The first important thing in Quality Engineering(QE) was "to clarify the objective function". I think that the second most important thing is “to set the evaluation method”. I think that this evaluation method should meet the following requirements.
- You should select a evaluation method with good reproducibility.
- You should not take any times for the evaluation.
- If the objective function cannot be evaluated, you should evaluate the alternative characteristic as the generic function.
- You should select evaluate dynamic characteristics rather than static characteristics
Regarding item 2: Those who have experienced the orthogonal array experiment can realize it, but if the evaluation procedure is complicated or time-consuming, it will affect the reproducibility and the burden on the experimenter will be heavy. You should perform multiple orthogonal array experiments for a short time than to perform one orthogonal array experiment for a long time.
Regarding item 3: As I explained an example that the objective function is that the liquid flows through a tube without leakage in QE3, it is necessary to consider how to evaluate "without leakage". You need to rack your brains. For example, when you immerse the tube into water and then you evaluate air bubbles or you evaluate colored water ooze out from tube inside . The ideal of "flowing liquid without leakage" is that the water put in the tube from one side comes out from the other side. For example, you attach liquid-filled container to one side of the tube and measure the time when all the liquid has dropped . Alternatively, you can put an electronic balance on the outlet of tube and measure the change in weight of liqid over time. Of course, there is also a method of directly measuring with a flow meter. Being able to graph the change in weight over time or the change over time in a flow meter may provide more useful information than simply measuring the liquid volume. For example, if the flow rate change is large, it contains information that there is energy loss.
Regarding item 3: See the following materials → QE4
p.1: At QE3, I talked about the input to the system and the output from the system. Plot the output (displacement, rotation speed, amount of precipitation, etc.) when the input energy (electric power, load, calorie, etc.) is plotted on the horizontal axis. For academic research, you use the least squares method to find an approximate expression and draw a line, as shown in the lower right. The graph have y-intercept. QE does not draw such approximate expressions. When the input is zero, the output should be zero, so draw a straight line (curve in some cases) that passes through the zero point. This is called the "zero point proportional formula". In rare cases, the coordinates that move the zero point may change. In some cases, it is evaluated by the "desired characteristic" (static characteristic) to see the deviation from the object value, but it is more useful to evaluate the "dynamic characteristic" in which the output is evaluated by changing the signal as much as possible. This is because the ideal may not be straight. In QE, the horizontal axis is called the “signal factor”.
p.2: In QE amount of evaluation is not the output itself, but the SN ratio(signal-to-noise). The formula is in the lower left and the graph is in the lower right. The blue line in the graph is ideal condition, and the smaller the energy loss, the smaller the denominator of the SN ratio and the larger the numerator, so the SN ratio increases. Increasing the SN ratio means approaching the ideal. The evaluation value of each line in the conventional design of experiments is the output itself, but the QE is the SN ratio. This is the feature that QE is excellent.