Professional Documents
Culture Documents
net/publication/260541900
CITATIONS READS
20 1,100
4 authors:
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Mark Kennedy on 16 December 2014.
O PTIMAL induction system electrical design including both air-gap, caused by the presence of a work piece.
the coil and power supply, requires accurate estimates of
Φw
the total system impedance, including the equivalent Φg
resistance of the work piece, the total equivalent inductance of Φc
the work-piece/coil [1], and the coil current required to induce
a given power (heating rate) in the work piece. The resonant sc
frequency for a given capacitor bank or the required pc dc
where δc is the electromagnetic penetration depth into the coil: The exclusion of the coil’s magnetic field from most of the
area occupied by the work piece (and in a similar fashion from
1 most of the coil itself) at high frequency, concentrates the
c , (8) lines of flux into the reduced area of the air-gap, resulting in
o r c f
an increased flux density as can be predicted with (2).
Equation (2) assumes that the total lines of flux produced by a
where σc is the electrical conductivity of the coil conductors, given coil at a fixed current are not changed by the reduction
μr the relative magnetic permeability with recommended in coil area (e.g. due to the increased internal magnetic
values of ~1.0000 for copper and aluminum and f the reluctance). This factor assumes negligible penetration of the
frequency of the applied coil current. Dc+δc can be assumed magnetic field into both the work piece and the coil in a
equal to the average coil diameter taken on the center-line of manner similar to Northrup’s (1). Equation (2) can be directly
the conductors at low frequency (i.e. the classical assumption compared to the square root of (1), as will be shown later. For
for the value of Dcs). Low frequency is defined as the complete coil ‘filling’, i.e. Dw = Dc, (2) becomes precisely
frequency where the electromagnetic penetration depth into equal to 1 (rather than 0.8½ or ~0.95 from Northrup’s
the coil conductor, δc is greater than or equal to the thickness, equation). For an empty coil, (2) reduces to the Nagaoka
tc for square or rectangular conductors or the diameter of the coefficient. Equation (2) assumes a uniform cylindrical shape,
coil conductor for round conductors, dc. but similar equations can be derived for alternate uniform
At high frequency, where δc ≤ tc or dc, it is necessary to shapes. Volumetric corrections can also be found for shapes
evaluate (4) at the ‘equivalent’ current sheet diameter by with non-uniform vertical profiles.
substituting Dc+δc into (4). Induction systems are typically Vaughan and Williamson’s assumption of negligible
designed using coils that have a conductor thickness greater electromagnetic penetration into both the coil and the work
than two times the electromagnetic penetration depth and are piece is reasonable for high frequency operation with large
hence usually operated at ‘high frequency’ according to this coils and work pieces. The present author’s postulated that at
definition. The experimental program described in the present
low frequencies, (2) could be improved using frequency
article was conducted under low frequency conditions.
correction as indicated below:
Equation (7) produces minor differences in the third
decimal place when compared to numerical solutions and
D 2
D 2
errors of less than half of one percent for coils of dimensions k N* k N 1 w w w . (9)
w
IV. INDUCED REAL AND REACTIVE POWER IN THE WORK PIECE 2(bei ' w ber w - bei w ber ' w )
and ( w ) , (18)
The Poynting vector [18] gives the energy flux at the ber 2 ( w ) bei 2 ( w )
surface of the work piece [W/m2], i.e.:
where ber, ber’, bei and bei’ are the real and imaginary parts
S Es H s . (10) of the modified zero order Kelvin Bessel functions and their
derivatives, the solutions to which can be found using
The real power, Pw, and the reactive power, Qw, induced in numerical solvers [14] or look-up tables [19]. The solutions to
the work piece, can both be derived starting from (10): (17) and (18) have been plotted in Fig. 2 as a function of the
dimensionless reference depth ξw.
Pw Dw l Re Es H z , (11) 1.2 100
1 90
and Qw Dw l Im Es H z , (12)
where L*coil is the inductance due to the flux within the turns The inductance of the coil L*coil in the presence of a work
of the coil itself in the presence of a work piece, and Ac is the piece can be found from (21) and (22), i.e.:
effective area occupied by both the coil turns and the coil flux
given by:
k N* 0 r N c2 Dc c Dc
2 2
L*
coil . (26)
lc 4
Dc c Dc2
2
Ac . (22)
4 When the diameter of the work piece, Dw is zero, k*N becomes
equal to kN, and (26) can then also be used to calculate Lcoil.
