The Taguchi design method is a statistical design technique that seeks to improve productivity by creating production systems that are insensitive to variations in the input factors or changes in the environment. This is achieved by modelling the outcome of the system as a response surface and using factorial experimental design techniques to identify areas of the response surface which are relatively flat to changes in inputs. This Engineering Initiation Award proposes to create an alternative to the factorial design methods by using spectral design methods. Spectral design of experiments is a technique whereby factors that are significant determinants of system behavior can be identified by analyzing the spectra of the system inputs and outputs. The resulting analysis is comparable to a traditional factorial experiment under classical design assumptions. The spectral approach has the additional property of yielding correct factor identifications even when a lag of unknown duration exists between the introduction of a change to an input and the corresponding change in the output of the system. Combining the Taguchi design method with spectral design of experiments will yield a method that can be used for productivity improvement in industries in which the Taguchi approach is currently not easily applicable, such as chemical engineering and other process industries.
