Locked lesson.

Upgrade

About this lesson

The DOE results can be used by problem solving teams, such as Lean Six Sigma project teams, to identify which factors provide the major contribution to the problem or problem performance. It can also be used to explain the expected benefit from implementing different types of solutions.

Exercise files

Download this lesson’s related exercise files.

DOE in Problem Solving.docx
60.9 KB DOE in Problem Solving - Solution.docx
59.9 KB

Quick reference

DOE in Problem Solving

In some instances, a DOE can help to identify root causes for a problem. If a problem solution requires a major redesign of a product, process or system, a DOE can be used to improve understanding of the critical control factors and to provide estimates of the level of improvement with the new solution.

When to use

DOEs can be used to identify contributing root causes when a problem is caused by common causes rather than a special cause. In that case the DOE is part of the investigation or measure phase of the project. DOE can also be used to aid in the development of a viable and effective solution when that solution requires a redesign. In that case, the DOE is part of the solution planning or improve phase of the project.

Instructions

Most problem solving projects will not need to do a DOE. The problem is based upon a special cause when that root cause is identified and eliminated or controlled, the problem goes away and normal operations are restored. However, some problem-solving projects may benefit from a DOE

When the root cause of the problem is the interaction of many normal factors in the business, a DOE can be used to create the design space equation for that system. The design space equation can then be solved for various circumstances and through that the combination of effects that create the problem can be discovered. In this case, it is usually best to do a full factorial DOE or at least a factorial with Resolution V or higher to ensure all the first order interaction effects are included. 

If the problem is in a manufacturing process and the concern is that the operator cannot control the process, a Taguchi DOE can analyze that through the inner and outer array. If the outer array – the Taguchi noise factors – are significant contributors to the problem performance, a system change is needed

When the problem solution is a major upgrade to an existing product, process or system, a DOE can help the team that is redesigning the system. At this point it is less about solving the problem and more about creating a new product, process or system. However, at least one of the response factors used needs to directly correlate with the problem so that it is clear that the solution is an improvement. The DOE at this time can generate data that illustrates the magnitude of the change. This data may be needed for a cost benefit analysis of the solution. I have found that using a DOE analysis at this point adds a significant level of confidence to the claims of improved performance due to the statistical rigor behind the DOE.

Hints & tips

  • DOE is not well suited to identify complex special cause problems. First, the control factors must be included in the analysis, but you don’t know what the problem is so you don’t know what factor to include. While you could include lots of factors – the test matrix will quickly get large and using a fractional factorial approach will confound interaction effects even though those are likely the most significant. Finally, the DOE will probably use two-level factors which assume linear behavior, but it is likely that the factors are not behaving linearly, so their true behavior will not be correctly modelled.
  • DOE is well suited to create the data that shows how the solution will improve the business situation. This will help the business case for the solution implementation.
  • Make sure any DOE related to problem solving has response factor that directly correlate with the problem behavior.
Login to download
  • 00:04 Hi, I'm Ray Sheen.
  • 00:06 A DOE can also be very helpful when working with a structured problem solving
  • 00:10 process, like Lean Six Sigma.
  • 00:14 Let's take a look at what DOE can do for us in this situation.
  • 00:18 One thing is that the DOE design space equation and sensitivity studies with
  • 00:23 multi-level factors can help to identify the limits for
  • 00:25 control factors, beyond which the product, process, or
  • 00:29 system starts to provide unacceptable performance.
  • 00:33 Also, DOE, especially Taguchi DOEs, can identify acceptable levels for
  • 00:38 manufacturing tolerance limits.
  • 00:40 And don't forget, one of the big advantages of DOE over trial and error
  • 00:44 and OFAAT, DOE can evaluate interaction effects between control factors,
  • 00:49 the others cannot.
  • 00:51 Finally, let's talk a little bit about the slope on the main effects plots and
  • 00:55 one of the ways we can use those to our advantage.
  • 00:59 A steep slope indicates that the factor
  • 01:02 has a very large effect on the performance.
  • 01:05 If I put that factor directly under the operator's control with the feedback
  • 01:09 system so that they can see how the process is currently performing at a given
  • 01:14 time, that control can be used as the primary process or system control.
  • 01:19 This will greatly simplify the overall system control methodology.
  • 01:23 While the items in the previous slide helps us to understand how to use
  • 01:26 DOE to identify ways to control the process and
  • 01:29 avoid problems, but what if the problem already exists?
  • 01:34 if you already know the value of the response variable that is the problem,
  • 01:38 the DOE design space equation can be used to isolate which factors or
  • 01:43 interactions could create that value in the response variable.
  • 01:47 And given the variability in the response factor, a Taguchi DOE can identify
  • 01:52 what types of variances can be expected within the Taguchi noise factors.
  • 01:57 This provides a reasonably good insight into process capability and
  • 02:01 that variance can then be compared to the allowable specification and
  • 02:04 tolerances to calculate the CP and CPK process capability indices.
  • 02:11 The DOE design space equation can be used to predict response factor performance for
  • 02:17 a variety of control factor settings.
  • 02:20 This will indicate which factors have a positive impact,
  • 02:23 which have a negative impact, and which have no impact.
  • 02:27 Another role of a DOE in problem solving project situations
  • 02:31 is the analysis of the proposed solution.
  • 02:34 When the solution is a new system, or a significantly modified system,
  • 02:38 there are often questions about the fragility of that system.
  • 02:42 These are reasonable questions to ask,
  • 02:44 since the old system was probably broken, that's why you're trying to fix it.
  • 02:48 It's reasonable to ask what problems the new system will have.
  • 02:52 The DOE design space equation can answer that question.
  • 02:55 It will identify which factors are significant and
  • 02:58 therefore need to be controlled.
  • 03:00 And the equation can be used to do what if analysis to predict performance under
  • 03:04 a variety of conditions.
  • 03:06 All of these will serve to answer the question about what types of
  • 03:09 problems the solution could have and
  • 03:12 what actions should be taken to prevent problems occurring with the solution.
  • 03:17 Another way in which I've used DOE on problem-solving projects
  • 03:20 is to let the DOE design space equation predict the performance
  • 03:25 once the improvement is in place.
  • 03:27 The benefit from the improved performance level can be evaluated and
  • 03:30 used in the cost benefit analysis that is presented for the different options.
  • 03:35 This provides some scientific and statistical rigor into those benefits, and
  • 03:40 not just wild guesses.
  • 03:42 Through the use of the DOE, you can illustrate the improvement
  • 03:45 by using the data from the refining and optimizing studies.
  • 03:48 There's nothing more powerful to convince someone of the need for
  • 03:52 change than an actual physical example that they can test with their own senses.
  • 03:57 A DOE will identify the factor settings to create those examples and
  • 04:01 it has the statistical rigor to assert that this is not just a lucky guess,
  • 04:06 this now represents the new normal.
  • 04:10 The value of using DOE in problem solving is that it gives us statistical confidence
  • 04:16 in both problem isolation and the adequacy of the problem solution.

Lesson notes are only available for subscribers.

PMI, PMP, CAPM and PMBOK are registered marks of the Project Management Institute, Inc.

Gift this course