Equation (20) does not include the correction factor (18) for
the phase angle between the magnetic and electric fields in the VI. EXPERIMENTAL
coil, which is simply assumed to be 1. The electromagnetic
An experimental program was conducted to determine the
penetration depth in (20) is also assumed to be much smaller
than the coil conductor thickness or the diameter (δc<<tc or accuracy of (9) for work pieces and coils of three different
dc). Equation (20) must therefore be acknowledged as being diameters, two different lengths, and two different electrical
only an approximation. conductivities. Data for the coils and work pieces are
The area used with (21) is not defined except using the summarized in Tables I and II and a representative picture is
empirical equation (22), making (21) also semi-empirical. shown in Fig. 3.
With some manipulation it can be shown that (20) and (21)- The impact of the reduced air-gap on the magnetic flux
(22), differ only by a factor of k*N, provided that Dc>>δc and density was measured directly using Hall Effect probes, the
δc ≤ tc. If k*N is taken equal to 1.0, then (20) and (21) are heating rates of the work pieces were determined by calorific
equivalent at ‘high frequency’. measurements using the water cooled work pieces. The total
At very high frequency, where δc<<tc, the coil reactive reactive power of the system (coil and work piece) was
power Qc can be safely assumed to be negligible. Relatively measured electrically. A schematic of the experimental
large discrepancies in Qc at high frequency, therefore have apparatus is shown in Fig. 4, providing construction details.
little impact on Qtotal. At low frequency (δc ≥ tc), Qc is not Power was supplied by a 45 kVA 50 Hz transformer, which
negligible. Equations (21) and (22) then give more reasonable delivered a sinusoidal current with minimal harmonics (as
results than (20), and (21) will therefore be used in the observed using a power quality analyzer). Power
calculations presented in the results and discussion sections of measurements (V, kA, kW, kVA, kVAr and power factor)
the present work. When δc ≥ tc, tc should be substituted for δc were taken using a power quality analyzer from Fluke, USA
in (22). For round tubing dc should be used instead of tc. (Fluke 43B), with a power measurement resolution of
A new equation is required to determine the reactive power 0.1 kW. Coil current measurements were made with an
associated with the flux in the air-gap of the coil, Qg: inductive current probe from Fluke, USA (i1000S), with an
accuracy of ±1% and a resolution of 1 A.
Qg I c 2 2 f Lg , (23) Measurements of the magnetic flux density were taken
using a Pacific Scientific OECO, USA (F.W. Bell model 6010
where Lg is the inductance of the air gap, and: Gauss meter). Standardized axial Hall probes, with a
measuring error of < ±1% for AC magnetic fields were used.
D 2 D 2 k N* Work piece IACS electrical conductivities were measured
Lg (Lc Lcoil ) c 2 w , (24) using a General Electric Inspection Technologies, UK
Dc kN (AutoSigma 3000) conductivity analyzer to within ±0.5%.
100% IACS conductivity is equivalent to 58 MS/m [21].
where Lc is the inductance of the empty ‘air-core’ coil and Work pieces were insulated to minimize heat losses, using
which can be found using (4) or (6), or measured
ultra-low thermal conductivity ceramic wool blankets, as
experimentally and Lcoil the inductance due to the flux within
shown in Fig. 4. Imbedded type K thermocouples (±1.5 K)
the turns of the coil itself in the absence of a work piece. The
provided reference temperatures to calculate the aluminum
use of calculated values based on (4) or (6), may result in a
6
electrical conductivity at elevated temperature for use in the temperature at 0.1 oC resolution, such that the precision error
analytical and FEM estimations. New equations to estimate in the delta-temperature was in the order of ±0.05 oC,
the electrical conductivity of aluminum alloys as a function of compared to a typical measured ΔT of 2 to 5 oC. The water
temperature, based on a single representative room flow rate was determined using data logged gain-in-weight
temperature reading, have been developed and presented and time. The total weight difference, over the period of each
elsewhere [23]. power reading, was then used to calculate the average flow,
which proved to have an error of <0.1%.
Fig. 3. Experimental work pieces taken from left to right, #1 to #3, and the 32
turn ’long’ induction coil #4 (132 mm average diameter, 218 mm long, and
(Dc+δc) / lc = 0.60). See Tables I and II for more details.
TABLE I
WORK PIECE DIMENSIONS AND
EXPERIMENTALCOMBINATIONS WITH COILS
(COIL #1 WITH WORK-PIECE #1 IS SHOWN AS “1-1” ETC).
Work Pieces 1 2 3
Alloy A356 6060 6060
Diameter, mm 75.0 95.0 95.0 Fig. 4. Schematic diagram of the water-cooled and highly insulated billet
Length, mm 130.0 130.0 260.0 heating apparatus.
Measured IACS Electrical Conductivity at 293 K, % 48.4 56.2 53.4
Penetration depth δw (mm) at 50 Hz and 293 K from (8) 13.43 12.47 12.79 VII. SUMMARY OF EXPERIMENTAL RESULTS
ξ w from (16) 3.948 5.388 5.252
φ (ξ w ) from (17) 0.8244 0.8640 0.8607 Eleven separate experiments and duplicates, were executed
Ψ(ξ w ) from (18) 1.0109 1.0044 1.0044 using a ‘high-low’ experimental design (22 sets of data).
