Bioelectromagnetics 36:544^550 (2015)
Letter to the Editor An Interlaboratory Comparison Program on ELF Electric and Magnetic Fields Measurements Performed in Greece: Second Round of the Scheme The second round of an interlaboratory comparison program for extremely low frequency electric and magnetic fields measurements was performed at the High Voltage Laboratory of the National Technical University of Athens (Greece). The 16 participating laboratories measured the following: (i) electric field produced by a scale transmission line; (ii) magnetic field produced by a medium voltage cable; and (iii) magnetic field and frequency at the center of a standard square coil and their delivered results were evaluated in all measurement scenarios with use of performance statistics z-scores. Deviations between z-scores based on usual estimators (mean value, standard deviation) and robust estimators (derived with the robust algorithm described by the International Organization for Standardization [ISO, 2005]) highlight improved performance of the robust algorithm. An overall comparison to measurement procedure and performance results of the first round proves effectiveness and necessity of the scheme. Improper instrumentation or calibration, instability of the field source and measurement position uncertainty are factors that may cause unsatisfactory performance of the participants. Bioelectromagnetics. 36:544–550, 2015. © 2015 Wiley Periodicals, Inc. Key words: exposure; human health; ELF EMF measurements; z-scores; error sources
INTRODUCTION Exposure to extremely low frequency (ELF) electric and magnetic fields often raises public concerns about possible negative effects on human health with attention focused on the alternating current (AC) power transmission systems, especially on high voltage overhead lines, because of the significant field levels in their vicinity. Measurement procedures with regard to exposure of human beings have been established by the International Electrotechnical Commission (IEC) [1998], IEC [2009], and the Institute of Electrical and Electronics Engineers (IEEE) [1994] standards. Reference levels for limiting exposure of the general public to fields produced by the operational power frequency of 50 Hz set by the International Commission on Non-Ionizing Radiation Protection (ICNIRP) [2010] are 5 kV/m for the electric field strength (E) and 200 mT for the magnetic flux density (B). It must be pointed out that the scientific background for the derivation of the exposure limiting guidelines is dosimetry, i.e., high resolution calculations of internal electric field strength values [ICNIRP, 2010]. To ensure compliance with the above limit values, ELF measurements in various parts of the power system are required. 2015 Wiley Periodicals, Inc.
Within this growing need for reliable and comparable laboratory results, participation in interlaboratory comparison (ILC) programs are an efficient method for assessing the adequacy and improving the performance of a laboratory. According to the International Organization for Standardization/International Electrotechnical Commission (ISO/IEC) [2005], each accredited test laboratory should participate in interlaboratory comparisons that cover all measurements lying within its scope of accreditation as a prerequisite for demonstrating traceability and proficiency and proving its technical competence. ILCs are defined as the organization, performance, and evaluation of calibration/tests on the same Conflicts of interest: None. *Correspondence to: Assistant Prof. Dr. Ioannis Gonos, National Technical University of Athens, 9 Iroon Politechniou Str, Zografou Campus, GR 15780, Athens, Greece. E-mail: [email protected] Received for review 2 September 2014; Accepted 6 March 2015 DOI: 10.1002/BEM.21913 Published online 6 April 2015 in Wiley Online Library (wileyonlinelibrary.com).
An Interlaboratory Comparison Program on ELF Fields Measurements
or similar calibration/test items by two or more laboratories under predetermined conditions [ISO/ IEC, 2010]. The type of ILC that must be applied to electric and magnetic fields measurements is a result comparison program where all participants measure the same field source, usually in normal working conditions [Nicolopoulou et al., 2012]. ILC programs described in this paper have been set up to comply with requirements of the International Laboratory Accreditation Cooperation (ILAC) [2000] guidelines, the ISO/IEC [1997], the ILAC [2005] policy, and the relevant Hellenic Accreditation System S.A. (ESYD) [2011] policy for participation of laboratories in proficiency testing schemes. Lack of available published proficiency testing program on ELF electric and magnetic fields measurements, organized either by national or by international bodies, led us in 2009 to the attainment of the first round of an ELF ILC program [Nicolopoulou et al., 2012] that covers two types of measurements (electric and magnetic field) described in IEC [1998]. Taking into consideration the need for continuous monitoring of laboratory performance as well as the requirement of ESYD for accredited laboratories to participate in proficiency testing programs at least every 4 years, the second round of the scheme was organized in February 2013.
