Beruflich Dokumente
Kultur Dokumente
net/publication/329420448
Correction to: Sensitivity study of hardness profile of 4340 steel disc hardened
by induction according to machine parameters and geometrical factors
CITATIONS READS
0 81
4 authors:
Some of the authors of this publication are also working on these related projects:
Combining approaches by projects and competencies towards successful student’s performances using project management - Application to CAD undergraduate
course View project
Relevance of collaborative and competitive approaches used to increase projects performance - mechanical engineering course View project
All content following this page was uploaded by Noureddine Barka on 17 December 2018.
Please note
If you would like to know when your article has been published online, take advantage
of our free alert service. For registration and further information, go to:
http://www.link.springer.com.
Due to the electronic nature of the procedure, the manuscript and the original figures
will only be returned to you on special request. When you return your corrections,
please inform us, if you would like to have these documents returned.
AUTHOR'S PROOF!
1 Article Title Sensitivity study of hardness profile of 4340 steel disc hardened by
induction by the variation of machine parameters and geometrical factors
2 Article Sub- Title
3 Article Copyright - Springer-Verlag London Ltd., part of Springer Nature 2018
Year (This will be the copyright line in the final PDF)
4 Journal Name The International Journal of Advanced Manufacturing Technology
5 Family Name Barka
6 Particle
7 Given Name Noureddine
8 Corresponding Suffix
9 Author Organization Université du Québec à Rimouski
10 Division Département de mathématiques, d’informatique et de génie
11 Address Rimouski G5L 3A1, Québec, Canada
12 e-mail noureddine_barka@uqar.ca
13 Family Name Khalifa
14 Particle
15 Given Name Mohamed
16 Suffix
Author
17 Organization Université du Québec à Rimouski
18 Division Département de mathématiques, d’informatique et de génie
19 Address Rimouski G5L 3A1, Québec, Canada
20 e-mail mohamed.khalifa@uqar.ca
21 Family Name Brousseau
22 Particle
23 Given Name Jean
24 Suffix
Author
25 Organization Université du Québec à Rimouski
26 Division Département de mathématiques, d’informatique et de génie
27 Address Rimouski G5L 3A1, Québec, Canada
28 e-mail jean_brousseau@uqar.ca
29 Family Name Bocher
30 Author Particle
31 Given Name Philippe
AUTHOR'S PROOF!
32 Suffix
33 Organization École de technologie supérieure
34 Division Département de génie mécanique
35 Address Montréal H3C 1K3, Québec, Canada
36 e-mail philippe.bocher@etsmtl.ca
37 Received 25 July 2018
38 Schedule Revised
39 Accepted 12 October 2018
40 Abstract An adequate induction heat treatment operation should always identify,
previously, the probable temperature distribution and hardness profile behavior
in specific regions of the treated component according to heating process
parameters. This paper presents an analysis of the effects of some geometrical
factor related to the component and the coil and machine parameters on
temperature distribution and case depth of an AISI 4340 low-alloy steel disc
heated by induction. A Comsol model was created, defined as a group of
process parameters followed by a mesh study. A Matlab algorithm coupled to
the simulation model was designed to handle a large number of simulations
and export temperature profile data. The case depth is then interpolated from
collected temperature data and a statistical analysis was developed to create the
hardness prediction model. The experimental tests conducted under the same
process parameters support the numerical model results and approve the
simulation, the prediction modeling, and the statistical study.