Coil 1 1-1 1-2 Duplicates are indicated by A’s and B’s in the detailed
Coil 2 2-1 2-2
Coil 3 3-1 3-2
experimental data summarized in the Appendix Tables I-III.
Coil 4 4-3 Calorific heating results are presented in Fig. 5, and
compared with the results of the 2D axial symmetric FEM
TABLE II model (the validation of which is described in detail elsewhere
EXPERIMENTAL COILS [23, 24]), and the analytical solution using (7), (9) and (14).
Short Short Short Long Detailed results are given in Appendix Tables I-II.
Coils Coil Coil Coil Coil
1 2 3 4
Total measured system reactive power is plotted in Fig. 6
Average Diameter, mm 132 142 155 132 and compared with the results of predictions from (15), (21),
Height, mm 106 109 108 218
Diameter to Height ratio 1.24 1.30 1.44 0.60 and (23) and (24). See Appendix Tables I-III for details.
Number of Turns 16 16 16 32 Measurements were also taken of the z-component of the
Short Coil Correction Factor k N from (5) 0.641 0.630 0.607 0.786
Electrically Determined IACS Conductivity, % 80 80 80 80 magnetic flux density of both ‘air-core’ coils and coils with
Penetration depth δ c (mm) from (8) at 50 Hz and 293 K 10.45 10.45 10.45 10.45 work pieces, in order to verify the FEM models and determine
Total Length of Leads, cm 66 24 20 58
if the magnetic fields behaved in accordance with (9).
Coil Resistance at Maximum Current Including Leads, Ω 0.01047 0.01051 0.01157 0.02086
Coil Inductance, μH 26.38 29.08 33.39 61.93
All experimental results have been rendered dimensionless
Coil Inductance calculated using (4) and (7), μH 26.48 29.43 34.18 63.05 (and unaffected by minor experimental current variations) by
Coil Inductance calculated using (4) and (7) and *, μH 25.49 28.40 33.05 61.09 dividing by (3). Representative results from ‘short’ Coil #1 are
*
Modified Nagaoka Coefficient k N for Work Piece 1 from
Equation (9)
0.720 0.700 0.669 presented in Fig. 7 for the ‘air-core’ case on the centerline,
*
Modified Nagaoka Coefficient k N for Work Piece 2 from and at a number of different radii. Experimental data are
0.783 0.755 0.718
Equation (9)
*
compared against the analytical Biot-Savart solution evaluated
Modified Nagaoka Coefficient k N for Work Piece 3 from
Equation (9)
0.870 at discrete intervals, and the FEM model on the centerline. At
* Using Rosa’s round wire correction formula [22]. the other radii, experimental results (shown by discrete
symbols) are compared to FEM results, which are drawn using
Water temperatures were measured using Type K green lines. Duplicate results are provided to show
thermocouples ‘matched’ to give identical readings at room experimental reproducibility.
7
2250
Condition Condition Condition Condition Condition Condition Condition Condition Condition Condition Condition
1 2 3 4 5 6 7 8 9 10 11
In Fig. 8, the magnetic flux density measurements for Coil
#1 at 40.5 mm radius are shown, with and without Work-
Heating Rate (Power), W
0.8
750 k * N = 0.72
250 0.7
1-1 1-1 1-2 1-2 2-1 2-1 2-1 2-1 2-2 2-2 2-2 2-2 3-1 3-1 3-1 3-1 3-2 3-2 4-3 4-3 4-3 4-3 k N = 0.64
0.65
Coil-Work Piece
Calorific FEM Equation (14)
0.6
0.55
Fig. 5. Measured calorific heating rate (power) comparison with 1D estimates,
0.5
using the frequency modified short coil correction factor (9) and analytical FEM 40.5 air-core 40.5 air-core
heating predictions of (14), and 2D axial symmetric FEM. Coil #1 with Work- 0.45
FEM 40.5 WP #1 40.5 WP #1
Piece #1 is shown as “1-1” etc.
0.4
0.00 0.20 0.40 0.60 0.80 1.00
11
Condition Condition Condition Condition Condition Condition Condition Condition Condition Condition Condition Dimensionless Coil Length, x
1 2 3 4 5 6 7 8 9 10 11
10
Fig. 8. Dimensionless flux density (z-component) as a function of the
Reactive Power, kVAr
9
dimensionless coil length for ‘short’ Coil #1 with ‘air-core’, and with Work-
8 Piece #1. Obtained values are compare with measured and FEM results at 40.5
7 mm radius with and without Work-Piece #1 [23].
