Abstract: The actual coal-rock is mainly in the form of assemblage, and the characteristics of the EMR generated during loading are clearly related to its stress state. In order to reveal the EMR generation mechanism of coal-rock assemblages, this study first analyzes the microstructure and element proportion of coal-rock assemblages, and explains the EMR generation process of different single coal-rock materials. Then, a time-varying model for the EMR of coal-rock assemblage is established, and its rationality and universality are verified by several uniaxial experiments. The results are as follows: In terms of structure, the number of original fractures in the coal seam of the coal-rock assemblage is significantly higher than that of the rock seam. And in terms of composition, while the total fraction of C, Si, and O in the coal-rock assemblage is close to 75% and they are the dominant components, the fractions of all three are significantly different in single coal-rock materials. Both of them increase the complexity of the EMR generation process for loaded coal-rock assemblies. The EMR of the loaded coal-rock assemblages is a composite signal, which is the result of the piezoelectric effect of the rock, the friction effect of multiple fractures in the coal seam, and the magnetization effect of ferromagnetic materials. This results in EMR having both fluctuation and nonlinear time-varying characteristics, that is, the volume composition ratio and loading rate have an impact on the distinctiveness of these two characteristics. The higher the proportion of the coal in the volume composition ratio, the stronger the fluctuation of the EMR, and the overall trend is closer to that of the loaded single coal. And the loading rate increases, the stability of the EMR decreases, with the average standard deviation decreasing by 33.01%, which is most obvious at 0.3 mm/min. At the same time, the fluctuation is most obvious within the first 60% of the loading time. The trend of the experimental results is in good agreement with the theoretical analysis, effectively validating the plausibility of the time-varying model for the EMR intensity of coal-rock assemblies. The above results are helpful to explain the EMR evolution law of coal-rock under different loading conditions, and also lay a foundation for study of multi-source signal coupling mechanism of coal and rock dynamic disasters.