Vous êtes sur la page 1sur 10

Measurement 43 (2010) 842851

Contents lists available at ScienceDirect

Measurement
journal homepage: www.elsevier.com/locate/measurement

Evaluating measurement and process capabilities by GR&R with four quality measures
Abbas Al-Refaie *, Nour Bata
University of Jordan, Amman-Jordan University street, Amman 11942, Jordan

a r t i c l e

i n f o

a b s t r a c t
This paper proposes a procedure for assessing a measurement system and manufacturing process capabilities using Gage Repeatability and Reproducibility (GR&R) designed experiments with four quality measures. In this procedure, a GR&R study is conducted to obtain replicate measurements on units by several different operators. The gage and part variance components are then estimated by conducting analysis of variance (ANOVA) on the GR&R measurement observations. Finally, the acceptance and rejection criteria of the precisionto-tolerance ratio (PTR), signal-to-noise ratio (SNR), discrimination ratio (DR), and process capability index (Cp or Cpk), are employed to assess the measurement and process capabilities. Three previously studied case studies are provided for illustration; in all of which the procedure provided efcient capability assessments at minimal computational and statistical efforts. In conclusion, the procedure proposed in this research using GR&R designed experiments provides valuable procedure and helpful guidelines to quality and production managers in assessing the capabilities of a measurement system and manufacturing process, and deciding the needed actions for improving performance. 2010 Elsevier Ltd. All rights reserved.

Article history: Received 13 August 2009 Received in revised form 16 October 2009 Accepted 26 February 2010 Available online 4 March 2010 Keywords: GR&R study PTR SNR DR Process capability indices ANOVA

1. Introduction Quality management efforts are often directed for zero defect production by reduction of variability. If a product is found nonconforming, it is usually claimed that the variability is attributed by process and thus improvement actions are implemented to enhance process capability. Unfortunately, such efforts may not necessarily result in improved process capability, because it is possible that the process is already capable enough, but there is no way of proving this due to inadequate measurement system. In addition, it may happen that the measurement system is already capable enough; however the measurement error is still unacceptable when compared to process variability. Therefore, investigating both the variabilities of a measurement system and a manufacturing process is

necessary before taking future improvement actions. Practically, a measurement system does not always produce the exact dimension of the part, but it gives measurements that are deviated from the true value by some error. In any activity involving measurements, some of the observed variability will be due to variability in the product itself, r2 p , whereas the rest will be due to the measurement error or gage variability, r2 g . The variance of the total observed measurements can be expressed as [1]
2 2 r2 Total rp rg

* Corresponding author. E-mail address: abbas.alrefai@ju.edu.jo (A. Al-Refaie). 0263-2241/$ - see front matter 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.measurement.2010.02.016

In many measurement system analysis studies, the gage is usually used to obtain replicate measurements on units by several different operators. Hence, two components of r2 g are frequently generated, including the repeatability and reproducibility. Repeatability, r2 Repeatability , represents the variation due to the gage itself when one operator uses the same gage to measure an identical quality characteristic of the same unit. Whereas, reproducibility, r2 Reproducibility , reects the variation caused by different operators using

A. Al-Refaie, N. Bata / Measurement 43 (2010) 842851

843

the same gage to measure identical quality characteristic of the same unit. That is, the gage variance could be expressed as
2 2 r2 g rReproducibility rRepeatability

