In order to gain knowledge of genetic diversity of released jute varieties(strains), and provide scientific evidence to the improvment of jute varieties. The yield characters of 26 jute varieties(strains) were investigated based on the principal of numerical classification, and the PCA(define) of 8 yield characters were analyzed. The scatter plot were drawn based on the first 3 principal(PC) and the clustering analysis was carried out based on the genetic coefficients. The results showed that the coffident variation of 8 yield characters ranged from 2.54% to 14.70%, 7 yield characters of Fiber bark thickness, Fresh bark weight per plant, Fresh bone weight per plant, Fresh bark dry rate, Dry bark weight per plant have a high coefficient of variation over 10%, there was variation potential on filed. Principle component analysis indicated that the first 3 PC accounted for 88.247% of variation among the strains, followed by Fresh bone weight per plant factor, Fresh bark dry rate factor, Branching height factor. According to 3 principal components factors score, 26 strains could be divided into 4 categories while D=2.15, with 18 strains belonging to the first cluster, 5 strains belonging to the second cluster, 1 strains belonging to the third cluster and 2 strains belonging to the fourth cluster. Each group had different characteristics, these varieties(strains) could be used for castor breeding with different purposes. The scatter plot of the former three principal components indicated the coefficients of the first three PCs of 9（09 Pin-11）, 11（89m5-83）, 12（09 Fan-7）, 19（Fuhuangma 1）, 21（Fuhuangma 3）, 25（Y007-10）were better than others. In general, cluster analysis and scatter plot based on PCs analysis together could make good comprehensive evaluation on jute strains, which could provide scientific evidence for the reasonable selection of germplasm resources for further jute breeding.