Recently, there have been a lot of JMPer What kind of preparations should we make before implementing experimental design? This topic has a wide range of topics, but most of our questions and answers often point to measurement system analysis. (Measure System Analysis, MSA) . After thinking about it, I discovered that what everyone wants to emphasize is not so much the analysis of the measurement system, but the issue of ensuring the quality of the data. Data is generally the result of measurement. Therefore, the validity of experimental data naturally needs to be verified by the measurement system. effectiveness is guaranteed. But there is actually no necessary causal relationship between the two. To put it another way, measurement system analysis is a basic and routine work. Regardless of whether experimental design is carried out, when you want to use data to drive quantitative decisions, shouldn’t the effectiveness of the measurement system be guaranteed by evaluation, verification and confirmation? ?
As a result, another more basic question arises, namely: how to pass JMP software to perform measurement system analysis? In this article, we will provide you with some guidelines for getting started.
First, it causes some users to use JMP conduct MSA One reason for the slight confusion may be that JMP for MSA Multi-platform support sometimes provides too many choices, making it difficult to choose. We're here to clarify.
In the current JMP , mainly provides two supports MSA functional platform ( picture -1) ,they are, respectively:
- Analysis> quality and process > Measurement system analysis
- Analysis> quality and process > variability / Counting tool drawing
Figure-1 JMP for MSA Multi-platform support
For the "Measurement System Analysis" platform, it is first based on EMP (Evaluating the Measurement Process) Methods for measurement system analysis. This concept was proposed not long ago and was proposed by the United States SPC expert Donald J. Wheeler Dr. Yu 1984 in his writings Evaluating the Measurement Process was first proposed in , and JMP The methods cited are derived from their 2006 published in EMP Ⅲ Using Imperfect Data (2006) in a book ( picture -2) Content and steps described.
Since this method has not yet been popularized in China and few people know it, therefore, in the JMP that we have come into contact with Among the user groups, the usage rate is low, so this article will not waste any time to elaborate on it. In addition, the platform actually also provides traditional " Gage R&R ” entrance to analytical methods, but this is not the same as the “variability” / The effect of entering the "Counting Tool Drawing" platform is the same, so you can refer to the following directly.
Figure-2 EMP Ⅲ Using Imperfect Data (2006)
At present, the mainstream measurement system analysis ideas in China are mainly based on the Automotive Industry Action Group (AIAG) The "Measurement Systems Analysis Reference Manual" is continuously updated.
Under this guideline, usually through 6 There are two conventional dimensions to evaluate the measurement system, which are stability (stability) , bias (bias) , linear (linearity) , resolution (discrimination) , repeatability (repeatability) and reproducibility (reproducibility) .
Among them, as shown in Figure-3 As shown, bias is the measure of system accuracy (accuracy) The most direct measure reflects the difference between the average value of the measurement results and the true value; repeatability and reproducibility are the accuracy of the measurement system. (precision) The measurements reflect the fluctuation of measurement results, and they are the most critical indicators for evaluating the quality of the measurement system. So, for “variability / For the "Counting Tool Drawing" platform, most of its functional items are related to this.
image 3 Bias and precision diagram
For the majority of Chinese users, among the above 6 Among these characteristics, the most familiar ones are “repeatability” and “reproducibility”, the so-called Gage R&R , almost all Six Sigma management-related teaching materials, training and projects are bound to touch upon it, so that many people over time just forget about it. Gage R&R almost equivalent to MSA , and this is also the focus of our article.
Here, "repeatability error" mostly refers to the same operator, using the same set of measuring equipment, making multiple measurements of the same characteristics of the same component within a short time interval, and the consistency of the results obtained, that is, emphasizing the consistency of all measurement results. Fluctuation conditions that are all under the same conditions as possible, also known as equipment variation (EV) ; In contrast, "reproducibility error" emphasizes the consistency of the results obtained from multiple measurements of the same characteristics of the same measurement component under various possible changing measurement conditions. In traditional cases, reproducibility Errors are mainly caused by different operators, also known as personnel fluctuations (appraiser variation, AV) .