2
2 g

The study used to measure the components of r is usually called a Gage Repeatability and Reproducibility (GR&R) study, which aims at determining whether the variability of the measurement system is small relative to the variability of the monitored process. The GR&R studies have been widely conducted to assess the measurement systems in several industrial applications; such as an injected molded plastic part [2], thermal impedance on a power module for an induction motor starter [3], and paddle dissolution test of a pharmaceutical industry [4]. Further, the two-way random effects analysis of variance (ANOVA) model [57] is commonly used to analyze the GR&R study and estimate these variability components. Consequently, this research adopts the GR&R designed experiments followed by ANOVA to estimate variance components. Evaluating a measurement system and manufacturing process capabilities is an important aspect of many quality and process improvement activities. The evaluation of the capability of a measurement is usually achieved by the use of appropriate quality measures, including the precision-to-tolerance ratio (PTR), signal-to-noise (SNR), and discrimination ratio (DR). On the other hand, the process capability is usually assessed using potential and actual process capability indices, Cp and Cpk, respectively. Several studies were conducted to evaluate measurement and process capabilities. Among them, Lin et al. [8] used Taguchi method to reduce the measurement variability, veried the improvement by conducting a GR&R study, and employed a process capability analysis to show the reduction effect on the product variability. Bordignon and Scagliarini [9] considered the problem of measurement error effects on the performance of process capability indices and presented the statistical analysis on the estimation of condence intervals for Cp with data contaminated with measurement errors. Majeske et al. [10] utilized the PTR, Cp, and correlation in repeat measurements for evaluating measurement system and manufacturing processes between manufacturers and suppliers. Larsen [11] analyzed the measurement system in a production environment using GR&R study with condence intervals and various test parameters, such as PTR, SNR, and Cpk. Hsu et al. [12] conducted a sensitivity analysis for the process capability index Cpmk in the presence of measurement error by considering a method for obtaining lower condence bounds and critical values for the true process capability. Li and Al-Refaie [13] conducted two GR&R studies to improve wooden parts quality through enhancing the measurement system capability. The PTR and SNR were employed to assess the adequacy of measurement system. Based on the above introduction, this paper proposes a procedure for assessing both the capabilities of a measurement system and manufacturing process utilizing GR&R designed experiments with four quality measures; including PTR, SNR, DR, and Cp or Cpk. ANOVA will be employed to estimate variance components. The quality

measures will be then used to evaluate the capabilities of a measurement system and manufacturing process. Three real case studies are provided for illustration. The remaining of this paper is organized as follows. Section two introduces the measurement system and a manufacturing process quality measures. Section three introduces the main steps of the proposed approach. Section four provides three illustrative case studies. Section ve summarizes conclusions. 2. Quality measures for capability assessment This section introduces the quality measures usually used for assessing the capabilities of a measurement system and a manufacturing process. 2.1. Assessing measurement system capability The quality measure usually used for assessing the measurement system is the PTR calculated as

PTR

6rg 6rg USLLSL T

where T represents the tolerance, and USL and LSL are the upper and lower specication limits, respectively. A gage is then judged capable if the PTR is less or equal to 0.1, whereas a gage is incapable if the PTR is greater than 0.3 [14]. It can be noted that the PTR for xed value of T depends only on rg . Another quality measure for assessing the capability of a measurement system, which relates the signal, the part variation, to the noise, the measurement error, and considers both rp and rg , is the SNR, which is calculated using the following equation:

SNR

p 2r p

rg

A SNR value of ve or greater indicates that the gage is adequate, whereas a value of less than two indicates inadequate gage capability [15]. An alternative to SNR dened by Wheeler [16] is the DR, which is expressed as

DR

s 2r 2 p 1 2

rg

Mader et al. [17] stated that a DR value of four or greater indicates an adequate measurement system, whereas a value of less than two indicates that the measurement system is inadequate. 2.2. Assessing the capability of a manufacturing process Process capability indices represent a class of quality measures for assessing manufacturing processes, which describe the ability of a process to produce parts that meet a predetermined level of production tolerance. The potential process capability index, Cp, enjoys a broad base of acceptance in industry, and is widely used for checking the capability of production processes. The Cp simply measures the spread of the specications relative to the six-sigma spread in the process, expressed as

844

A. Al-Refaie, N. Bata / Measurement 43 (2010) 842851

Cp

T 6r p

For an existing process, the minimum acceptable value of the Cp index is 1.67 [1]. However, the potential Cp index does not take process centering into account. Another measure for assessing process capability is the actual process capability index, Cpk, which takes both the magnitude of process variance and the process departure from the target into consideration. The Cpk is expressed as

where SSP, SSO, SSPO, and SSRepeatability represent the sum of squares (SS) of part, operator, interaction between part and operator, and random error, respectively. The variance components in Eq. (9) can then be estimated as follows. ^2 The r P is calculated as

^2 r P

MSP MSPO on

12

C pk min

  USL l l LSL ; 3rp 3rp

The MSP is the mean squares for the part, and MSPO is the mean squares for the interaction between part and ^2 operator, respectively. The r O is calculated using

where l represents the process mean. A manufacturing process is said to be capable if the Cpk index is at least 1.33 [18]. 3. The proposed procedure The proposed procedure is divided into the three following stages. 3.1. Stage 1: conducting a GR&R study Assume that GR&R study require p randomly selected parts chosen from the process to be evaluated. Each of the o randomly selected operators measures every part n times using the same measurement device. Let yijk represent the k replicate measurement of part i measured by operator j. A reasonable measurement model for GR&R study is [19]

