Resumo

NARANJO-PALACIOS, Fernando; RIOS-LIRA, Armando Javier; PANTOJA-PACHECO, Yaquelin Verenice  e  TAPIA-ESQUIVIAS, Moisés. Fractional Taguchi orthogonal designs. Ing. invest. y tecnol. [online]. 2020, vol.21, n.2, e1621.  Epub 13-Nov-2020. ISSN 2594-0732.  https://doi.org/10.22201/fi.25940732e.2020.21n2.011.

The design of experiments is a tool used to discover how different variables in a process come into play to obtain a product. There are two main approaches to perform experimentation, the classic approach and the Taguchi approach. Taguchi experiments are orthogonal arrays that specialize in estimating main effects and control by noise interactions, leaving in second place control by control interactions. The Taguchi orthogonal arrays were designed in such a way that a specific array can be used for different numbers of factors, for example, the L32 is used when there are 16 to 31 factors and requires 32 experiments. When the number of available columns exceeds the number of factors that we wish to investigate, the remaining columns are used commonly to estimate interactions. Nevertheless, in cases in which the experimenter is interested only in estimating main effects, running the full array could be unnecessary and expensive. This research proposes a method to fractionate the Taguchi orthogonal arrays L8, L12, L16 and L32 in such a way that the fraction generated helps only to estimate main effects and the remaining runs can be added only in cases in which they are required. The proposed approach is based in exhaustive search and uses as selection criteria the D-optimality, variance inflation factors (VIF) and the general balance metric (GBM). Only two-level Taguchi orthogonal arrays were considered for this research. The results of this research translate into significant savings in resources, reduction in experimentation time and reduction in the number of runs.

Palavras-chave : Experiments; design; robust; Taguchi; fractions.