PERFORMANCE STATISTICS The basic concept of interlaboratory comparisons is evaluation of laboratories through a statistical indicator of their competence (performance statistic). The initial stage of statistical analysis is to define the best available estimations for the true ^ and the value of the measurand (assigned value m) dispersion of the measurements (standard deviation for the proficiency assessment s^ ). As in the first round of the scheme [Nicolopoulou et al., 2012], estimators were calculated in each test level by applying the iterative robust algorithm described in ISO [2005] (Annex C, Algorithm A) to laboratory measurements. The main advantage of this algorithm is robustness, which is the persistence of the method under perturbations or conditions of uncertainty. Thus, produced robust statistics have a reasonably small bias and are resistant to errors in the results due to deviations from assumptions (e.g., normality of data). This is of high importance for electric and magnetic fields measurements, which are dominated by numerous uncertainty factors. Robust estimators were then used to calculate
545
performance statistic z-score from the relationship: z¼
b xm b xm ¼ bs s b
ð1Þ
where x is measurement of the laboratory,^x is robust mean, and ^s is robust standard deviation. Results were evaluated according to the following regions of the z-score values [ISO, 2005]: when | z| 2, performance of the laboratory is satisfactory; when 2 < |z| < 3 accuracy and correctness of the measurement are questionable and the performance statistic is a “warning signal;” and when |z| 3, the performance of the laboratory is non-satisfactory and the performance statistic is an “action signal.” An “action signal” is a strong indication of problems in the measuring performance of the laboratory and further investigation is required for determination of error factors, which affect measurement quality. “Warning signals” are milder indicators of inaccuracy sources and raise concerns mainly if they reappear in various test levels (or various test rounds). DESCRIPTION OF ILC SCHEME Participants-Measuring Equipment Sixteen laboratories participated in this interlaboratory comparison procedure. Participants were randomly named Laboratories 1–16. For data confidentiality reasons, each group was aware of its own number, but did not know numbers of other groups. Participants conducted measurements using the following equipment: eight teams used NARDA (EFA300: New York, NY); three teams used NARDA (EFA-3, New York, NY); two teams used NARDA (PMM 8053: Milano, Italy); one team used ENERTECH Consultants (EMDEX LITE: Campbell, CA); one team used ENERTECH Consultants (Standard EMDEX II: Campbell, CA); and finally, one team used Mascheck Electronik (ESM-100: Bad W€orishofen, Germany). Measurement Procedures The ELF ILC program was carried out in three steps at the High Voltage Laboratory of the National Technical University of Athens (Greece) (NTUA). Photographs of measurement positions for each test stage are presented in Figures 1–3. During the first test phase, electric field strength (E), created by a high-voltage transmission line, was measured. A setup with a scale transmission line supplied with 35 kV was properly formed at the NTUA laboratory. Fourteen laboratories (two teams did not participate in this test), recorded electric Bioelectromagnetics
546
Ztoupis et al.
Fig. 1. Setup for measurement of electric field strength under an overhead transmission line.
Fig. 3. Setup for measurement of magnetic field strength and frequency in center of a standard coil.
field strength at a height of 1 m at three selected positions at specific distances from the transmission line. During the second test phase, magnetic flux density (B), generated by a double-loop cable carrying 500 A, was measured. For each current value, 16 measurement groups recorded magnetic field at three selected positions at a height of 1,0 m above ground level with a sensor and a fieldmeter. During the third test phase, magnitude and frequency of a uniform magnetic field, generated by a standard 1 1 m2 square coil, was measured. The multiturn magnetic field coil was connected via an interface unit with an AC source, which provided required value of current, free of harmonics. Specifically for frequency measurement, most participants reported the same frequency value resulting in a zero robust standard deviation in the first iteration. To enable execution of the algorithm and taking into consideration that the exact frequency value was set at 127 Hz, the organizer considered acceptable measurement values within the range from 126 to 128 Hz. Thus, the initial value of the standard deviation was calculated as follows: Sf ¼ Fig. 2. Setup for measurement of magnetic flux density produced by a medium voltage cable. Bioelectromagnetics
128 126 pffiffiffi ¼ 1:414 2
ð2Þ
An Interlaboratory Comparison Program on ELF Fields Measurements
According to the calibration certificate of the AC source, the magnetic flux density (T) in the center of the square coil is given by the equation: B ¼ m0 coilfactor I
ð3Þ
where I is total current through the coil. m0 ¼ 4p 107 m kg s2 A2 pffiffi coilfactor ¼ 2pl2 reductionfactor, a constant that incorporates geometry (l ¼ 1m is the side of the square) and losses of the coil (reduction factor). During measurements, the square coil was supplied by the AC source with a fixed current value I ¼ 1.12A 2% (95%CI). Applying a coil factor ¼ 0.84 derived from the AC source calibration certificate, the resulting magnetic flux density is B ¼ 1.184 mT 2%. Note that reported B4 values are very close to theoretically calculated magnetic flux density, with small deviations being attributed mainly to small deviations of measurement position and current uncertainty of the AC power source.
