11th International Symposium on Measurement and Quality Control 2013, September 11-13, 2013, Cracow-Kielce, Poland

MSA PLANNING - A PROPOSITION OF A METHOD

Jan REWILAK1

1 Cracow University of Technology, Mechanical Faculty, Production Engineering Institute,

Al. Jana Pawła II 37, 31-864 Cracow, [email protected]

Abstract: (250 Words) World-wide known and used quality management

Measurement System Analysis (MSA) is a mandatory system specification for automotive industry ISO/TS 16949
element of quality management systems in automotive [2] requires that “Statistical studies shall be conducted to
industry. Suppliers has to comply both with ISO/TS 16949 analyze the variation present in the results of each type of
and meet customer requirements specified in APQP measuring and test equipment system. This requirement
(Advanced Product Quality Planning), PPAP (Production shall apply to measurement systems referenced in the
Part Approval Process) and MSA manuals. control plan. The analytical methods and acceptance criteria
Manufacturing companies (not only in automotive industry) used shall conform to those in customer reference manuals
perform MSA studies for all measuring instruments on measurement system analysis”. The above mentioned
specified in Control Plans. There are variety of methods MSA-4 manual is most often the one agreed between
available, with gauge R&R as the most popular, to conduct customer and supplier as a MSA reference (it is also a
assessment of capability of a measuring system. However, common practice to develop company specific MSA
although the assessment has to be performed periodically, procedures, basing on MSA-4 [1] manual).
ISO/TS 16949 requires that all measurement systems
none of the above mentioned standards and manuals
included in control plan have MSA performed with
suggest any method for MSA planning. The paper describes
acceptable results. The most common guidelines on how to
a few propositions on how the problem can be resolved develop and manage control plans is - also coming from
basing on a risk associated with each measurement system. U.S. automotive industry - APQP (Advanced Product
The first proposed method is based on simple risk analysis, Quality Planning) [3], which addresses MSA stating that
which can be easily run together with developing a Control “The specified monitoring and measuring devices and
Plan. It uses process capability index (required or achieved) methods should be used to check the control plan identified
and characteristics impact on product/process function for characteristics to engineering specification and be subjected
making a decision on MSA frequency and scope. to measurement system evaluation during or prior to the
The second method assumes that a process FMEA (Failure significant production run. Where special gages, fixtures,
Mode and Effects Analysis) is developed. The method uses test equipment or devices are required per the control plan,
performed FMEA to establish priorities for MSA. Basing on verify gage repeatability and reproducibility (GR&R)…”
the priorities, frequency and scope of MSA can be planned. The third important manual developed by U.S.
automotive industry, addressing MSA, is PPAP (Production
Keywords: MSA, GR&R, APQP, FMEA Part Approval Process) [4], which define documents
required from a supplier to demonstrate successful project
1. INTRODUCTION implementation. Submitted documents should also comprise
MSA records. PPAP [4] states that “The organization shall
MSA (Measurement System Analysis) methods are have applicable Measurement System Analysis studies, e.g.,
widely used in various industries in order to identify and gage R&R, bias, linearity, stability, for all new or modified
assess extra variation (noise) generated in manufacturing gages, measurement, and test equipment.”.
process control and product inspection by a selected Originated from automotive industry, the above
measurement system, This variation, i.e. measurement guidelines spread across various manufacturing industries,
errors, if not controlled, may result in false decisions being adopted as a good practice standard in MSA.
regarding to product or process quality. Standard calibration Applications can be found in machine building, electronics,
procedures do not address the issue of measurement system aircraft, domestic appliances manufacturing, etc.
variation. Therefore, to reduce risk of false decisions based
on process and product measurements, many manufacturing 2. PROBLEM STATEMENT
companies use MSA not only at a measurement system
validation phase (before manufacturing process for a new 2.1 Resources for MSA
project is approved and started) but also after launching a Vast majority of manufacturing companies performing
production process. MSA studies report deficiencies of human resources
The MSA standard procedures originate from necessary to address all measuring systems referenced in
automotive industry OEMs. The standard procedures are control plans. Resources are necessary not only to properly
defined as mandatory quality system requirements for select and perform MSA study, but also to plan, conduct and
suppliers, who are referred to the so-called supplier validate (through another MSA study) corrective actions in
manuals, with MSA-4 [1] from U.S. automotive industry case of unsatisfactory results. The paper does not directly
being the most commonly used MSA good-practice standard
address a problem of resources necessary for MSA.
all over the world, apparently not only in automotive
However, application of the proposed method of MSA
industry.
planning can contribute to more efficient usage of available
resources for conducting MSA.
 11th International Symposium on Measurement and Quality Control 2013, September 11-13, 2013, Cracow-Kielce, Poland