41 Keywords separated Induction heating - Simulation - Experimental validation - Sensitive study -
by ' - ' 4340 steel disc
42 Foot note
information
AUTHOR'S PROOF! JrnlID 170_ArtID 2892_Proof# 1 - 23/10/2018
4
5 Sensitivity study of hardness profile of 4340 steel disc hardened
6 by induction by the variation of machine parameters
7 and geometrical factors
Q1 8
Q2 Mohamed Khalifa 1 & Noureddine Barka 1 & Jean Brousseau 1 & Philippe Bocher 2
9
10 Received: 25 July 2018 / Accepted: 12 October 2018
11 # Springer-Verlag London Ltd., part of Springer Nature 2018
F
12 Abstract
O
13 An adequate induction heat treatment operation should always identify, previously, the probable temperature distribution and
O
14 hardness profile behavior in specific regions of the treated component according to heating process parameters. This paper presents
15 an analysis of the effects of some geometrical factor related to the component and the coil and machine parameters on temperature
PR
Q3 16 distribution and case depth of an AISI 4340 low-alloy steel disc heated by induction. A Comsol model was created, defined as a
17 group of process parameters followed by a mesh study. A Matlab algorithm coupled to the simulation model was designed to handle
18 a large number of simulations and export temperature profile data. The case depth is then interpolated from collected temperature data
19
D
and a statistical analysis was developed to create the hardness prediction model. The experimental tests conducted under the same
20 process parameters support the numerical model results and approve the simulation, the prediction modeling, and the statistical study.
TE
21 Keywords Induction heating . Simulation . Experimental validation . Sensitive study . 4340 steel disc
EC
22
R
24 For many years, induction heating has shown its success in the currents with rapid cooling (quenching) to increase the surface 32
O
25 metallurgical industry, his remarkable productivity, energy ef- hardness of the steel [4–6]. The process consists of creating 33
26 ficiency, and capacity to manipulate complex parts and vari- high heat intensity at concentrated region on the surface of the 34
C
27 ous metal shapes has made it a popular tool in many industrial metal part. An alternative current characterized by its intensity 35
N
28 heat treatment applications like melting, injection molding, and frequency is supplied to an inductor coil. A magnetic 36
29 brazing, and surface hardening [1–3]. Steel hardening by field, which generated immediately inside the inductor coil, 37
U
8 mm
TM
13], by simulation with validation [14–16], and experimenta- 76
tion with planning strategy [17–20]. These studies are carry out 77
TE2
to analyze various mechanical and thermal effect on a wide and 78
Gap different part shape, using electromagnetic induction process, in 79
Width
F
54 power density (PM), and duration of the heating process (tH) of combination parameters chosen by several methods includ- 84
O
55 while geometrical parameters concern the shape, the size, di- ing Taguchi method and synthesized with statistical model in- 85
56 mensions, and relative positions of the treated object with an cluding regression and ANOVA techniques with verification by 86
O
57 induction coil [1, 10]. The most important geometrical and confirmation tests. The chosen combination parameters include 87
PR
58 machine parameters to consider in any induction heat applica- either machine parameters, like power, heating time, and speed 88
59 tion are the size of the treated object and its position beside the [11, 12, 15, 20], either machine and quenching parameters [18, 89
60 induction coil, where the most important machine parameters 19], either machine parameters and metallurgical factors [17] or 90
61 D
are current frequency, current density, and heating time [10, 11]. only geometrical part versus induction coil parameters [14]. 91
62 The study of induction heat treatment process has several This literature review demoted that there are no previous works 92
TE
63 difficulties at the simulation and experimental level. There is a that have combined both simulation and experimentation in 93
64 difficulty in managing the complex electromagnetic and statistical analyzes that include studying both of the machine 94
65 95
EC
heating physics that includes several parameters and coupled parameters with the main geometric factors of the process.
66 physical phenomena. Moreover, material property data are not However, the aim of this study is to develop a model that 96
67 accurate and vary dissimilarly with the temperature. Measuring describes the hardness profile as a function of both geometrical 97
R
68 adequately the temperature and the machine current during and machine parameters based on a L81 orthogonal array. First, 98
69 heating process is a hard and challenging experimental task a 2D asymmetric model was developed, mesh studied, and then 99
R
70 due to the quick heating rate. Researchers have proposed ex- validated by experimental data. Second, a Matlab algorithm 100
O
71 perimental and statistical methods that may be relevant to solve was developed and coupled with Comsol to carry out the sim- 101
72 such problems. To make a good statistical study, it is necessary ulations with different set of parameter’s configurations. 102
C
73 to choose the right parameters, to conduct well the experimental Finally, prediction equations of temperature profile and hard- 103
N
74 tests and to analyze adequately the obtained results. There is ness depth were obtained by statistical analysis and partial ex- 104
75 105
U
researchers that proposed this approach by simulation [10, 12, perimental validation of extracted data.
