Abstract: Aiming at the problems of rare faulty samples, low accuracy of fault prediction and difficulty in maintenance and repair of liquid rocket engine test stand，based on the analysis of the advantages and disadvantages of standard RVM, a fault prediction ARVM model with strong adaptive ability is proposed. In order to test this model, the thrust vector of the high-altitude simulation test stand of a certain orbit control engine is predicted using the flow rate and combustor pressure as input parameters. The prediction result shows that, ARVM method can effectively track the trend of the thrust vector parameters, and obtain a higher prediction accuracy and model sparseness. This method has certain theoretical value and engineering application significance for fault prediction and maintenance of complex systems.

Key words: Liquid rocket, Engine test-stand, ARVM, Fault prediction