Fault state detection apparatus

A method of training a machine learning algorithm for a fault state detection apparatus, the method comprising:

• providing a plurality of non-fault state condition monitoring data and associated ground truth information;
• providing a plurality of fault state condition monitoring data and associated ground truth information, the providing comprising generating a subset of the plurality of fault state condition monitoring data from one or more non-fault state condition monitoring data, and wherein the generating of fault state conditioning monitoring data in the subset of the plurality of fault state condition monitoring data comprises transforming non-fault state condition monitoring data to fault state condition monitoring data, wherein the transformation of the non-fault state condition monitoring data to the fault state condition monitoring data comprises adding a hot spot to a non-fault state infrared image, and wherein the hot spot is added at a random position within the non-fault state infrared image and alters the non-fault state infrared image;
• implementing a machine learning algorithm;
• training the machine learning algorithm on a basis of training data sets including the plurality of non-fault state condition monitoring data and the associated ground truth information and the plurality of fault state condition monitoring data and the associated ground truth information, wherein the training data sets mitigate imbalances between training data reflecting healthy operations and training data representing faulty operations;
• receiving condition monitoring data comprising vibration data;
• analyzing, by the trained machine learning algorithm, the received condition monitoring data to determine if the received condition monitoring data is associated with a fault state, wherein the plurality of non-fault state condition monitoring data comprises vibration data, the plurality of fault state condition monitoring data comprises vibration data, wherein the transformation of the non-fault state condition monitoring data to the fault state condition monitoring data further comprises increasing a vibration amplitude peak, wherein the non-fault state condition monitoring data are represented as a plurality of amplitudes in different frequency bins, wherein increasing the vibration amplitude peak comprises increasing an amplitude in one of the different frequency bins, and wherein the associated ground truth information of the plurality of fault state condition monitoring data includes a type of fault;
• and
• determining the type of fault that led to a respective fault state data of the plurality of fault state condition monitoring data.