2 Formulation 106
Parameters PM tH WP GS fr t2:2
Level 72.6 0.5 6.5 2.5 200 t2:3
Unit kW s mm mm kHz t2:4
Fig. 2 Temperature evolution according machine power
AUTHOR'S PROOF! JrnlID 170_ArtID 2892_Proof# 1 - 23/10/2018
B¼∇ A ð8Þ
where A is the magnetic vector potential. Since the field is time- 143
144
harmonic, assumption of harmonically oscillating currents with 145
single frequency make possible to write Eq. 2 in complex form 146
such that 147
∇ E ¼ jω ∇ A ð10Þ 153
149
148
∇ ðE−jωAÞ ¼ 0 ð11Þ 151
152
156
The curl is zero, the term E − jωA will be equal to the gra- 157
155
154
dient of a scalar function φ such that 158
F
113 161
O
∂B After multiplying with σ and inserting Eq. 5, Eq. 10 could 162
160
159
0
112
111
116 Faraday s law ∇ E ¼− ð2Þ be written as: 163
O
∂t
0
119 Gauss s law for magnetism ∇ :B ¼ 0 ð3Þ
PR
115
114 J ¼ −σ∇φ−jσωA ð13Þ
0
166
117
118 Ampere s law ∇ H ¼ J ð4Þ The current density could be written as the sum of an in- 167
165
164
122
duced current Ji and an imposed external source current in the 168
123
121
120 D
With E, D, H, and B describes respectively the electric field
induction coil, J0 such that: 169
124 intensity (V/m), electric flux density (C/m2), magnetic field
TE
125 intensity (A/m), and magnetic flux density (T). ρcharge is the J i ¼ −jσωA ð14Þ 172
126 electric charge density (C/m3) and J is the conduction current
J e ¼ −σ∇φ ð15Þ
127 density (A/m2) and could be expressed by Ohm’s law: 170
171
EC
175
J ¼ σE ð5Þ Inserting Eqs. 7, 8, and 11 in Eq. (4) gives 176
174
173
130
R
1
131
129
128 E and D, H, and B are related using Eqs. 6 and 7. ð∇ ∇ AÞ ¼ J ð16Þ 179
μ
R
134 D ¼ εE ð6Þ 1
O
138
136
135 With σ, ε, and μ represents, respectively, the electrical con- The resolution of Eq. 14 made it possible to determine the 183
181
180
N
139 ductivity (S/m), permittivity (F/m), and magnetic permeability current density as a function of magnetic vector potential A 184
185
U
140 (H/m) of the material. These parameters are temperature depen- which in turn could be deduced by resolving Eq. 17. The
141 dent and the material is considered to be homogenous. Since B amount of heat generated inside the metal part due to joule 186
142 satisfies a zero divergence condition, equation becomes heating is evaluated using the following equation. 187
F
2 where hc is the convection coefficient, ε is the emissivity, and 207
208
O
J
Qind ðW Þ ¼ ∫v dV ð18Þ σs is the Stefan-Boltzmann constant. 209
σ
O
190
191
PR
189
188 The heat generated by induction is introduced into the heat
192 equation in order to calculate the temperature distribution in
3 Simulation model 210
193 the part. In induction heating, the heat transfer is described by
194 Fourier’s equation and it is given by [1]:
D A disc made from low-alloy steel (AISI 4340) is placed inside 211
∂T 8 × 8 mm square coil. The geometrical parameters to be con- 212
TE
ρC ¼ ∇ :ðk ∇TÞ þ Qind ð19Þ
∂t sidered are the width of the disc (mm) and its relative distance 213
with the coil (gap (mm)). The model is represented in Fig. 1. 214
195
196 where T is the temperature (K). ρ, C, and k are non-linear
EC
202 ation heat flux losses qc and qr between the workpiece and
point on the edge of outside surface. Due to symmetrical 221
203 open air are defined respectively as
C
properties of the disc, the two edges will have the same tem- 222
206 qc ¼ hc ðT s −T a Þ ð20Þ perature distribution so TE1 will be equal to TE2. 223
N
qr ¼ εσs T s −T a
4 4
ð21Þ
U
205
204 Zone 1 Zone 2 Zone 3 Zone 4
H ardn ess
HH
HI
HL
dH Depth
dL
dI
Fig. 6 Temperature distribution at the edge and medium profile Fig. 7 Typical hardness profile by induction heating
AUTHOR'S PROOF! JrnlID 170_ArtID 2892_Proof# 1 - 23/10/2018
F
O
O
PR
D
TE
224 3.1 Determination of power under typical will decrease rapidly while getting closer to the surface medi- 231
225 configuration um and the disc center [7]. The mesh size is very refined and it 232
