In order to stand out from the competition, businesses have little choice but to constantly redesign their products, or develop new products to gain a foothold in other markets. However, with each new product or design comes a change to processes, which brings a certain element of risk.
While there is value in prototyping and product validation testing, this can be costly, and in many cases problems won’t be detected until late in the development process. To lower the risk, engineers have to use a number of analysis and statistical tools – evaluating how every change could affect the output. Designment of experiments (DOE) is one such tool.
What is Design of Experiments?
Design of Experiments is a statistical tool used by engineers to evaluate the effect of single or multiple changes to a process or design. With this knowledge, they can design a product that satisfies customer needs and meets or exceeds quality standards.
The effective use of DOE methodology can significantly reduce the number of test runs or trial builds—saving project time and uncovering hidden issues in the process. With DOE, a project team is able to identify the factors that have the most and least impact on output.
Common Terms and Concepts
There are numerous terms used in DOE methodology—these are some of the most common:
Controllable input factors – also known as x factors, these are the input parameters that can be modified in the manufacturing or production process. When cooking pasta, for example, these will include the quantity of pasta and water that is used.
Uncontrollable input factors – the input parameters that can’t be changed, such as the temperature of the kitchen, or the temperature of a factory where products are manufactured. It’s important to understand how these factors may affect output.
Responses (or output measures) – the elements of the process outcome that gage the effect of the input factors. For example, the taste and texture of the pasta.
Hypothesis testing – determines the significant factors using statistical methods. A hypothesis statement can produce one of two outcomes: a null or an alternative. The null hypothesis applies when the status quo is true, while the alternative hypothesis is true when the status quo is not valid.
Blocking – an experimental technique designed to avoid any unwanted variations in the input or process, e.g. manufacturing a product with the same equipment to avoid any equipment variations.
Replication – using the same combination of factors more than once, in order to estimate the amount of random error in the process.
The Basic Flow for Design of Experiments
The process for a design of experiments is typically as follows:
Define the objective(s)
Gather knowledge on the process
Select your variables
Assign levels to the variables
Conduct tests or experiments
Data analysis and conclusions
When to Use Design of Experiments
Design of Experiments can be used during a Kaizen or Rapid Improvement Event (RIE), or as part of a wider Lean Six Sigma DMAIC improvement project. It is typically used in two stages of a DMAIC project:
During the ‘Analyze’ phase of a DMAIC project, DOE is a useful tool for identifying the root cause of a problem. Using DOE, an improvement team can examine the effects of various inputs on the output.
During the ‘Improve’ phase, DOE can be used to develop a predictive equation, allowing a what-if analysis to be conducted. The project team can then test different ideas to help determine the optimum settings for achieving the best output.
Determining the Appropriate Design
There are different types of experiment design, each involving a varying amount of factors and interactions.
One Factor at a Time (OFAT)
This approach entails exploring each single factor independently. It can work when there are just a few factors or interactions; however, it isn’t the most efficient approach for large factor sets, as you may miss the complex interactions uncovered by more sophisticated designs.
Two-Level Factorial Design
Two-Level Factorial Design is used in the majority of experiments because it is simple, versatile and can be used for many factors. As the name suggests, the factors are varied at two levels – low and high. This design is favoured because it is smaller than other designs but still allows for interactions to be detected.
Full Factorial Design
The most comprehensive approach to DOE, Full Factorial Design results in experiments where at least one trial is conducted for all possible combinations of factors and levels. Although it means no interactions can be missed, the thoroughness of the approach means it is quite expensive and time-consuming for experiments with multiple factors.
Partial Factorial Design
Given the time and cost implications of full factorial design, many organizations tend to lean towards partial, or fractional, factorial design. These experiments only evaluate a subset of the possible permutations of factors and levels.
Design Better Experiments for Your Organization
Design of Experiments (DOE) is just one of the many statistical tools you will be introduced to with Juran’s Lean Six Sigma Black Belt program, while our Master Black Belt program takes a candidate’s knowledge and expertise in the methodology to the next level.
Please contact us to discover more about our expert-led programs and certifications, or enrol today to start driving quality in your organization.
