Reliability DOE
Unlike other design types, reliability designs are specifically intended to handle life data. Only one response (typically failure time) is measured, but the designs can accommodate data sets that include suspensions (right censoring) and/or uncertainty as to when the units failed (interval and/or left censoring), in addition to complete data sets in which all of the units under test failed and the failure time for each unit is known.
In contrast to traditional DOE techniques, which assume that response values at any treatment level are normally distributed, reliability DOE (R-DOE) uses the Weibull, lognormal or exponential distribution to analyze data. You can publish the fitted distribution as a model for use in other applications, if desired. In addition, the kinds of results and plots available may differ from those that are available with DOE using standard response data.
Tip: To perform R-DOE, select any design type in a standard design folio and set the Response Type for one of the responses to Life Data. (See Adding, Removing and Editing Responses.)
This section explains how to perform reliability DOE, including:
The ReliaWiki resource portal provides more information about R-DOE at http://www.reliawiki.org/index.php/Reliability_DOE_for_Life_Tests.
Tip: If you wish to analyze life data from a prior experiment, considering using the free form folio, which does not require that you create an experiment design.
Available Reliability Designs
The available designs for reliability DOE are:
One factor reliability designs allow you to test a single factor using from 2 to 20 levels to determine if the factor has an effect on the product's reliability. Note that the factor in a one factor design is treated as a qualitative factor. Therefore, predictions cannot be made for factor levels that are not tested, and the designs cannot be optimized.
Factorial reliability designs allow you to test multiple factors at two levels each to investigate the effect of the factors on the product's reliability.
In factorial reliability design, only the linear effects of the quantitative factors are studied. Response surface method (RSM) reliability designs allow you to study the quadratic effects of the factors (i.e., effects that differ depending on the level of the factors), making them well-suited to predictive modeling and optimization. For example, RSM reliability designs can be used to determine the factor settings that will optimize the reliability of a component.
For more information about how to use the design types, please consult the documentation on design folios.
