Design of Experiments and Analysis of Variance
Product testing is a common part of reliability practitioner's work, which may also involve experimenting intended to improve the product design or some of its characteristics. This chapter deals with the problem of assessing the combined effects of multiple variables on a measurable output or other characteristic of a product, by means of experiments. When designs have to be optimized in relation to variations in parameter values, processes, and environmental conditions, particularly if these variations can have combined effects, we should use methods that can evaluate the effects of the simultaneous variations. For example, it might be necessary to maximize the power output from a generator, and minimize the variation of its output, in relationship to rotational speed, several dimensions, coil geometry, and load conditions. All of these could have single or combined effects which cannot all be easily or accurately computed using theoretical calculations.
Statistical methods of experimentation have been developed which enable the effects of variation to be evaluated in these types of situations. They are applicable whenever the effects cannot be easily theoretically evaluated, particularly when there is a large component of random variation or interactions between variables. For situations when multiple variables might affect an output, the methods are much more economical than performing separate experiments ...