Search results

To improve accuracy and efficiency of multi‐fault recognition and localization for large‐scale system such as satellite.

First, fault propagations of a system are modeled by a digraph, which composes of nodes and arcs. Each arc is associated with information about propagation probability and propagation strength. Then, based on consistency‐based theory and semantic theory of abstractions, hierarchical diagnosis model of a system is built. Finally, according to a two‐way hierarchical diagnosis strategy, two incorporated algorithms are adopted which are the Lagrangian relaxation algorithm and the “method of propagation strength”.

Hierarchical model can greatly improve efficiency of diagnosis compared with un‐hierarchical one. The combined qualitative and quantitative knowledge can improve fault resolution.

The propagation probability and propagation strength must been known.

The method shows its superiority when it is applied to complex system such as spacecraft.

A novel hierarchical framework for large‐scale system multi‐fault diagnosis, which include some new ideas and algorithm is put forward.

The purpose of this paper is to resolve the problem of complex system diagnosis with unobservable events such as Hall thruster.

Fault class is defined by unobservable transitions subset. Every fault class is departed into four levels of alarm. The level of alarm for a fault class is identified by constructing basis reachability graph and computing j vector and basis marking. Certain factor values of diagnosis results are given by introducing two fuzzy operators. A two‐directional reasoning strategy is adopted which are forward reasoning and backward reasoning.

Using basis reachability graph and notions of basis marking and j vector, fuzzy Petri net with unobservable transitions can resolve the diagnosis problems with high accuracy and efficiency.

The relationship between unobservable events must be known.

The method shows its superiority when it is applied to unobservable systems such as spacecraft.

A novel fuzzy Petri net framework for unobservable system multi‐fault diagnosis is put forward which introduces some new ideas.