Statistical Experimentation History

February 15th, 2010

Tags: experiment design.statistical experiments, designed experiments, Design of Experiments, DOE

Experimental Design was developed in the 1920s by Sir Ronald A.Fisher of England. His techniques were first applied in agriculture.

Dr.Genichi Taguchi has been one of the primary contributors to upgrading these experimental design methods for use in industry and design applications. Dr.Taguchi, an engineer, has developed a very powerful way to help improve the quality of products while simultaneously lowering costs. Between 1950 and 1970 Dr. Taguchi's methods of experimental design were developed at the Electrical Communication Laboratories (E.C.L.), the Japanese counterpart of Bell Laboratories. A notable application of these techniques was the development of a switch relay device. In 1971, the E.C.L. beat Bell Labs to market with this device, completing the project with one- fifth of Bell's personnel and one-fiftieth of its budget. Bell Labs invited Dr.Taguchi to explain his methods in 1972. A few years after the switch relay's introduction, Western Electric stopped production of the device and now imports them solely from Nippon Telephone and Telegraph.1 Ford Motor Company embraced Dr.Taguchi's methods in 1980 and formed the American Supplier Institute, where Dr.Taguchi is based, in 1981. Since that time, companies in many different industries, including ITT, Hewlett Packard, 3M, AT&T, Texas Instruments, and Sheller Globe, have begun to use these methods. Dr.Taguchi is considered to be one of the leading engineers of this century. He has received four Deming Prizes for his contributions to Japanese Quality. As a consultant to industry, he was awarded the Willard F. Rockwell, Jr. Medal at the 1986 International Congress on Technology and Technology Exchange. This prestigious award is given to only two individuals each year, with no more than one award per continent.

Among the explanations suggested for Japan's post-war industrial success are Japanese methods of management, the use of statistical process control, and the application of Just-In-Time manufacturing techniques. However, the real key to their success has been designing the quality into the product using Dr.Taguchi's System of Experimental Design. The method is best described as an engineering tool with a statistical base. This approach is concerned with gains in productivity. Cost effectiveness is stressed, rather than statistical rigor. In the world of manufacturing, the classical assumptions of a detailed hypothesis, normality, or homogeneity of variance are generally impractical.

In manufacturing, cost savings are realized by the reduction of scrap, lowering of inspection costs, and minimizing rework losses. These savings are achieved through process improvements and variation reduction. Design cost savings are realized by reducing the delivery cycle and minimizing engineering design changes. Reducing total Product cost is the ultimate goal.

Industrial experiments consist of three different groups of key elements. First are factors such as time, temperature, and speed. Second are the levels for these factors, such as one minute versus two minutes, or 100 degrees versus 200 degrees. Third, is the outcome or quality characteristic being measured, or evaluated, like surface finish or cost. Selection of each of these elements is an important step in developing a well-designed experiment.

    The experimental design process flows through the following steps:
  1. Define the problem
  2. Determine the objective
  3. Brainstorm
  4. Design the experiment
  5. Conduct the experiment and collect the data
  6. Analyze the data
  7. Interpret the results
  8. Verify the predicted results

These steps do not guarantee a successful experiment, but they do force the experimenter to proceed in a logical manner. All experiments conducted in this manner provide useful information, although some of them require a second experiment to achieve the desired improvement. Designing a successful experiment often requires a team of people familiar with the process or design. The members of the team contribute "a priori" knowledge, which helps to facilitate a well-designed experiment.

Dr.Taguchi's methods provide a means for minimizing the effect of factors that can't be controlled, by controlling the factors that are controllable. Thus, the process or product is made robust in the face of uncontrollable factors. Dr.Taguchi calls these uncontrollables noise factors. A noise factor causes definite variation, but can't be eliminated from the design or the manufacturing process.

Dr.Taguchi's experimental methods provide required information in a cost-effective manner for sound engineering decisions. Also, factors are identified which do not impact the quality of the process or product but can provide additional cost savings. Reproducible results are the key strength of Dr.Taguchi's methods. These techniques can improve quality without incurring capital and material cost increases. An important benefit is the separation of the vital few from the trivial many.

Industrial experiments have too many variables with different characteristics for the cost-effective use of the classical full factorial experiment methods. Dr.Taguchi modified these experiment methods for manufacturing and design applications for cost effectiveness and efficiency. By doing so, he has provided the engineering community with a powerful, applicable, and useful tool for making sound engineering decisions.

Process Predictability Management and EducateVirtually.com have assisted numerous companies over the past 26 years in the design and successful application of design of experiment methods. We have been directly involved in more than 450 experiments. These experiments ranged from new product development to solving product and process issues that were deemed impossible to resolve.

Want to learn more? Become a member of Educate Virtually and have access to over 100 courses. Subscriptions are either monthly or annual for extra savings. Sign Up Today!

Leave a Comment