2.2 MSA Planning referred to as special) are characteristics associated with

MSA manual [1] requires that measurement capability of product appearance (e.g. uniform color) or behavior (e.g.
measurement systems should be evaluated taking into silent operation) influencing customer pleasure, not
consideration both a short and a long term behavior of the affecting any functions.
Each characteristics present different levels of non-
measuring system. This requirement incurs a necessity of
conformance risk in terms of probability. Looking at
planning MSA studies, not only at manufacturing process
requirements for capability indices, based on safety area
validation phase, but also during serial production phase to between process variation (usually estimated as six standard
control stability and periodically re-validate measuring deviation spread for normal distribution) and characteristics
systems referenced in control plan (which also may be specification limits, expressed in process standard deviations,
subject to changes). one can group quality characteristics into 3 subgroups
There is no standard method developed for this purpose. determined by standard Cpk (or Ppk) limits 1,33 and 1,67.
Author of the paper reviewed about 50 manufacturing Those two criteria, when connected, form a simplified
companies (in Poland) with respect to MSA planning, risk matrix – Table 1. (no function gradation), Table 2. (with
coming into conclusion that very few of them (approx. 5%) distinction between main and secondary functions).
used other than “once a year” approach. Those few tried to
perform MSA studies more often for critical characteristics, Table 1: SRM for grouping quality characteristics by risk.
but did not pay any attention to MSA method selected, using Cpk / Ppk
≥ 1,67 < 1,67 < 1,33
only standard GR&R analysis, ignoring other methods, e.g. Impact
control charts. Safety / Extreme
There is a growing need to develop a method of MSA legal requirements risk
planning, easy and little time consuming, applicable in Functions Moderate
industrial conditions, using available data regarding product (main / secondary) risk
and process addressed. The purpose of the method would be Appearance/ Negligible
to give guidelines on MSA planning in term of methods behavior risk
selected and frequency of studies, so that deficient resources
devoted to perform MSA are more efficiently consumed in Table 2: SRM for grouping quality characteristics by risk.
terms of reducing risks connected with incapable or unstable Cpk / Ppk
≥ 1,67 < 1,67 < 1,33
measurement systems. Impact
Safety / Extreme
3. SOLUTIONS PROPOSED legal requirements risk
Main High
The basis for the proposed MSA planning method are functions risk
the following assumptions: Secondary Moderate
1. actual or required process capability/performance indices functions risk
are known for characteristics being measured, i.e. short Appearance/ Negligible
term capability Cp, Cpk and long term process behavior risk
performance Pp, Ppk after launching a production,
2. personnel responsible for planning MSA is aware of Each measured characteristics belongs to a relevant cell
product functions and their connections with measured in the matrix. Depending on risk perception (or risk
characteristics, elimination strategy in a project or in an organization), each
3. up-to-date PFMEA (process FMEA) is available, cell can be given a relevant, descriptive risk assessment (e.g.
4. the main criterion for deciding on MSA frequency and low risk, high risk, etc.). Then one can define reasonable
method is a risk connected with a measured requirements regarding MSA for each cell. Thus, each
characteristics in terms of non-conformance potential characteristics can be associated with specific MSA
effects and its probability, requirements.
5. all gauges undergo effective calibration program.
The method is proposed in two basic variants: Table 3: Requirements for MSA planning – example.
• Variant A. Simplified risk matrix (SRM). Cpk/Ppk
≥ 1,67 < 1,67 < 1,33
• Variant B. PFMEA based MSA planning. Impact
Very often,
Safety / legal Often, rigorously
4. MSA PLANNING BASED ON SRM very rigorously
Moderate frequency, Often,
Functions
Each measured characteristics affects product quality. rigorously rigorously
Basing on failure effects severity ranking recommended in Appearance/ Rarely, Often,
None
FMEA in automotive industry [5] one can group quality behavior simplified simplified
characteristics into 3 or 4 subgroups. The most important are
special characteristics, as defined by ISO/TS 16949 [2]. The MSA requirements for each cell can be set by
They typically include all safety related characteristics (or selecting the MSA method of measurement system
relating to legal requirements) and those affecting product capability assessment and/or stability control, approved
functionality (with possible distinction between main and limits for MSA indices (e.g. %GR&R) and frequency of
secondary functions). Less (but still important, rarely MSA activities. To make the method easy for industrial
 11th International Symposium on Measurement and Quality Control 2013, September 11-13, 2013, Cracow-Kielce, Poland

