotive and other interconnected systems. The converted time series data enables direct application in various analytical frameworks, fostering innovation and enhancing the reliability of fault detection solutions.

Funding

This research was sponsored in part by Oak Ridge National Laboratory’s (ORNL’s) Laboratory Directed Research and Development program and by the DOE. There was no additional external funding received for this study. This research used birthright cloud resources of the Compute and Data Environment for Science (CADES) at the Oak Ridge National Laboratory, which is supported by the Office of Science of the U.S. Department of Energy under Contract No. DE-AC05-00OR22725. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of this dataset.