Accuracy and Reproducibility of Temperature Measurement by Thermocouples

Transcription

1 THE AMERICAN SOCIETY OF MECHANICAL ENGINEERS 345 E. 47 St.. New York, N.Y GT-44 ry The Society shall not be responsible for statements or opinions advanced in papers or in dis- (; cussion at meetings of the Society or of its Divisions or Sections, or printed in its publications. M Discussion is printed only if the paper is published in an ASME Journal. Papers are available ]^C from ASME for fifteen months after the meeting. Printed in USA. Copyright 1989 by ASME Accuracy and Reproducibility of Temperature Measurement by Thermocouples T. P. WANG Manager of Research Thermometry and QA Wire and Cable Division Thermo Electric Company, Inc. Saddle Brook, New Jersey ABSTRACT The accuracy of base metal thermocouples in initial condition and after service is reviewed in conjunction with the accuracy and reproducibility of measurement and factors affecting calibration. Lot calibration data on insulated thermocouple wire and metal sheathed thermocouples are examined on a statistical basis. Some stability test data on K, JET and E thermocouples in air atmosphere are presented. Calibration data by standard methods are compared with those obtained by data acquisition and freeze point measurements. Roundrobin calibration of J and K thermocouples is included for the determination of the uncertainty of calibration by J and K thermocouples. By integrating the information on initial calibration and stability of the thermocouple with the accuracy of measurement, one is in a position to select the best system for close temperature control throughout a process. INTRODUCTION An analogy can be drawn between time keeping by watches and temperature mesurement by thermocouples. In either case, the accuracy requirement is different for different applications. Watches which meet our daily needs would not be accurate enough for Olympic judges who must check the speed of runners or swimmers to the nearest one hundreth of a second. For temperature measurement in heat treating applications, a standard type K thermocouple would serve heat treaters very well in most occasions. However, this standard type K thermocouple is not satisfactory for a 1, C thermal treatment which requires the temperature to be within a couple of degrees throughout the process. Whether it is in on time keeping or temperature measurement, the watch or the thermocouple should maintain its accuracy throughout the period of use. In order that the above statements on accuracy be meaniful, the accuracy and reproducibility of the measuring system itself must be known. It is the objective of this paper to provide and examine the answers to the following questions: (1) How accurate are thermocouples when they are brand new? (2) How accurate are thermocouples during and after service? (3) What is the accuracy and reproducibility of the measurement system? In order to answer the above questions, it is necessary to treat a large number of thermcouple calibration data on a statistical basis. The statistical terms, X and O (sigma), which we use throughout this paper, are defined as follows: X = Average = F.X./n Eq. (1) i where n is number of measurement of X. 1 2 = Standard Deviation = kxi - X) Eq.(2) - 1 where X.- X is the deviation of an i individual measurement from the average Prior to presentation of test results, a discussion on the types of thermocouples and definitions on calibration and reproducibility are in order. TYPES OF Commonly used thermocouples have been standardized and letter designated. The standardized thermocouples can be subdivided into two groups: base metal thermocouples (types J, K, T, E and N) Presented at the Gas Turbine and Aeroengine Congress and Exposition-June 4-8, 1989-Toronto, Ontario, Canada Downloaded From: on 12/31/218 Terms of Use:

2 and precious metal thermocouples (types R, S and B). The standardized thermocouples referred in this paper are described as follows. The type J thermocouple, iron versus constantan (45 Ni/55Cu based alloy), is used extensively in the C to 76 C range in both oxidizing and reducing atmospheres. The type K thermocouple (Ni/1Cr vs Ni/5 (Si, Al, Mn), for example, Chromel-Alumel*, can be used up to 1,26 C in oxidizing atmosphere. The type T thermocouple, copper-constantan (also a 45 Ni/55Cu alloy but with slightly different composition and calibration than type J constantan) can be used up to 37 C in both oxidizing and reducing atmospheres. The type T thermocouple is used extensively for cryogenic application. The type E thermocouple consists of the type K positive thermoelement versus the type T negative thermoelement, is used primarily for power generation because of its large thermoelectric power. The type E thermocouple can be used up to 87 C in oxidizing atmosphere. where n TK couple is the initial deviation of the couple in degree temperature and L E/oT is the thermoelectric power of the K couple Initial calibration tolerance(2) is defined as the allowable deviation of the thermocouple (or extension wire) in its initial condition as supplied by the wire manufacturers from the EMF - temperature tables. The initial calibration tolerance of standardized thermocouples are published in several technical bulletins (3) (4). The initial calibration tolerances of the standard grade and the special (premium) grade type K thermocouple are shown in the bottom graph of Fig. 1. The tolerance limits for the standard grade are.2 C (+4 F) or + 3/4% of temperature. The tolerance limits for the special (premium) grade are.1 C ( F) or +.4% of of temperature. Thermocouple AD meets the standard grade limits. Thermocouple BC meets the special (premium) limits because of the better matching of the positive and the negative thermoelements. INITIAL CALIBRATION AND INITIAL CALIBRATION TOLERANCE The initial calibration of a thermocouple is a measure of its accuracy in initial condition. The initial calibration of an individual thermocouple wire or a thermocouple is expressed as its deviations in millivolts ( n E) or degree temperature ( \ T) at various test temperatures. The top and middle graphs of Fig. 1 illustrates the initial deviations of a set of type K positive the thermoelements (KP) from NBS** values (1) of KP and the initial deviations of a set of type K negative thermoelement (KN) from NBS values of KN. When a thermocouple wire producer makes a heat of thermocouple alloy, he aims to have all the reels of thermocouple wire from this heat to have zero deviation from NBS table. The thermal EMF of an entire heat of thermocouple alloy usually falls within a band. This is attributed to the overall deviations of chemical analyses of the heat from the desired composition and the inhomogeneity of composition of the heat. The initial deviation of the type K couple, L E K couple in millivolts is equal to the algebraic difference between E and E as shown in Eq. (3). E z LL STANDARD TOLERANCE >BC (PREM) E 2 U Z NBS LL - = t ^ a2 a SPECIAL OR _6 PREMIUM TOLERANCE 1.4 nek couple = ne KP - LE KN Eq. (3) where LEKP and LEKN are the initial deviations of KP and KN in millivolts respectively The initial deviation of a thermocouple can also be expressed in degrees C or F. It can be obtained by the following equation: L TK couple = L EK couple Eq. (4) ne nt *Chromel-Alumel, Trademark, Hoskins Manufacturing Co **NBS (National Bureau of Standards) changed its name to National Institute of Standards and Technology (NIST) in Aug FIG.1 INITIAL CALIBRATION OF TYPE K THERMOCOUPLE WIRE AND TYPE K THERMOCOUPLE REPEATABILITY AND REPRODUCIBILITY Repeatability is the agreement of data among successive tests on the same material using the same test equipment and methods. Reproducibility is the agreement of data among successive tests on the same material using different test equipment and/or methods. 2 Downloaded From: on 12/31/218 Terms of Use:

3 L LOT CALIBRATION OF Single bare thermocouple wire is supplied in reels. The calibration of the wire from top and bottom (front and back) ends of the same reel could be a few microvolts different from each other. This difference is higher at higher temperatures. Finished thermocouple wires in insulated form are also supplied in reels. They are usually a fraction of a degree or even a degree different from end to end. We can reasonally expect that thermocouples made from the same reels could be a few degrees different from one another in their initial condition. It is quite evident from Fig. 1 that two thermocouples could be as much as 8 C from each other at 1, C and still be within premium requirement at that temperature. Lot calibrations on various standardized thermocouples are shown in Figs. 2 through 7. Calibration data is obtained from the file of our Calibration Laboratory. LThe top and bottom range of initial deviation (AT) and standard deviation ( O ) at various test temperatures are shown in these graphs. (a) Type J Thermocouples The top and bottom end calibration data of over 3 reels (total length 45, meters) of QTW-JJ, a 2 gauge glass insulated premium grade type J thermocouple wire, are shown in Fig. 2. The calibration temperatures were 15 C, 36 C and 48 C. All were within premium grade limits. The Standard deviations at these temperatures are.39,.39 and.6 C respectively. These sigma values still represent a significant spread even though the calibration of the entire lot is within premium range. The initial lot consisted of more than five reels of starting thermocouple stock. Therefore, three lots A, B and C, about 1, meters each, were selected from the initial lot and their calibration data is shown in Fig. 3. The top and bottom ranges of initial deviation of each lot were well within the premium range. And the Q values were about one quarter of a degree. Therefore, we can obtain one group of thermocouples with every close initial calibrations to one another by judicious selection from a large lot based on calibration data..2 a-1 z U A NCE -1 Z a -1.2 FIG. 3 LOT CALIBRATIONS OF TYPE J INSULATED Lot calibration data of premium grade 34 stainless steel sheathed type J Ceramo* is shown in Fig. 4. All material meets premium requirement. The top and bottom ranges of initial deviation ( AT) are within the positive half and the tolerance range. The C values are well within 1/4 C of all test temperatures up to 48 C. FIG.2 LOT CALIBRATION OF 45, M OF INSULATED TYPE J *Ceramo, trademark, Thermo Electric Co., Inc. for metal sheathed mineral insulated thermocouple 3 Downloaded From: on 12/31/218 Terms of Use:

4 RANCE z z_ a 1 _1 _ b _ Z b FIG. 4 LOT CALIBRATION OF TYPE J METAL SHEATHED FIG. 5 LOT CALIBRATION OF TEN 1/8" METAL SHEATHED TYPE K (b) Type K Thermocouples Like type J thermocouples the type K thermocouples are also supplied in either insulated (with glass or ceramic fibers) or metal sheathed forms. The calibration data of one of 3.2 mm (1/8") diameter Inconel** sheathed MgO insulated premium grade type K thermocouples are shown in Fig. 5. The top and bottom ranges of LT at all test temperatures are well within type K premium limits. It should be noted that the initial calibration tolerance of both the type K and the type J thermocouples are the same, except that the maximum temperature of operation of type K couples is much higher than that of type J thermocouple. At lower temperatures, the values of the K couples, about 1/4 C, are comparable to those of the J thermocouples. However, at temperatures above 7 C, the values for K couples are considerably higher. (c) Type T Thermocouples The initial calibration tolerance of type T couples is tighter than those of the J and the K thermocouples. It is + 1 C or + 3/4% of temperature for the standard grade and ± 1/2 C or **Inconel, trademark, International Nickel Co. +.4% of temperature for the premium grade. The lot calibration of over 3, meters of insulated premium grade type T thermocouple wire is shown in Fig. 6. The calibration of all the wire is well within premium limits. The type T couples can be used for temperature measurement with high degree of accuracy. The J values at 12 C and below are only several hundreths of a degree. At 3 C the Jvalue of.21 is comparable to lot calibration of the J couple. It is believed that this can be reduced by selection in the same manner as previously described. (d) Type E Thermocouples The initial calibration tolerance is C or + 1/2% for the standard grade and + 1 C or +.4% of temperature for the premium grade. The calibration data of a 3 meter lot of insulated 3 gauge premium grade type E thermocouples are shown in Fig. 7. All wire is within premium limits and the Ovalues are comparable to those of the J thermocouples at the same temperature ranges. STABILITY OF CALIBRATION OF When in use, the calibration of a thermocouple will change from its initial condition. The magnitude and direction of this change depends on temperature, time of exposure, diameter of wire, method of protection of the thermocouple wire, depth of 4 Downloaded From: on 12/31/218 Terms of Use:

5 Q LANCE z b FIG. 6 LOT CALIBRATION OF 34 REELS (3, M) OF TYPE T FIG. 7 LOT CALIBRATION OF 3 M INSULATED TYPE E immersion and the environmental condition to which the thermocouple is exposed. Most of the previous publication on stability tests of base metal thermocoules were on type K and type N thermocouples both in the form of bare wire as well as in metal sheathed mineral insulated thermocouples. (5) (6) (7) (8) Very little information was on J thermocouple (9) and probably none existent on type E and type T thermocouples. The following are short term stability tests conducted at the Company using purchased bare thermocouple wire and our own manufactured products. Type K Thermocouples The stability tests of 16 gauge (1.3 mm or.51" diameter) type K thermocouple at test temperatures of 194 C (2 F), 115 C (21 F), 125 C (22 F) and 126 C (23 F) for 24 hours and beyond in air atmosphere are shown in Fig. 8 (a). Tests at 194 C and 115 C were terminated at 24 hours. Tests at 125 C and 126 C were carried out to 48 hours. For the 126 C tests, thermocouple failure (open circuit) occurred after the 24 hour test. Test results showed that all calibration changes were in the positive direction. The higher the test temperature, the larger the magnitude of the change. The calibration change of C after 24 hours at 1,94 C should be acceptable to most processes. However, the calibration changes of the 1.3 mm bare wire type K couples at temperatures higher than 194 C would be too large in operations where close temperature control throughout the process is necessary. The changes in calibration of 3.2 mm (1/8") diameter type K Ceramos at the same test temperature of 194, 115, 125 and 126 C in air atmosphere for 48 hours were all within.5 C. Therefore, these 3.2 mm type K Inconel sheathed MgO packed thermocoules can be operated up to 126 C for 48 hours in air atmosphere. Unlike the bare wire thermocouples where the slope of calibration change with time was quite large, a small change in calibration of the metal sheathed thermocouples took place within the initial 8 hours. The changes in the next 4 hours were quite gradual in all four cases. It appears that this trend may extend for a period of time. However, the exact change cannot be predicated with certainty. It must be measured. Type J Thermocouple and Type E Thermocouple The stability test results of type J and type E thermocouples for 1 hours at 6 C in air atmosphere are shown in Fig. 9. (1) The type J 5 Downloaded From: on 12/31/218 Terms of Use:

