The control criticality class is probably the first feature that will be pronounced by the spare parts (SPs) logistics practitioners, but it has been found that no systematic and well-structured procedures exist to evaluate the control criticality. This paper presents a systematic and scientific approach to identify the criticality classes of SPs under highly uncertain environment with limited data and information. By using group-discussing and anonymous questionnaire methods, the criteria to evaluate criticality classes of SPs are proposed. A practical and reliable algorithm integrated of AHP, fuzzy comprehensive evaluation and grey relational analysis (GRA) is utilized to convert the qualitative description to quantitative data. Subsequently, its effectiveness and superiority are tested by a practical example. Moreover, case studies are rare for the evaluation of SPs control criticality in large power plants. We also conducted a case study to test the proposed model. Empirical findings suggest that the proposed model is successful in correcting unreasonable criticality classes setting of SPs that are important to inventory control in a famous power plant in China.