The main construction disaster of tunnel crossing granite fault zone is the inrush of highpressure and rich water, which is extremely dangerous and destructive. In order to effectively solve the core problem that large errors exist between the predicted values and actual values of tunnel water inflow in theoretical calculation, model test and numerical simulation, 13 independent disastercausing factors for granite fault water inrush in engineering geology, hydrogeology and construction design are systematically identified, and a fine prediction method of tunnel fault water gushing based on optimized FAHPTOPSIS method is proposed. The innovations of the method mentioned above are using average dominance to optimize the judgment matrix of FAHP, which can solve the problem of unbalanced weight distribution caused by mutual exclusion of traditional FAHP ranking method. Meanwhile, the TOPSIS is used to replace fuzzy comprehensive evaluation method, which avoids the problem that the fuzzy comprehensive evaluation method relies on a large number of sample data and model training redundancy, and greatly improves the prediction accuracy of tunnel water gushing. The optimized FAHPTOPSIS method is applied to the prediction of water gushing, and the fine prediction of water gushing grade and water inflow in 6 fault water gushing sections is realized. The prediction results show that: (1) There is medium risk of water gushing in S1, S4, S5 and S6, and large risk of water gushing in S2 and S3; (2) The maximum relative error between predicted water inflow and actual water inflow in 6 fault water gushing sections is 14.8%, and the average relative error is 8.87%, which meets the requirement of engineering prediction accuracy.