6 Z_ a TIME IN HOURS V (b) 1/8" METAL SHEATHED 5...,...,...,......,..._. +5 Z_ 5 +5 U Z_ Q -5 '2a5 c X 115 c D 194'C i^ TT TIME IN HOURS FIG. 8 STABILITY TESTS OF 16 GAUGE BARE WIRE VERSUS 1/8" METAL SHEATHED TYPE K IN AIR -15 -i i TIME IN HOURS FIG. 9 STABILITY TESTS OF TYPE J AND TYPE E AT 6 C IN AIR thermocouple changed about 1/2 C in the negative direction, while the type E thermocouple changed slightly less than 1 C in the positive direction. Both couples were very stable during this time period at 6 C in air. Again, the exact change of calibration of these thermocouples for extended period of 6 C or at higher temperatures need to be determined in the actual environment. ACCURACY AND REPRODUCIBILITY OF MEASUREMENT (a) Accuracy & Reproducibility Of Measurement On Insulated Iron Constantan Thermocouple The accuracy and reproducibility of calibration on two separate sets of 2 gauge (.8 mm or.32" dia.) insulated iron constantan thermocouple wire QTWO-JJ are listed in Table I. In the previous test, (11) which consisted of 5 duplicate tests plus three data acquisition system (DAS) and NBS tests, the Q values were.12 C or less at all three test temperatures. And, the average value X agreed with NBS data to less than.15 C. In the present test (12) on another lot of wire, the calibration data of the three duplicate tests were in very good agreement with the two DAS test results. The 6Tvalues of these five tests are all within.13 C, which is essentially the same on the previous test. The X values of the present test are plotted in Fig. 1 together with the freeze point measurements. The test data agreed to.4 C. (b) Round-Robin Calibrations Of Type J Thermocouples Round-robin calibrations were conducted by a total of siz parties, including two outside the U.S., on consecutive lengths from a reel of 2 gauge insulated premium grade iron constantan thermocouple QTW-JJ. TE's test results (No. 7) are included in the shaded band. The calibration results, the X and values of these tests and NBS calibrations are shown in Fig. 11. At 37 C or below, the calibration data of all agreed with NBS data to within.5c. At 56 C, the largest deviation from NBS data was close to 1 C. The X and c7values of all tests are shown in the bottom of Fig. 11 in comparison with NBS data. At 37 C and below, X agreed with NBS data to.25 C. The CT values were less than.25 C. At 56 C, the agreement was only to.5 C. The agreement was very good since the uncertainty claimed by NBS was +.1 C below 5 C and + 1 C up to 12 C for base metal thermocouples.(131 6 Downloaded From: on 12/31/218 Terms of Use:

7 I Table I Accuracy And Reproducibility Of Calibration Of 2 Gauge Insulated Iron Constantan Thermocouples Material Used: Two separate lots of QI -JJ, 2 gauge glass insulated premium grade iron-constantan thermocouples Test 1 Previous data (11) 8 tests plus NBS pt in C 94 C 37 C 56 C O U +.5 = NBS.15.5 Test 2 This study: 3 duplicate tests, 2 data acquisition and freeze points of tin, zinc and aluminum Test 15 C 37 C 56 C l DAS l +.3 DAS Freeze Point Measurement Sn Zn Al Measured, C Listed, C T, C (c) Round-Robin Calibration Of Type K Thermocouple Round-robin calibrations were conducted by six parties, including two from abroad, on consecutive lengths of 16 gauge (1.3 no or.51" dia.) type K bare wire standard which was calibrated by NBS. TE's results (No. 7) were obtained by calibrating this NBS standard with another NBS calibrated type K standard. The test results are shown in Fig. 12. At 816 C or below, the agreement was within +.5 C. At 194 C, the range was + 1 C. This was close to the uncertainty claimed by NBS at this temperature. The Q and X values of all calibration tests were plotted at the lower position of Fig. 12. It is interesting to note that X, the average values of FIG. 1 CALIBRATION OF 2 GAUGE GLASS INSULATED IRON CONSTANTAN BY STANDARD METHOD, DATA ACQUISITION AND FREEZE POINT MEASUREMENTS all tests are very close to NBS data at all test temperatures. (d) Comments On Uncertainty Of Test The calibration data from U.S. concerns who participated in these two round-robin tests are all traceable to our NBS. The calibration data from two foreign concerns were referenced to their own respective national standards laboratories. it interesting to observe that all their calibration data is in agreement to within 1 C with the US data which is based on our NBS. It is highly improbable that the national standards laboratories from three different countries make wrong calibrations to the same magnitude and direction. Therefore, the inclusion of calibration data from these two parties and their agreement with all others attest not only to the accuracy of calibration of the national standard agencies of the three countries, but also to the accuracy of calibration of all participants. (e) The Factors Affecting Calibrations Of Thermocouples Numerous factors affecting the calibration of thermocouples and methods of prevention or correction were covered by a recent publication. (14) They will not be repeated in this paper. fi Downloaded From: on 12/31/218 Terms of Use:

