This paper focuses on the geological adaptive control of tunneling boring machine (TBM). To deal with the issue that the geological condition is uncertain, clustering analysis is used to identify geological types based on two important indices indicating the strength of rock. Based on the results from above, the principles of variation and range of tunneling parameters under different rock condition are determined. Furthermore, considering several vital performances of TBM during tunneling operation, a multi-objective optimization problem is proposed. In the light of non-dominated sorting and crowded distance evaluation concepts in non-dominated sorting genetic algorithm-II, the proposed multi-objective optimization problem is solved by using differential evolution algorithm. Based on the practical construction data, the simulation results show that the proposed method is effective in improving the performance of tunneling operation of TBM, compared with unoptimized operating actions from current systems.