6
VIII. DISCUSSION
5
4
Differences between the calorifically measured work piece
3
heating rate and the value calculated using (9) and (14) are
1-1 1-1 1-2 1-2 2-1 2-1 2-1 2-1 2-2 2-2 2-2 2-2 3-1 3-1 3-1 3-1 3-2 3-2 4-3 4-3 4-3 4-3 shown in Fig. 5 to be from <1% for Coil #4 with Work-Piece
Coil-Work Piece
#3 to <10% for Coil #2 with Work-Piece #2, with a typical
Experimental Equations (15), (21), (23) with (24) difference of ~5% for all coils and work pieces. The
magnitudes of the differences in the work piece power depend
Fig. 6. Measured reactive power comparison with 1D analytical equations on the work piece and on the coil geometries, as would be
(15), (21) and, (23) with (24). Coil #1 with Work-Piece #1 is shown as “1-1”
etc. expected when using a 1D semi-empirical correction factor,
such as the modified Nagaoka coefficient (9). The longer Coil
0.8 #4 and larger diameter Work-Pieces #2 and #3, yielded lower
0.75 differences, i.e. errors were minimized as the modified
Dimensionless flux density, Bz,r / B∞
k N = 0.64
0.7
Nagaoka coefficient approached a value of 1, as would also be
expected from an empirical correction factor. Further analysis
0.65
on work piece heating has been presented elsewhere [25].
0.6 The measured total coil/air-gap/work-piece reactive power
0.55 is compared with the sum of the analytical predictions found
0.5
using (15), (21), as well as (23) and (24) in Fig. 6. Results
indicate minor differences in the total from 0.4 % to 1.4%,
0.45
with a typical difference of ~0.8%. The break down of the
0.4 components of the total reactive power, and detailed
0 0.2 0.4 0.6 0.8 1
Dimensionless coil length, x
experimental data, can be found in Appendix Table III. The
measured coil ‘air-core’ inductances were used to calculate
the air-gap values given in Appendix Table III, using (23) and
FEM Center-Line FEM 40.5
(24), as this procedure gave the minimum error. Using the
FEM 50.5 FEM 59
Center-Line Center-Line duplicate
theoretical inductance values predicted by (4) and (7) would
40.5 duplicate add ~1.5% to the differences.
40.5
50.5 Coil #1 Biot-Savart Law (Dcs/lc)=1.24 The overall agreement in the total system reactive power
59 must be attributed to the accuracy, with which the modified
short coil correction factor estimates the average flux density
Fig. 7. Dimensionless flux density (z-component) as a function of
dimensionless coil length and radial position [mm] for ‘air-core’ Coil #1. of a coil containing a work piece. Given the predominance of
Obtained values are compare with measured, analytical and FEM results [23]. the coil and the air-gap contributions to the total system
reactive power, i.e. ~87% of the total, it would be assumed
8
from (21) to (26) that k*N has an error of < ±1.4% (i.e. the internal cooling area available in the water cooled aluminum
largest error indicated in Appendix Table III). billets, it would not have been possible to adequately cool the
It is not possible to directly determine the accuracy of the work pieces at higher frequency. Virtual ‘experiments’ were
estimates of the individual components of the reactive power therefore conducted to compare model predictions made using
presented in Appendix Table III, i.e. Qc, Qg and Qw, from the the analytical model, with those made by 2D axial symmetric
experimental data. As Qg represents ~75% of the total modeling as a function of operating frequency.
inductance in these experiments, the low overall difference of Equations (4) and (7) have been used to predict both the
<1%, shown in Appendix Table III, implies that Qg has been low and high frequency inductances of the coils. This gives a
estimated to similar accuracy using the measured ‘air-core’ pure model vs. model comparison. Equation (4) has been
inductance values, (24) and (25). Examination of the evaluated at the ‘equivalent current sheet diameter’ at low
similarity of (14) and (15), would suggest that if the real frequency, i.e. Dc+dc, and at high frequency, i.e. Dc+δc.
The analytical and FEM model reactive power predictions
power of the work piece, Pw, can successfully be modeled by
are compared in Tables III and IV for the short Coil #1 with
the use of (14), that (15) should provide an estimate of the
Work-Piece #1, and for the long Coil #4 with Work-Piece #3.
work piece reactive power, Qw, of similar accuracy, i.e. ~1-
Experimental data for 50 Hz is presented for reference.
10% error.
