Taguchi's method enriches design of experiments methods by bringing a considerable improvement to the complete and fractional factorial designs detailed in the dossier.
.
Its aim is to simplify the experimental protocol in order to highlight the effects of factors on the response, which may be a variable in an agri-food process or even the measurement of product quality.
Taguchi's method is characterized by a significant reduction in the number of trials, while maintaining good precision. It places the model as a key element of the experimental design strategy. The experimenter is free to choose the factors and interactions to be studied according to the model he or she proposes, in close alignment with his or her objectives.
The place of each factor in the Taguchi design is important, and is chosen according to the difficulty of realizing the factor in the experiment. This place in the plan will enable the factor that is the most difficult to realize to make the fewest possible level changes. In this way, factors can be grouped by degree of difficulty of realization.
Taguchi's method initially met with success in the industrial sectors and, in particular, in the agri-food sector, and then aroused the interest of the statistical community for wider development and study.
In this file, we begin by defining orthogonal tables with respect to a model. We also give some properties concerning these tables, which we will later use to choose the Taguchi table that best suits the problem under study. We then study Taguchi's method, showing how to choose the appropriate orthogonal table from among the existing ones, and how to calculate the model coefficients needed to highlight the effects of the factors under study.
The case where factors form 2 groups of different natures (internal factors and external factors) is addressed by studying the product design crossing two experimental designs and calculating the coefficients of the product model.
Some Taguchi tables are given as examples in the Bibliography and Appendices tab.