Hose-Vibration Dataset: Triaxial ESP32–ADXL345 Recordings for Six-Class Operational State Classification

This dataset contains triaxial vibration measurements collected from two flexible water hose of 1 inch (wide) and 0.5 inch (narrow) diameter under five operational states. The signals were recorded using an ESP32–ADXL345 sensor module mounted with a rigid 3D-printed clamp to ensure consistent mechanical coupling. The objective of this dataset is to support research on lightweight vibration-based diagnostics, condition monitoring, and small-footprint classification models. The dataset is provided as a single combined CSV file consisting of normalized and label-corrected time-series windows, ready for machine learning workflows. Contents The dataset includes: Three-axis acceleration signals (columns: x, y, z) Window-level class labels derived by majority voting over raw samples Balanced representation across all classes Class Definitions (6 Classes) Class ID Description 0 Normal flow (no fault) 1 Twisted 2 Blocked 3 Loose Join 4 Minor leak 5 Large leak Data Collection Setup Sensor: ADXL345 accelerometer Controller: ESP32 microcontroller Mount: Custom rigid clamp to eliminate placement variability Environment: Indoor laboratory with steady water flow (≈1.5–2.0 bar) Pipes Used: Wide-radius and narrow-radius hose geometries Use Cases This dataset is suitable for: Lightweight deep learning models (e.g., 1D CNNs) Classical ML baselines (e.g., Random Forest, SVM) Domain adaptation / cross-condition generalization Few-shot learning for IoT sensing Vibration-based fault detection research