To take full advantages of much health information in the practical engineering, a combination method is presented, which is used to assess health condition of complex system. The process of combination assessment based on multi-types of evidences is analyzed firstly. Standardized methods of different types of uncertainty indexes are proposed, and different weights are given to different indexes by means of calculating information entropy. Then, the basic belief transformation method of interval gray number indexes is proposed. Finally, all types of evidences are synthesized in the same recognition framework. The application analysis shows the feasibility and effectiveness of the proposed model.