applications it is suggested to use a (heuristic) rule of thumb on PFMEA). Moreover, each measurement system
- the higher is the risk, the more rigorous MSA requirements referenced in PFMEA as failure or cause detection, is
should be (similar to FMEA approach). precisely associated with corresponding process step and
Table 3. shows an example of SRM filled with specific controlled characteristics. Therefore, replacing SRM in the
MSA requirements for MSA planning (for the purpose of first proposed solution by risk matrix built on Severity and
clarity, content of the cells has been detailed in Table 4). Occurrence indices, it is possible to plan MSA taking
additional advantage of developed PFMEA. Moreover, one
Table 4: Requirements for MSA planning – specification. can consider updates of MSA plan as PFMEA is
Specification periodically reviewed and updated, reflecting process
MSA planning improvement. Thus, MSA plan changes can follow process
Stability -
requirements Capability improvements.
Control Chart
Very often, Xbar-R; n=5, Linearity study (n=5); Obviously, there is no need to make MSA specification
very sampling every GR&R- ARM method: different for each of 100 cells of the above mentioned risk
rigorously 4 hours (3 times 10 parts x 3 trials / operator, matrix (Severity x Occurrence). Instead, rather few (but
/ shift) GR&R ≤ 10%; monthly precisely bordered) risk areas can be defined, as it is
Often, Xbar-R; n=3, Bias study; GR&R-ARM recommended in process optimization based of PFMEA [6].
rigorously sampling daily; method: 10 parts x 3 trials An Example is shown in Fig. 1.
per op., %GR&R ≤ 10%;
every 3 months O 10
Moderate X-mR; Bias study; C 9
frequency, sampling GR&R-ARM method: C 8 Rigorous & frequent
rigorously weekly; 5 parts x 3 trials / operator, U 7 MSA
%GR&R ≤ 10%; R 6
every 6 months R 5
Often, X-mR; GR &R- RM or ARM E 4 Moderately freq.
simplified sampling daily; method: 5 parts x 3 trials N 3 & rigorous MSA
per op., %GR&R ≤ 20%; C 2
every 3 months Simplified & rare MSA
E 1
Rarely, None GR&R- RM or ARM 1 2 3 4 5 6 7 8 9 10
simplified method: 5 parts x 2 trials S E V E R I T Y
per op., %GR&R ≤ 20%;
every 12 months Fig. 1: MSA planning guidelines based on PFMEA risk
None Capability and short term stability assessed areas recommended for process improvement [6].
only at the PPAP phase.
Another possible way of taking advantage of PFMEA for
While risk description in SRM (Table 3) has to fit MSA planning is by process risk ranking. Sorting failures
company risk perception or risk management strategy, MSA with their causes by Severity (Sev), Occurrence (Occ) and
specification has to be adjusted to available resources Detection (Det) one automatically obtains a ranking of
capable of setting up and running control charts and process detection controls, i.e. the ranking of measurements
conducting measurement system capability studies (Table 4). systems addressing characteristics with the highest effect
Attempts of such adjustments performed by author showed risk for external and internal customers (Sev), with the
deficiencies of resources for MSA in companies trying to highest probability of a cause (Occ), sorted at the and by
optimize MSA planning. There are two alternatives in such risk of not detecting a non-conformance (Det). Such a risk
a case: analysis approach has been recently recommended [6, 7, 8]
1. acquire necessary resources (proactive approach, as opposed to criticized traditional RPN (Risk Priority
recommended), Number) strategy, in which risk had been measured as
2. adjust MSA planning specification to available resources, multiplication of all risk indices, giving misleading
reducing frequencies and, if necessary, methods, in a recommendations for process improvement. Similar effect is
proportional manner to retain dependence between MSA obtained if RPN is replaced by SOD index, which is a
plan and risk identified in SRM - Tables 1,2,3 (reactive product of combining Sev, Occ and Det digits [5]. Example
approach). of sorting effect is shown in Table 5.
The latter can be achieved e.g. by iterative process of Please note, that the same inspection method (see Table
reducing the plan and summing up number of MSA studies 5 – methods X and E) can be mentioned in PFMEA more
and number of control charts samples to collect during a than once (typically: final inspection).
predefined period of time (e.g. one year) and comparing it Each detection method referred in PFMEA is connected
with company actual capabilities (depending on resources with relevant measurement system (specified in more detail
devoted for MSA), until both match each other. in Control Plan), Thus, according to the process risk ranking,
measurement system ranking can be developed. being a
5. MSA PLANNING BASED ON PFMEA basis for MSA planning. The higher is a measurement
system position in the ranking, the stricter MSA
Assuming that a company has PFMEA in place, one requirements should apply, both in terms of rigor and
finds in PFMEA all measurement systems referenced in frequency. Evidently, defining separate MSA requirements
control plans (according to [2] control plan should be based for each ranking position is not reasonable. Instead, rational
 11th International Symposium on Measurement and Quality Control 2013, September 11-13, 2013, Cracow-Kielce, Poland

