Experiments and Design

Quick reference

• 00:04 Hello, I'm Ray Sheen.
• 00:05 Let's start our discussion of Design of Experiments by setting a baseline for
• 00:10 the use of experiments in the design process.
• 00:14 First let's take a minute to agree on what we mean by experiments.
• 00:18 Experiments are investigative research.
• 00:21 By that we mean that we are using the experimental process to uncover new
• 00:25 understanding or confirm a conjectured hypothesis.
• 00:28 At the end of the experiment,
• 00:30 there should be a conclusion based upon the experimental results.
• 00:34 As part of this investigation,
• 00:36 we control a number of factors known as independent factors or variables.
• 00:40 And then we have one or more dependent factors or
• 00:43 variables that we measure during the experiment.
• 00:46 This dependent factor is often something that we want to understand or control.
• 00:51 By conducting the experiment, we determine how the independent factors impact
• 00:56 the observed result we see in the dependent factor.
• 00:59 We can begin to understand to what extent the dependent factor depends upon
• 01:03 the independent factors.
• 01:05 This can be useful in several ways.
• 01:07 If we gain a thorough understanding of the relationship, we can create an accurate
• 01:12 mathematical model that can be used for predicting performance.
• 01:16 The experimental results may be used to expand knowledge of a subject, or
• 01:20 may be used to demonstrate cause and effect between the independent factors and
• 01:24 the dependent factor.
• 01:25 With this understanding, we can tune product and process designs for
• 01:30 optimal performance.
• 01:31 This leads us to the question of when we should use experimental design.
• 01:37 And there are some obvious situations where it would add value.
• 01:40 One is the creation of a new generation or a category of products.
• 01:43 Experimental design helps to understand the performance characteristics of
• 01:47 the new product.
• 01:48 Another time is when doing a technology change or upgrade to a product or process.
• 01:53 Experiments will provide the information needed to understand the new capabilities,
• 01:57 model the performance, and then optimize the product or
• 02:00 process with this new technology.
• 02:02 A third occasion is when changing an existing product or
• 02:05 process to resolve a problem.
• 02:07 This is often the reason why Lean Six Sigma projects use Design of Experiments.
• 02:11 The changes to some element of the product or process is an independent factor
• 02:15 that will affect the overall product or process performance.
• 02:19 The experiments are used to optimize the product or
• 02:22 process with that new technology.
• 02:24 This is normally needed when the problem being solved is a common cause problem.
• 02:28 That means that the problem is inherent in the existing design and
• 02:32 a new design is required to eliminate or control the problem.
• 02:35 The experiments support the creation, verification, and
• 02:39 documentation of this new design.
• 02:41 While experiments can be very helpful in the design process,
• 02:45 they don't fit every situation, so we need to know when not to use them.
• 02:49 Experiments can be costly and time-consuming.
• 02:52 I'll make the assumption that your boss wants an answer quickly and
• 02:55 doesn't wanna spend a lot of money to get there.
• 02:57 So if I can answer the question without experiments, it's often best to do so.
• 03:01 Let's look at a few instances where experiments are not necessary.
• 03:06 If the project you're working on is tasked with solving some problem and
• 03:10 you've determined that the problem is due to a special cause,
• 03:13 you don't need to experiment.
• 03:14 You need to get rid of the special cause.
• 03:17 So for instance, if you determined that the reason the machine was making bad
• 03:21 parts was that the tooling used in the machine is broken, fix the tools.
• 03:25 You don't need to do experiments to understand how badly the tools were
• 03:29 broken.
• 03:30 Just fix it and then the process returns to the normal condition.
• 03:34 Another case could be a design modification that doesn't really modify
• 03:38 the design.
• 03:39 It could be design change to do a line extension.
• 03:42 Adding one more color option to the product selection or
• 03:45 translating the product documentation into another language.
• 03:48 This may expand the product offering, but
• 03:50 it does not change the product performance.
• 03:53 So let's wrap up this discussion by introducing the idea that there are many
• 03:57 ways to conduct a set of experiments.
• 03:59 Scientists, designers, and
• 04:01 inventors have been using experiments since the dawn of man.
• 04:04 When faced with a technical problem or a challenge, someone has tried or
• 04:08 experimented with several options until they find one that worked.
• 04:12 Sometimes, experiments have quickly provided knowledge or
• 04:15 insight into what works.
• 04:17 But many times, it is not, and numerous experiments were needed.
• 04:21 Thomas Edison and his staff conducted thousands of experiments when working on
• 04:25 inventing the light bulb.
• 04:27 Many of these created light, but not to the level of brightness,
• 04:30 length of operation, and fragility that would be needed for a commercial product.
• 04:35 However, each experiment added to the base of knowledge.
• 04:38 When we consider the approach to the experimental process and the methods used
• 04:42 to plan the experiments, they fall into one of four general categories.
• 04:46 The trial and error or lucky guess method, which is the most common.
• 04:51 The one-factor-at-a-time or OFAAT method,
• 04:53 which is often referred to as the scientific method of experimentation.
• 04:57 Then there is the full factorial design of experiments, which determines the envelope
• 05:02 of performance for the product or process by looking at the combination of min and
• 05:07 max settings of all the independent factors.
• 05:09 Finally, there is the fractional factorial design of experiments.
• 05:13 This technique uses a statistical analysis of a subset of
• 05:16 the full factorial DOE to estimate and optimize the design.
• 05:20 Each of these are viable, and in fact,
• 05:23 may be the preferred approach under certain project conditions.
• 05:26 We'll consider each one in the next few lessons of this course.
• 05:31 Experiments have historically been a part of the design process.
• 05:34 But as we saw, there are several different ways to organize your experiments.

Lesson notes are only available for subscribers.