During the cased-based reasoning procedure of requirement mapping, the randomness and subjectiveness of weight setting always lead to the poor accuracy of results affected by irrelevant noise. A new method of CBR weight setting has been put forward to solve it based on ANN, which defined the relevant weight from the aspects of feature sensitivity, activity, saliency, relevance and the transforming mechanism of features input also been described. Based on it, a new requirement mapping procedure module has been put forward oriented to CBR and verified with the design of the hydraulic workbench, which gives a new method for product requirement mapping.