^2 r O

MSO MSPO pn

13

where MSO is the mean squares for operator. The estimated as

r2 PO is
14

^2 r PO

MSPO MSRepeatability n

The MSRepeatability is the mean squares of the random er^2 ror. Finally, the r Repeatability is calculated as

^2 r Repeatability MSRepeatability

15

If any variance component has a negative value, then the GR&R study sets its value to zero. 3.3. Stage 3: assessing the measurement and process capabilities Once the variance components are estimated by ANOVA, the capabilities of a measurement system and manufacturing process can be easily assessed. As mentioned earlier, the PTR and SNR/DR are the quality measures for assessing the capability of a measurement system. The rg can be written in terms of PTR criterion given by Eq. (3) as

yijk

8 > < i 1; 2; . . . ; p x Pi Oj POij eijk j 1; 2; . . . ; o > : k 1; 2; . . . ; n

where x is the true value of the measurement. The parameters Pi, Oj, (PO)ij, and eijk are all assumed independent random variables for the effects of the parts, operators, the interaction between parts and operators, and random error, and are normally distributed with means zero and 2 2 2 2 variances r2 p , rO , rPO and rE , respectively. The rTotal of any observation is calculated as
2 2 2 2 r2 Total rp rO rPO rE

rg 6 rg

0:1T 6 0:3T > 6

if PTR 6 0:1; if PTR > 0:3;

the gage is capable 16 the gage is incapable

The SNR expressed by Eq. (4) provides a relationship between rg and rp as follows:

2 Typically, the r2 Repeatability is equal to rE , whereas the 2 r is calculated as the sum of rO and r2 PO . Then according to Eq. (2), the r2 g can be estimated as 2 Reproducibility

rp P p rg if SNR P 5; the gage is capable rp < 2rg if SNR < 2; the gage is incapable
2 p

17

2 2 2 r2 g rO rPO rE

10
Plotting inequalities (16) and (17), the (rg , rp ) space is shown in Fig. 1a. If the DR given by Eq. (5) is employed instead of SNR for assessing the measurement capability, the relationship between rg and rp becomes

3.2. Stage 2: conducting ANOVA for GR&R measurement observations The two-way random effects ANOVA model is commonly used to estimate the variance components in the analysis of a GR&R study. Based on Eq. (9), the total sum of squares (SSTotal) in the measurements is divided into the following component parts [17]:

rp P rp

r 15 rg if DR P 4; the gage is capable 2 r 3 < rg if DR < 2; the gage is incapable 2

18

SSTotal SSP SSO SSPO SSRepeatability

11

Inequalities (16) and (18) are plotted in the (rg , space is shown in Fig. 1b.

rp )

A. Al-Refaie, N. Bata / Measurement 43 (2010) 842851

845

Fig. 1. Assessing the measurement system. (a) PTR and SNR criteria. (b) PTR and DR criteria.

From Fig. 1, the following results are noticed, including: 1. The PTR criterion is satised in regions R1, R2, and R3, whereas the SNR (DR) criterion is satised in regions R1, R4, and R7. Among all regions, the measurement system is only deemed capable in region R1, since both criteria are satised. 2. Region R2 satises the PTR criterion, but the SNR (DR) can be improved by reducing the rg . 3. Although region R3 satises the PTR criterion, the rg is still large compared with rp and hence the SNR (DR) criterion is not met. Thus, actions to improve the accuracy of the gage should be taken. 4. Region R4 indicates that the rg should be reduced to satisfy the PTR criterion. Despite this, the rp is large compared to rg , hence the SNR (DR) criterion is satised. Also, a reduction in rg is needed in region R5 in order to satisfy both criteria and improve the gage capability. 5. The gage in region R6 is not sufciently capable to measure the product and hence the gage must be changed in order to improve the measurement system capability. 6. Region R7 indicates that the measurement system is inadequate, as the PTR criterion is not satised. Even though, the SNR (DR) criterion is satised since the rp is large compared to rg . Consequently, it is suggested to reduce both the rg and rp in order to improve the PTR and SNR (DR). 7. The measurement system in regions R8 and R9 deem inadequate since PTR and SNR (DR) criteria are not satised. As a result, the gage must be changed in order to improve the measurement system capability. From the above results, it is noted that the PTR is insufcient quality measure for evaluating measurement system. Moreover, the SNR and DR provide almost similar results. Consequently, each of these two quality measures will be used with PTR to assess a measurement system capability and provide a relationship between gage and parts variances. In order to approve a production process, the Cp index given by Eq. (6) shall be satised. The rp can be expressed in terms of T and Cp as follows:

Alternatively, the given by Eq. (7) as

rp can be expressed in terms of Cpk

rp 6

minfUSL l; l LSLg ; 3C pk

for a capable process 20

In order to evaluate the capabilities of the measurement system and manufacturing process, inequalities (19) and (20) are incorporated in the (rg , rp ) space in Fig. 1. The resulting plots are depicted in Fig. 2. Similarly, using the DR the corresponding (rg , rp ) space is depicted in Fig. 3. From Figs. 2 and 3, the following results are obtained: (a) The PTR and SNR (DR) criteria are satised in region R1, however, the rp is still large that makes the process incapable. Hence, it is suggested to reduce the rp . (b) The gage and the process capabilities are adequate in region R2 since it is the only region that satises the three quality measures. (c) Region R3 satises the PTR criterion, but the SNR (DR) can be improved by reducing rg . However, actions should be taken to improve the gage accuracy in region R4, because the PTR criterion is satised and the SNR (DR) criterion is not. (d) Regions R5 and R10 indicate that both rg and rp should be reduced in order to improve the PTR and the process capability and have the same level of SNR (DR). (e) Region R6 suggests that both rg and rp should be reduced in order to improve the PTR, SNR (DR), and the process capability. However, regions R7 and R8 require a reduction in rg to improve the accuracy of the gage. (f) The gage in region R9 is not sufciently capable to measure the product and hence the gage must be changed in order to improve the measurement system capability. (g) Region R11 and R12 suggest changing the measurement system by conducting a designed experiment in order to identify the most signicant factors affecting the process and the gage capability. (h) Region R13 and R14 indicate that the process is capable but the measurement system should be changed to improve the gage capability.

rp 6

T ; 6C p

for a capable process;

19

846

A. Al-Refaie, N. Bata / Measurement 43 (2010) 842851

Fig. 2. Assessing the measurement system and process capabilities. (a) PTR, SNR and Cp criteria. (b) PTR, SNR and Cpk criteria.

Fig. 3. Assessing the measurement system and process capabilities. (a) PTR, DR and Cp criteria. (b) PTR, DR and Cpk criteria.

The above results are summarized in Table 1 and will be utilized in evaluating the capabilities of a measurement system and process capabilities for the following three case studies. 4. Illustrations Three real case studies, which were investigated previously in literature, are considered to illustrate the proposed procedure for assessing the capabilities of a measurement system and manufacturing process. 4.1. Assessing measurement system and drilling process capabilities Li and Al-Refaie [13] conducted two GR&R studies to assess the capability of the measurement system for drilling process of wooden parts. The drilling process has diameter specications of 26.15 0.75 mm. The diameter T is 1.5 mm. Three operators (o = 3) were randomly selected to measure 10 parts (p = 10), and each part was measured 3 times (n = 3). The collected measurement observations for the drilling machine output are displayed in Table 2.

ANOVA is conducted for the measurement observations in Table 2 then the results are displayed in Table 3. For illustration, the variance components can be estimated as ^2 follows. The r P is calculated using Eq. (12) as

^2 r P

MSP MSPO 0:028858 0:001118 0:003082 33 on

^ p equals 0.0555. The r ^2 and accordingly, the r O is calculated from Eq. (13) as

^2 r O

MSO MSPO 0:003858 0:001118 0:000091 10 3 pn

An estimate of the

r2 PO in Eq. (14) is

^2 r PO

MSPO MSRepeatability 0:001118 0:000668 3 n 0:000150

^2 The r Repeatability , according to Eq. (15), equals

^2 r Repeatability MSRepeatability 0:000668

A. Al-Refaie, N. Bata / Measurement 43 (2010) 842851 Table 1 The regions obtained from relating PTR, SNR (DR) and Cp/Cpk indices. Region R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 Satised measures PTR, SNR(DR) PTR, SNR(DR), Cp (Cpk) PTR, Cp (Cpk) PTR, Cp (Cpk) SNR(DR) None SNR(DR), Cp (Cpk) Cp (Cpk) Cp (Cpk) SNR(DR) None None Cp (Cpk) Cp (Cpk) Unsatised measures Cp (Cpk) None None SNR(DR) Cp (Cpk) Cp (Cpk) None None SNR(DR) PTR, Cp (Cpk) PTR, Cp (Cpk) PTR, SNR(DR), Cp (Cpk) PTR PTR, SNR(DR) Improvement suggestions/decisions The gage is capable but the process capability can be improved by reducing Both the gage and the manufacturing process are capable

