Based on the idea of Multilayer information fusion technology, and from the reality of equipment fault diagnosis, We established neural network evidence fusion fault diagnosis system which is based on information fusion technology. Neural network has good nonlinear mapping ability, and d-s evidence theory has unique advantages in the expression of uncertainty. Both of the two methods have been widely used in the field of fault diagnosis. That is, through the effective combination of the fault feature information, use the seeds of neural network from different sides for equipment fault diagnosis of preliminary, then applying the preliminary diagnosis to Dempster - Shafer theory evidence for decision fusion. The diagnosis example indicates that, after fault feature information fusion, the credibility of the diagnostic increased significantly, and can effectively improve the diagnosis rate.