In Fig. 7 the results of the ‘air-core’ axial flux density TABLE III
measurements for Coil #1 are compared with results of both ‘SHORT’ COIL #1, WORK-PIECE #1, AND CONDITION #1
the 2D axial symmetric FEM model, and the results of the (1001.0 A, COIL CONDUCTIVITY 4.3*107 S/m)
FEM
Biot-Savart law on the coil centerline, indicating the expected Reactive
radial and axial variation in the magnetic flux density. All Electrical Analytical Electrical- Power for Analytical-
measured magnetic flux densities were taken with a 6 mm Reactive Reactive Analytical round FEM
diameter probe and the results have therefore been plotted Frequency Power Power Difference wire Difference
(Hz) (VAr) (VAr) (%) (VAr) (%)
allowing for the average 3mm offset. Agreement between the 10 N/A 1.53E+03 1.54E+03 0.8
analytical, FEM and measured values in Fig. 7 is in the order 20 N/A 2.99E+03 2.95E+03 1.4
of ±1-2%. This is however, not the case near the ends of the 35 N/A 5.06E+03 4.91E+03 3.0
50 7.20E+03 7.12E+03 1.09 6.83E+03 4.1
coil, where the non-axial symmetry becomes most apparent, or 500 N/A 6.50E+04 6.22E+04 4.2
near to the coil tubing, where local variations caused by the 5000 N/A 6.10E+05 5.82E+05 4.5
round conductors (and any stacking imperfections) can be 50000 N/A 5.97E+06 5.63E+06 5.6
observed. FEM results for the outermost measurable radius 500000 N/A 5.93E+07 5.56E+07 6.2
Average: 3.7
clearly show that the coil does not behave like a perfect
‘current sheet’, and the local flux density variations caused by TABLE IV
the round tubing can be seen. An extremely tight mesh was ‘LONG’ COIL #4, WORK-PIECE #3, AND CONDITION #11
used to ensure high accuracy during both low and high (892.6 A, COIL CONDUCTIVITY 4.1*107 S/m)
frequency simulations [24]. FEM
Reactive
The close agreement of the FEM predictions with the Biot- Electrical Analytical Electrical- Power for Analytical-
Savart law and measured values, gives strong validation to the Reactive Reactive Analytical round FEM
2D axial symmetric modeling technique used. The details and Frequency Power Power Difference wire Difference
development of the FEM model are discussed elsewhere [23, (Hz) (VAr) (VAr) (%) (VAr) (%)
50 1.05E+04 1.06E+04 1.22 1.01E+04 4.5
24]. All calculations have been performed using the 500 N/A 8.67E+04 8.50E+04 2.0
commercial code of COMSOL® 4.2. 5000 N/A 7.63E+05 7.57E+05 0.7
The Nagaoka coefficient (0.64) is plotted in Fig. 7 for 50000 N/A 7.29E+06 7.23E+06 0.8
reference, and is fortuitously equal to the average of the 500000 N/A 7.18E+07 7.10E+07 1.1
Average: 1.8
measured values for the ‘air-core’ coil.
In Fig. 8 the 2D axial symmetric FEM model results are
The portion of the total reactive power contributed by the
compared with the measurements for Coil #1 with and without
Work-Piece #1, showing a 10% increase in flux density with coil and work pieces falls from ~10% at 50 Hz, to ~1% each
the presence of the work piece. The Nagaoka ‘short coil’ at 50 kHz. At high frequency the contributions of both coil
correction factor from (7), and the modified Nagaoka ‘short and work piece therefore become negligible, and the total
coil’ correction factor from (9) are indicated (0.64 and 0.72 reactive power becomes linear with frequency as predicted by
respectively, or a 12.5% increase). Differences of 0.9-3.6% (23). Errors in either coil or work piece reactive power
are observed comparing the discretely measured experimental estimates are then of little consequence.
values to the equivalent FEM results, with a typical difference Examination of Tables III and IV at either 50 kHz or 500
of 2.0%. kHz, gives an indication of the error in the magnetic field
estimation using (9) relatively independent of errors in the coil
IX. COMPARISON OF ANALYTICAL AND FEM REACTIVE and work piece fluxes. From Table III, one would expect an
POWER MODEL PREDICTIONS VS. FREQUENCY error in the order of 6% for the ‘short’ Coil #1, and from
Heating efficiency increases with frequency for a given Table IV an error in the order of 1% for the ‘long’ Coil #4.
work piece, as shown previous in Fig. 2. With the small The excellent agreement for the ‘long coil’ is not surprising as
9
it is more ‘ideal’ than the ‘short coil’, with a correction factor ACKNOWLEDGMENT
closer to unity and a more homogenous magnetic field in both The present study was carried out as part of the RIRA
the axial and radial directions. (Remelting and Inclusion Refining of Aluminium) project
Comparisons of real power as a function of frequency funded by the Norwegian Research Council (NRC) - BIP
calculated using (14), and using the 2D FEM model, were Project No. 179947/I40. The industrial partners involved in
previously published for conditions identical to Tables III and the project are: Hydro Aluminium AS, SAPA Heat Transfer
IV [25]. The results showed typical differences of 4.3% and AB, Alcoa Norway ANS, Norwegian University of Science
0.8% respectively, and are in good agreement with the and Technology (NTNU) and SINTEF Materials and
reactive power calculations presented here. Chemistry. The funding granted by the industrial partners and
the NRC is gratefully acknowledged.
