Control Charting Process

Quick reference

• 00:04 Hi, I'm Ray Sheen.
• 00:06 We keep looking at control charts and talking about the use of control charts,
• 00:10 but let's take a few moments and consider how we actually build a control chart.
• 00:16 Process is quite simple.
• 00:18 Start by selecting what process parameter you wish to control.
• 00:21 It should be something that is critical to product or
• 00:23 process performance, such a critical dimension or cure time.
• 00:27 Or it could be a parameter that provides important insight into how the process
• 00:31 is operating, such as a hold time for calls that are waiting in the call center.
• 00:35 Next, based upon the data type, determine which control chart to use.
• 00:40 I'll talk more about this in the next few slides.
• 00:43 3rd, start collecting the data and calculate the appropriate statistics.
• 00:47 This will vary a little based upon the type of chart, but
• 00:50 we're talking about descriptive statistics like mean, range, standard deviation.
• 00:55 Once you have these, you can calculate the control limits for your data set.
• 00:58 4th, create your preliminary control chart,
• 01:02 I say preliminary because we may need to make some process changes.
• 01:06 But if your process is already under statistical control and you're making
• 01:10 conforming product, your preliminary chart will be your final chart.
• 01:15 Step 5 is to establish statistical control,
• 01:17 that means eliminate any special causes.
• 01:20 Until this is done, you can't create a final control chart.
• 01:23 Depending upon what's happening in your process,
• 01:26 you may need to remove several special causes.
• 01:28 6th is to establish process capability, by that I mean to calculate Cpk or sigma.
• 01:34 This will determine if your process is producing conforming results.
• 01:38 And based upon this, you may need to center your process or
• 01:41 possibly even making a change to reduce common cause variation.
• 01:45 On the step 7 in creating the final SPC control chart for the process.
• 01:49 By now your process is stable and capable.
• 01:52 So you're ready for step 8,
• 01:54 which is to maintain process control by monitoring the control chart and
• 01:58 taking appropriate action whenever you see indications of a special cause.
• 02:03 Let me give you some pointers about data and
• 02:05 data collection for SPC control charts.
• 02:08 First is the question of the amount of data to collect.
• 02:11 Remember, we want to get at least 30 data points to stabilize our control limits.
• 02:15 I suggest that with low volume processes, you try to use every data point.
• 02:20 This is sometimes called census data.
• 02:22 It basically means that you're using a sample or sub group size of one.
• 02:27 Now if you need a large sub group to address some abnormality,
• 02:31 well, do what you need to do.
• 02:33 If your process is a high volume process, you can use subgroups to sample data and
• 02:38 still obtain primal data points.
• 02:40 Of course, you could still use census data it just means that it's lots and
• 02:44 lots of data being collected.
• 02:45 If your SPC is automated, that may not be a problem for you.
• 02:50 The next question is, how frequent to collect data?
• 02:52 Well, of course if you're collecting census data,
• 02:55 meaning every point, you collect it whenever a process is running.
• 02:59 However, if you are collecting sample or subgroup data, you have a choice to make.
• 03:03 I view this as a risk decision.
• 03:05 If the process is not performing particularly well,
• 03:08 I will collect data more frequently, say, every hour if it's a high volume process.
• 03:12 However, if the process is performing very well and has been doing so for
• 03:16 a long time, then I may use a less frequent interval such as once a day.
• 03:20 Part of this decision is also based upon the cost to collect the data,
• 03:24 it's automated, the cost to collect is almost free, so do it more frequently.
• 03:28 But if the data collection is manual, the cost maybe very high and
• 03:32 that could also be a reason to collect less frequently.
• 03:36 The last point is to be very clear on your data definition of what is a defect or
• 03:41 a defective unit and an opportunity.
• 03:44 The control chart that you use will depend upon what you're counting and
• 03:47 what you're measuring.
• 03:49 So be sure it is clearly defined.
• 03:52 Now I keep talking about different types of control charts,
• 03:55 lets look at the major ones that we will be working with.
• 03:59 To decide what type of control chart you should use,
• 04:01 you start by first being clear about what type of data you are working with.
• 04:05 When data is variable data, we'll use either the I-MR chart or
• 04:10 the Xbar-R or an Xbar-S type of chart.
• 04:13 The differences between these is the size of the sample of the subgroup of data.
• 04:18 If you're looking at every datapoint, then you're using the I-MR chart.
• 04:22 If you're using samples, then a sample size less than 10 uses Xbar-R and
• 04:27 one greater than 10 uses Xbar-S.
• 04:31 But you may have recalled that I said that much of our data is attribute,
• 04:35 not variable data.
• 04:36 Fortunately, Shewhart developed control charts for that type of data.
• 04:40 There are four attribute data control charts.
• 04:43 And they vary depending upon whether you are counting defects or defectives.
• 04:48 Or whether your sample size is consistent or if it varies from sample to sample.
• 04:53 The final two control charts we will discuss are Time Weighted charts.
• 04:57 That means that the data point is given an added weight in the analysis
• 05:00 depending upon whether it was recently collected or it is an old data point.
• 05:05 We'll talk about those in more detail when we get to them.
• 05:09 Well, control charting process really is straight forward.
• 05:12 Define the data, collect the data,
• 05:14 plot the data, stabilize the process and then control the process.

Lesson notes are only available for subscribers.