Factor Selection

Quick reference

• 00:04 Hello, I'm Ray Sheen.
• 00:06 And as we discussed in the last lesson, one of the two steps that must be
• 00:10 completed at the beginning of a DOE study, is selecting the factors for the study.
• 00:15 What do we mean when we say factors?
• 00:17 Factors are the data elements of the study, both the input and the output.
• 00:22 Carefully consider the factor selection because this will
• 00:26 determine the limits of your data collection and analysis.
• 00:29 One type of factor is the response or output factor.
• 00:33 This is what will be measured at the end of each run and
• 00:36 should directly relate to the objective of the study.
• 00:39 This is where we will see improvement or degradation of performance.
• 00:43 There are also a set of input or control factors.
• 00:46 These are the things that are set as part of the test item or
• 00:50 test process configuration.
• 00:52 These factors represent the items being studied to determine their impact on
• 00:56 the product, process or system performance.
• 00:58 Often, there are also a set of constants that are used to minimize noise or
• 01:02 uncertainty.
• 01:03 So for instance, the same operator does all the tests,
• 01:06 the same data set is used and the same procedure is followed.
• 01:10 These constants will improve the DOE study results because they eliminate
• 01:15 sources of statistical noise and variation.
• 01:17 Let's look a little deeper at the Y factor or response variable.
• 01:22 There are several questions that you should ask yourself when selecting the Y
• 01:26 factors.
• 01:27 These considerations include, whether the variable was qualitative or quantitative?
• 01:31 We definitely prefer quantitative so
• 01:33 that we can measure the level of the response factors during each run.
• 01:37 Next is that output or response factor directly tied to the study goals.
• 01:42 This ensures that the analysis will meet the study objective.
• 01:46 After that, is the output currently under statistical control?
• 01:49 What we mean by that is whether the output is currently predictable.
• 01:53 If it's not under control, there maybe many factors that
• 01:56 are influencing performance that we are not aware of.
• 01:59 The DOE will probably need to be more complex to try and
• 02:02 capture all of these factors.
• 02:04 If it is under control,
• 02:05 we can focus on just the factors that we're being asked to study.
• 02:09 Another concern is, does the output vary over time?
• 02:12 If it drifts, we need to account for that drift in our study design.
• 02:16 And, are there multiple outputs?
• 02:19 Many statistical analysis software programs that support
• 02:22 DOE can support multiple outputs.
• 02:24 However, when optimizing, it's important to know which is more important.
• 02:29 Another attribute is, what is the objective for the performance?
• 02:33 Is it to minimize, maximize, or center on a specific target of variance reduction?
• 02:38 These answers will drive the optimization portion of the analysis.
• 02:43 Also you need to consider the anticipated range for the output.
• 02:46 With this knowledge, an appropriate measurement system can be selected
• 02:51 that does not saturate at the extremes of performance.
• 02:54 And while talking about measurement systems,
• 02:56 how much change in the output do you want to detect?
• 02:59 The discrimination of the measurement system needs to be able to tell
• 03:03 the difference between small changes in performance.
• 03:06 And finally, is the measurement system that you have chosen adequate?
• 03:10 Is there a measurement system analysis that shows that this system provides
• 03:14 repeatable and reproducible measurements for
• 03:17 the nature of the experiments that will be done?
• 03:20 If not, find a new system.
• 03:22 So now let's look at the control factors,
• 03:24 the items that will be the x variables in our design space equation.
• 03:28 The control factors are the design attributes or
• 03:31 process settings that are being investigated as part of the study.
• 03:35 The DOE will determine what effect the variation in these factors has on product,
• 03:39 process, or system performance.
• 03:41 So to do this, factors must be controllable during the study.
• 03:45 Either they are controlled at the time of test sample preparation or
• 03:49 controlled by the operator as settings in the experimental runs.
• 03:52 They are controlled to the appropriate level for each test configuration.
• 03:57 Normally these factors are elements of the product, process, or
• 04:00 systems control methodology.
• 04:02 The results of the DOE study will be to establish targets and
• 04:05 limits for those elements that will ensure the desired performance.
• 04:09 This is why there are normally either design features that are selected at
• 04:14 the time of design, or product and
• 04:15 process control features that are selectable by the operator or user.
• 04:20 Let's consider some criteria that can be used while selecting these factors.
• 04:24 One criteria is to be sure that you have a factor associated with everything that you
• 04:28 want to study.
• 04:29 The study is to determine the impact and inner relationship of the control factors on
• 04:34 the response factor.
• 04:35 It provides no information about those factors that are not part of the study.
• 04:40 These input or control factors can be either quantitative or qualitative.
• 04:45 Quantitative factors are those with a continuous range of settings.
• 04:49 The DOE will use two or more of those settings to determine sensitivity.
• 04:54 The qualitative factors are usually category factors,
• 04:57 such as material from supplier A or supplier B.
• 05:00 There is no range in between those, just one category or the other.
• 05:05 Both types of control factors can be used in the DOE.
• 05:08 Finally, there are three principles we apply when selecting control factors,
• 05:13 practical, feasible, and measurable.
• 05:15 Practical looks at the factor from a business perspective.
• 05:19 Does it make sense to control that factor?
• 05:21 Is the business willing to place it under control, or
• 05:24 does it require too much money and effort to control this factor?
• 05:28 If it is impractical, then the factor becomes a noise element in the study.
• 05:33 Second is feasible.
• 05:34 It's the factor one that is controllable during normal operations.
• 05:38 Either up front in the design process or
• 05:41 once in use by the manager of the process or controllable by the operator or user.
• 05:46 When controlling it, does the factor create safety or regulatory concerns?
• 05:51 If so you probably want to set that factor into the safe zone and not change it.
• 05:56 Finally, the factor needs to be measurable.
• 05:59 That means we need to know what its value was when the experimental
• 06:03 run was occurring.
• 06:04 Ideally, you can directly measure or
• 06:06 observe the factor at the start of the run and throughout the run.
• 06:09 In some cases, you may not be able to directly measure the factor,
• 06:13 but you can indirectly measure it through some other parameter,
• 06:17 such as electrical current or a temperature.
• 06:20 Practical, feasible,
• 06:22 and measurable control factors create a DOE study that is biased toward success.
• 06:27 Since factors is what we are setting in the study,
• 06:30 we need to ensure that the appropriate factor selection occurs.
• 06:35 That way we will have a study analysis that gives us meaningful results.

Lesson notes are only available for subscribers.