For designing, executing and analyzing tests structured methods were developed by Sir Ronald A. Fisher in the first half of the 20th century. We generally refer to them with the term Design Of Experiments or it's well known abbreviation DOE. The objective of DOE is to quantify the effect of parameter settings on the resulting characteristics or responses of a product or process and doing this with maximum efficiency. As seen in the figure below during testing we need to take disturbances (generally called noise) into account.
Within the global field of DOE QS Consult offers two approaches that are being used in industry: classical design of experiments and Taguchi methods. In the following paragraphs we will explain the two systems and introduce the training courses that QS Consult offers on DOE.
Each approach has its proponents but QS Consult is convinced that an intelligent integration of tools and the correct use of the right tool at the right time will give the best results. That is why in most of our DOE courses elements of the two approaches can be found.
Course on Classical Design Of Experiments
This is a course that everyone who is involved with setting up tests and analyzing test results should follow. The basic course (1 day) contains the following subjects:
- Problem definition
- Selection of factors and levels
- Interaction, randomizing and blocking.
- Full factorial test designs at two levels
- Creating a test protocol
- Executing tests
- Analyzing the results
- Conclusions and final report
People that want to go further with DOE can follow an extra day. In this advanced course we explain fractional factorial designs, designs for more than 2 levels and analysis of variation. We also introduce EVOP (EVolutionary OPeration) and test designs for mixtures. In this course we show the possibilities of MINITAB as supporting software for DOE.
During the course (basic and advanced) participants perform and analyse real tests, both with paper helicopters and with a special catapult.
Taguchi Quality Engineering
Dr. Genichi Taguchi has played a major part in making Design Of Experiments popular in industry. But it is wrong to think that Taguchi methods are just another way of doing tests. It is a complete philosophy comprising several aspects and referred to as “Quality Engineering”. The most important ideas are:
- A high quality product will cause a minimum loss to society (expressed in money) during it's lifetime. The relation between this loss and a technical characteristic is given, by the loss function.
To minimize the loss we need to produce on average on target with minimum variation.
- Quality has to be designed into products and processes. Much more effort is needed in Off Line Quality Control to prevent problems in production.
- All forms of noise (within-product noise, changes in external conditions, neighboring systems, …) influence a system and cause deviations from the ideal situation. To eliminate or reduce these influences we need to design robust products end processes. The robustness of a system is the ability to function stable under changes noise conditions.
- Looking for parameter settings that increase robustness (parameter design) is an essential part of the Taguchi methods. Robust design is a method aimed at designing a system that operates as close as possible to ideal function under varying noise conditions.
Within the quality world there is a lot of controversy between proponents of Taguchi and classical DOE. QS Consult has a different view on this, related to the following quote by W.E. Deming: “each tool is useful, as long as one knows it's limitations" We are convinced of the power of the integration of methods and that is practiced in our Taguchi course.
For those who have followed our course on classical DOE we have created an expansion to Taguchi methods that takes 1,5 day. For people that are new to DOE but want to start with Taguchi we propose our 3 day course. In both cases a lot of exercises are made during the training session.