Design of Experiments (DoE) and the Analysis of Variance (ANOVA) techniques are economical and powerful methods for determining the statistically significant effects and interactions in multivariable situations. DoE may be utilized for optimizing product designs, as well as for addressing quality and reliability deficiencies. Within the DoE framework, the practitioner may explore the effects of a single variable or analyze multiple variables. A multivariable analysis is often referred to as a “factorial experiment.” An example of how DoE would be used in a DoD context that is relevant to reliability engineering includes supplier selection for the use of COTS items, GFE, or other non-developmental devices that may be used in a given acquisition program. A common approach would be to test a sample of bearings, for instance, from each of four different vendors until they are run to failure. Using DoE, one can quantify the variation between samples and determine if the variation is statistically significant, or if perhaps it is only an artifact of the variations within the populations from which the samples were randomly drawn. Of course, the ultimate goal is to make an inference regarding the variation in product quality across vendors, as well as within the individual populations.
Factorial experiments may be extended to analyze two or more sources of variation. Leaks of O-ring seals within hydraulic systems, for example, can result from multiple factors, such as, excessive oil pressures, extreme temperatures, or installation error. DoE methods may be used to analyze these sources of variation, determine interactions that may occur between them, and identify the statistical significance thereof. There is a great deal of literature covering the subject, and a number of different DoE techniques have been developed.
The Taguchi Method is a popular approach that is tailored to the requirements of the engineering design. The method consists of three phases system design, parameter design, and tolerance design. The system design phase focuses on the material concept and determining the required performance, quality, and reliability characteristics. The parameter design stage focuses on refining parameter values to optimize the desired characteristics of the system relative to the sources of variation and the interactions between them. The final stage, tolerance design, highlights the effects of random variation, e.g., manufacturing and production processes, to determine if further design margin or optimization is needed. Of course, the value of reaching a given level of design margin is constantly assessed in terms of investment cost versus the current level safety, quality, performance, reliability etc. that has been achieved.