Three years ago Measurement Siege Lion Consulting on the calculation method of GRR report on MAX resistance. Three months ago, there was also technical Yuan Consulting to find out how to give the GRR tolerance of shape tolerance . Three days ago, a colleague asked how to define the tolerance in the Min size in the GRR report. It can be seen that this is indeed a problem that has troubled everyone for a long time.
Today, Measurement Siege Lion will discuss with you the definition of unilateral tolerance dimensions in the GRR report, let us become the terminator of this problem together.
Measurement Siege Lion has already introduced the definition, classification and calculation process of GRR in the article "Why do GRR Analysis". I will not repeat it here, we will directly get to the topic.
is mentioned in the article "Why do GRR analysis" that for the upper and lower tolerance sizes of , the final value of the GRR is 16.68% of the pink font base data in the figure below, and for the shape tolerance and one-side tolerance MAX and MIN, the final judgment value of the GRR is 21.13% of the red data on the left of the pink font base data.
Why?
Let's first understand the analysis method of GRR:
There are 4 GRR analysis options in the GRR report of the mean extreme difference method Average and method, which are:
1. Based on part variation
uses the part variation method based on part variation. This method does not require size tolerance . Even if the tolerance is entered, it will not have any impact on RR. This analysis method can be used in this way.
2. Based on part tolerance (Base on parts spec)
uses part tolerance analysis method, so you need to fill in the dimensional tolerance. This analysis method is generally not used for single-side dimensions.
Among the four GRR analysis options for the mean extreme difference method, there are normal dimensions with upper and lower tolerances. We generally choose the default "Base on parts spec" for form and position tolerance and single-side tolerance MAX and MIN, we can choose "Base on parts variation"
In the MSA manual, there are two calculation formulas for GRR and CP, one is based on process deterioration and the other is based on tolerance range. The calculation formulas of these two formulas are obviously not convertible to each other. Their application conditions depend on the purpose of MSA's analysis.
3. Base on process variation
is used for process control, distinguish between ordinary causes and special reasons and process variation
4. Baseon process PP
is used for product quality and bad judgment, distinguish between qualified or not
. So what is ordinary causes and special reasons?
Common reasons: refers to the reason why the process has a stable and repeatable distribution process that occurs in a controlled state. Ordinary reasons are manifested as accidental reasons for stabilizing the system. The output of the process can only be predicted if the common cause of the process deterioration exists and does not change.
Special reasons: (usually also called the identifiable cause) refers to the cause of deterioration that does not always act on the process, that is, when they appear, it will cause the distribution of (the whole) process to change. Only use special reasons to be detected and taken, otherwise they will continue to affect the output of the process unpredictable.
During the work process, various factors that affect the results may play a mutated role. We often summarize Man (personnel), Machine (machine tooling), Material (raw materials), Methods (operation methods), Milieu (environment), and measurement (Measurement), these are all factors in the process. Or the reason for the worsening results. Through practice, statistician Shaw Hart found that we can divide these causes into two categories: ordinary causes and special causes. According to this classification to control the process, a large number of disputes can be solved (mainly distinguish whether the leader is responsible or the on-site operator is responsible to solve the problem). The conclusion is:
When the process only causes deterioration due to ordinary reasons, the problem is mainly solved by the leader;
When there are special reasons for the process, the problem is mainly discovered and solved by the on-site operator.
In the GRR report of the ANOVA method (or "variance analysis method"),
For the conventional dimensions of upper and lower tolerances, we need to enter the tolerance bandwidth = upper tolerance - lower tolerance.
, for example, size 0.35+/-0.05. The tolerance bandwidth is
upper tolerance 0.05-lower tolerance -0.05=0.05-(-0.05)=0.05+0.05=0.10
So we input 0.1 in the GRR report of the ANOVA method, and calculate the GRR value:
based on part tolerance (PERCENT OF TOLERANCE) = 7.27%
Also, we assume that this is a single-sided control dimension Max0.35 for assembly parts, and its GRR value is
Based on the overall variation of the part (PERCENT OF TOTAL VARIATION) = 58.76%
At this time, the Tolerance in the report should be left blank, as shown in the figure below:
Leave blank Tolerance. We found that there is no data based on the PERCENTOF TOLERANCE column, which means that there is no tolerance bandwidth involved in the calculation, so there is no tolerance percentage.
(The P/T ratio metriccannot be applied to one-sided specifications. Use the %RR to assess theacceptability of the measurement system.)
In the standard GRR analysis of variance method ANOVA method report, a note will generally be added to the tolerance cell, indicating that P/T (Parts/Tolerance) is not applicable to GRR analysis of unilateral tolerances. (T does not exist, how to calculate it?)
But many readers are still confused. The GRR analysis of Max0.35, the tolerance column does not fill in the tolerance column, which seems a bit unaccustomed. Of course, some customers who do not know how to read GRR reports. So how should I deal with such a client?
here to measure the siege lion, that is, for Max0.35 or Min0.35 or contour degree 0.35, fill in the cell directly with the tolerance. At this time, the report will be complete, but remember to indicate that PERCENT OF TOTAL VARIATION = 58.76% is the final GRR value. And please ignore the existence of (PERCENTOF TOLERANCE) = 2.08%. As shown in the figure below
However, how to select a sample to meet the GRR of the unilateral tolerance to reach within 30% or less?
Let’s give you another chestnut. As shown in the figure above, the GRR has reached 58.76%. Let’s assume that we reduce the parts1 of the above data by 0.02, and parts6 is increased by 0.01, and parts9 is reduced by 0.01, so that the variation between 10 samples is increased to above 0.03. Let’s take a look at the data calculation results.
I'll go! It actually dropped from 58.76% to 13.64%, which is so unreasonable!
We then use MIN0.10 as a chestnut. In fact, this is a natural principle. The natural principle is: the variation between
samples must reach more than 1/10 of the unilateral tolerance, that is, the deviation must reach 0.01 (if it is MIN0.35, it is 0.035).
The variation within the same sample and between groups should be controlled at about 1/10 of the total variation between samples, that is, it should be controlled at the μ level range. For example, the variation between samples is 0.01, then the deviation between groups within the group is controlled at about 0.001~0.003mm.
Why do you dare to use such a rigorous word as "left and right" when measuring siege lions?
That's because, 30 groups of GRR data have a ratio of 0.001~0.003. For example, if there are more than 25 groups to reach a deviation of 0.003, then it is definitely difficult to reach within 20%. If there are more than 25 groups to control the deviation of 0.001, it is basically impossible. You can see at a glance that you are playing with a monkey--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Therefore, this ratio is randomly dynamic.
(Inscription by Zhu Rongji, the National Bureau of Statistics)
Through the above examples, for engineering analysts such as QE, SQE, MQE, PD, etc., understanding the GRR calculation principle of unilateral tolerance is equivalent to solving a long-term problem.
