Uncertainty Measure And Fusion Of Uncertain Information In The Evidence Theory Framework

Uncertain information exists widely in practical engineering applications such as air combat situation assessment.Uncertain information includes the fuzziness represented by fuzzy sets,the ambiguity information because of discord or non-specificity in semantic inconsistency and the incomplete information in the open world assumption.Dempster-Shafer evidence theory-based uncertain information processing theories and methods have been widely researched among researchers.Currently,most of the researches focus on the closed world assumption.More and more researchers are paying attention to the uncertain information processing in the open world assumption with the increasing importance of the open world features of uncertain information.This dissertation researches on the classification of uncertain information,the category of uncertain information should be extended to three classes which are fuzzy information,ambiguity information and incomplete information in the open world assumption from the closed world assumption where there are only two categories of uncertain information including fuzzy information and ambiguity information.After that,this dissertation proposes a union framework for uncertain information processing in the open and closed world assumption.This dissertation proposes new methods for evidence generating,evidence measure,evidence fusion and applications in the open and closed world framework.The contribution of this dissertation mainly includes the following four aspects.(1)To measure the uncertainty of uncertain information in the closed world framework,this dissertation proposes the weighted belief entropy based on the scale of the frame of discernment and the improved belief entropy based on the non-specificity of the body of evidence.Both the weighted belief entropy and the improved belief entropy can take into consideration of more available uncertain information in the body of evidence,while the improved belief entropy can measure the difference among body of evidence with respect to the non-specificity.The proposed methods can be more accurate while measuring the uncertain degree of ambiguity information,which can be helpful to enhancing the accuracy and ability of uncertain information fusion in practical application.The new uncertainty measure-based conflict data fusion methods for sensor data-based fault diagnosis as well as for the case study in decision-making verify the rationality and effectiveness of the proposed methods.(2)To address the uncertain information in the closed world,this dissertation proposes a new method for measuring the ambiguity degree of experts’ assessments as well as fusing the ambiguity information.Firstly,the proposed method addresses the conflict feature and non-specificity characteristic of the knowledge from experts based on the ambiguity measure.Then,the assessments of experts after measuring its uncertainty can be fused by design a proper evidence fusion method.Finally,an improved risk evaluation method based on failure mode and effects analysis(FMEA)is proposed and applied in the aircraft turbine rotor blade,the experimental result verifies the availability and effectiveness of proposed methods.(3)This dissertation proposes a new method for uncertainty measure of evidence function with respect to the incomplete uncertain information in the open world assumption,named the extended belief entropy for the open world(EBEOW).According to the literature,EBEOW is the first method designed for measuring the uncertainty of evidence in the open world assumption.Thus,it can measure the uncertain degree of evidence function in the open world assumption which is beyond the ability of those methods for uncertainty measure in the closed world.Numerical examples and an application of evidence measuring and fusing in fault diagnosis verify the rationality and value of the proposed measure.(4)This dissertation proposes a new method to address the uncertain information with incomplete frame of discernment(FOD)in the open world assumption.The proposed method can address the new items or unknown item in the incomplete FOD in the open world assumption.If the mass value of the empty set is non-zero,then the uncertain information processing is under the incomplete FOD in the open world assumption.Otherwise,the uncertain information processing is under the complete FOD thus it can be degenerate to the closed world assumption.To illustrate and verify the availability and effectiveness of the proposed method,an improved risk priority number of an improved FMEA method for uncertain,unknown or incomplete risk factors is proposed in the open world assumption.Finally,to address the dynamic uncertain information in the air combat situation assessment,the improved belief entropy-based air combat situation assessment method in the closed world assumption and the EBEOW-based air combat situation assessment method in the open world assumption are proposed respectively.Simulation results verify the theoretical significance of the proposed methods and its value in engineering applications.