grouping would be suggested. An example is shown in 6. CONCLUSSIONS

Table 6., where three groups were defined, somewhat
resembling a concept of ABC method. The paper intends to start discussion about MSA
planning as a vexed problem in manufacturing companies
Table 5: Risk ranking as a result of PFMEA. with management systems according to ISO/TS 16949 [2].
The following problems related to MSA planning would

Detection - Cause

Detection - Failure
need to be further included in proposed methods in further
developments:
Failure Effects

Failure Cause
Failure Mode

Occurrence
1. A few identical or similar gauges are assigned to the
Operation

Detection
same measurement task (characteristics).
Severity

2. There are significant differences between gauges, affect-

SOD
ting a probability of measurement system falling into of
out of control state: a gauge robustness and frequency of
Developed PFMEA for operations O1, O2, O3
measurements (e.g. 100% inspection vs sampling).
O1 A B 7 C 4 D E 5 745 3. Differences in time of measurement cycle affects time of
O1 A B 7 F 2 G E 4 724 MSA test, which could affect MSA planning in terms of
O2 H I 4 J 7 K X 9 479 a test specification (see – Table 4.) and resources.
O2 M N 9 P 3 Q X 6 936 4. Apart from measurement systems for variable data, there
O2 M N 9 Q 5 R X 3 953 can also be measurement systems for attribute data in
O3 T U 9 V 4 W X 2 942 place (e.g. go-nogo gauges, visual inspection) with other
PFMEA after SOD sorting MSA methods (Cross Tab – Cohen Kappa, Signal
O2 M N 9 Q 5 R X 3 953 Detection, Analytic [1]) to include in MSA plan.
O3 T U 9 V 4 W X 2 942 Despite the above considerations, the proposed methods
O2 M N 9 P 3 Q X 6 936 – heuristic and simple, matching widely used quality tools in
O1 A B 7 C 4 D E 5 745 industry, with no special qualifications necessary from
O1 A B 7 F 2 G E 4 724 quality engineers and requiring no extra experimental data –
O2 H I 4 J 7 K X 9 479 would radically increase benefits from MSA by rational
planning.
Table 6: Risk ranking as a result of PFMEA.
No. Measurement REFERENCES
MSA requirements
Systems
1 R (O2), X (O2, O3) Very high [1] Measurement System Analysis Reference Manual. Chrysler
2 W (O3) (SCC* daily, GR&R monthly) Group LLC, Ford Motor Company, General Motors
3 Q (O2) High Corporation. 4th Edition, 2010.
[2] ISO/TS 16949:2009 “Quality management systems -
4 D, E (O1) (SCC weekly, GR&R quarterly
