Considering the problems of “difficult acquisition of test data and diverse failure mechanism” of the radar seeker of a missile, we proposed a competition failure state prediction model related to degraded failure and sudden failure based on the analysis to the characteristics of the degraded data and failure data.Aiming at the characteristics of small sample, nonlinearity, and uncertainty of the degraded data, we used Relevance Vector Machine (RVM) to regress and predict the distributed parameter sequence.Then the kernel parameter and embedded dimensions were both optimized via Quantum-Behaved Particle Swarm Optimization(QPSO) and Hannan-Quinn(H-Q) criterion.Taking into account the correlation between the sudden failure and the degraded failure, a position-scale model was introduced to describe the correlation between the sudden failure distribution parameters and the degraded data, thus to predict the failure state of the next stage.Prediction result of an example verified the feasibility and rationality of the model.