A first industrial machine (110) provides historical event data (130) to a model generation computer (101) that generates a prediction model. A behavior prediction computer (102) uses the model (150) in combination with past and current data (140) from a second machine (120) - the industrial machine under supervision (IMUS) - and provides feedback to the IMUS. Both machines (110, 120) have common properties. Generating the prediction model comprises to obtain event features and event vectors, to cluster the vectors and to assign clusters to machine states, and to identify probabilities of machine state transitions. Features, vectors and clusters are processed by processing techniques, with some of the techniques are natural language processing techniques (NLP).
展开▼