For the problem of high conflicting evidence combination which exists in fault diagnosis of large manufacturing equipment, an improved fault diagnosis algorithm based on information fusion is proposed. To begin with, the algorithm tries to obtain average belief with the weighting factors based on the credibility of each evidence. With the relative distance of each evidence, a dynamic parameter is constructed, which can reflect the conflicting intensity of evidence, and then the algorithm obtains dynamic weight coefficients. Finally, iteration fusion is adopted with the weighted evidence according to D-S evidence theory. Meanwhile, in order to make the evidence reflecting the fault features of the equipment objectively, the mass functions of the evidence are acquired from the similarity among the patterns. Experimental result indicates this method can reduce the conflicts among evidences effectively, which has high recognition rate for large manufacturing equipment fault and shows good practical value.