Analysis of Signals and Noises from Unreplicated Factorial Experiments via Graphical Procedures

Statistical experimentation IS the key to make the process robust to environmental noises, to produce the products ~t a lower cost and to enhance the productivity. Analysis of means and dispersions are rational extensions of the Shewhart control chart for comparing products from different design configurations in off-line quality improvement. In this article, we utilize these graphical procedures to analyze signals and noises from an unreplicated factorial experiment to improve the control of a textile manufacturing process. The method of maximum likelihood is employed to estimate the mean and variance functions for predicting the average and variance of yarn strengths. Finally, we exercise a constrained optimization procedure to find the best combination of controllable process variables.