VIBRATION DATA IUT DOUALA
Description
This dataset contains vibration signals collected from a three-phase asynchronous electric motor operating under various health and fault conditions. The data were acquired from an experimental test bench developed at the Electrotechnical Laboratory of the IUT of Douala. Vibration measurements were obtained using a high-sensitivity piezoelectric sensor (approximately 100 mV/g, bandwidth 0.5 Hz–10 kHz) mounted on the motor shaft. The analog signals were digitized using an Arduino Uno microcontroller equipped with a 10-bit analog-to-digital converter (ADC), producing values ranging from 0 to 1023. Data acquisition was performed at a sampling frequency of 5 kHz. The dataset consists of approximately 1.6 million data points distributed across five operating conditions: normal operation and four mechanical fault types (ball fault, ring fault, bearing fault, and shaft fault). Each fault condition is further characterized by two severity levels (“speed 1” and “speed 2”), resulting in nine distinct states. Each condition contributes approximately 200,000 data points. The dataset is structured into four main variables: time (timestamp), vibration signal, fault type, and severity level. For machine learning applications, the data are provided in CSV format and organized into training, validation, and test subsets, ensuring a balanced and representative distribution of all classes. To enhance realism and robustness, Gaussian white noise has been added to the signals, simulating industrial operating environments. The dataset is suitable for applications in fault diagnosis, condition monitoring, and predictive maintenance using machine learning and deep learning techniques.