zl Ely Designation: E2782 11 Standard Guide for Measurement Systems Analysis (MSA)* ‘This standart i ove wndor the ied devignation L278, the number immo felling the deignation indica the yar of vigil len 1, Seope LL This guide presents terminology, concepls, and selected methods and fornnulas useful for measnrement systems analy sis (MSA). Measuremen described as a hadly of theory and methodology that applies to the non-desimctive measurement of the physical properties of ‘mannfactnred objects 12 Units—The system of units for this guide is not speci- fed. Dimensional quantities in the guide are presented only as illustrations of calenlation methods and are not binding on products or test methods treated. 3 This standard does not purport to address all of the safety concerns, if any. associated with its use. Tt is the responsibility of the user of this standard to establish appro- priate safety and health practices and determine the applica- bility of regulatory limitations prior to use. systems analysis may be hinadlly 2. Referenced Documents 2.1 ANIM Standards? E177 Practice tor Use of the ‘Terms Precision and Bias in ‘ASIM ‘Test Methods £456 Terminology Relating (0 Quality and Statistics 1586 Prachce lor Calculating and Using Basic Statistics BIS8/ Prachce for Use of Control Charts in Statistical Process Control 3. Terminology 3.1 Defintions—Unless otherwise noted, terms relating 10 quality and statistics are defined in Terminology E456, 3.1.1 accepted reference value, n—a value that serves as an agreed-upon reference for comparison, and which is derived as: (1) a theoretical or established value, based on scientific principles, (2) an assigned of certified value, based on experi ‘memtal work of some national or international organization, or (3) a consensus of certified value, based on collaborative This gue our te ation oC ASTAL Cm 11 om Cui smd Stns and the diet reeponibldy of Subcommiles L110 on Ie Method valuation and Quality Con Curent eon approved Nox. 13, 2011, Publishal Febuary 2012 DOE Ter rfoonced ASTM stander, vst the ASTM website, wirmastm.org oF age STAI Caster Serie at servoeasim ang. For nm! Rank STM SC le mal te sans Lael Sui Pon ‘as of revise the yor ola rvson. A nimbar in parece nis hey st anova. ‘pong speon a} meson ma soto ahungs ence tel son experimental work under dhe auspices of a sciemtifie or engineering group. £177 3.1.2 calibrarion, n—process of establishing 4 relationship between a mmusuzement device andl x kugwa standard value(s). 3.1.3 guye, n—deviee used as part of die wexsumeuent process (0 ubtui u nueusurement result 3.14 measuremem process, n—pascess used. tw ussiga a uuber to a property Of un object ur ether physical entity VAT Piserrsion—The (em “uegstnement system!” is Sometinnss used iv pice of measurement pracess, (See 31.6) 315 mensuerauent result —auuiber ussivie la 8 properly casune 1 iy used in of 1 object or other ply sical eulily being u SAAT Pisenysinn—The woud “umessinen tesa the same sens 316 meusurement sytem, n—the collec nl metlnds, noma eff, envinonnes associated devices, aul thee objects tha ate the: pase of proc 317 measurement sustems mnulsis (MSA). n—any of a tuber af specialized methods useful for sty ing, a measire= meat system and its properties 32 Definitions of Terms Specific to This Stanatar. 32.1 appmiser. n—the person who usesa gage or measnre= ‘ment system. 3.22 discrimination ratio, n—satistical ratio calenlated from the statistics from a gage R&R sindy that measures the number of 97% confidence intervals, constnicted fiom gage R&R variation, that fit within six standard deviations of tme object variation 3.2.3 distinct product categories, n—alterate meaning of the discrimination ratio. 3.24 gage consistency, n—constancy of repeatability vati= ance over a period of time. 3.24.1 Discussion—Consisteney. means that the variation vwithin measurements of the same object (or group of objects) under the same conditions by the same appraiser behaves in a state of statistical control as judged, for example, using a control chart. Sec Practice £2587. 3.23 gage performance curve, n—eurve that shows the probability of gage acceptance of an object given its real value or the probabihty that an object's teal measure mects. a requirement given the measurement of the object P hada, softwarehp cera2—n1 3.26 gage R&R, and reproducibility 3.2.7 gage resolution, -—degree to which a gage can discriminate between differing objects. 3.2.71 Discussion—The smallest difference between two objects that a page is capable of detecting is referred t0 as its finite resolution property. For example, a linear scale graduated {in tems of an inch is not capable of discriminating between objects that differ by less than 0.1 in, (0.25 om), 3.28 gage stability, n—absence of a change, dri, or erratic behavior in bias over a period of time. 3.28.1 Diseusston—Stability means that repeated measure- ‘ments of the same object (or average of a set of objects) under the same conditions by the same appraiser behave in a state of statistical contol as judged for example by using a control chant technique. See Practice E2587, 3.2.9 linearin, n—dilference or change in bias Uroughout he expected operating range of a gage or measurement syste. 