The Application of Different Distance Functions in DS Evidence Theory

Jie Chen,Yiliang Li,Mengwei Zhao,Ke Huang,Kunren Wu

DOI: https://doi.org/10.1109/icet55676.2022.9824199

2022-01-01

Abstract:DS evidence theory is a widely used uncertainty inference algorithm for decision fusion that can fuse this complementary and redundant information to reduce the uncertainty of the system. Evidence conflict is the most common problem in the application of DS evidence theory, and how to resolve the conflict is the key to obtaining reasonable and reliable decision results. To resolve the evidence conflicts, numerous scholars have used different distance functions to improve the DS combination rule to effectively manage the conflicting evidence. However, there are only few relevant studies that address the performance of different distance functions for handling evidence conflicts. Therefore, this paper discusses the performance of five distance functions (Euclidean distance function, Manhattan distance function, Minkowsky distance function, Jffreys distance function and Camberra distance function) to handle evidence conflicts by performing simulations in a fault diagnosis system for typical mechanical vibration signals.