Reviewing the Taguchi Experiment

You can use the Experiment tab to view the full Taguchi experiment, based on the selected control factors and control technique, the selected noise factors and noise technique, and the selected signal factor. The first column in the table displays the run number (Run#) and nested experiment number (Exp#). The remaining columns in the table display the experiment data for each factor. If the system type is static, the control factors are listed first, followed by the noise factors (C, N). If the system type is dynamic, the signal factor is listed first, followed by the control factors and then the noise factors (S, C, N). The columns are labeled with both the factor name and the factor type (signal, control, noise, or control-noise if the factor is both a control and a noise factor).

Isight does not display the experiment on the Experiment tab in the following cases:

  • No control factors and no noise factors are defined. Isight displays a message to select control/noise factors on the Control/Factors tab or the Noise/Factors tab.

  • The control matrix or the noise matrix requires manual generation. Isight displays a message to manually generate the matrix on the Control/Matrix tab or the Noise/Matrix tab.

  • The noise matrix requires manual generation and one or more noise factors is also a control factor. In this case the noise matrix must be regenerated for each control run. This can be expensive, and with randomness the experiment generated will not match the actual executed experiment. Isight displays a message indicating that the experiment will not be displayed for this case.

In addition, Isight displays a warning message if there is undesirable overlap in the experimental values for a control-noise factor. For example, if the control levels for a factor X are 79 and 80, and the noise factor levels for X are set to a diff of +/–1.5 from the control factor levels, the resulting experiments would be:

(C,N)

Value

(1,1)

77.5

(1,2)

80.5

(2,1)

78.5

(2,2)

81.5

As can be seen in the example, the range for control experiment 2 overlaps the range for control experiment 1. The overlap may be undesirable; therefore, Isight highlights it on the Experiment tab. If desired, you can remove the overlap by increasing the control factor range or by decreasing the noise factor range before executing the model.

Isight displays a validation warning on the Taguchi component if any undesirable overlap in levels for control-noise factors are saved in the Taguchi component configuration.