Beam hardening is a common phenomenon in the application of industrial computerized tomography (ICT). It can cause different grayscale values in components with the same density, which affects the threshold segmentation and later reconstruction of components seriously. Herein, study was conducted on the distribution law of grayscale values under the effect of beam hardening, so as to segment the CT scanning data of coal and rock samples under the beam hardening effect accurately. Through the study, it is found that the change in grayscale value on the grayscale beam could truly reflect the change in component density, which is proved by theoretical derivation. Based on this, the threshold segmentation method was proposed for the grayscale beam. Specifically, the threshold segmentation method of grayscale beam is to discrete the three-dimensional grayscale data volumes through CT reconstruction as one-dimensional grayscale beams at first, then segment and multivalue them with the appropriate full-threshold segmentation method according to the number of target component types, and finally gather the multivalued one-dimensional data volume for three-dimensional data. Thereby, the components can be distinguished through the different values in the three-dimensional data that represent different components. Herein, the threshold segmentation method of grayscale beam was used to segment and reconstruct the scanning data of coal-rock combination under the effect of 6 beams, with the validity of this method verified. The research results could provide a reference for the accurate segmentation of CT scanning data of non-homogeneous coal, rock and other materials.