Based on knowledge decision method, an intelligent diagnosis system for robot is introduced. The system combines intelligence decision technology and knowledge of robot failures diagnosis. One dimension structure of data-base is proposed for the knowledge-base. Each failure symptom has its exclusive record number therefore, the complicated nature language is changed to simple code description respective, and avoids to searching problem for the complicated nature language. The arithmetic of knowledge decision is proposed in the intelligence system for robot failures diagnosis. Considering the factors of failure event probability, and diagnosis knowledge validating weight value, methods are used which is compositor for each failure, and the DFS (deep first search) strategy to validate each failures. By this means, the efficiency of the intelligence system for failures diagnosis is improved.