These two fluctuations are usually expressed in terms of variance or standard deviation. The smaller the fluctuation, the better the measurement system. The traditional analysis method is often combined with the fluctuation of the production process, and is achieved through variance analysis that considers random effect factors. The process of variance decomposition is shown in Figure-4 shown.
Figure-4 Gage R&R Schematic diagram of traditional analysis methods
Finally, two important indicators for evaluating the accuracy of the measurement system are derived:
Among them, %Gage R&R Focus on evaluating the measurement effect of the measurement system on the overall process variation, emphasizing the measurement system's ability to resolve production process improvements;
%P/T Focus on evaluating the measurement effect of the measurement system against product specifications, emphasizing the measurement system's ability to resolve tolerance limits.
%Gage R&R and %P/T are two different aspects to evaluate the performance of the measurement system. A good measurement system should make both indicators small enough at the same time. -1 It is the general reference judgment standard for these two metrics.
Table-1 Gage R&R reference judgment standard
At this point, if you can understand the above analysis ideas and logic, then, in JMP With the help of modern statistical analysis software, you only need to collect and organize Gage R&R The observation data required for analysis can be entered into the software, and all subsequent calculation processes can be left to the software.
Next, we will use a simple small case to analyze Gage R&R The operation process of traditional analysis methods is demonstrated, and the output results are briefly analyzed.
"Use a vernier caliper to measure the shaft diameter of a certain machined component. The known tolerance requirement is 45.5±0.5 , for measurement system accuracy analysis:
- A representative sample of 10 parts and numbered
- Pick 3 at random Inspectors measure each part with their own calipers
- After shuffling the order of the parts, the second test Second-rate
- After shuffling the order of the parts, the third place was tested Second-rate
- The measurement process diagram and recording data table are shown in Figure-5 Shown"
Figure 5 Measurement process diagram and recording data sheet
JMP Gage R&R Analysis steps:
- Combined with JMP Data management related platform, easily import and organize data as shown in the figure -6 The data table structure shown
Figure-6 Gage R&R Data entry and organization
- Press Figure-7 Settings, start "Analyze > quality and process > variability / "Counting Tool Drawing" Platform
Figure-7 "Variability / "Counting Gage Drawing" Startup Platform Settings
- As shown in Figure-8 As shown, select the red triangle next to "Variability Gage" > Gage research > Measuring tools RR , enter the tolerance interval, for this example 46.0-45.0=1
Figure-8 Start "Measuring Tools" RR "analyze
Gauge R&R Brief analysis of analysis output report:
Figure-9 Measuring tools R&R Analyze output report
From Figure-9 Gage shown R&R From the output report, the repeatability variance component is approximately 0.0011 , the reproducibility variance component is approximately 0.0035 , two important indicators to evaluate the accuracy of the measurement system %Gage R&R=33.56% , %P/T=40.56% , all greater than 30% , Therefore, the accuracy level of this measurement system is not high, and measures should be taken to improve it, and the improvement can first start with reducing the reproducibility error.
Further in-depth analysis can also be found that in the reproducibility variation, "inspector* "Part" variation (0.3116)> "Inspector" variation (0.1708) , which is plotted with the variability of "measured value" and the mean "measured value" - Inspectors *" part " interaction diagram ( Gage research > Measuring tools RR picture > Mean chart ) presented in 3 of an inspector 4 , 5 , 10 It is consistent with the phenomenon of large differences in the measurement results of parts.
Therefore, if the characteristics of these components are discussed in depth during measurement standard operation training, it is possible to reduce the reproducibility error and thereby improve the accuracy of the measurement system.
Figure-10 Variability plot for "measured value" and mean "measured value" - Inspectors *" part " interaction diagram
The above is what we will introduce in this issue about MSA The introductory content is actually about MSA There is still a lot of basic knowledge, such as measurement system analysis for destructive experiments, measurement system analysis for counting data, measurement system analysis using standard parts or fixed parts, and how to evaluate and determine the measurement system when the two major accuracy indicators are inconsistent. etc. On the one hand, these contents are related to software operation, and on the other hand, they also need to start from the methodology itself. We will discuss it with you in subsequent articles.
