In this paper, a new dynamic model was proposed for identifying the rock hardness during the process of roadway tunnelling, thereby regulating the speed of the driving motor and the torque of the cutting head. The presented identification model establishes a multi-information feature database containing vibration signals in the y-axis, acoustic emission signals, cutting current signals, and temperature signals. Subsequently, we obtain the membership functions (MFs) of the given multiple signals with the amount of feature samples according to the principle of minimum fuzzy entropy. Furthermore, a rock hardness identification model was established based on multi-sensor information fusion and Dempster-Shafer (D-S) evidence theory. To prove the accuracy of the proposed model, an identification experiment was carried out through the cutting of a poured mixed rock specimen with five grades of hardness. As a result, the proposed identification model recognizes the rock hardness accurately for fifteen sampling points, which indicates the significance of the method with regard to the dynamic identification of rock hardness during the process of roadway tunnelling, and further provides data support for adjusting the speed of the cutting head adaptively, thereby achieving high efficiency tunnelling.