The Design Of Experiments

Design of Experiment

The design of experiments A design of experiments (DOE) process improves upon the one-change-atatime process (see the previous section) by planning out all possible dependencies in advance. By defining in advance what experiments to perform in the DOE process, you put yourself in a position to catch multiple attributes that work together to affect the quality of your product or service. Inexpensive software packages for performing a DOE are available. Use your favorite Internet search engine to see what you can find. To perform a DOE, use the following steps. For this process, we stick with the pizza-dough example from the previous section. You want to sell your pizza dough as a mix for consumers to use at home, and, of course, you want to make the best possible mix. Many factors may affect the production of your pizza dough mix, but for this example, you’re concerned only with the recipe. 1. Identify the input and output factors that the experiments will measure. Start with a limited number of factors that you think will impact the quality of your pizza dough mix: flour, yeast, and salt. 2. Define for each input value a number of levels for which the output value is known. Your current standard recipe calls for 500 grams of flour, 30 grams of yeast, and 2.5 grams of salt. You then need to decide how to vary each of these factors. For example, you can vary flour from a low of 400 grams to a high of 600 grams; yeast from 20 grams to 40 grams; and salt from 1.5 grams to 3.5 grams. 3. Create an experiment plan that includes the input-level values defined. In this case, you try different combinations of flour, yeast, and salt, mix and bake the crust, and then have a taste tester judge the quality of the crust on a scale of 1 to 10 (with 10 being best).

1

Design of Experiment 4. Perform the experiments for each input level and measure the output. In other words, you create a matrix of the selected combinations of input values and levels. For instance, Mix 1 — with 400 grams of flour, 20 grams of yeast, and 1.5 grams of salt — yields a quality score of 7. Mix 2 — with 500 grams of flour, 20 grams of yeast, and 1.5 grams of salt — yields a score of 6. Mix 3 — with 500 grams of flour, 40 grams of yeast, and 2.5 grams of salt yields a score of 8. 5. Look for differences between the output values for the different levels of the input changes. These differences are due either to the input value by itself or to the input value acting in combination with another input. Mix 3 received the best score, but you may want to do more experiments to determine whether other factors play a role in the quality of the dough. Make sure that each department involved in the process of creating your product or developing your service is involved in identifying the input values and levels (design, production, shipping, service, and so on). For the DOE method to work, you must perform an experiment for each possible combination of input values and levels. If this seems impractical for your business, you may need to focus on the values you think are most critical; however, the fewer experiments you do, the closer you get to guessing. References: 1) Larry Webber and Michael Wallace. Quality Control for Dummies. Wiley Publishing, Inc. 2007.

2