EVALUATIONçz-SCORES Two different methods for definition of the ^ and s^ and calculations of the respective estimators m z-scores were applied as follows: estimators defined as mean value m and standard deviation s of measurements, and estimators defined as robust mean ^x and robust standard deviation ^s of measurements. The z-scores of all laboratories for the electric and magnetic field measurements calculated with both methods are comparatively depicted in Tables 1 and 2, respectively. Warning and action signals are highlighted. DISCUSSION AND CONCLUSION The purpose of the presented analysis was to describe organization and execution of the second round of an ELF ILC scheme and to assess its overall function by detecting weaknesses of predetermined measurement procedures and inaccuracy factors within laboratories. Factors of Inaccuracy Based on calculations of z-scores, evaluation of the performance of participating laboratories can be determined as follows: Tables 1 and 2 show that participants demonstrated a better performance when measuring magnetic flux density, as no |z | 3 occurred. Warning signals that Laboratories 15 and 16 (Table 2) received
547
could be attributed to lack of a recent calibration (Laboratory 16) or to false processing of recorded value. It should be pointed out that Laboratories 13 and 15 used the same measuring instrument, but a field sensor of a different diameter. Absence of |z| > 2 values for Laboratory 13 implies a bigger sensor provides higher measurement accuracy. All laboratories achieved a satisfying performance at the B4 measurement. This proves how strongly evaluation of a laboratory is affected by stability of the field source both regarding level of emitted field and measurement position. Laboratory 9 reported a non-acceptable value for frequency measurement at the B4 position despite accurate recording of corresponding magnetic field strength (Table 2). An inappropriate setting/adjustment of the instrument could be a possible error source. As expected, measurements of electric field strength, which is perturbed by the presence of the operators and nearby metallic objects, are prone to deviations. Repetitive |z| > 2 values that Laboratory 12 received in all electric field measurements (Table 1) indicate a systematic error (bias) caused by the specific instrument, which requires presence of the operator very close to equipment, thus resulting in a distortion of electric field. When the field meter is not connected via optical fiber to the sensor, presence of the operator who is holding the measurement device perturbs electric field distribution, whereas magnetic field measurements can be influenced only due to increase of measurement position uncertainty. To minimize deviations of measurement position as much as possible, exact points where the tripod had to be placed were marked on the floor, as seen in Figures 1 and 2, and the center of the magnetic coil was marked with two crossed stripes (Fig. 3). Also the base and height of the tripod were fixed by the organizer before execution of measurements. Warning signals of Laboratories 7 and 9 (Table 1) are possibly a result of a lack of proper instrument calibration and/or operator errors. Although it showed an overall satisfying performance, Laboratory 1 was evaluated with a “warning signal” for its E1 measurement because of lack of calibration of recording sensor for the last 4 years. Proper calibration (calibration body, repetition interval, etc.) arises as a factor that highly influences measurement quality. Non-accredited calibration laboratories with high calibration uncertainties may provide certificates of questionable accuracy. Furthermore, calibration should cover frequency range and measurement range applied in measurements. An additional point that requires attention is Bioelectromagnetics
548
Ztoupis et al.
TABLE 1. z-Scores (Calculated With Both Methods) for Measurements of Electric Field Strength. The Shaded Areas Refer to Non-Satisfactory z-Scores: the Light-Shaded Cells Contain “Warning Signals” i.e. 2 < |z| < 3 and More Intensely Shaded Cells Contain “Action Signals” i.e. |z| 3. z-scores E1
E2
E3
Laboratory
m, s
b x ;bs
m,s
bx ;bs
m, s
bx ;bs
1 2 3 4 5 6 7 8 9 10 11 12 13 14
1.715 0.103 0.224 0.025 0.522 0.605 2.196 0.052 1.203 0.152 0.200 1.718 0.062 0.261
2.121 0.010 0.167 0.093 0.558 0.918 3.001 0.057 1.450 0.073 0.136 2.375 0.044 0.217
1.262 0.214 0.152 0.288 0.014 0.528 2.418 0.300 0.678 0.195 0.171 2.102 0.257 0.165
2.480 0.267 0.153 0.405 0.156 0.851 4.632 0.428 1.392 0.233 0.187 3.781 0.347 0.176
0.308 0.276 0.224 0.026 0.688 0.177 0.276 0.047 1.434 0.109 0.250 3.164 0.146 0.172
1.332 0.403 0.248 0.340 1.627 0.944 0.403 0.278 3.842 0.093 0.326 9.820 0.016 0.093
application of the correction factor of the field meter provided by the calibration certificate. Laboratory accreditation is not always adequate on its own; it should be followed by recent and proper calibration of equipment. Measurement accuracy always depends on experience of the operator and of the laboratory (i.e., how often laboratory conducts this specific measurement.)
Comparison of Two Methods to Calculate z-Scores In various test levels, disagreement between two methods for calculation of the z-scores was observed. All warning/action signals of the first method (m, s) (Tables 1 and 2) were also detected with robust algorithm (Tables 1 and 2). On the contrary, there
TABLE 2. z-Scores (Calculated With Both Methods) for Measurements of Magnetic Flux Density. The Shaded Areas Refer to Non-Satisfactory z-Scores: the Light-Shaded Cells Contain “Warning Signals” i.e. 2 < |z| < 3 and More Intensely Shaded Cells Contain “Action Signals” i.e. |z| 3. z-scores B2
B1
B3
B4
f4
Laboratory
m, s
b x ;bs
m, s
bx ;bs
m, s
bx ;bs
m, s
b x ;bs
m, s
bx ;bs
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
0.764 0.377 0.437 0.593 0.939 1.025 1.554 1.729 0.025 0.606 0.363 0.575 0.271 0.013 2.338 0.154
0.935 0.353 0.566 0.596 0.986 1.083 1.826 1.877 0.044 0.610 0.483 0.575 0.378 0.058 2.710 0.246
0.147 0.191 0.896 0.039 1.177 1.340 0.299 0.580 0.039 1.970 0.701 1.168 0.558 0.352 0.462 2.036
0.154 0.198 0.914 0.044 1.189 1.354 0.308 0.583 0.044 1.992 0.715 1.189 0.561 0.352 0.473 2.069
0.517 0.037 0.750 0.951 0.564 1.796 0.654 1.676 0.417 1.868 0.577 1.158 0.453 0.894 0.373 0.043
