Dynamic Multiple Fault Diagnosis Based on HMM and BPSO

For systems needed short diagnosis delay, dynamic multiple faults diagnosis (DMFD) is put forward for the systems of demanding diagnosis quickly. In this paper, a method of DMFD based Hidden Markov Mode (HMM) is established for the system´s inner states transform and the corresponding external observing sequence, thus the inner states transform could be recovered from the external observing sequence with the decoding algorithm of HMM, which belongs to NP completeness problems. This paper decomposes original DMFD problem into several separable sub problems, and solves each of them with binary particle swarm optimization algorithm (BPSO). It is shown from the application examples that the system´s real-time health status could be evaluated at a high correct ratio with this method.