3.2.10 measurement errar, n—error incurred in the process of measurement, 3.2.10.1 Diseussion—As used int this guide, meusurement ermor includes one or both of R&R types of error 3.2.11 repeautitity conciiions, n—in a gage R&R. study, conditions in which independewt wewsuncusents are obinned ot identical objects, or a group of ubjects, by dhe sume operator using the saume mucasuremenl system within short intervals of ie. 39.111 Disemssinn—As used in this yuidle, epewabiiey is ftw wefeene to 38 24 Variation ox BV 32.1 seprantueihitty comtitinns, a—is n gnge RAR siuly, ccaditions in whicte iepenlent (estes ee aban wih {Ue Sane: wethad, idk test items by diferent aperatons FID Diseussiom—As use in this poi, vepwontiility is offen sefeuread th as appraiser vavialion ov AV This tenn fly sea in Dnosket seus nt activ: B77 combined effect of page repeatability 4. Significance and Use 1.1 Many types of measurements are made routinely in research oreanizations, business and industry, and government Typically, data are generated from effort or as observational studies. From such d ‘management decisions are made that may have wide reaching social, economic, and political impact ‘and decision ‘making go hand in hand and that is why the quality of any ‘measurement is important—for data originate from a measure ‘ment process. This guide presents selected concepts and ‘methods useful for describing and understanding the measure ‘ment process. This guide is not intended to be a comprehensive survey of this topic. 4.2 Any measurement result will be said to originate from a ‘measurement process or system, The measurement process will consist of a number of input variables and general conditions that affect the final value of the measurement. The process variables, hardware and software and their properties, and the ‘human effort required 10 obtain a measurement constitute the ‘measurement process. A measurement process will have sev eral properties that characterize the effect of the several variables and general conditions on the measurement results. It isthe propenies of the measurememt process that are of primary interest in any such study. The term “measurement systems analysis” or MSA study ig used to describe the several methods used to characterize the measurement process, Now 1—Sauple statistics discussed in ths guide ae ws desasibed in Practice #2986; conteolchast methodologtas ae as deeenbed mn Practice 5. Characteristics of a Measurement System (Process) 5.1 Measurement has been defined as “the assignment of ‘numbers 10 material objects to represent the relations existing among them with respect to paricular proper, The number assigned to some panicular propery serves to represent the relative amount of this property ascociaed with the object concemed.” (1)° ‘5.2 A measurement system may’be described asa collection of hardware, software, procedures and methods, human effor, environmental conditions, associated devices, and the objects that are measured forthe purpose of producing a measurement In the practical working of the measurement system, these factors combine to cause variation among measurements of the same object that would not be present if the system were perfect, A measurement system can have varying degrees of cach of these factors, and in some cases, one or more factors ‘may’ be the dominant conmiburor to this variadon, 52.1 A measurement system is like a manufactring pro- cess for which the product is a supply of numbers called measurement results, The mensurement system uses input factors and a sequence of steps to produce a result. The inputs are just varying degrees of the several factors described in 3.2 including the objects being measured. The scquence of process steps are that which would be described in a method or procedure for producing the measurement. Taken as a whole, the various factors and the process steps work collectively 10 fon the measurement systemprocess 53 An importa consideration in analyzing any measure~ ‘ment process i is imeraction with time, This gives tse 10 the propertcs of stability and consisteny. A system that is suble find consistent is one that is predicwble, within limits, over a period of time. Such a system lus properties that do not deteriorate with time (at least within some set time period) and is said to be ina stu of suaistical como, Statistical contro, stabil’ and consistency, and predictably have the sume aig im this sense, Measurement systin instability and inconsisteney will cause further added overall variation over a period of tine ‘$3.1 In general, insabiliy is a common problem in meae surement systems. Mechanical and electrical components ‘wear or degrade with time, han effort may exhibit increas ing faigue with time, software and procedures may change ‘vith time, environmental variables will vary with time, and so fon. This, meusuremen system stbilty is of primary cone com in any ongoing