0.454 0.039 0.692 0.898 0.655 1.917 0.593 1.794 0.504 1.991 0.515 1.110 0.388 0.839 0.306 0.121
1.791 1.208 0.458 0.441 1.041 1.283 1.208 0.218 1.041 1.071 0.908 – 0.291 0.683 0.291 –
1.612 1.075 0.403 0.403 0.940 1.142 1.075 0.188 0.940 0.967 0.806 – 0.268 0.605 0.268 –
0.258 0.417 0.417 0.258 0.258 0.379 0.258 0.258 3.403 0.417 0.471 – 0.258 0.258 0.258 –
0.805 1.302 1.302 0.805 0.805 1.184 0.805 0.805 10.631 1.302 1.471 – 0.805 0.805 0.805 –
Bioelectromagnetics
An Interlaboratory Comparison Program on ELF Fields Measurements
were multiple cases where warning/action signals of the second method had a satisfactory evaluation (|z| < 2) when using the first method. Additionally, some action signals (|z| 3) derived with the robust algorithm were detected only as warning signals with the first method. It is obvious that the first calculation method (m, s) was not particularly reliable, since remote values (i.e., results that deviate significantly from the rest of the measurements) are taken into account when calculating estimators. Resulting standard deviation s of results in most test levels was higher compared to robust estimator and, therefore, in most cases, z-scores of the first method tended to be smaller (Tables 1 and 2). As expected, zscores calculated with the robust iterative algorithm were a more stable indicator, less affected by the presence of distant values. Comparison With First Round of ILC The lessons learned from the first round concerning necessary actions needed for improvement of the overall procedure were implemented in this second scheme, which therefore proved to be very effective. The main alterations in measurement process in this new round are, at first implementation of measurement of electric/magnetic field values at different distances from a fixed source rather than at different voltage/current levels and, secondly, the addition of the third measurement phase which includes a frequency measurement. In this round, particular emphasis was given to maintaining stability of measured fields by keeping both voltage and current sources as stable as possible. This was achieved for each of the three measurement stages by using a voltage stabilizer, a current clamp in combination with a digital multimeter for continuous monitoring of the current, and use of appropriate software, which automatically adjusts coil current, respectively. Absence of other major spectral components at the test site allowed organizers to ask laboratories to deliver results in only one of two measurement modes (either broadband or band pass filter modes as provided by measuring equipment). This enabled a significant increase in number of participants in this round, considering time limitation on the current transformer and required time for execution of measurements. The influence of harmonics in the network voltage was reduced with the use of voltage stabilizers in the first two measurement stages and at the third measurement stage by use of an AC power source free of harmonics.
549
All participants of the first ELF ILC scheme (four accredited laboratories consisting of five measuring teams) were also present in this second round. A comparison between rounds of the scheme using an aggregated performance statistic would not be suitable because both test levels and measuring teams, defined as the unique combination of equipment and operator, were not identical. An attempt to examine only influence of equipment by counting number of test levels where | z| > 2 shows the majority of laboratories that participated in the first round (where no one had been evaluated with a |z| > 2) received warning and/or action signals in this second round. This is a conclusion of high importance as it proves necessity of the scheme, along with the fact that even accredited laboratories with recently calibrated instruments received |z| > 3. Ioannis Ztoupis,Eleni Nicolopoulou,Ioannis Gonos*, Ioannis Stathopulos HighVoltage Laboratory, School of Electrical and Computer Engineering, National Technical University of Athens, Athens, Greece
Efthymios Karabetsos Non-Ionizing Radiation Office, Greek Atomic Energy Commission (EEAE), Agia Paraskevi, Greece
REFERENCES ESYD. 2011. ESYD policy relevant to PT Schemes and Interlaboratory Comparisons. Athens, Greece: Hellenic Accreditation System S.A. PDI/02/01. ICNIRP. 2010. Guidelines for limiting exposure to time varying electric and magnetic fields (1 Hz–100kHz). Health Phys 99:818–836. IEC. 1998. Measurement of low-frequency magnetic and electric fields with regard to exposure of human beings—Special requirements for instruments and guidance for measurements. Geneva, Switzerland: IEC. 61786. IEC. 2009. Electric and magnetic field levels generated by AC power systems—Measurement procedures with regard to public exposure. Geneva, Switzerland: IEC. 62110. IEEE. 1994. (Revision of IEEE 644-1987). IEEE standard procedures for measurement of power frequency electric and magnetic fields from AC power lines. New York, NY: IEEE. 644. ILAC. 2000. Guidelines for the requirements for the competence of providers of proficiency testing schemes. Silverwater, Australia: International Laboratory Accreditation Cooperation. G13. ILAC. 2005. ILAC Policy for participation in national and international proficiency testing activities. Silverwater, Australia: International Laboratory Accreditation Cooperation. P9. Bioelectromagnetics
550
Ztoupis et al.
ISO. 2005. Statistical methods for use in proficiency testing by interlaboratory comparisons. Geneva, Switzerland: International Organization for Standardization. 13528. ISO/IEC. 2005. General requirements for the competence of testing and calibration laboratories. Geneva, Switzerland: International Electrotechnical Commission. 17025. ISO/IEC. 2010. Conformity assessment-General requirements for proficiency testing. Geneva, Switzerland: International Electrotechnical Commission. 17043.
Bioelectromagnetics
ISO/IEC. 1997. Proficiency testing by interlaboratory comparisons Part 1: Development and operation of proficiency testing schemes. International Electrotechnical Commission: Geneva, Switzerland. 43–1. Nicolopoulou EP, Gonos IF, Stathopulos IA, Karabetsos E. 2012. Two interlaboratory comparison programs on EMF measurements performed in Greece. IEEE Electromagnetic Compatibility Magazine Vol. 1, Q.2: 50–59.