847

rp

The manufacturing process is capable, but the SNR (DR) can be improved by reducing rg The manufacturing process is capable, and actions to improve the accuracy of the gage should be taken Both rg and rp should be reduced in order to improve the PTR and Cp (Cpk) Both rg and rp should be reduced in order to improve the PTR, SNR(DR), and the process capability The manufacturing process is capable, and a reduction in rg is suggested to improve the accuracy of the gage The manufacturing process is capable, and a reduction in rg is suggested to improve the accuracy of the gage The manufacturing process is capable, but the gage is not sufciently capable and must be changed Both rg and rp should be reduced in order to improve the PTR and the process capability and have the same level of SNR (DR) The gage must be changed The gage and the manufacturing process are incapable, and the gage must be changed The manufacturing process is capable, but the measurement system should be changed to improve the gage capability The manufacturing process is capable, but the measurement system should be changed to improve the gage capability

Table 2 The GR&R study at initial stage for case study 1. Part number Operator 1 Measurements (mm) M1 1 2 3 4 5 6 7 8 9 10 25.60 25.58 25.61 25.57 25.57 25.72 25.66 25.72 25.51 25.61 M2 25.61 25.62 25.58 25.63 25.56 25.67 25.63 25.66 25.46 25.55 M3 25.57 25.63 25.61 25.62 25.59 25.68 25.59 25.69 25.46 25.57 Operator 2 Measurements (mm) M1 25.56 25.65 25.62 25.57 25.64 25.68 25.61 25.68 25.47 25.62 M2 25.62 25.64 25.59 25.63 25.62 25.67 25.63 25.67 25.44 25.59 M3 25.59 25.63 25.61 25.59 25.67 25.64 25.66 25.73 25.49 25.64 Operator 3 Measurements (mm) M1 25.57 25.67 25.62 25.59 25.68 25.65 25.69 25.72 25.47 25.61 M2 25.59 25.66 25.64 25.62 25.62 25.63 25.64 25.67 25.54 25.57 M3 25.62 25.64 25.67 25.63 25.65 25.67 25.68 25.69 25.49 25.62

Finally, the
2 g 2 O

r2 g can be estimated from Eq. (10) as


2 PO 2 Repeatability

r r r r

0:000091 0:000150 0:000668 0:000909


^ g is then equals to 0.0301. The r In order to check the adequacy of the gage and the process, the T value (=1.5) was substituted into inequality (16), which gives

Finally, assuming that the process is centered, then the potential and actual process capability indices are equal. Therefore, in order to assess the process capability, the T and the Cp index values are substituted into inequality (19), which results in

rp 6

T 1:50 0:149; 6C p 61:67

for a capable process

rg 6 rg >

0:1T 0:11:50 0:025; 6 6 0:3T 0:31:50 0:075; 6 6 5

for a capable gage

for an incapable gage

Using the SNR criterion, inequality (17) gives

rp P p rg 3:535rg ; for a capable gage


2

rp < 2rg 1:414rg ; for an incapable gage

The above inequalities are plotted in the (rg , rp ) space ^ p of 0.0301 and ^ g and r as shown in Fig. 4. Inserting the r 0.0555, respectively, in Fig. 4, the corresponding region is R8. Table 1 suggests only a reduction in rg to improve the measurement system, because the process seems highly capable. From ANOVA results, it is noticed that the part-to-part variation has the largest contribution (=77%) to the total variation. Moreover, the repeatability and reproducibility variances contribute about 73% and 27%, respectively, of the gage variation. Thus, in order to improve the measurement system capability, the gage accuracy should be improved or replaced to reduce

848

A. Al-Refaie, N. Bata / Measurement 43 (2010) 842851 Table 4 The ANOVA for the GR&R study at improvement stage for case study 1. Source of variation ANOVA results Part Operator Partoperator interaction Repeatability Total variation DF 9 2 18 60 89 SS 1.05150 0.00398 0.01386 0.02727 1.09661 MS 0.116833 0.001990 0.000770 0.000454 P-value* 0.000 0.103 0.066

Table 3 The ANOVA for the GR&R study at initial stage for case study 1. Source of variation ANOVA results Part Operator Partoperator interaction Repeatability Total variation DF 9 2 18 60 89 SS 0.259721 0.007716 0.020129 0.040067 0.327633 MS 0.028858 0.003858 0.001118 0.000668 P-value* 0.000 0.054 0.070

Variance component Gage R&R study results Total gage R&R 0.000909 Repeatability 0.000668 Reproducibility 0.000241 Operator 0.000091 Partoperator 0.000150 interaction Part-to-part 0.003082 Total variation 0.003991
*

Standard deviation 0.0301 0.0258 0.0155 0.0095 0.0122 0.0555 0.0632


*

Variance component Gage R&R study results Total gage R&R 0.000600 Repeatability 0.000454 Reproducibility 0.000146 Operator 0.000041 Partoperator 0.000105 interaction Part-to-part 0.012896 Total variation 0.013496 Assume 5% level of signicance.

