In this paper to further enhance the measuring precision of vibration signal in the process of rotary machine fault diagnosis, a new data fusion method is presented on the basis of correlation function to determine the weighted value. The method doesn’t require knowing the prior knowledge about sensors, and the weighted value of sensors can be determined on the basis of the correlation measure of real-time data tested in the fusion process. It gives greater weighted value to the greater correlation measure of sensor signals, and vice versa. Through simulation analysis, the dynamic adaptability, precision, and fault-tolerance of this method are better than the traditional weighted averaging method. Furthermore, this method has good performance of anti-jamming due to the correlation measures between noises and effective signals are usually small. Finally, rotor dynamics and integrated fault diagnosis system is taken as an example to verify the feasibility and advantages of this method with respect to multi-sensor vibration signal processing.