Letter to the Editor An Interlaboratory Comparison Program on ELF Electric and Magnetic Fields Measurements Performed in Greece: Second Round of the Scheme The second round of an interlaboratory comparison program for extremely low frequency electric and magnetic fields measurements was performed at the High Voltage Laboratory of the National Technical University of Athens (Greece). The 16 participating laboratories measured the following: (i) electric field produced by a scale transmission line; (ii) magnetic field produced by a medium voltage cable; and (iii) magnetic field and frequency at the center of a standard square coil and their delivered results were evaluated in all measurement scenarios with use of performance statistics z-scores. Deviations between z-scores based on usual estimators (mean value, standard deviation) and robust estimators (derived with the robust algorithm described by the International Organization for Standardization [ISO, 2005]) highlight improved performance of the robust algorithm. An overall comparison to measurement procedure and performance results of the first round proves effectiveness and necessity of the scheme. Improper instrumentation or calibration, instability of the field source and measurement position uncertainty are factors that may cause unsatisfactory performance of the participants. Bioelectromagnetics. 36:544–550, 2015. © 2015 Wiley Periodicals, Inc. Key words: exposure; human health; ELF EMF measurements; z-scores; error sources
INTRODUCTION Exposure to extremely low frequency (ELF) electric and magnetic fields often raises public concerns about possible negative effects on human health with attention focused on the alternating current (AC) power transmission systems, especially on high voltage overhead lines, because of the significant field levels in their vicinity. Measurement procedures with regard to exposure of human beings have been established by the International Electrotechnical Commission (IEC) [1998], IEC [2009], and the Institute of Electrical and Electronics Engineers (IEEE) [1994] standards. Reference levels for limiting exposure of the general public to fields produced by the operational power frequency of 50 Hz set by the International Commission on Non-Ionizing Radiation Protection (ICNIRP) [2010] are 5 kV/m for the electric field strength (E) and 200 mT for the magnetic flux density (B). It must be pointed out that the scientific background for the derivation of the exposure limiting guidelines is dosimetry, i.e., high resolution calculations of internal electric field strength values [ICNIRP, 2010]. To ensure compliance with the above limit values, ELF measurements in various parts of the power system are required. 2015 Wiley Periodicals, Inc.
Within this growing need for reliable and comparable laboratory results, participation in interlaboratory comparison (ILC) programs are an efficient method for assessing the adequacy and improving the performance of a laboratory. According to the International Organization for Standardization/International Electrotechnical Commission (ISO/IEC) [2005], each accredited test laboratory should participate in interlaboratory comparisons that cover all measurements lying within its scope of accreditation as a prerequisite for demonstrating traceability and proficiency and proving its technical competence. ILCs are defined as the organization, performance, and evaluation of calibration/tests on the same Conflicts of interest: None. *Correspondence to: Assistant Prof. Dr. Ioannis Gonos, National Technical University of Athens, 9 Iroon Politechniou Str, Zografou Campus, GR 15780, Athens, Greece. E-mail: [email protected] Received for review 2 September 2014; Accepted 6 March 2015 DOI: 10.1002/BEM.21913 Published online 6 April 2015 in Wiley Online Library (wileyonlinelibrary.com).
An Interlaboratory Comparison Program on ELF Fields Measurements
or similar calibration/test items by two or more laboratories under predetermined conditions [ISO/ IEC, 2010]. The type of ILC that must be applied to electric and magnetic fields measurements is a result comparison program where all participants measure the same field source, usually in normal working conditions [Nicolopoulou et al., 2012]. ILC programs described in this paper have been set up to comply with requirements of the International Laboratory Accreditation Cooperation (ILAC) [2000] guidelines, the ISO/IEC [1997], the ILAC [2005] policy, and the relevant Hellenic Accreditation System S.A. (ESYD) [2011] policy for participation of laboratories in proficiency testing schemes. Lack of available published proficiency testing program on ELF electric and magnetic fields measurements, organized either by national or by international bodies, led us in 2009 to the attainment of the first round of an ELF ILC program [Nicolopoulou et al., 2012] that covers two types of measurements (electric and magnetic field) described in IEC [1998]. Taking into consideration the need for continuous monitoring of laboratory performance as well as the requirement of ESYD for accredited laboratories to participate in proficiency testing programs at least every 4 years, the second round of the scheme was organized in February 2013.
