Abstract: To evaluate the tunnel rock quality, six parameters reflecting the general properties of rock engineering was selected to build the decision table, which was evaluated by extenics theory and expert examination, and rough sets theory was applied to reduce the original decision table and to analyze the relative importance of every parameter. Finally, the reduction results are transformed into rules, which are used as input of the BP neural networks. Combining rough sets theory with artificial neural networks, then the evaluation model of tunnel rock quality was established. Through the case study, the model can efficiently simplifies the networks structure, reduces the networks training period and has better study efficiency and can more precisely reflect the engineering characteristics of tunnel rock．

Key words: ]rock quality evaluation, rough sets, data reduction, artificial neural networks