measurement eff. ‘4 There are several basic propenies of measurement systems that are widely recognized among practitioners, These > Thelin mumbor ie paremhees fer oe Hf teens athe end af thw standardlp ceme—a1 are repeatability, reproducibility, linearity, bias, stability, con- sistency, and resolution. In studying one or more of these properties, te final result of any such study is some assessment of the capability of the measurement system with respect to the property under investigation, Capability may’ be cast in several ‘wavs, and this may also be application dependent. One of the primary objectives in any MSA effort is to assess variation attributable to the various factors ofthe system. All of the basic, propenies assess variation in some form. ‘S41 Repeatability is the variation that results when a single object is repeatedly measured in the same way, by the same appraiser, under the same conditions (sce Fig. 1). The tem precision” also denotes the same concept, but “repeatability” {s found more often in measurememt applications. The term conditions” is sometimes combined with repeatability 10 denote “repeatability conditions” (sze Terminology E46) ediae precision is also used (lor see Price E177). The user of a mwusurement slull decide wha constitutes "repeatability conditions” mn conditions” for the given appl repeatability condiions for MSA will be as syster or “inwerinediae pre tion, Typical described above $42 Reproducibilily is defined ay the vuriaion sumoug wenage values us devermined by several appraisers whem wusuring Ue Sune group of objects using ideulical meusune wat systems under de sauue conditions (see Fig, 2), Lin a bbrouder seuse. Uhis any be taken as variation in average values cal or Selecta at wgualom Fant one population, 30 cased using, Several syste 54.2.1 Reproducibility tof samples, cider ide Taungyene iy include different equipment urement conditions, This lms ingerpeetations Teas willy canalitivnss” anal shall Ie efi val sal system (In Practice ny vation ) eseuce: between ova reference: val fin an ohjest, often called nverage value af a sample « under a fixed set of conditions 544 Linearity is the change in bias over the operational range of the measusement sysiem. Ifthe bias is changing as a Function of the object heing measued, we wonkd say that the system is not linear Tineasity can also he intespreted to mean that an instrument response is linearly related to the character- istic heing measured, coped easier,” nul the P the alijecsts) True Value 429 Repeatability HIG. 1 Repeatability and Bias Concepts Reproducibility ‘Measurements by Operator | / i FIG. 2 Repro ity Concept 5.45 Stability is variation in bias with time, usually a drift or trend, or erratic behavior. 3.4.6 Consistency is the change in repeatability with time, A system is consistent with time when the standard deviation of the repeatability error remains constant. When a measurement system is stable and consistent, we say that it is a state of statistical control. 47 ‘he resolution of a measurement system has to do with its ability to discriminate between different objects. A system with high resolution is one that is sensitive to smal changes from object to object. Inadequate resolution may result 1n identical measurements when the same object is measured several tines under identical conditions, In this scenano, the measurement device 1s not capable of prcking up variation as a result of repeatability (under the conditions detined). Poor resolution may also result in identical measurements when dilfering objects are measured. In this scenano, the objects themselves are 190 close in true magnitude for the system to distinguish among, 5.4./.1 Resolution plays an important role m measurement 1n general, We can imagine the output of a process that 15 1m statistical control and follows a normal distribution with mean, sk and standard deviation. c. Based on the normal distnbution. the natural «pread of the process is bi. Suppose we measure objects from this process with a perfect gage except for its finite resolution property. Suppose further that the eae we are using is “eradhated” as some fraction, I/k, of the 6 natural process spread (integer 4). For example, if f ~ 1, then the ‘natural process tolerance would span four araduations on the sage; if f= 6, then the natural process spread would span six fAraduations on the gage. It is clear that, as & increases, we ‘would have an increasingly better resolution and would be more likely 10 distinguish between distinct objects, however close their magnitudes: at the opposite extreme. for small fewer and fewer distinct objects from the process would be distinguishable, In the limit, for lange &, every object from this process, woul! be distinguishable, 5.4.7.2 In using this perfect gage, the finite resolution property plays a role in repeatability. For very large k, the resulting standard deviation of many objects from the process would be nearly the magnitude of ihe true object standard deviation, o. As & diminishes, the standard deviation of the ‘measurements would increase asa result ofthe finite resolution propery. Fig. 3 illustrates this concept for a process cemtered at and having o = 1 for k= 4. 