The Data Fusion Method Research Based On D-S Evidence Theory

Dempster-Shafer evidence synthesis theory is widely used in the fields of information fusion, detection diagnosis and artificial intelligence because of its advantage of expressing and combining in the uncertainty problem. But it will get the counter-intuitive answer even the wrong answer in the case of high conflict evidence. To solve this issue, there is an evidence synthesis method base on the cosine distance. The method uses the cosine distance and threshold value to identify the conflicts then correct the conflict evidence by the correlation between evidences. The non-conflicting will be reserved. Finally, the fusion calculation is completed by using Dempster-Shafer evidence rule.There is another evidence synthesis method base on the information entropy and evidences clustering. The method uses the information entropy to identify the evidence similarity matrix and get the conflict evidence. Then the evidence source is divided into two clusters by clustering value. They are non-conflict evidences and conflict evidences. In each cluster, the method uses the Dempster-Shafer evidence rule to synthesize the evidences. After that, it uses the weighted coefficient to synthesize the evidence clusters and gets the final result.The experiments show that the two evidence synthesis methods can effectively calculate the evidences source with high conflict evidence especially in the case of the evidences source with multiple conflicting evidences. At last, let these two evidence synthesis methods into the automotive intelligent fault diagnosis system. Through the practical application cases, it is proved that the two methods are feasible.