Statistical Process Control

Quick reference

• 00:04 Hi, I'm Ray Sheen.
• 00:06 Another control approach that is often part of the solution for a Lean Six Sigma
• 00:11 project is the use of control charts based upon statistical process control.
• 00:16 Now we have an entire course on statistical process control and
• 00:19 the creation of control charts.
• 00:21 So in this module I wanna focus more on how to use those
• 00:25 control charts in your process.
• 00:28 I'll start with a review of process control and process performance.
• 00:33 There are four states of process performance and control.
• 00:37 These four states are a combination of the conditions of a process that is in
• 00:41 statistical control or not.
• 00:43 And whether the process currently is delivering 100% good parts, or
• 00:47 completing some portion of the output in an out of specification condition
• 00:51 resulting in scrap or rework.
• 00:52 Let's look at the implication of these four conditions.
• 00:56 When the process is in statistical control and delivering 100% good results,
• 01:01 it is operating in an ideal state, the results are good, and
• 01:04 thanks to the process control, they're likely to stay good.
• 01:07 Next is the case where the process is in statistical control, but
• 01:11 it's creating some process results that are outside the customer's specification.
• 01:15 This is known as threshold performance,
• 01:18 because it is in statistical control process output is predictable.
• 01:22 The organization can decide whether the cost of the defective output is such that
• 01:26 a project to reduce variation would be worthwhile.
• 01:29 The question is, what are the costs and performance thresholds?
• 01:33 The answer will determine what should be done.
• 01:36 The next condition is a process that is not in statistical control and
• 01:39 is producing some defective outputs.
• 01:42 This will lead to high costs, delays, and often chaos.
• 01:46 The only good news here is that everyone knows there is a problem.
• 01:50 You probably have inspection and testing in place to screen for good and
• 01:53 bad product, and you can probably get support to improve things.
• 01:57 The final case is when the process is not in statistical control, but
• 02:01 it's not currently producing any defective units.
• 02:04 This will often lead to a false sense of security.
• 02:07 The output is good from a customer perspective, but
• 02:10 the problem is that the process is unpredictable.
• 02:14 While it's good today, it may be much different and out of control by tomorrow.
• 02:18 Then everyone will be angry and
• 02:20 questioning, what happened, how did you lose the recipe?
• 02:23 Well, the facts were that you never had the recipe to begin with,
• 02:27 you were just lucky.
• 02:29 So let's take a moment and look at process control, control charting, and
• 02:32 control limits.
• 02:34 Recall that a process is in statistical control when
• 02:37 there are no special cause issues present.
• 02:39 The process is stable with a predictable mean, and predictable standard deviation.
• 02:44 A control chart is designed so as to discriminate between special cause and
• 02:49 common cause variation within the process.
• 02:51 The mean value and control limits are used to help the process operators and
• 02:55 managers know when a special cause is injected into the process.
• 02:59 If the process outputs are between the control limits,
• 03:02 the process is normally operating under control.
• 03:05 One thing to keep in mind, control limits are not spec limits,
• 03:09 the specification may be wider or narrower than the control limits.
• 03:14 Also the customer target and spec limits may even be above or
• 03:17 below the process mean and control limits.
• 03:20 In fact, let's look at the anatomy of a control chart, and what that means for
• 03:24 managing the process.
• 03:26 The horizontal axis is time, and the units or
• 03:29 data points are a sequential plot of the process output.
• 03:33 The plot includes the mean, or in some cases median, and the control limits,
• 03:37 which are the upper and lower bounds for the data, when the process is in control.
• 03:42 These limits and the mean are used to determine whether there is
• 03:45 only common cause variation present, or if there is special cause.
• 03:49 We discussed how to calculate these in the statistical process control course.
• 03:53 You can see we are showing here two special causes.
• 03:56 One is a value above the upper control limit.
• 03:58 The other is more than eight consecutive points below the mean,
• 04:02 which indicates that the mean has shifted.
• 04:05 While these charts reflect the voice of the process, and
• 04:08 what we can expect from the process, it's easy to compare these
• 04:11 points to the customer requirements so that we know which of the four states of
• 04:15 the quadrant I discussed in the first slide apply to our condition.
• 04:19 Finally, let's talk about the strategies for
• 04:22 using control charts in a Lean Six Sigma project.
• 04:25 For starters let me say that we use these in several stages of the Lean Six Sigma
• 04:30 project.
• 04:30 We often think of these in the control stage as a way to maintain the improve
• 04:34 status, but we can use them earlier if we have the data.
• 04:39 We can use them in the measure stage to establish base line performance
• 04:42 as part of our as is process,
• 04:45 that will also indicate which quadrant we are currently operating in.
• 04:48 They can be very helpful during the analysis stage to help isolate any
• 04:52 problems, and can provide insight into the possibility
• 04:55 of a process output that is really a combination of several processes.
• 05:00 We can also use them in the improve stage to demonstrate the impact of
• 05:04 the improvement, by comparing the improve stage control
• 05:07 charts to the baseline control charts created during the measure stage.
• 05:11 One final comment about the use of control charts in the control stage.
• 05:15 Control charts aren't needed for every parameter on every process.
• 05:18 Use them wisely, but use them sparingly.
• 05:22 A proliferation of control charts on every possible measure
• 05:25 just leads to people becoming numb to them.
• 05:28 Decide what are the key control parameters for your process, your design of
• 05:32 experiments could help with that, and just put control charts on those items.
• 05:37 Control charts, when well understood and applied to appropriate measures,
• 05:41 will help you get your process into the ideal state and keep it there.

Lesson notes are only available for subscribers.