A smartphone-based vibration dataset for induction motor fault diagnosis under different speed and load conditions
Description
This dataset provides smartphone-based vibration measurements collected from a three-phase squirrel-cage induction motor operating under multiple fault conditions, rotational speeds, and load states. The primary objective of the dataset is to support research on low-cost and accessible condition monitoring and fault diagnosis of induction motors using inertial sensors embedded in consumer-grade smartphones. Vibration signals were acquired using an iPhone 15 Pro Max equipped with a tri-axial accelerometer, with data recorded via the Sensor Play mobile application. Data were recorded along three raw acceleration axes (gx, gy, gz) as well as gravity-compensated linear acceleration components (guserx, gusery, guserz). All signals were sampled at 100 Hz, and each measurement corresponds to a continuous recording of 15 minutes. The smartphone was firmly attached to the motor housing using double-sided adhesive tape to ensure consistent coupling during data acquisition. The dataset includes six motor health conditions: (i) Healthy operation (H), (ii) Insufficient bearing lubrication (B1), (iii) Severely insufficient bearing lubrication (B2), (iv) Cracked bearing outer ring (B3), (v) Voltage imbalance (V), and (vi) Broken rotor bar (R). Measurements were conducted at three operating speeds (30 Hz, 40 Hz, and 50 Hz) under both loaded and unloaded conditions, covering all combinations of fault type, speed, and load (36 operating scenarios in total). For each fault class, data corresponding to the six speed–load combinations are provided, and each recording is available in both .mat and .csv formats. As a result, each class folder contains 12 files, consisting of paired .mat and .csv files for the same measurement. File names follow a standardized convention: Class_Load_SpeedHz, where the load condition is indicated by “1” (loaded) or “0” (unloaded). This organization facilitates straightforward access to the data and enables seamless integration into signal processing, machine learning, and statistical analysis workflows. The dataset is intended for researchers working on vibration-based fault diagnosis, signal processing, and machine learning applications in electrical machines. By relying solely on smartphone inertial sensors, it offers a practical and cost-effective alternative to conventional industrial accelerometer-based datasets, enabling benchmarking, comparative studies, and the development of accessible motor condition monitoring solutions.