3.4.7.3 The illustration from Fig, 3 is a system capable of discriminating objects into groups no smaller than 1.So inhp cera2—n1 ions FH within the Natural fer Process Spread” ‘width so that a frequency distnibution of measured objects from this system will generally haye four bins, ‘This means four distinct product values can be detected. Using Fig. 3 and the theoretical probabilities from the normal distnbution, it 15 possible (0 calculate the vanance of the measured values tor various values of &. In this case, the variance of the measured ately 1.119 or 11.9% lamer than the true ince, ‘The standard deviation is, therefore, 1.098 or 3.8% larger SATA This illustrates the important role that resolution plays in measurement in general and an MSA study in particular. There is a subtle interaction between the degree of resolution and more general repeatability and other measure ‘ment effects. In extreme cases of poor resolution, an MSA study may not be able to pick up a repeatability effect (all objects measured yield the same value), For an ideal system, for varving derees of finite resolution as described in 5.4.7, there will be a component of variance as a result of resolution alone. For positive integer value, &, when the smallest mea surement unit for a device is With of the Ga true natural process range, the standard deviation as a result of the resolution effect may be determined theoretically (assuming a normal distribution). Table 1 shows the effect for selected values off. 54.7.5 A common rule of thumb is for a measurement device to have a resolution no greater than 0.6c, where o is the tue natural process standard deviation. This Would give us TTARLF 1 Retsavior of the Mecsneren! Varianes: an! Starr Deviation for Selected Finite Rezolution Property, k, True Prox Variance is 1 7 Tear ro2okiion ‘ov dao fo veriance component resalion i tieo7 o1807 30102 5 ‘100000 0.00000 020204 6 sre: osrai 024002 ° 1.03540 oo3619 osae30 2 4.00598 0.00599 oraz 10 graduation divisions within the true Ge natural process: mits, In that particular case, the resulting variance of all ‘measurements would have increased by approximately 1.9 % (Table 1, &= 10), 5.5 MSA isa broad class of activities that studies the several properties of measurement svstems, either individually, or some relevant subset of propenies taken collectively. Much of this activity uses well known methods of classical statistics, ‘most notably experimental design techniques, In classical statises, the term variance 1s used to denote variation ina set of numbers. It is the square of the standard deviation. One of the primary goals in conducting an MSA study is to assess the several variance components that may be at play. Each factor ‘will have its own variance component that contributes to the overall variation. Componems of variance for independent variables are additive, For example, suppose y is the result of st measurement in which three independent Gictors are at phy. Suppose dha the three independent factors are, ¥,, and x. A simple model for the linear sunn of the three couponents is y xy" Xp Xy The Variance of the overall sum, y, given the variances of ue components is a-oj+ol+ot w 5.5.1 We say that each variance on the riaht is a component of the overall variance on the let. This model is theoretical; in practice, we do not know the tme variances and have’ 10 estimate their values from data, ‘3.5.2 Statistical methods allow one 10 estimate the several variance components in MS. Sometimes the analyst may only bbe imerested in one of the components, for example, repeat ability. In other cases, it may’ be two or more components that may be of interest. Depending on how the MSA study’ is designed, the variance componems may be estimable free and clear of each other or combined as a composite measure. Several widely used basic models and associated statistical (eehmigues are discussed in Section 6, 6. Basic MSA Methods 6.1 Simple Repeerubitity—Simple repeatability may be evaluated using al least (wo measurements of each of seven dbjects by a single appraiser ules identical canaliigns, Thee Sinuplest such ony uct ah cisinel abjoets say n, and (we Ly use as of ects object, Tet yi aa yg be the (wo mmersurements uf abject i Rach umssur Contauningted will repeatability ere compare wel may Tae writen bya rey a 6.1.1 In this mode, the y, values are the observed measure~ rmeants of objec, i, measurement, the x, values are considered. the “true” or reference values of the objects being measured; and ¢, 18 the repeatability error associated with object, f, and measurement. j. The difference d, between twe measurements of the same oiyeet may be wntten as: 4 y4 Beane & “® 6.1.2 If the error terms can be considered normally disirb- wed, then the paired differences, the cs, will possess a normal distribution. Generally, the repeatability error term, e, is, assumed to have a mean of © and some unknown variance 6°,