Standard deviation 0.0245 0.0213 0.0121 0.0064 0.0102 0.1136 0.1162

Assume 5 % level of signicance.

Fig. 4. A plot for PTR, SNR and Cp index for case study 1.

repeatability variance. Further reduction in reproducibility variance can be achieved by training operators. A second GR&R study was conducted by Li and Al-Refaie [13] after conducting improvement actions. The corresponding ANOVA results are summarized in Table 4. ^ p are found equal to 0.0245 ^ g and r From this table, the r and 0.1136, respectively. In Fig. 4, this point corresponds to region R2. Utilizing the guidelines shown in Table 1, this region indicates the adequacy of the measurement system and the approval of drilling process for production. Similar results are obtained using the DR instead of SNR for this assessment. These results are consistent with those obtained by Li and Al-Refaie [13]. 4.2. Assessing measurement system and injected molded plastic process capabilities This case study was considered by Tsai [2] to assess a measurement system of an injected molded plastic part.

Hamada and Weerahandi [6] considered this GR&R study and assessed the measurement system by generalized inference methodology. A GR&R study was conducted for this assessment, and the data set is displayed in Table 5. These data are the deviations of one dimension of the plastic parts from a nominal value of 685 mm. Two operators were randomly selected, from those who normally use the measuring machine, to measure 10 parts. Each part was loaded twice into the xture. The USL and LSL are 685.5 and 684.5 mm, respectively, and hence, the T equals 1 mm. Table 6 displays the ANOVA results, where ^ p are calculated as 0.0155 and 0.0160, ^ g and r the r respectively. Substituting the known terms into inequalities (16) and (17) give, respectively,

rg 6 0:019; for a capable gage rg > 0:058; for an incapable gage

A. Al-Refaie, N. Bata / Measurement 43 (2010) 842851 Table 5 Measurements for GR&R study for case study 2. Part Number Operator 1 Measurements (mm) M1 1 2 3 4 5 6 7 8 9 10 0.289 0.311 0.295 0.301 0.265 0.298 0.273 0.276 0.328 0.293 M2 0.273 0.327 0.318 0.303 0.288 0.304 0.293 0.301 0.341 0.282 Operator 2 Measurements (mm) M1 0.324 0.340 0.335 0.304 0.289 0.305 0.287 0.275 0.316 0.300 M2 0.309 0.333 0.326 0.333 0.279 0.299 0.250 0.305 0.314 0.297

849

and assuming that the process is centered, inequality (19) gives

rp 6 0:099; for a capable process


The above inequalities are plotted in (rg , rp ) space as shown in Fig. 5. It is found that the (0.0155, 0.0160) point corresponds to region R4. From Table 1, this region indicates that the process can be approved for production. However, only the SNR criterion is not satised, because the rp is small relative to rg . The total variation consists of 51.5% due to part-to-part variation and 48.5% due to measurement error. The repeatability contributes about 68% of the gage variation. Accordingly, actions are needed to improve the gage accuracy, mainly in the variability due to repeatability. Similarly, using the DR instead of SNR the estimated point corresponds to region R4 in Table 1. This result is similar to that obtained by Hamada and Weerahandi [6]. 4.3. Assessing measurement system and optical comparator process capabilities Lin et al. [8] studied the gear diameter measurements using an optical comparator, and employed the Taguchis robust design method to study the effects of several measurement factors simultaneously and yield to the optimum measurement practice for reducing the measurement variability. Two GR&R studies were conducted on the initial and optimized measurement practices to verify the predicted improvement. The gear diameter has specications of 3.03 0.03 mm. The diameter T is 0.06 mm. The true gear diameter (l) is 3.015 mm. Ten gear parts were measured twice by each of two operators. The initial GR&R study observations are displayed in Table 7. ANOVA is con^ g and ducted and the results are displayed in Table 8. The r ^ p are found equal to 0.0037 and 0.0104, respectively. r Substituting the known terms into inequalities (16) and (17) give, respectively,