8 _ +.5. Z_ TEMPERATURE IN 'C TEMPERATURE IN 'C U Z TEMPERATURE IN `C TEMPERATURE INC FIG. 11 ROUND ROBIN CALIBRATION OF 2 GAUGE TYPE J FIG. 12 ROUND ROBIN CALIBRATION OF 2 GAUGE TYPE K SUMMARY & CONCLUSIONS 1. The initial calibration of thermocouples depends on the match of the positive and the negative thermoelements. Closer adherence to EMF - temperature tables and minimum variation of calibration among thermocouples can be achieved by combination and matching of homogeneous material in the initial stage as well as judicious selection after calibration of the finished thermocouples. 2. Proper protection of thermocouple wire by either metal sheathing or by insulating with glass or ceramic fiber improves the stability of calibration of thermocouples. 3.2 mm metal sheathed type K thermocouples can be operated to 126 C (23 F) in air for 5 hours with + 3 C change of calibration from initial condition. Protected type E and J thermocouples can be used in air for 1 hrs with less than + 1 C change from initial. 3. Round-robin calibration data showed that the uncertainty of calibration of type J and K thermocouple wire is within C up to 194 C. The uncertainty values at lower temperatures are less than + 1 C. 4. In operations where temperature must be held to within a few degrees throughout the process, the initial calibration and the stability of the thermocouple in actual environment must be considered in conjunction with the accuracy and reproducibility of the measuring system and factors affecting calibrations. In addition, all information should be integrated together in selection of the best available system to assure the success of the process. ACKNOWLEDGMENT This writer wishes to thank those, including our colleagues from Thermo Electric Canada and Thermo Electric International (Holland), who participated in the round-robin calibrations. He also acknowledges the encouragement and approval from D. Frank for publication of this paper. Freeze point measurements by D. Bediones, and X calculations by A. Wells, the supply of J and E thermocouple stability data by J. Richmond and assistance in the preparation of manuscript and line drawings from D. Sisti, B. Lamert, R. Vuoncino, D. Gloede, K. Sheldon and J. Murphy are also acknowledged. 8 Downloaded From: on 12/31/218 Terms of Use:

9 REFERENCES 1. NBS 125 Supplement 1 (1975) 2. ANSI MC96.1 (1982) 3. ANSI MC96.1 (1982) ibid 4. ASTM 23 (1987) 5. Wang, T.P. "Thermocouples For High Temperature Applications" Proceedings of ISA/87 Conference, Advance In Instrumentation, Vol. 42 Part 3, p. 1615, Wang, T.P. & Starr, C.D. "Oxidation Resistance And Stability of Nicrosil Nisil in Reducing Atmospheres" Temperature, Its Measurement and Control in Science and Industry, Vol. 5, p. 1147, Anderson, R.L., Lyons, J.D., Knollie, T.G., Christie, WH and Eby, R. "Decalibration of Sheathed Thermocouples," Temperature, Its Measurement and Control in Science and Industry, Vol. 5, P. 977, Bentley, R.E. and Morgan, T.L., "Ni-Based Thermocouples in Mineral Insulated Metal Sheathed Format = Thermo Electric Instabilities to 11 C, J. of Physics E. Vol. 19, p. 262, NBS RP1278, Dahl, (1939) 1. Richmond, J.R., Thermo Electric engineering report (1988). 11. Wang, T.P., Martincavage, J.R., and Bediones, D. "Precision Calibration of Thermocouples and RTD's and A New Calibration Laboratory With Computerized Data Acquisition System" ISA Proceedings Instrumentation and Control Systems Conference & Exhibit p. 17, D. Bediones Calibration Report 1/1/ NBS "Calibration Services Users Guide" (1986). 14. Wang, T.P. And Wells, A., "Thermocouple System Technology and Applications For Gas Turbines," Proceedings of ASME International Symposium on Turbo-Machinery, Combined Cycle Technologies and Cogeneration, Vol. II, p. 35, Downloaded From: on 12/31/218 Terms of Use: