Abstract: An evaluation method is proposed on FNN basis for those products which are featured with a big amount of information on assembling process with complicated influencing factors and lots of parameters relying on human judgment/reasoning especially the deficiencies of most of existing evaluation methods to them. According to the characteristic type of a product to assemble, the method shall be done in such a sequence, i.e., construct a database of all assembly information; process fuzzily the conventional 3-level BP network; set up an FNN model for assembling capacity evaluation of which both the inputs and weights have been fuzzed up. Then the weight distribution for optimum neuronal continuity of assembling process are available after a training to the evaluation network in accordance to preset rules. An exemplification shows the applicability and stability of the FNN model.