Table 6 The ANOVA for the GR&R study of case study 2. Source of variation ANOVA results Part Operator Partoperator interaction Repeatability Total variation DF 9 1 9 20 39 SS 0.011749 0.000648 0.002538 0.003270 0.018205 MS 0.001305 0.000648 0.000282 0.000164 P-value* 0.016 0.164 0.148

Variance component Gage R&R study results Total gage R&R 0.000241 Repeatability 0.000164 Reproducibility 0.000077 Operator 0.000018 Partoperator 0.000059 interaction Part-to-part 0.000256 Total variation 0.000497
*

Standard deviation 0.0155 0.0128 0.0088 0.0042 0.0077 0.0160 0.0223

Assume 5 % level of signicance.

rp P 3:535; for a capable gage rg rp < 1:414; for an incapable gage rg

rg 6 0:001; for a capable gage rg > 0:003; for an incapable gage

Fig. 5. A plot for PTR, SNR and Cp index for case study 2.

850 Table 7 The GR&R study at initial stage for case study 3. Part number Operator 1 Measurements (mm) M1 1 2 3 4 5 6 7 8 9 10 3.045 3.037 3.021 3.017 3.048 3.046 3.039 3.033 3.048 3.041 M2 3.041 3.038 3.024 3.012 3.049 3.045 3.034 3.035 3.041 3.049

A. Al-Refaie, N. Bata / Measurement 43 (2010) 842851 Table 9 The ANOVA for the GR&R study at improvement stage for case study 3. Source of variation ANOVA results Part Operator Partoperator interaction Repeatability Total variation DF 9 1 9 20 39 SS 0.004118 0.000003 0.000035 0.000058 0.004214 MS 0.000458 0.000003 0.000004 0.000003 P-value* 0.000 0.379 0.275

Operator 1 Measurements (mm) M1 3.048 3.031 3.023 3.011 3.042 3.042 3.033 3.038 3.034 3.046 M2 3.046 3.038 3.022 3.016 3.048 3.043 3.035 3.037 3.032 3.045

Variance component Gage R&R study results Total gage R&R 0.0000034 Repeatability 0.0000029 Reproducibility 0.0000005 Operator 0.0000000 Partoperator 0.0000005 interaction Part-to-part 0.0001134 Total variation 0.0001168
*

Standard deviation 0.0018 0.0017 0.0007 0.0000 0.0007 0.0106 0.0108

Table 8 The ANOVA for the GR&R study at initial stage for case study 3. Source of variation ANOVA results Part Operator Partoperator interaction Repeatability Total variation DF 9 1 9 20 39 SS 0.004078 0.000027 0.000171 0.000161 0.004437 MS 0.000453 0.000027 0.000019 0.000008 P-value* 0.000 0.262 0.052

Assume 5% level of signicance.

Finally, using the Cpk index to assess the process capability, inequality (20) gives

Variance component Gage R&R study results Total gage R&R 0.0000139 Repeatability 0.0000080 Reproducibility 0.0000059 Operator 0.0000004 Partoperator 0.0000055 interaction Part-to-part 0.0001085 Total variation 0.0001224
*

Standard deviation 0.0037 0.0028 0.0024 0.0006 0.0023 0.0104 0.0111

rp 6 0:0038; for a capable process


Fig. 6 plots the corresponding (rg , rp ) space, where this point corresponds to region R11. According to Table 1, this region indicates that both the measurement system and the process capability are inadequate and incapable, respectively. From ANOVA results, it is noticed that about 89% of the total variation is due to part-to-part variation and 11% due to measurement error. Moreover, the repeatability variance contributes about 58% of the total gage variation. A second GR&R study was performed after the process was optimized. ANOVA results are tabulated in Table 9, ^ p are found 0.0018 and ^ g and r where the calculated r 0.0106, respectively. According to Fig. 6, the corresponding region is R5, which indicates that the measurement system had been improved but the process capability is still

Assume 5% level of signicance.

rp P 3:535; for a capable gage rg rp < 1:414; for an incapable gage rg

Fig. 6. A plot for PTR, SNR and Cpk index for case study 3.