PERFORMANCE STATISTICS The basic concept of interlaboratory comparisons is evaluation of laboratories through a statistical indicator of their competence (performance statistic). The initial stage of statistical analysis is to define the best available estimations for the true ^ and the value of the measurand (assigned value m) dispersion of the measurements (standard deviation for the proficiency assessment s^ ). As in the first round of the scheme [Nicolopoulou et al., 2012], estimators were calculated in each test level by applying the iterative robust algorithm described in ISO [2005] (Annex C, Algorithm A) to laboratory measurements. The main advantage of this algorithm is robustness, which is the persistence of the method under perturbations or conditions of uncertainty. Thus, produced robust statistics have a reasonably small bias and are resistant to errors in the results due to deviations from assumptions (e.g., normality of data). This is of high importance for electric and magnetic fields measurements, which are dominated by numerous uncertainty factors. Robust estimators were then used to calculate
545
performance statistic z-score from the relationship: z¼
b xm b xm ¼ bs s b
ð1Þ
where x is measurement of the laboratory,^x is robust mean, and ^s is robust standard deviation. Results were evaluated according to the following regions of the z-score values [ISO, 2005]: when | z| 2, performance of the laboratory is satisfactory; when 2 < |z| < 3 accuracy and correctness of the measurement are questionable and the performance statistic is a “warning signal;” and when |z| 3, the performance of the laboratory is non-satisfactory and the performance statistic is an “action signal.” An “action signal” is a strong indication of problems in the measuring performance of the laboratory and further investigation is required for determination of error factors, which affect measurement quality. “Warning signals” are milder indicators of inaccuracy sources and raise concerns mainly if they reappear in various test levels (or various test rounds). DESCRIPTION OF ILC SCHEME Participants-Measuring Equipment Sixteen laboratories participated in this interlaboratory comparison procedure. Participants were randomly named Laboratories 1–16. For data confidentiality reasons, each group was aware of its own number, but did not know numbers of other groups. Participants conducted measurements using the following equipment: eight teams used NARDA (EFA300: New York, NY); three teams used NARDA (EFA-3, New York, NY); two teams used NARDA (PMM 8053: Milano, Italy); one team used ENERTECH Consultants (EMDEX LITE: Campbell, CA); one team used ENERTECH Consultants (Standard EMDEX II: Campbell, CA); and finally, one team used Mascheck Electronik (ESM-100: Bad W€orishofen, Germany). Measurement Procedures The ELF ILC program was carried out in three steps at the High Voltage Laboratory of the National Technical University of Athens (Greece) (NTUA). Photographs of measurement positions for each test stage are presented in Figures 1–3. During the first test phase, electric field strength (E), created by a high-voltage transmission line, was measured. A setup with a scale transmission line supplied with 35 kV was properly formed at the NTUA laboratory. Fourteen laboratories (two teams did not participate in this test), recorded electric Bioelectromagnetics
546
Ztoupis et al.
Fig. 1. Setup for measurement of electric field strength under an overhead transmission line.
Fig. 3. Setup for measurement of magnetic field strength and frequency in center of a standard coil.
field strength at a height of 1 m at three selected positions at specific distances from the transmission line. During the second test phase, magnetic flux density (B), generated by a double-loop cable carrying 500 A, was measured. For each current value, 16 measurement groups recorded magnetic field at three selected positions at a height of 1,0 m above ground level with a sensor and a fieldmeter. During the third test phase, magnitude and frequency of a uniform magnetic field, generated by a standard 1 1 m2 square coil, was measured. The multiturn magnetic field coil was connected via an interface unit with an AC source, which provided required value of current, free of harmonics. Specifically for frequency measurement, most participants reported the same frequency value resulting in a zero robust standard deviation in the first iteration. To enable execution of the algorithm and taking into consideration that the exact frequency value was set at 127 Hz, the organizer considered acceptable measurement values within the range from 126 to 128 Hz. Thus, the initial value of the standard deviation was calculated as follows: Sf ¼ Fig. 2. Setup for measurement of magnetic flux density produced by a medium voltage cable. Bioelectromagnetics
128 126 pffiffiffi ¼ 1:414 2
ð2Þ
An Interlaboratory Comparison Program on ELF Fields Measurements
According to the calibration certificate of the AC source, the magnetic flux density (T) in the center of the square coil is given by the equation: B ¼ m0 coilfactor I
ð3Þ
where I is total current through the coil. m0 ¼ 4p 107 m kg s2 A2 pffiffi coilfactor ¼ 2pl2 reductionfactor, a constant that incorporates geometry (l ¼ 1m is the side of the square) and losses of the coil (reduction factor). During measurements, the square coil was supplied by the AC source with a fixed current value I ¼ 1.12A 2% (95%CI). Applying a coil factor ¼ 0.84 derived from the AC source calibration certificate, the resulting magnetic flux density is B ¼ 1.184 mT 2%. Note that reported B4 values are very close to theoretically calculated magnetic flux density, with small deviations being attributed mainly to small deviations of measurement position and current uncertainty of the AC power source.
