Electronic nose (E-Nose) technology was applied in this study to detect the aroma profile of abdomen, claw, leg meat and spawn of , which were farmed in Yangcheng Lake, Songjiang and Chongming region within different grades. Results showed that the aroma profile of four edible parts of Yangcheng-Lake-crabs in three grades (special, first and second) could be distinguished well. Soft independent modeling of class analogy (SIMCA) method was applied to establish the recognizing model of Yangcheng-Lake-crabs either on one single part or on multiple parts. By a comparison, the model based on multiple parts turned out to be more effective. Rejection ratio of none-Yangcheng-Lake-crabs reached 100 percent. Meanwhile, a partial least squares regression (PLSR) model was also developed to evaluate the correlation between E-Nose responses and grade scores of . Correlation coefficient was 0.96, manifesting the PLSR model was able to predict grade scores of unknown crab samples.