Abstract:In order to reasonably classify the line grade and distribute limited resources of maintenance, fusing detection data sources of geometrical status which were obtained by existing means, the track geometry comprehensive index (TGCI) was proposed, the initial weights of parameters of TGCI were determined by AHP, and a weight sensitivity analysis for parameters of TGCI was conducted by using the perturbation method. The quantitative study on the relationship between changes in weights of parameters and percentage of different line grades, and initial weights were modified. The results show that changes in weights of TQI has the most significant effect on line grade classification. The weight range of TQI should be no less than 0.2, the weight range of data overrun of dynamic detection should be no more than 0.45, and the internal weight distribution ratio of each parameter is 1∶1 and 3∶2. The weight of vehicle detection should be less than 0.3, and the weight of locomotive shaking monitor meter should not be less than 0.1.