00272 ù0cbbf4ea84dc6ff85e81a7c4f53ac6

274

Montgomery

these packages and have usually found that far better choices of designs for the problem at hand could have been chosen.

In selecting the design, it is also important to keep the experimental objectives in mind. In many engineering experiments, we already know at the outset that some of the factor levels will result in different values for the response. Consequently, we are interested in identifying which factors cause this difference and in estimating the magnitude of the response change. In other situations, we are interested in demonstrating uniformity. For example, two production conditions A and B may be compared, A being the standard and B being a morÄ™ cost-effective altemative. The experimenter will then be interested in demonstrating that, say, there is no difference in yield between the two conditions.

When a finaÅ‚ test plan has been produced, it is important to carefully evaluate each run in the experiment to ensure that it is a feasible combination of process variables. A good way to do this is to have the team examine each run, evaluate the combination of factor levels for that run, and then estimate what the response value (for one of the key responses) will be for that run. This produces a set of "artificial" data that can be analyzed by the experimenters to make surÄ™ that all team members understand how the analysis of the actual data will be performed.

Performing the Experiment

When conducting the experiment, it is vital to monitor the process carefully to ensure that everything is being done according to plan. Errors in experimental procedurÄ™ at this stage will usually destroy experimental validity. Up-front planning is crucial to success. It is easy to underestimate the logistical and planning aspects of running a designed experiment in a complex manufacturing or research and development environment. These issues should be carefully evaluated and acceptable Solutions identified and implemented before the start of the experiment.

Data Analysis

Statistical methods should be used to analyze the data so that results and conclusions are objective rather than judgmental in naturÄ™. If the experiment has been designed correctly and if it has been performed according to the design, then the statistical methods required are not elaborate. There are many excellent software packages designed to assist in data analysis. Ideally, if a Computer software package has been used to assist in design selection and