In visual inspection for aesthetics features we can use sensory analysis procedures to assess quality. Some methods have been developed to help inspectors in determining if a product meets or no a given quality criteria, like: the criteria / level table, tree-like evaluation table and corrected hierarchical evaluation table. These methods use linear correlation to link evaluation and decision. In some cases, however, the linear correlation cannot be applied because the relationships between the multiple attributes which are considered in the analysis have strong non-linear interactions. In this work, we proposed the use of neural network to make the correlation between evaluation and decision. This methodology was applied in the S.T. Dupont Company and the experiments show that neural network can successfully correlate complex interactions. As result, we were able to validate the evaluation table which was defined by the quality managers of this company.