Multisensor Condition Monitoring Dataset from a Sulzer Industrial Centrifuge at Arab Potash Company: A One-Year Time Series for Predictive Maintenance Research

This paper presents a publicly accessible industrial condition monitoring dataset collected from a Sulzer centrifuge at the Arab Potash Company (APC) in Jordan. The twelve-month dataset (September 2022-August 2023) was acquired using an ABB 800xA distributed control system (DCS) sampling four sensor channels: torque (% of nominal load), vibration velocity (mm/s RMS), and dual bearing temperatures (°C, solid and liquid ends) at one-minute intervals (≈0.0167 Hz), yielding approximately 525,600 observations per channel. The preprocessing pipeline included timestamp synchronisation, duplicate removal, and linear interpolation for short gaps; records exceeding DCS alarm thresholds (vibration ≥12 mm/s, torque ≥85%, temperature ≥85 °C) were retained and labelled as candidate anomalies rather than deleted, to preserve event realism. The dataset is internally consistent, plant-specific, and suitable for exploratory benchmarking in industrial AI, digital twin development, and prognostics and health management (PHM) research.