Dempster-Shafer　(D-S)　theory　of　evidence　is　widely　used　in　many　real　application　systems.　It　can　not　only　deal　with　imprecise　and　uncertain　information　but　also　combine　evidences　of　different　sensors.　Therefore　it　plays　an　important　role　in　multi-sensor　reports＇　combination　in　fault　diagnosis.　However,　when　the　evidences　highly　conflict　with　others,　Dempster＇s　combination　rule　may　lead　to　a　counter-intuitive　result　and　come　to　a　wrong　conclusion.　It　is　inevitable　to　handle　conflict　in　fault　diagnosis.　This　paper　proposes　a　new　method　to　address　the　issue.　Deng　entropy　function　is　adopted　to　measure　the　information　volume　of　evidences.　Evidence　distance　is　introduced　to　represent　the　compatibility　of　evidences.　An　improved　combination　method　considering　both　the　uncertainty　of　evidences　and　the　conflict　degree　of　the　system　is　proposed.　The　proposed　method　can　deal　with　conflicting　evidences　efficiently.　An　application　in　fault　diagnosis　is　illustrated　to　show　the　efficiency　of　the　new　method　and　the　result　is　compared　with　that　of　other　methods.　Besides,　and　example　in　IRIS　based　on　information　fusion　is　given　to　validate　the　accuracy　of　the　proposed　method.