Abstract: In this paper, we analyzed the dynamic behavior of the ball, and then derived the correlation between the preload and the friction coefficient of linear rolling guide. As the preload cannot be directly measured and was hard to be adjusted once of the linear rolling guide was assembled, we designed a novel preload adjustable loading device mainly consisting of an ultra-precision flat jaw forceps (HERBERT MPV160A), two extrusion blocks, two cylindrical blocks, a separated slider (THK SHS35V) and a pressure sensor (FUTEK-LCA300). Combined with the main worktable and the S-type force sensor (FUTEK-LSB350) and the ball screw feeding system, a synchronous online measurement system of both the preload and the preload drag force was constructed. In such a system, the slider was fixed on the ultra-precision flat jaw forceps and the preload can be adjusted and measured by the pressure sensor (FUTEK-LCA300); the rail was connected with the ball screw nut through the main worktable and the force sensor (FUTEK-LSB350), and can move back and force along the screw axis with the nut when the servo motor rotated, during which the preload drag force of the linear rolling guide can be measured by the force sensor (FUTEK-LSB350). Accordingly, the preload and the preload drag force can be measured simultaneously, based on which the friction coefficient under different preloads and running speeds can be obtained.　In the experiment, the friction coefficient of linear rolling guide THK SHS-35V was measured under four different preloads (1 600 N, 3 100 N, 6 100 N and 9 600 N) and five different running speeds (1 200 mm/min, 2 400 mm/min, 3 600 mm/min, 4 800 mm/min and 6 000 mm/min). The experimental results showed that, the friction coefficient increased with increasing preload at a certain speed, while firstly decreased and then increased with increasing speed at a certain preload. Due to the geometry error of the rolling element and raceway, the preload and preload drag force were always changing when the position of slider on the rail varied. The higher the preloading level, the lower the varying range of the preload and preload drag force. The detailed test data were illustrated in the appendix, and the following main conclusions can be drawn as follows.　Under a certain preload, the friction coefficient firstly decreased and then gradually increased as the running speed increased. For the linear rolling guide of THK SHS-35V used in this paper, which was lubricated at a speed of less than 6 000 mm/min and a viscosity grade of 100 cst, the friction coefficient varied from 0.002 885 to 0.004 325. When the preload increased to about 15% of the dynamic load rating, the friction coefficient at each running speed became stable and was around 0.004 3.　Under a low preloading level, the running speed had a great influence on the friction coefficient of the linear rolling guide. While the preloading level increased, the influence of the running speed on the friction coefficient was decreased. Therefore, when the speed of the linear rolling guide was high, the preload should be set at a relatively high level to obtain a more small and stable friction coefficient. For the linear rolling guide THK SHS-35V used in this paper, the preload should be set at a preloading level higher than 10% of the dynamic load rating.　More importantly, there existed an approximate linear relationship between the friction coefficient and the preload of the THK SHS-35V linear rolling guide, which can be approximated as μ=1.014×10⁻⁷ F_H+0.002 827.　In particular, this study was carried out under a lubrication state with a kinematic viscosity of 100 cSt, and did not consider the influences of frictional heat generation, lubricating oil properties on the friction coefficient, so further research is needed.