EVALUATIONçz-SCORES Two different methods for definition of the ^ and s^ and calculations of the respective estimators m z-scores were applied as follows: estimators defined as mean value m and standard deviation s of measurements, and estimators defined as robust mean ^x and robust standard deviation ^s of measurements. The z-scores of all laboratories for the electric and magnetic field measurements calculated with both methods are comparatively depicted in Tables 1 and 2, respectively. Warning and action signals are highlighted. DISCUSSION AND CONCLUSION The purpose of the presented analysis was to describe organization and execution of the second round of an ELF ILC scheme and to assess its overall function by detecting weaknesses of predetermined measurement procedures and inaccuracy factors within laboratories. Factors of Inaccuracy Based on calculations of z-scores, evaluation of the performance of participating laboratories can be determined as follows: Tables 1 and 2 show that participants demonstrated a better performance when measuring magnetic flux density, as no |z | 3 occurred. Warning signals that Laboratories 15 and 16 (Table 2) received
547
could be attributed to lack of a recent calibration (Laboratory 16) or to false processing of recorded value. It should be pointed out that Laboratories 13 and 15 used the same measuring instrument, but a field sensor of a different diameter. Absence of |z| > 2 values for Laboratory 13 implies a bigger sensor provides higher measurement accuracy. All laboratories achieved a satisfying performance at the B4 measurement. This proves how strongly evaluation of a laboratory is affected by stability of the field source both regarding level of emitted field and measurement position. Laboratory 9 reported a non-acceptable value for frequency measurement at the B4 position despite accurate recording of corresponding magnetic field strength (Table 2). An inappropriate setting/adjustment of the instrument could be a possible error source. As expected, measurements of electric field strength, which is perturbed by the presence of the operators and nearby metallic objects, are prone to deviations. Repetitive |z| > 2 values that Laboratory 12 received in all electric field measurements (Table 1) indicate a systematic error (bias) caused by the specific instrument, which requires presence of the operator very close to equipment, thus resulting in a distortion of electric field. When the field meter is not connected via optical fiber to the sensor, presence of the operator who is holding the measurement device perturbs electric field distribution, whereas magnetic field measurements can be influenced only due to increase of measurement position uncertainty. To minimize deviations of measurement position as much as possible, exact points where the tripod had to be placed were marked on the floor, as seen in Figures 1 and 2, and the center of the magnetic coil was marked with two crossed stripes (Fig. 3). Also the base and height of the tripod were fixed by the organizer before execution of measurements. Warning signals of Laboratories 7 and 9 (Table 1) are possibly a result of a lack of proper instrument calibration and/or operator errors. Although it showed an overall satisfying performance, Laboratory 1 was evaluated with a “warning signal” for its E1 measurement because of lack of calibration of recording sensor for the last 4 years. Proper calibration (calibration body, repetition interval, etc.) arises as a factor that highly influences measurement quality. Non-accredited calibration laboratories with high calibration uncertainties may provide certificates of questionable accuracy. Furthermore, calibration should cover frequency range and measurement range applied in measurements. An additional point that requires attention is Bioelectromagnetics
548
Ztoupis et al.
TABLE 1. z-Scores (Calculated With Both Methods) for Measurements of Electric Field Strength. The Shaded Areas Refer to Non-Satisfactory z-Scores: the Light-Shaded Cells Contain “Warning Signals” i.e. 2 < |z| < 3 and More Intensely Shaded Cells Contain “Action Signals” i.e. |z| 3. z-scores E1
E2
E3
Laboratory
m, s
b x ;bs
m,s
bx ;bs
m, s
bx ;bs
1 2 3 4 5 6 7 8 9 10 11 12 13 14
1.715 0.103 0.224 0.025 0.522 0.605 2.196 0.052 1.203 0.152 0.200 1.718 0.062 0.261
2.121 0.010 0.167 0.093 0.558 0.918 3.001 0.057 1.450 0.073 0.136 2.375 0.044 0.217
1.262 0.214 0.152 0.288 0.014 0.528 2.418 0.300 0.678 0.195 0.171 2.102 0.257 0.165
2.480 0.267 0.153 0.405 0.156 0.851 4.632 0.428 1.392 0.233 0.187 3.781 0.347 0.176
0.308 0.276 0.224 0.026 0.688 0.177 0.276 0.047 1.434 0.109 0.250 3.164 0.146 0.172
1.332 0.403 0.248 0.340 1.627 0.944 0.403 0.278 3.842 0.093 0.326 9.820 0.016 0.093
application of the correction factor of the field meter provided by the calibration certificate. Laboratory accreditation is not always adequate on its own; it should be followed by recent and proper calibration of equipment. Measurement accuracy always depends on experience of the operator and of the laboratory (i.e., how often laboratory conducts this specific measurement.)
Comparison of Two Methods to Calculate z-Scores In various test levels, disagreement between two methods for calculation of the z-scores was observed. All warning/action signals of the first method (m, s) (Tables 1 and 2) were also detected with robust algorithm (Tables 1 and 2). On the contrary, there
TABLE 2. z-Scores (Calculated With Both Methods) for Measurements of Magnetic Flux Density. The Shaded Areas Refer to Non-Satisfactory z-Scores: the Light-Shaded Cells Contain “Warning Signals” i.e. 2 < |z| < 3 and More Intensely Shaded Cells Contain “Action Signals” i.e. |z| 3. z-scores B2
B1
B3
B4
f4
Laboratory
m, s
b x ;bs
m, s
bx ;bs
m, s
bx ;bs
m, s
b x ;bs
m, s
bx ;bs
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
0.764 0.377 0.437 0.593 0.939 1.025 1.554 1.729 0.025 0.606 0.363 0.575 0.271 0.013 2.338 0.154
0.935 0.353 0.566 0.596 0.986 1.083 1.826 1.877 0.044 0.610 0.483 0.575 0.378 0.058 2.710 0.246
0.147 0.191 0.896 0.039 1.177 1.340 0.299 0.580 0.039 1.970 0.701 1.168 0.558 0.352 0.462 2.036
0.154 0.198 0.914 0.044 1.189 1.354 0.308 0.583 0.044 1.992 0.715 1.189 0.561 0.352 0.473 2.069
0.517 0.037 0.750 0.951 0.564 1.796 0.654 1.676 0.417 1.868 0.577 1.158 0.453 0.894 0.373 0.043
