Abstract: A fault prediction method based on clustering particle filter is put forward to reduce computing cost and predict system faults in time.The k-means clustering algorithm is introduced into the particle filtering,and all the particles are classified into some clusters.The centroid of each cluster regarded as the representative of the particle cluster is updated each time.The number of particles needed to be iteratively computed is decreased sharply,thus the computing cost of particle filtering is reduced.Simulation results verify the validity of the presented clustering-particle-filter-based fault prediction method.