Measurement Systems

Quick reference

• 00:04 Hi, I'm Ray Sheen.
• 00:05 Measurement systems are used to collect and record data measurements.
• 00:10 Let's look at some of the problems we have with measurement systems.
• 00:15 >> I want to start by describing how important it is to have a measurement
• 00:18 system analysis tell you whether you have good data.
• 00:21 I'll do this by using Lean Six Sigma project to illustrate this point.
• 00:25 Lean Six Sigma, like most other business processes, relies on data.
• 00:30 During the measurement phase of the Lean Six Sigma project, data is collected to
• 00:34 quantify the performance of the system and set a baseline.
• 00:37 With this data, the magnitude of the problem is understood.
• 00:41 During the analyze phase, data is used to investigate possible causes and
• 00:45 test hypotheses.
• 00:46 Root causes for the problem are confirmed with data,
• 00:49 demonstrating the impact of the root cause on the product or process performance.
• 00:54 In the improve phase,
• 00:55 the new solution that corrects the problem must prove that it works with data.
• 00:59 Tests of the solution are conducted to demonstrate that the problem has been
• 01:03 reduced or eliminated.
• 01:05 If the data does not support that conclusion,
• 01:07 the solution is not considered to be completely effective.
• 01:11 Finally, during the control phase, the solution is in place.
• 01:14 Now a control system is established that will allow process operators and
• 01:18 managers to monitor the performance using data.
• 01:21 Based upon the data recorded in the control system,
• 01:24 a response plan is also established and put in place.
• 01:27 So as to keep the product or process stable and in control.
• 01:31 As you can see, data is a necessary component in each phase.
• 01:35 Without data that is accurate and
• 01:37 reliable, Lean Six Sigma projects are likely to prove ineffective.
• 01:41 This same principle holds true in almost all business processes.
• 01:45 If we can't trust the data, we will make bad decisions or no decisions.
• 01:51 So let's look at the elements of an observed measurement, and
• 01:54 see where the problems might come in.
• 01:56 Let's start by understanding what happens when we do a measurement.
• 02:00 An observation is made using a combination of the human senses,
• 02:03 such as sight or hearing, and the physical measurement equipment or sensors.
• 02:08 This observation is compared to a standard to determine how many units or
• 02:12 increments of the standard is represented.
• 02:15 The observed measurement is then expressed as a quantity of the standard units.
• 02:20 Be it time, dimension, temperature, weight, color or any other items for
• 02:24 which a standard has been established.
• 02:28 That means that there are three essential components that
• 02:31 might go into an observed measurement.
• 02:34 Of course, there's the true value of what is being measured,
• 02:38 then there's the measurement standard that is being used.
• 02:41 But also,
• 02:42 there's the element of measurement error based upon the measurement system.
• 02:46 This error can have many different contributing causes.
• 02:49 Ideally, it is so small to be insignificant.
• 02:53 And the observed measurement of the item compared to the standard
• 02:56 will provide an accurate expression of its true value.
• 03:00 But if the measurement error is large or
• 03:02 unpredictable, the observed measurement cannot be trusted.
• 03:07 The measurement error has many components, and
• 03:09 we will look at each of these in more detail later.
• 03:12 Precision is the variation or uncertainty from measurement to measurement, and
• 03:17 includes both repeatability issues and reproducibility issues.
• 03:21 Accuracy involves the calibration of the measurement system with the standard.
• 03:25 And there are several aspects that can contribute to accuracy issues.
• 03:29 Stability reflects the variability in the performance over time.
• 03:33 Linearity considers whether the accuracy is the same
• 03:36 in all regions where measurements occur.
• 03:38 And discrimination involves the number of gradations in the system performance,
• 03:43 can it detect small differences?
• 03:46 Let's look at the impact of measurement system errors on the observed measurement.
• 03:50 And for simplicity, I'm only addressing common cause errors.
• 03:54 Special cause errors can create even more uncertainty.
• 03:57 In an ideal world, the uncertainty from the measurement process is small and
• 04:02 easily buried in the normal variation of the product or process being measured.
• 04:06 Resulting in a very low impact on the observed measurement.
• 04:10 However, if the measurement error grows, it can take on a much larger impact.
• 04:15 Or if the measurement error introduces a bias or shift in the measurement,
• 04:19 a wrong conclusion can be reached about the item being measured.
• 04:23 Another problem is a measurement system with very poor precision and resolution.
• 04:27 It creates a high impact on the observed measurement because of the tremendous
• 04:30 uncertainty caused by the measurement error.
• 04:34 What are the results of these second two scenarios?
• 04:37 Well first, there is a wrong decision about the quality of the parts or
• 04:40 products.
• 04:41 Good ones are rejected and bad ones may be accepted.
• 04:44 This adds to the immediate scrap and rework costs, but
• 04:47 there is another problem with this.
• 04:49 The uncertainty it creates in the actual performance or quality characteristic
• 04:54 leads to additional testing and analysis that adds costs and delays.
• 04:58 Both are bad for business.
• 05:01 Well, that was for measurement of the product quality.
• 05:03 There is also a problem with wrong decisions about problem solving.
• 05:07 Inaccurate data may cause a Lean Six Sigma or
• 05:10 problem solving team to waste time fixing a non-issue.
• 05:14 While at the same time not recognizing the true root causes of the problem.
• 05:18 And there is also a problem with measurements used to manage and
• 05:22 control processes.
• 05:23 If there are significant measurement errors, the process control is likely to
• 05:28 make the process unstable, as it chases non problems and misses the actual ones.
• 05:33 All this adds variation and uncertainty, which are both bad for
• 05:37 efficient business management.
• 05:40 >> The measurement process includes more than just the item being measured and
• 05:43 the standard it's measured against.
• 05:45 There's an entire system, and that system just like all processes and
• 05:50 systems, is prone to variation and error.

Lesson notes are only available for subscribers.