I start posting about the commentary of Quality Engineering(QE). QE is the "Taguchi Methods" developed by Dr. Genichi Taguchi. I knew that there was a tool called "Taguchi Methods" about 30 years ago. We started to study in earnest about 10 years ago, and I am feeling the splendor of QE. Unfortunately, I have a frustration that I cannot teach young developer fully how great QE is. In the future, I would like to introduce the greatness of QE little by little. The more skilled you are, the greater your skepticism about QE. From my experience, I believe that the people who believe in QE and practice it succeed. People who make excuses for not having time to experiment by QE will fail. Also, some people misunderstand that QE is the same as the Design of Experiments（DE）. QE is not a religion, but if you believe it and then you will be saved.
Please click the attachment. → QE1
p.2: In the case that you are trying to obtain the optimum condition by experiment, if you only have two parameters (for example, temperature and pressure), you can check the effect of two parameters by changing the numerical values. The circle-mark in the left figure is within the allowable range, and the x-mark is outside the allowable range. If there are four or more parameters as shown in the figure on the right, it is difficult to represent them in the figure. In the case of multi-dimensions, DE and QE are most effective analysis method.
p.3: Both DE and QE method use the assignment table called “Orthogonal array” to conduct experiments. If you change two level of three parameters (X, Y and Z) , you need to perform 8 (23) combinations of experiments.
Performing four experiments (No,1, 4, 6, and 7) using the L4 orthogonal array has the same effect as performing eight combined experiments. If the experiments No.1 to No.8 are vertices of a rectangular parallelepiped, these shadows can be casted on the XY plane, YZ plane and ZX plane and be four points on each three planes. You can intuitively imagine that experiments No.1, 4, 6 and 7 represent as eight experiments. There are 81 (34) combinations of 3 levels for 4 parameters. If you use the L9 orthogonal array, you can obtain the same effect as 81 ways in 9 experiments. Since there are 4 parameters, 81 vertices are on a four dimensional shapes and cannot be drawn. See the figure on the right. The shadows of the 81 vertices are represented by nine circle positions. And then you can know nine experiments represent as 81 experiments.
Today is up to here. See you next week.