1 July 1999 in Journal of Quality Technology
Rational subgroups of size n = 1 are frequently encountered in process monitoring and control. The Shewhart control chart for individuals is often used in these situations. It is well-known that the in-control average run length (ARL) of this chart is 370.4 under the assumption that the observations are selected at random from a normal population. When the assumption of normality is violated, the ARL of the individuals control chart is adversely affected. We show that an exponentially weighted moving average (EWMA) control chart can be designed so that it is robust to the normality assumption, that is, so that the in-control ARL is reasonably close to the normal-theory value for both skewed and heavy-tailed symmetric non-normal distributions. The EWMA chart also performs quite well in detecting shifts in the process mean.
1 January 2003 in Journal of Quality Technology
Computer-generated designs are useful for situations where standard factorial, fractional factorial or response surface designs cannot be easily employed. Alphabetically-optimal designs are the most widely used type of computer-generated designs, and of these, the D-optimal (or D-efficient) class of designs is extremely popular. D-optimal designs are usually constructed by algorithms that sequentially add and delete points from a potential design by using a candidate set of points spaced over the region of interest. We present a technique to generate D-efficient designs using genetic algorithms (GA). This approach eliminates the need to explicitly consider a candidate set of experimental points and it can be used in highly constrained regions while maintaining a level of performance comparable to more traditional design construction techniques.
1 January 2002 in Journal of Quality Technology
In process robustness studies, it is desirable to simultaneously minimize the influence of noise factors on the system and to determine the levels of controllable factors that will optimize the overall response or outcome. A methodology for evaluating designed experiments that involve both controllable and uncontrollable, or noise, factors is outlined and presented in this paper. Two variance expressions are developed for evaluating competing experimental design strategies. The maximum, average, and minimum scaled prediction error variances resulting from the models developed are displayed visually on variance dispersion graphs. The scaled prediction error variances account for mean model errors as well as variation transmitted to the process by noise variables.
27 August 2002 in Quality and Reliability Engineering International
1 October 2003 in Journal of Quality Technology
In a mixture experiment, the design factors are mixture components whose proportions are varied, and the response variables are assumed to depend only on these component proportions. In addition to the mixture components, the experimenter may be interested in other variables that can be varied independently of one another and of the mixture components. We consider the case where one or more of these variables is a noise variable, or a variable that cannot be controlled in practice. We develop models for these robust mixture formulation problems. We then derive mean and variance functions and illustrate their use in formulation optimization. Cases of uncorrelated and correlated noise variables are addressed.
1 April 2010 in Journal of Quality Technology
In experimental testing, it is desirable to select combinations of input factors that yield optimal results given an objective specified by a practitioner. Often this objective involves minimizing the uncertainty in a parameter or prediction estimate. To determine the experimental design to achieve such an objective, it is necessary to understand the relationship between the testing variables and the response. Alphabetic optimal designs are commonly used in such applications based largely on the ease of their construction with advanced statistical software. However, many publications cite concerns with the overall robustness of such designs to small departures in parameter estimates. This paper examines the issue of model parameter sensitivity to the selection of an accelerated life test based on the UC-optimality criterion, minimizing of the prediction variance at the usage condition, using a generalized linear model framework. We will examine the trade-off implications of choice of experimental design, sample size, and censoring.