IPC-TM-650 EN 2022 试验方法-- - 第50页

A graph showing only the relationship between the measurement variability and the spec l imits is shown below . Here the normal curve represents only the variability of a given me asurement. The curve clearly consumes ha…

The next table in the scorecard is shown below. It compares the measurement variation to the speciﬁcations for this prod-

uct. This table will only have numbers in it if there are two speciﬁcations. In this example, there are two specs, so the table

is complete.

The numbers in the table are percentages. The calculations show what portion of the speciﬁcation width is used up by mea-

surement error. The total measurement variability, called S

R&r

, consumes 55% of the speciﬁcation width. This is higher than

is considered acceptable. A good measurement would consume less than 10% of the spec width. An unacceptable measure-

ment would consume more than 33% of the spec width. Note that this conclusion is conﬁrmed by the dialog box at the bot-

tom of the table.

The graph plainly shows that an unacceptable number of the gaskets measured will fail, and that this failure rate is driven

by the measurement error. The process capability must also be improved, but the measurement must be improved before one

can be sure that measurements will be precise enough to reﬂect true product performance. To help diagnose the cause of the

measurement variation, the repeatability and reproducibility estimates are included. In this example the measurement varia-

tion is roughly equally split between within and between test condition sources. Therefore there may be several sources of

variation, with no one dominant cause.

A graph is often more compelling than the numbers. If one graphs the total observed variation and process variation as nor-

mal curves and superimpose the spec limits, the following graph would emerge:

This graph is prepared by creating normal curves which center on the overall product average, and extend out three times

the applicable standard deviation in each direction. For the total observed variation, the total variation was used. For the

process capability curve the product variability was used.

Comparison to Specifications

(Only used when both specifications exist)

% Repeatability

39.6067 Percent of specification width used by repeatability error

39.1143

Percent of specification width used by reproducibility error

% Measurement R&r 55.6652

Percent of specification width used by measurement error

Compared to the specification limits, the precision of

this measurement:

Needs improvement

% Reproducibility

0.2 0.4 0.6 0.8 1.0 1.2 1.4

Spec

limits

Total

observed

variation

Process

capability

January

2003

Users

Guide

A graph showing only the relationship between the measurement variability and the spec limits is shown below. Here the

normal curve represents only the variability of a given measurement. The curve clearly consumes half the spec width. Place-

ment of the curve in the spec interval is arbitrary. The curve is formed by drawing it with a total width of six times the

R&r

Sometimes there is only one spec, and the table above will be blank. For those situations, the next table compares measure-

ment variability to the total observed variation.

The numbers in the table are again percentages, but this time percentages of total variation. In this example, the measure-

ment variation is 23% of the overall total observed variation. The same guidelines apply, except measurements would con-

sume less than 10% of the total variation. Unacceptable measurements would consume more than 30%. This example falls

in the gray zone in between; not good enough, with improvement recommended.

The ﬁnal metric in the scorecard provides a measurement tolerance. The stated value forms a 99% conﬁdence interval. The

tolerance is shown in this example:

A ‘‘conﬁdence interval’’ is one of those strange statistical concepts that ought to be easy to describe, but somehow become

garbled when translated from equation to English. Let us try to state it clearly:

We make a measurement to estimate the actual, but unknown, performance point of the product. There is a 99% chance that

the actual performance will be within ± the measurement tolerance of the measurement.

In this example, any given gasket measurement will have a tolerance of ± 0.111 mm. The real gasket thickness is within ±

0.111 mm of our measurement.

The ﬁnal step is to determine lessons learned from the MSA and document any changes to the test procedure. If the evalu-

ation indicates the test procedure needs to be improved, these improvement projects should be undertaken as soon as pos-

sible.

0.2 0.4 0.6 0.8 1.0 1.2 1.4

Spec

limits

Measurement

error

Comparison to Total Variation

% Repeatability 2.76653 Percent of total variability used by repeatability error

% Reproducibility

2.69817

Percent of total variability used by reproducibility error

% Measurement R&r

5.46469

Percent of total variability used by measurement error

Compared to the total variability, the precision of

this measurement:

Is adequate

Measurement Tolerance

Units

On the basis of this study, there is a 99% chance that

this interval around the measured product performance

will contain the actual product performance level.

+/- 0.111243938

Users

Guide

January

2003

Document

and

Correct

Measurement Systems Analysis For Variables Data:

Measurement Precision Calculator

Version 1.0, August 2002

Introduction

Welcome to the Measurement Precision Calculator. This workbook will help

perform the calculations for Measurement Precision Studies as described

in the IPC test method IPC-TM 1.9.

This software is useful in cases where the measurement data is recorded

as variables data and recoreded on a continuous scale.

Please refer to the test method before beginning the measurement precision study.

The test method defines the terms and equations used in detail.

Also refer to the User Guide for detailed instructions..

This workbook includes the following

- A blank spreadsheet to enter measurements

- A spreadsheet with example data

- A form for manual data entry, if desired.

- Further information on the K factors used in the calculations

The title cells on the spread sheet have been color coded to aid understanding.

The following is the color key:

Color

General title

Cells for user to enter data

Title cells for test condition related data

Title cells for data related to repeated readings

Title cells for sample related data

Title cell for calculated data

Usage