A benchmark dataset of electrical signals from a permanent magnet synchronous generator for condition monitoring

The proper monitoring of sensitive components in rotating electrical machines plays a critical role in preventing internal faults that may lead to irreversible damage and unplanned shutdowns. Offshore wind power generation is increasingly adopting permanent magnet synchronous generators (PMSGs) because of their high efficiency and low maintenance requirements. However, internal short-circuit faults remain a challenge and require effective fault detection strategies. Inter-turn and inter-winding faults, in particular, may not cause immediate damage but can evolve over time, leading to severe equipment failures. These failures may require generator shutdowns, resulting in significant financial and operational losses. This dataset provides high-resolution electrical measurements from a PMSG under healthy and faulty conditions, supporting the development and validation of related diagnostic and control strategies. We collected data from a laboratory test bench that allows controlled insertion of internal faults, such as short-circuits between turns and windings. The generator, connected to the grid via a power converter, was monitored using an Imperix B-Box RCP system loaded with control algorithms developed in Simulink. Signals were sampled at 20 kHz and recorded through the Imperix Cockpit, with each test lasting three seconds and capturing pre-fault, fault, and post-fault conditions. This structure enables users to study transient responses, steady-state behavior with faults, and system recovery. The dataset comprises 225 .mat files covering 24 fault cases and one healthy case, each tested under three torque levels and three rotational speeds. The selected operating conditions reflect typical points of a 15-megawatt offshore wind turbine. In addition to the raw data, the dataset includes a Python interface to facilitate visualization. The dataset can support diverse applications, such as validating analytical models of PMSGs, benchmarking fault detection algorithms, and generating synthetic data for further testing. It may also serve as a practical tool in electrical engineering education, especially in courses focused on wind energy systems and fault analysis.