"BiWheel-IMU-Fault"

"The BiWheel-IMU-Fault dataset contains synchronized tri-axial accelerometer and gyroscope time-series data collected from low-cost MPU6050 IMUs rigidly mounted on opposing wheels of a test vehicle. Data were acquired at ~100 Hz using an ESP32-based logger during controlled drives at 10\u201320 km\/h on varied road conditions, including normal operation and induced wheel-specific faults (flat tires, imbalances, and alignment issues).The dataset includes:Raw sensor signals (Ax, Ay, Az, Gx, Gy, Gz) from paired left\/right wheelsProcessed differential (bilateral) features, such as variance ratios in vertical acceleration (\u0394std(Az)) and cross-wheel gyroscope discrepanciesLabeled data windows for anomaly detection and specific fault typesSupporting files: raw CSVs, feature-extracted datasets, and visualization scriptsThis resource enables research on lightweight, edge-based unsupervised machine learning (Isolation Forest and undercomplete Autoencoders) for proactive wheel fault monitoring. Its key innovation is bilateral sensing, which suppresses shared environmental noise (road bumps, vibrations) through relative comparisons, achieving improved accuracy and significantly fewer false positives compared to single-wheel approaches."