Particular requirements for the application of ISO 9001:2008
5 G (O1) for automotive production and relevant service part
Moderate (GR&R yearly)
6 K (O2) organizations”. ISO, 2009.
* abbreviation SCC here denotes: Stability Control Chart [3] Advanced Product Quality Planning and Control Plan.
Chrysler Corporation, Ford Motor Company, General
As a guideline, ABC analysis, referred to as a variation Motors Corporation.2nd Edition, 2008.
of Pareto analysis [8] could be used to divide the risk [4] Production Part Approval Process. DaimlerChrysler
ranking into subgroups A, B, C (typically - A: 20% of the Corporation, Ford Motor Company, General Motors
ranking – highest risks, B: 30% of the ranking – medium Corporation. 4th Edition 2006.
risks, C: 50% of the ranking – low risks, the subgroups [5] Potential Failure Mode and Effects Analysis Reference
addressing 80%, 15% and 5% of the risk potential effects Manual. Chrysler Group LLC, Ford Motor Company,
respectively). Of course, typical subgroups share (20%, 30%, General Motors Corporation. 4th Edition, 2008.
50%) and contribution (80%, 15%, 5%) can be adjusted to a [6] Verband der Automobilindustrie. “Quality Management in
the Automobile Industry. Product and Process FMEA”.
company specific risk management approach, products,
VDA 4.2. 2nd Edition 2006.
legal and customer requirements, etc. Other values,
[7] SAE International. “Potential Failure Mode and Effects
recommended in quality control, for example include: Analysis in Design (DFMEA), Manufacturing and Assembly
A:15%/65% (share/effects), B:20%/20%, C:65%/15% [9]; Process (PFMEA)”. Surface Vehicle Standard. J1739, 2009.
A:20%/70%-80%, B:30%/15%-20%, C:50%/5%-10% [10]. [8] John S. Oakland, Roy F. Followell, “Statistical Process
The above described method faces the same problem of Control. A practical guide”, p.62, Butterworth Heinemann,
resources, as described in chapter 4. The solutions can also 1995.
be similar to those mentioned: estimate and acquire [9] Szczepańska K., „Metody i techniki TQM”, p.98, Oficyna
necessary resources for MSA plan or adjust MSA plan to Wydawnicza Politechniki Warszawskiej, 2009.
match available resources. Both approaches require relevant [10] Miller P., „Systemowe zarządzanie jakością. Koncepcja
models, which are not addressed by this paper. systemu, ocena systemu, wspomaganie decyzji”,
Wydawnictwo Difin, Warszawa 2011.

Powered by TCPDF (www.tcpdf.org)