Research On Dynamic Fusion Based On Evidence Reasoning And Evidence Updating Rules

The classic Dempster-Shafer evidence combination rules can only fuse evidences that is "static" or "symmetric", and they never consider the time of evidence obtained and dynamic changes of the relationship of fusion result. So they are not suitable for the dynamic fusion of multi-source information, such as fault diagnosis and state estimation and so on. As so far, most of the existing studies have focused on proposing kinds of the fusion rules or algorithms, but paying little consideration on how to give the appropriate evaluation index to measure the performance of the static or dynamic fusion algorithm. How to solve these problems effectively is the key to realize the dynamic fusion of multi-source information by using evidence theory. To solve these problems, this paper studied the dynamic fusion method based on evidence reasoning and evidence updating rules, and applied it to solve the problem of dynamic fusion in the on-line fault diagnosis and state estimation of the system. The main work is as follows:(1) Fault diagnosis by combining static and dynamic fusing strategies. In static fusing strategy, Dempster rule is used to obtain static fused evidence and belief degree convergence index of static evidence is defined based on distance of evidence. In dynamic fusing strategy, the updating rule with conditional linear combination is used to obtain updated global diagnosis evidence and dynamic belief degree convergence index of updated evidence is defined based on S function. In the two strategies, the optimal learning method is given based on the corresponding convergence index function, and we can obtain the optimal value of the corresponding parameters. Finally, diagnosis decision-making can be made via the updated evidence. In diagnosis experiment on a rotor test bed, the proposed method is compared with some classical information fusion method to show the effectiveness of the proposed diagnosis method.(2) Dynamic fault diagnosis method based on interval-valued belief structure. The interval-valued belief structure(IBS) is proposed to solve the imperfect and coarse measurement of fault information by the single-valued belief structure(BBA). The diagnosis evidence is constructed as interval-valued belief structure(IBS), which is more applicable than the single-valued belief structure(BBA) to describe fault information. Then, the proposed evidential updating strategy can generate the global diagnosis evidence (the updated IBS) by recursively updating the previous (old) evidence with the incoming (new) evidence. Finally, the fault experiments of machine rotor show the effectiveness of the proposed updating method.(3) Fusion estimation method based on evidential reasoning rule. This method regards the dynamic system equations and the actual observations of the system states as two information sources. The random set description of evidence and the extension principle of random set are used to recursively generate state evidence and observation evidence respectively from the two information sources and to propagate them in system equations. At each time step, the ER rule is used to fuse the two pieces of evidence in observation domain and then the fused result is transformed to state domain by the extension principles. Finally, we can obtain state estimation value by Pignistic expectation. The method is shown to have better performances in an application to liquid level estimation of industrial level apparatus than does the Nassreddine's method.