In the textile industry, cloth is made from many different brands of yarns, while yarns are made from various different kinds of cotton. Naturally, we want to achieve a full-fledged utilization of combining different brands of cotton to produce yarns with the best quality. Such a combination of different brands of cotton in the manufacturing process is called cotton allocation. In this regard, it is often encountered a problem of how to select a substitute (from the warehouse) for a brand of cotton in the cotton allocation that is currently in use but running out of the raw material. A model that uses Fuzzy Dynamic Integrated Judgment (FDIJ) provides a very useful tool in resolving this problem. In view of several factors in cotton allocations, the FDIJ is made for cotton selection according to the data feedback from previous production runs. In , we explained how the data feedback can be stored in a fuzzy dynamic relationship dictionary which generates a fuzzy dynamic relation to be used in the learning ability in the problem of cotton allocation. It should be noted that finding a fuzzy weight set in the FDIJ is an inverse operation of fuzzy integrated judgment. In applications, it requires an initial fuzzy weight set (or the so-call the original image) to make the whole process work. The main objective of this paper is to explore a practical approach of obtaining the optimal fuzzy weight set for building a model of FDIJ whether or not the exact original image exists.