Abstract:

Fault prediction needs to complete two mapping processes, one is from state space to trend space, and the other is from trend space to fault space. To make the latter mapping results more accurately, a fault mapping model based on correction high dimensional cloud is proposed on the basis of multivariate normal distribution theory. Entropy and hyper entropy are replaced by covariance entropy and covariance hyperentropy in the correction cloud model, which are used to describe the correlation of some domains. The fault mapping model is constructed by the multi-rule-based reasoning system composed of the correction high dimensional cloud generators. Finally, the trend prediction values are mapped to the fault space accurately to generate the final prediction results. In order to verify the validity of the model, the simulation and analysis of rudder damage are performed. The results demonstrate the proposed model has high mapping precision and efficiency.