Information Fusion Method Based On Interval BPA Evidence Theory Research And Application

Evidence theory is widely used by virtue of its advantages in uncertainty information processing.The fusion rules of classical evidence theory are only applicable to the synthesis of single-valued evidence,but in the context of practical engineering applications,single-valued evidence often loses a lot of useful information,so interval BPA evidence theory has become a hot topic of current research.Interval BPA evidence not only conforms to conventional human thinking,but also can retain more useful information and fuse decisions to obtain more accurate results.However,the current research on this theory,especially the generation of interval BPA,normalization of interval BPA,and fusion of conflicting interval evidence,still needs further improvement.To address the above issues,this paper conducts a study on interval BPA evidence theory.Firstly,to address the problem that the classical single-valued BPA is coarse in its description of uncertainty information and will lose a lot of useful information,resulting in unreasonable generated single-valued BPA,an interval BPA generation method based on the bell-type affiliation function is proposed.The method introduces bell-type affiliation function as the target attribute feature model,and generates interval BPA by measuring the match between the test sample and the training sample.Finally,the method is applied to a fault diagnosis example and compared with the single-value BPA generation method,and the experimental results prove the effectiveness and accuracy of the proposed interval BPA generation method.Secondly,an interval BPA normalization method based on an optimization factor is proposed to address the problems of information loss and incomplete utilization of the traditional interval BPA normalization method.The method is based on the definition of the minimum Euclidean distance between the original interval BPA and the standard interval BPA to find the optimization factor,which is used to optimize the original interval BPA and perform global normalization to obtain the standard interval BPA.The validity and superiority of the method is verified.Finally,a method of fusing conflicting interval BPA evidence based on correction coefficients is proposed for the fusion of conflicting interval BPA evidence.The method first obtains the credibility and falsity based on the similarity and conflict degree among individual evidences,then obtains the correction coefficients by combining the credibility and falsity analysis,and then uses the correction coefficients to fuse the evidences after correction,and gives the selection method of single-value BPA in order to facilitate decision-making.Finally,the feasibility,validity and practicality of the method are demonstrated using several classical conflict calculations,and the practicality of the method is demonstrated by applying it to the interval BPA evidence fusion obtained from the fault diagnosis example.