CONCLUSIONS
The present investigation has shown that (i) the magnetic REFERENCES
field produced by a ‘short’ induction coil is lower in [1] O. Lucia, J. M. Burdío, I. Millán, J. Acero, and D. Puyal, "Load-
magnitude than one would expect from the use of a ‘long coil’ Adaptive Control Algorithm of Half-Bridge Series Resonant Inverter
formula such as (3), and (ii) that the magnetic field is strongly for Domestic Induction Heating," Industrial Electronics, IEEE
influenced by the presence of a work piece and the size of the Transactions on, vol. 56, pp. 3106-3116, 2009.
air-gap. The improved ‘short coil’ correction factor (9), has [2] O. Heaviside, Electrical Papers Volume 1., London: Macmillan & Co.,
been proposed and experimentally validated at 50 Hz, for pp. 353-416, 1892.
application to axial symmetric induction heating. Good [3] E.F. Northrup, "High-Frequency Induction Furnace," US patent
agreement was achieved between measured reactive power, 1,694,792, 1928.
and analytical model estimates at 50 Hz (typically <1% [4] M. W. Kennedy, S. Akhtar, J. A. Bakken, and R. E. Aune, "Review of
difference). Classical Design Methods as Applied to Aluminum Billet Heating with
Analytical and 2D axial symmetric FEM model reactive Induction Coils," EPD Congress 2011, Hoboken: John Wiley & Sons,
power estimates were in good to acceptable agreement at 2011, pp. 707-722.
frequencies from 50 Hz to 500 kHz, with ‘typical’ differences [5] H. Nagaoka, "The inductance coefficients of solenoids," Journal of the
of less than 4% for the ‘short coil’, and 2% for the ‘long coil’, College of Science, vol. 27, pp. 18-33, 1909.
over the entire range of frequencies. [6] C. Burch and N. Davis, An Introduction to the Theory of Eddy-Current
The use of (9) reduces the magnitude of the errors in Heating, London: E. Benn Limited, 1928.
heating and reactive power estimations for the work pieces [7] J. Vaughan and J. Williamson, "Design of Induction-Heating Coils for
from a factor of two in comparison with making a ‘long coil’ Cylindrical Nonmagnetic Loads," Transactions of the American
assumption, i.e. assuming k*N = 1.0 for a ‘short coil’, to a Institute of Electrical Engineers, vol. 64, pp. 587-592, 1945.
‘typical’ 5% error. This also represents a significant [8] R. Baker, "Design and calculation of induction heating coils'," AIEE
improvement over results obtained using Vaughan and Trans, vol. 57, pp. 31-40, 1957.
Williamson’s original equation (2) at low frequency, which [9] C. Tudbury, Basics of Induction Heating Volume 1., New York: John F.
gives a typical error of ~15% at 50 Hz for the cases Rider, 1960.
summarized in Appendix Table II. [10] J. Lavers and P. Biringer, "An Improved Method of Calculating the
The improved ‘short coil’ correction factor can be used to Power Generated in an Inductively Heated Load," IEEE Transactions
estimate the average z-component of the magnetic flux density on Industry Applications, pp. 273-278, 1974.
of a ‘short coil’ containing a work piece, with an error of [11] A. Boadi, Y. Tsuchida, T. Todaka, and M. Enokizono, "Designing of
<1%, based on the observed errors in the reactive power at 50 Suitable Construction of High-Frequency Induction Heating Coil by
Hz. Using Finite-Element Method," Magnetics, IEEE Transactions on, vol.
Model estimates from high frequency imply errors of up to 41, pp. 4048-4050, 2005.
about 6% for a ‘short coil’ with a relatively large air-gap (Coil [12] C. Carretero, O. Lucia, J. Acero, and J. Burdio, "Computational
#1 and Work-Piece #1) when also predicting rather than Modeling of Two Partly-Coupled Coils Supplied by a Double Half-
measuring the high frequency ‘air-core’ inductances. Bridge Resonant Inverter for Induction Heating Appliances," Industrial
The improved accuracy of (9) should allow for more Electronics, IEEE Transactions on, 10.1109/TIE.2012.2202360, pp. 1-
accurate design of axially symmetric heating systems, e.g. 15, [Online] 4 June 2012.
billets and cylindrical furnaces, and metal flow control [13] L. Lorenz, "Ueber die Fortpflanzung der Electricität," Annalen der
systems, e.g. mold – ElectroMagnetic Stirring (EMS) in Physik, vol. 243, pp. 161-193, 1879.
continuous casting of steel, in the absence of FEM [14] R. Weaver. http://electronbunker.ca/DL/NumericalExamples01.ods
simulations. [Online], 23 March, 2013.