is fixed at 0.2 mm that generate 33,677 elements and 135,072 233
EC
226 Using induction-heating systems to harden the part requires degrees of freedom. The heating time (TH) is fixed at 0.5 s and 234
227 changing its surface layer microstructure by heating it to tem- the width (WP) is fixed at 6.5 mm. Finally, the gap (GS) is fixed 235
228 236
R
peratures that are above or inside the austenitization range. The at 2.5 mm. The outer surface exactly into the austenite trans-
229 temperature distribution after heating will not be uniform be- formation temperature will not ensure that the heat will prop- 237
R
230 tween the surface layer and the inner portion of the disc, and it agate sufficiently along the internal part of the part and that will 238
O
quency is fixed at 200 kHz. The last parameter to tune is the 241
N
F
the induced currents are distributed at external surface. 286
250 and 65 kW. To ensure that a transformation will occur inside
O
251 the part, and that 100% austenite will be produced at treated
252 287
O
area, the temperature used for austenitization must be above 3.3 Temperature and induced current distribution
253 Ac3, surface medium temperature of TM = 900 °C will be con-
PR
254 sidered sufficient to meet these conditions. Considering all Figure 5 shows the distributions of the total current density 288
255 geometrical and machine parameters, the most flexible input and the temperature after a heating time of 0.5 s in the case of 289
256 parameter is power delivered by the machine into the induction high-frequency heating. The initial density of the current in 290
257 D
coil because it can be changed softly and precisely from the the inductor (J0) is adjusted to also have a maximal tempera- 291
258 simulation software by the machine operator with no risk of ture of 1100 °C. The currents are concentrated, in this case, on 292
TE
259 error. All the machine and geometrical parameters, except the a thin layer in the edges of the inductor and in the part because 293
260 input power, will be fixed into reference level as shown in of the depth of skin. The temperature profile is profounder at 294
EC
261 Table 2. The input power values vary from 10 to 100 kW in the edges than the median plane since the currents are more 295
262 the simulation with a step of 10 kW and very smooth mesh size concentrated at the corners of the inductor. The preliminary 296
263 of 0.1 mm to see in which point the surface medium tempera- results clearly show the presence of the two electromagnetic 297
R
264 ture TM will exceed 900 °C. effects, the skin effect and the edge effect. These results also 298
265 299
R
The edge and medium temperature increase with power show that the temperature distributions are a direct conse-
266 and could attempt very high levels that exceed the melting quence of the currents induced in the part and this profile 300
O
267 point from 80 kW. The edge temperature increase faster than has a considerable effect on the final hardness profile. 301
268 302
C
the medium temperature due to the edge effect phenomena. The temperature is at its maximum on the surface and de-
269 Edge temperature exceeds the value of 900 °C starting from crease rapidly toward the disc center. Edge temperature is 303
N
270 57 kW power machine value while medium temperature ex- slightly higher than in the middle, but still both have the same 304
U
271 ceeds it with 66.5 kW input power value. The reference values temperature profile for every heating step time. In the HF case, 305
272 are given in the Table 1. the temperatures are distributed over a larger area on the edge 306
F
O
307 compared to the median plane. In addition, the maximum 0.50 s is approximated to 850 °C, it is possible to appreciate the 333
308 temperature values record a clear offset of 280 °C between hardened depth for both the MF and HF cases on the edge and 334
