Method and apparatus for classification and prediction of health of a physical system

A diagnostic system comprising:

• a sensor module configured to measure an excitation of an operating physical system and measure a response of the operating physical system to the excitation, and
• a computing apparatus configured to process a trained classification model that receives, as input, the measured excitation and response from the sensor module, and produces an output indicative of a predicted health of the operating physical system,
• wherein the classification model is trained by:
• obtaining simulated healthy measurements by a physics-based simulation model of a physical system from measured input excitations to the physical system,
• training a generator model to generate realistic healthy measurements from the simulated healthy measurements and the measured input excitations, the training comprising tuning the generator model based on an error between a generated realistic healthy measurement and a corresponding real measurement obtained from the physical system in response to each input excitation,
• using the trained generator model to generate realistic faulty measurement data from simulated faulty measurements obtained by the physics-based simulation model, and
• training the classification model using a training dataset augmented by the realistic faulty measurement data generated by the generator model.