It is hoped that the 1D empirical correction factor proposed [15] H. Wheeler, "Simple Inductance Formulas for Radio Coils,"
in the present work, i.e. equation (9), will lead to the Proceedings of the IRE, vol. 16, pp. 1398-1400, 1928.
development of more exact 2D correction factors, or to a [16] D. Knight, "Part 1: Solenoid inductance calculation", October, 2012,
proper theoretically derived correction factor for use with high http://www.g3ynh.info/zdocs/magnetics/Solenoids.pdf, [Online], 23
precision analytical modeling. March, 2013.
10
[17] E. J. Davies, Conduction and Induction Heating. London: Peter APPENDIX TABLE II
SUMMARY OF WORK PIECE HEATING DATA COLLECTED
Peregrinus on behalf of the Institution of Electrical Engineers, pp. 95-
97, 1990. Average Work Work Work Work
Average Aluminum Piece Piece Piece Piece
[18] J. H. Poynting, "On the Transfer of Energy in the Electromagnetic Water Cp Delta T Aluminum Electrical Power Power Power Power
Field," Philosophical Transactions of the Royal Society of London, vol. Flow Water Water Temperature Conductivity Calorific FEM Equation (9) Equation (2)
Condition (g/s) (J/g/K) (K) (K) (MS/m) (W) (W) (W) (W)
175, pp. 343-361, 1884. 1-A 76.1 4.1942 1.99 310.2 26.62 636 623 659 732
[19] N. McLachlan, Bessel Functions for Engineers, Gloucestershire: 1-B 75.8 4.1941 1.99 310.3 26.62 634 623 659 732
2-A 76.7 4.1938 3.03 325.6 29.09 975 976 1035 1167
Clarendon Press, 1955, pp. 215 - 230. 2-B 70.8 4.1937 3.21 324.4 29.21 954 970 1029 1161
[20] N. Stansel, "Induction Heating - Selection of Frequency," American 3-A 68.9 4.1684 1.89 309.2 26.70 544 537 571 629
3-B 69.7 4.1685 1.87 309.1 26.72 543 536 570 627
Institute of Electrical Engineers, Transactions of the, vol. 63, pp. 755- 4-A 23.2 4.1709 5.53 336.6 24.65 536 541 576 637
759, 1944. 4-B 22.6 4.1710 5.66 337.4 24.59 534 542 577 638
5-A 36.4 4.1707 5.36 340.8 27.69 813 840 888 994
[21] Copper Wire Tables Circular No. 31: US Bureau of Standards, 1913. 5-B 35.9 4.1707 5.42 340.9 27.69 811 840 888 994
6-A 68.8 4.1691 2.78 321.2 29.43 798 825 873 974
[22] E. B. Rosa and F. Grover, "Formulas and Tables for the Calculation of
6-B 69.4 4.1691 2.75 321.0 29.46 795 824 872 974
Mutual and Self Induction," Scientific Papers of the Bureau of 7-A 30.8 4.1936 3.19 317.7 26.02 411 399 432 473
7-B 30.6 4.1936 3.17 317.7 26.03 407 398 431 472
Standards, vol. No. 169, pp. 5-231, 1916.
