Abstract: Rock mass quality classification is a non-linear and complex uncertain system problem.Its classification index parameters are uncertain, fuzzy and stochastic.On the basis of fully considering the correlation of rock mass quality classification indexes, the reliability measurement weighting theory based on index distance is established, and the accurate weights of each index are calculated, which reduces the risk of large deviation between the evaluation results and the actual results caused by errors or errors in the actual process.According to set pair theory, the traditional cloud model is improved, and the concept and algorithm of set pair cloud model are proposed.Through simulation analysis of 25 sets of sample data, it is found that the evaluation results coincide with those of approximate ideal solution ranking method and fractal interpolation method, which proves the reliability of the algorithm.The model is applied to the evaluation of rock mass quality in Xilinhot fluorite mine.The predicted results are consistent with the actual situation, which shows that the algorithm has practical engineering value and can be applied to the classification of rock mass quality in underground engineering.At the same time, the algorithm can judge the authenticity of the index data according to the classification function of index reliability, which can provide theoretical guidance for rock engineering parameter acquisition.