0.454 0.039 0.692 0.898 0.655 1.917 0.593 1.794 0.504 1.991 0.515 1.110 0.388 0.839 0.306 0.121
1.791 1.208 0.458 0.441 1.041 1.283 1.208 0.218 1.041 1.071 0.908 – 0.291 0.683 0.291 –
1.612 1.075 0.403 0.403 0.940 1.142 1.075 0.188 0.940 0.967 0.806 – 0.268 0.605 0.268 –
0.258 0.417 0.417 0.258 0.258 0.379 0.258 0.258 3.403 0.417 0.471 – 0.258 0.258 0.258 –
0.805 1.302 1.302 0.805 0.805 1.184 0.805 0.805 10.631 1.302 1.471 – 0.805 0.805 0.805 –
Bioelectromagnetics
An Interlaboratory Comparison Program on ELF Fields Measurements
were multiple cases where warning/action signals of the second method had a satisfactory evaluation (|z| < 2) when using the first method. Additionally, some action signals (|z| 3) derived with the robust algorithm were detected only as warning signals with the first method. It is obvious that the first calculation method (m, s) was not particularly reliable, since remote values (i.e., results that deviate significantly from the rest of the measurements) are taken into account when calculating estimators. Resulting standard deviation s of results in most test levels was higher compared to robust estimator and, therefore, in most cases, z-scores of the first method tended to be smaller (Tables 1 and 2). As expected, zscores calculated with the robust iterative algorithm were a more stable indicator, less affected by the presence of distant values. Comparison With First Round of ILC The lessons learned from the first round concerning necessary actions needed for improvement of the overall procedure were implemented in this second scheme, which therefore proved to be very effective. The main alterations in measurement process in this new round are, at first implementation of measurement of electric/magnetic field values at different distances from a fixed source rather than at different voltage/current levels and, secondly, the addition of the third measurement phase which includes a frequency measurement. In this round, particular emphasis was given to maintaining stability of measured fields by keeping both voltage and current sources as stable as possible. This was achieved for each of the three measurement stages by using a voltage stabilizer, a current clamp in combination with a digital multimeter for continuous monitoring of the current, and use of appropriate software, which automatically adjusts coil current, respectively. Absence of other major spectral components at the test site allowed organizers to ask laboratories to deliver results in only one of two measurement modes (either broadband or band pass filter modes as provided by measuring equipment). This enabled a significant increase in number of participants in this round, considering time limitation on the current transformer and required time for execution of measurements. The influence of harmonics in the network voltage was reduced with the use of voltage stabilizers in the first two measurement stages and at the third measurement stage by use of an AC power source free of harmonics.
549
All participants of the first ELF ILC scheme (four accredited laboratories consisting of five measuring teams) were also present in this second round. A comparison between rounds of the scheme using an aggregated performance statistic would not be suitable because both test levels and measuring teams, defined as the unique combination of equipment and operator, were not identical. An attempt to examine only influence of equipment by counting number of test levels where | z| > 2 shows the majority of laboratories that participated in the first round (where no one had been evaluated with a |z| > 2) received warning and/or action signals in this second round. This is a conclusion of high importance as it proves necessity of the scheme, along with the fact that even accredited laboratories with recently calibrated instruments received |z| > 3. Ioannis Ztoupis,Eleni Nicolopoulou,Ioannis Gonos*, Ioannis Stathopulos HighVoltage Laboratory, School of Electrical and Computer Engineering, National Technical University of Athens, Athens, Greece
Efthymios Karabetsos Non-Ionizing Radiation Office, Greek Atomic Energy Commission (EEAE), Agia Paraskevi, Greece
REFERENCES ESYD. 2011. ESYD policy relevant to PT Schemes and Interlaboratory Comparisons. Athens, Greece: Hellenic Accreditation System S.A. PDI/02/01. ICNIRP. 2010. Guidelines for limiting exposure to time varying electric and magnetic fields (1 Hz–100kHz). Health Phys 99:818–836. IEC. 1998. Measurement of low-frequency magnetic and electric fields with regard to exposure of human beings—Special requirements for instruments and guidance for measurements. Geneva, Switzerland: IEC. 61786. IEC. 2009. Electric and magnetic field levels generated by AC power systems—Measurement procedures with regard to public exposure. Geneva, Switzerland: IEC. 62110. IEEE. 1994. (Revision of IEEE 644-1987). IEEE standard procedures for measurement of power frequency electric and magnetic fields from AC power lines. New York, NY: IEEE. 644. ILAC. 2000. Guidelines for the requirements for the competence of providers of proficiency testing schemes. Silverwater, Australia: International Laboratory Accreditation Cooperation. G13. ILAC. 2005. ILAC Policy for participation in national and international proficiency testing activities. Silverwater, Australia: International Laboratory Accreditation Cooperation. P9. Bioelectromagnetics
550
Ztoupis et al.
ISO. 2005. Statistical methods for use in proficiency testing by interlaboratory comparisons. Geneva, Switzerland: International Organization for Standardization. 13528. ISO/IEC. 2005. General requirements for the competence of testing and calibration laboratories. Geneva, Switzerland: International Electrotechnical Commission. 17025. ISO/IEC. 2010. Conformity assessment-General requirements for proficiency testing. Geneva, Switzerland: International Electrotechnical Commission. 17043.
Bioelectromagnetics
ISO/IEC. 1997. Proficiency testing by interlaboratory comparisons Part 1: Development and operation of proficiency testing schemes. International Electrotechnical Commission: Geneva, Switzerland. 43–1. Nicolopoulou EP, Gonos IF, Stathopulos IA, Karabetsos E. 2012. Two interlaboratory comparison programs on EMF measurements performed in Greece. IEEE Electromagnetic Compatibility Magazine Vol. 1, Q.2: 50–59.
