Adaptive multi-model diagnosis using Monte Carlo method

Commonly, the models of hybrid system switch according to a finite state Markov chain with known transition probabilities. For state estimation of hybrid system with unknown transition probabilities, an adaptive estimation algorithm is proposed based on Monte Carlo particle filtering. The proposed algorithm assumes that the prior distribution of unknown transition probabilities follows Dirichlet distribution. First, a set of random samples of model sequence is achieved by sampling. Second, the prior transition probabilities are calculated by the frequency of model transitions in model sequence samples. Third, the posterior estimation of transition probabilities is achieved via measurement update and selection. Finally, the posterior estimation of state and model probability is obtained by particle filtering. In the state monitoring and multiple faults diagnosis of a class of hybrid system, the proposed method has been proved to be very effective.