Understory herbaceous species diversity plays an important role to maintain ecosystem services in arid and semi-arid areas of northwest China. Exploring its related influencing factors can provide a theoretical basis for the sustainable management of artificial ecological public welfare forests in the eastern Qilian Mountains. In this study, coniferous pure forest, broadleaf pure forest, coniferous and broad mixed forest and mixed coniferous forest (planted forest) located in the eastern Qilian Mountains were selected as the research objects, and a multivariate parameter prediction model and structural equation non-parametric effect model of herbaceous species diversity were constructed on the basis of correlation analysis. The relationship between multiple factors (stand spatial structure, stand non-spatial structure, and environmental factors) and the degree of influence on the diversity of understory herbaceous species were explored. The results showed that: (1) among them, the number of Coniferous and broad-leaved mixed forest species was the largest, with a total of 132 species in 129 genera in 39 families. And in the coniferous and broad-wide mixed forest, Elymus dahuricus, Lolium perenne, Fragaria vesca, Carex spp. are the dominant species. (2) Pearson correlation analysis was indicating that mean diameter at breast height, mean tree height, angle index, mingling degree, neighborhood comparison, and slope were significantly correlated with herbaceous species diversity (P<0.05). (3) Multivariate parameter prediction model for herbaceous species diversity was indicating that the diameter at breast height was the key factor influencing the Simpson diversity index. Slope was a key factor influencing the Shannon-Wiener diversity index. Angle index and mingling degree were the key factors affecting Pielou's diversity index. Mean diameter at breast height and mean tree height were the key factors affecting richness. (4) The structural equation of herbaceous species diversity was not effect-based model indicating that stand structure was the key factor influencing herbaceous species diversity (P<0.001), and environmental factors were important factors influencing herbaceous species diversity (P<0.05). Among them, angular scale and mingling degree were the main stand spatial structural factors affecting the diversity of herbaceous species, the mean tree height and mean breast diameter were the main stand non-spatial structural factors affecting the diversity of herbaceous species, and Slope was a major environmental factor affecting the diversity of herbaceous species. Moreover, slope not only had a direct effect on herbaceous species diversity, but also has an indirect effect by affecting stand structure. Therefore, in the subsequent forest vegetation restoration measures, the spatial structure of forest stands should be considered to form randomly distributed heterogeneous multi-layer mixed forests to increase photosynthetic use efficiency and follow the theory of niche complementarity, so as to improve herbaceous diversity.