Design of experiments:
Improving processes with experimental design
By Jackie Graham, Ph.D.
Managing Director, PQ Systems, Australia
Experimental design is a hot topic these days, gaining increasing attention around the world. This column will offer a series offering a practical approach to experimental design. It will explain the buzz words and where you can really use this technique to improve processes. Let’s start with what design of experiments is all about.
The reason for the popularity of experimental design is that when applied correctly it is effective; the techniques will literally tell you how to optimize a process. It will help you determine the actual settings to use to achieve your desired outcome; whether that desired outcome is productivity or quality based, or even if it is both. The way the approach achieves these outcomes is by conducting various experiments on your process, and the ensuing analysis then reveals how to optimize your process. Before we get too carried away, let’s look at how these wonderful techniques developed.
In the 1920s the techniques of experimental design were pioneered by R. A. Fisher of Rothamstead Agricultural Research station in the UK. He first used them to optimize the yield of crops. In the 1930s his techniques were adopted by the British textile industry. Gradually they spread from one industry to the next. Chemical industries in the USA and Western Europe started to use the approach in the 1940s and 1950s. During the 1960s and 1970s Genichi Taguchi in Japan devised a new approach focusing on the key aspects of a process. Ellis Ott in 1967 added another assessment technique.
These were the true pioneers of experimental design. Of course, with the addition of the computer, the calculations and use of experimental design has increased. PQ Systems has a product specifically developed for experimental design called DOEpack. You can learn more about it at www.doepack.com.
Today experimental design is used in all types of manufacturing industries and in some service industries too. In fact, design of experiments is seen as a key tool of the Six Sigma approach to improving your organization. So, how do we start? The key is to begin with a relatively simple process, develop a good understanding and then apply it in more complex situations. In this series of articles in e-line I am going to take an example through project selection, working out what experiments to complete, and then the analysis techniques. First, let’s get familiar with the terms used in experimental design. What follows is a graphic that you can use for any process.
Let’s define what is meant by this diagram.
Inputs – This is any input into a process that is used to produce the product. For example, if we were looking at plastic molding, the input would be resin. If we were looking at a pressed part, the input may be steel coil. There may be one single input or several. If we were to look at my mother’s process to make her magnificent chocolate cake, there would be many ingredients, such as flour, sugar, butter, eggs, and cocoa.
Outputs – This is what is produced by the process, such as a plastic molding, a pressed part, or my mother’s chocolate cake. There may be more than one output produced by a process. For example, when crushing rock at a quarry, the products will be different graded aggregates, and a dust all from one process.
Factors – These are anything that can have an impact on the process. When considering plastic molding, factors include temperature and pressure. When baking a chocolate cake, it would include oven temperature and mixing time. Some of the factors can be controlled; others cannot and are referred to as uncontrollable factors.
Controllable factors – These are factors that can be varied and controlled by you or the operator of the process. For example, the temperature setting on a plastic molding machine can be varied and controlled by the operator.
Uncontrollable factors – These are factors that cannot be controlled by the operator. For example, if one of the factors is humidity from the natural environment, this is beyond the control of the operator.
What we are aiming to do in experimental design is to determine the inputs and factors that most influence the process. Further, experimental design then defines the settings of the factors and the types of inputs to use to get the optimum out of the process. We can use experimental design to determine how to operate the process to reduce variation. Also, we can work out how to minimize the impact of the uncontrollable factors on the process.
Experience in many different organizations has shown that processes are operated differently. One shift comes in and their immediate action is to change the settings on equipment – to their optimum setting! When another shift comes in, they do the same thing. The product varies and changes as the shifts change. Sometimes these changes are easy to pick up in a control chart. However, how do you determine the optimum settings for the equipment? Which shift is right? Are any of them right? This is where the techniques of experimental design can solve the questions and get processes working at their very best! Get the most out of every process by applying design of experiments.
In the next article in this series, we will look at how to start using design of experiments.
Copyright 2008 PQ Systems.
Please direct questions or problems regarding this web site to the Webmaster.