In dynamic mechanical measurements, due to the accuracy limitations of the sensor itself, environmental noise, and changes in measurement conditions, the measurement data of a single sensor often contains errors, which will affect the performance evaluation and fault diagnosis of the mechanical system. To address this problem, this paper proposes a dynamic error modeling and compensation method based on sensor fusion. This method uses the system identification method to dynamically model the sensor. On this basis, multi-sensor fusion technology is introduced to comprehensively process the data of different types of sensors. Error compensation technology is used to separate, correct and suppress errors in the fused data to further reduce measurement errors. The sensor fusion-based approach significantly reduces measurement errors in dynamic mechanical measurements. Specifically, the MPE value based on sensor fusion is between 0.05 mm and 0.24 mm, while the MPE value based on a single sensor is between 0.25 mm and 0.45 mm. Sensor fusion technology reduces the error introduced by a single sensor by integrating the data of multiple sensors. In addition, the measurement time based on sensor fusion is always lower than that of the single sensor method in different dynamic machines, showing that it also has advantages in measurement time. This paper not only improves the accuracy and efficiency of mechanical measurement, but also provides a means of error control for the fields of intelligent manufacturing and precision engineering, helping to promote the development of industrial automation technology.