For multiple-input multiple-output(MIMO) systems whose dynamic processes have conspicuous time-delay and inertia, the conventional fuzzy control method is hard to establish control rules and the control effect can not be satisfied. In this paper, a predictive control based on the decentralized fuzzy inference(DFIPC) method is proposed for MIMO control objects. A cluster of a decentralized fuzzy inference modules corresponding to the outputs of a control object are established. Each decentralized fuzzy inference module contains a set of decentralized fuzzy inference units, and decentralized fuzzy inference are carried out based on the deviation between the expectation in the future time and the predicted value of each output, respectively. In the time domain, the inference results are weighted and synthesized based on the dynamic response degree, and thus the equivalent control inputs of the system are generated. Further, the control increaments of system inputs are generated by weighted synthesis of equivalent control inputs. The difficulty of establishing the control rules of the fuzzy inference system in processing multi-input information is resolved. Simulation experiments are proceeded on MIMO objects with obvious time-delay and inertia to verify the effectiveness and adaptability of the proposed method.