On stochastic optimization: Taguchi Methods (TM) demystified; its limitations and fallacy clarified

The intent of this paper is to clarify and straighten out important misperc eptions. Recently, the need for Stochastic Optimization in Design and Manuf acturing Engineering has received increased recognition. The only technique belonging in the realm of Stochastic Optimization that has been practiced somewhat in industry is Taguchi Methods(TM). (Taguchi Methods(TM) and Robus t Design(TM) are registered trademarks of the American Supplier Institute f or Taguchi methodology. In this paper, Taguchi Methods or methodology is us ed in the singular and without the trademark symbol, and mostly its acronym , TM, is used.) However, Taguchi Method (TM) has been incorrectly promoted by many as THE work process for attaining "a robust design," thus the use o f Robust Design(TM) to stand for TM! This has caused great confusion since much else is needed for designing a robust product. More serious is that si mple mathematical fundamentals behind TM have not been fully clarified so t hat its limitations and deficiencies would be obvious. Thus even greater co nfusion exists amongst people who "learn" or hear of TM or are intended pra ctitioners. Because of our extensive work on HPD (Holistic Probabilistic De sign), which holistically treats Stochastic Optimization and thus includes comprehensively and rigorously addressing robustness, we have gone beyond T M capabilities. As byproducts, we can now demystify TM in a simple manner, show its limitations dearly, and clarify TM's fallacy which essentially inv alidates it as a technique! (C) 2000 Elsevier Science Ltd. All rights reser ved.