8-A 29.4 4.1930 4.20 327.2 25.31 518 508 549 602
[23] M. W. Kennedy, S. Akhtar, J. A. Bakken, and R. E. Aune, "Analytical 8-B 29.1 4.1931 4.25 327.6 25.28 520 508 549 602
9-A 75.2 4.1946 2.04 310.5 30.63 642 618 658 729
and FEM Modeling of Aluminum Billet Induction Heating with 9-B 75.2 4.1947 2.04 310.4 30.64 643 617 657 728
Experimental Verification,",Light Metals 2012, Hoboken: John Wiley 10-A 75.2 4.1930 5.98 323.8 27.95 1884 1888 1911 2042
10-B 75.0 4.1931 6.03 323.8 27.95 1894 1881 1903 2034
& Sons, 2012, pp. 269-275. 11-A 74.0 4.1943 2.40 299.6 30.29 746 713 721 769
[24] M. W. Kennedy, S. Akhtar, J. A. Bakken, and R. E. Aune, “Analytical 11-B 73.8 4.1943 2.37 299.4 30.31 736 713 722 770
Average Difference from Calorific Results: 2.3 5.4 15.6
and Experimental Validation of Electromagnetic Simulations Using
COMSOL®, re Inductance, Induction Heating and Magnetic Fields,”
COMSOL Conference 2011 - Proceedings CD, Burlington, MA: APPENDIX TABLE III
COMSOL, pp. 1-9, 2011. COMPARISON OF MEASURED AND
CALCULATED REACTIVE POWER
[25] M. W. Kennedy, S. Akhtar, J. A. Bakken, and R. E. Aune, "Improved
Air Work Total
Short Coil Corection Factor for the Induction Heating of Billets," 3rd Total Coil Gap Piece Calculated Absolute
International Symposium on High-Temperature Metallurgical Measured Reactive Reactive Reactive Reactive Difference
Reactive Power Power Power Power Measured-
Processing, Hoboken: John Wiley & Sons, 2012, pp. 373-382. Power Eq. (21) Eq.'s (23) and (24) Eq. (15) Equation (19) Calculated
Condition (kVAr) (VAr) (VAr) (VAr) (VAr) (%)
1-A 7.20 0.83 5.45 0.82 7.10 1.35
APPENDIX TABLE I 1-B 7.20 0.83 5.45 0.82 7.10 1.40
DETAILED EXPERIMENTAL DATA 2-A 6.34 0.95 4.14 1.21 6.31 0.51
2-B 6.30 0.95 4.12 1.21 6.28 0.36
Average Average 3-A 7.70 0.80 6.12 0.71 7.63 0.95
3-B 7.63 0.80 6.10 0.71 7.61 0.29
Total Average Total Average 4-A 7.60 0.79 6.00 0.73 7.52 1.07
Work Average Average Apparent Total Reactive Measured 4-B 7.60 0.79 6.01 0.73 7.52 1.01
5-A 6.90 0.90 5.03 1.05 6.97 1.02
Coil Piece Current Voltage Power Power Power Power
5-B 6.90 0.90 5.03 1.05 6.97 1.02
Condition (#) (#) (A) (V) (kVA) (kW) (kVAr) Factor 6-A 6.90 0.90 5.05 1.02 6.98 1.16
6-B 6.90 0.90 5.05 1.02 6.98 1.16
1-A 1 1 1001.3 13.25 13.30 11.13 7.20 0.84
7-A 7.38 0.69 6.10 0.54 7.32 0.81
1-B 1 1 1001.0 13.25 13.30 11.10 7.20 0.84 7-B 7.30 0.69 6.08 0.53 7.30 0.06
2-A 1 2 1028.0 13.30 13.71 12.16 6.34 0.89 8-A 9.30 0.87 7.69 0.69 9.24 0.66
8-B 9.30 0.87 7.68 0.69 9.24 0.70
2-B 1 2 1025.4 13.25 13.60 12.03 6.30 0.89 9-A 7.10 0.80 5.61 0.77 7.18 1.12
3-A 2 1 973.3 13.36 13.00 10.50 7.70 0.81 9-B 7.10 0.80 5.60 0.77 7.17 0.96
10-A 10.50 1.56 6.66 2.25 10.47 0.28
3-B 2 1 972.3 13.35 13.00 10.50 7.63 0.81
10-B 10.45 1.55 6.63 2.24 10.43 0.19
4-A 2 1 965.2 13.29 12.82 10.40 7.60 0.81 11-A 4.03 0.61 2.60 0.85 4.05 0.66
4-B 2 1 965.5 13.30 12.85 10.38 7.60 0.81 11-B 4.00 0.61 2.60 0.85 4.05 1.37
Average 7.35 0.88 5.49 0.96 7.34 0.82
5-A 2 2 992.0 13.32 13.20 11.30 6.90 0.85 Fraction of total reactive power: 12.0 74.7 13.1
5-B 2 2 992.0 13.31 13.20 11.30 6.90 0.85
6-A 2 2 994.0 13.30 13.23 11.30 6.90 0.85 In the appendix the data are presented by ‘Conditions’ where
6-B 2 2 994.0 13.29 13.20 11.30 6.90 0.85
A and B are replicates of each experimental data set. “1-A”
7-A 3 1 876.0 13.42 11.77 9.20 7.38 0.78
and “1-B” are therefore replicates of Condition #1, etc. Coil
7-B 3 1 874.3 13.40 11.70 9.18 7.30 0.78
8-A 3 1 983.6 15.24 15.00 11.80 9.30 0.79
numbers and work piece numbers for each Condition are only
8-B 3 1 983.4 15.25 15.00 11.80 9.30 0.79 given in Appendix Table I.
9-A 3 2 909.5 13.54 12.30 10.10 7.10 0.82
9-B 3 2 908.8 13.52 12.30 10.10 7.10 0.82
10-A 4 3 893.5 24.10 21.53 18.80 10.50 0.87
10-B 4 3 891.8 24.05 21.45 18.75 10.45 0.87
11-A 4 3 557.8 14.12 7.90 6.80 4.03 0.86
11-B 4 3 558.0 14.12 7.90 6.80 4.00 0.86