A. Al-Refaie, N. Bata / Measurement 43 (2010) 842851

851

inadequate. That is, improving measurement process has reduced the product variability, which is reected in higher values of Cpk, which is less than 1.33. Consequently, actions need to be taken to improve the capability of the manufacturing process. Similar results are obtained using DR instead of SNR. These results are also consistent to the results obtained by Lin et al. [8]. 4.4. Research results The proposed procedure successfully assessed the measurement system and process capability in the above three case studies. The following results are obtained. First, the proposed procedure incurs minimal computational efforts and statistical knowledge, and provides consistent assessment with those obtained in previous studies with complicated methods. This may attract practitioner to utilize this procedure in assessing the measurement system and manufacturing process capabilities in a wide range of manufacturing applications. Second, the PTR is insufcient criterion for evaluating a measurement system capability. Hence, the SNR or DR criterion is used to asses the measurement system capability and relate the gage error with process variation. Third, the SNR and DR criteria provide similar results when combined with PTR and process capability indices. Consequently, either criterion can be used in assessing a measurement and process capabilities. Fourth, the GR&R designed experiments provide helpful information regarding the signicant effects and the needed actions toward improvement; such as actions needed for improving gage accuracy, training operators, or reducing process variations given in Table 1. 5. Conclusions This research proposes a procedure for evaluating the capabilities of a measurement system and a manufacturing process utilizing GR&R designed experiments with four performance measures, including PTR, SNR, DR, and process capability indices. Three case studies were provided for illustration, in all of which the proposed procedure provides efcient assessment of the measurement and process capabilities with minimal computational and statistical knowledge. Hence, this procedure should provide helpful guidelines to practitioners in the assessment of a measurement system and manufacturing process capabilities in a

wide range of manufacturing applications and resolves the usual conicts arise about the reasons of product/process variability. References
[1] D.C. Montgomery, Statistical Quality Control: A Modern Introduction, sixth ed., Wiley, New York, 2009. [2] P.F. Tsai, Variable gauge repeatability and reproducibility study using the analysis of variance method, Quality Engineering 1 (1988) 107115. [3] R.F. Houf, D.B. Berman, Statistical analysis of power module thermal test equipment performance, IEEE Transactions on Components, Hybrids and Manufacturing Technology 11 (4) (1988) 516520. [4] Z. Gao, T. Moore, A.P. Smith, W. Doub, B. Westenberger, L. Buhse, Gauge repeatability and reproducibility for accessing variability during dissolution testing: a technical note, AAPS PharmSciTech 8 (4) (2007) (article 82). [5] S.B. Vardeman, E.S. VanValkenburg, Two-way random-effects analyses and gauge R&R studies, Technometrics 41 (3) (1999) 202 211. [6] M. Hamada, S. Weerahandi, Measurement system assessment via generalized inference, Journal of Quality Technology 32 (3) (2000) 241253. [7] S. Senol, Measurement system analysis using designed experiments with minimum ab Risks and n, Measurement 36 (2004) 131141. [8] C.Y. Lin, C.L. Hong, J.Y. Lai, Improvement of a dimensional measurement process using Taguchi robust designs, Quality Engineering 9 (4) (1997) 561573. [9] S. Bordignon, M. Scagliarini, Statistical analysis of process capability with measurement errors: the case of Cp, Statistical Methods & Applications 10 (2001) 273285. [10] K.D. Majeske, R.W. Andrews, Evaluating measurement systems and manufacturing processes using three quality measures, Quality Engineering 15 (2) (2002) 243251. [11] G.A. Larsen, Measurement system analysis in a production environment with multiple test parameters, Quality Engineering 16 (2) (2003) : 297306. [12] B.M. Hsu, M.H. Shu, W.L. Pearn, Measuring process capability based on Cpmk with gauge measurement errors, Quality and Reliability Engineering International; 23 (2007) 597614. [13] M.-H. Li, A. Al-Refaie, Improving wooden parts quality by adopting DMAIC procedure, Quality and Reliability Engineering International 24 (2008) 351360. [14] AIAG: Automotive Industry Action Group (2002). Measurement Systems Analysis. third ed., Detroit, MI. [15] R.K. Burdick, C.M. Borror, D.C. Montgomery, A review of methods for measurement systems capability analysis, Journal of Quality Technology 35 (4) (2003) 342354. [16] D.J. Wheeler, Problems with gauge R&R studies, ASQC Quality Congress Transactions, Nashville (1992) 179185. [17] D.P. Mader, J. Prins, R.E. Lampe, The economic impact of measurement error, Quality Engineering 11 (4) (1999) 563574. [18] F. Spring, S. Cheng, A. Yeung, B. Leung, Process capability indices a review, 19922000 discussion, Journal of Quality Technology 34 (1) (2002) 2327. [19] R.K. Burdick, A.E. Allen, G.A. Larson, Comparing variability of two measurement processes using R&R studies, Journal of Quality Technology 34 (1) (2002) 97105.