O
309 the edge and middle (Fig. 6). The heated zone on the edge the median plane. 335
PR
310 reaches 260 °C at a depth of 4 mm and 230 °C at the median
311 plane at the same depth. The temperature curve following the
312 depth at the end of heating makes it possible to predict the 3.4 Case depth deduction from temperature curves 336
313 hardened region and suggests that the region near the edges is
D
314 transformed into hard martensite while that at the median The simulated case depth is deduced from critical transforma- 337
TE
315 plane of the part does not reach the temperature necessary to tion temperature distribution across the disc assuming that a 338
316 martensitic transformation. Similarly, if the temperature of rapid cooling (quenching) is done perfectly after the heating 339
317 600 °C is assumed a legitimate assumption, the depths affect- process to form a new and hard martensitic microstructure. 340
EC
318 ed are 2.1 mm at the edge and 1.6 mm at the median plane. The transformation temperatures are identified to be the criti- 341
319 In order to interpret a temperature profile into a hardness cal temperature Ac1 (825 °C) corresponds to the apparition of 342
343
R
320 profile, the critical austenitization temperatures Ac1 and Ac3 that first austenite germs. Ac3 (850 °C) corresponds to the temper-
321 characterize the heated region after 0.5 s must be considered. ature at which the first germs of austenite appear. Finally, the 344
R
322 Thus, the temperature Ac1 characterizes the onset of formation critical temperature (Tr) fixed at 640 °C is assumed to be the 345
O
323 of the austenite while the temperature Ac3 characterizes the end lowest temperature that affects the part microstructure [22]. In 346
324 of the martensite returned (initial microstructure) and all the fact, this temperature characterize the depth where the hard- 347
C
325 regions heated above this temperature are austenitized to ness recovers the initial value in the part core. To retrieve the 348
N
326 100%. The region between the Ac1 and Ac3 temperatures is a corresponding critical depth di for each critical temperature 349
327 mixture of austenite and unconverted returned martensite. The Ac1, Ac3, and Tr, a linear interpolation function was used 350
U
328 final temperature distribution has a direct impact on the hardness between the closest higher temperature and closest smaller 351
329 profile in the part and its evolution as a function of depth (sur- temperature values with their corresponding depth values 352
330 face hardness and hardened depth). If the assumption that all from the simulation results (Fig. 7). High surface hardness is 353
331 regions heated above Ac3 (100% austenite) become martensite related to martensite formation, which in turn is dependent 354
332 lasts after cooling and that this temperature for a heating time of upon heating to austenite range and cooling rate. Therefore, 355
the simulation results and temperature distribution could re- 356
veal the shape and size of the hardened region and case depth. 357
t6:1 Table 6 Temperature at
t6:2 the medium point Parameters Edge Medium
Table 7 Contributions on the case depth at the edge t7:1
obtained by simulation
t6:3 (°C) PM 45.26% 32.92%
Parameter dH dL dI t7:2
t6:4 TH 31.84% 40.05%
t6:5 WP 1.11% 2.44% PM 35.39% 34.54% 28.22% t7:3
t6:6 GS 18.25% 17.48% TH 48.3% 49.6% 59.26% t7:4
t6:7 PM × TH 0.92% 2.12% WP 0.74% 0.73% 0.69% t7:5
t6:8 PM × GS Insignificant 0.97% GS 12.91% 12.48% 9.47% t7:6
t6:9 Error 2.61% 4.07% Error 2.66% 2.66% 2.36% t7:7
AUTHOR'S PROOF!
JrnlID 170_ArtID 2892_Proof# 1 - 23/10/2018
F
Fig. 11 Simulated temperatures versus predicted temperature at middle will be calculated in three levels, giving 81 possible con- 406
and edges
figurations and corresponding to the 81 simulations. 407
O
The classical technique for handling simulation is 408
O
358 3.5 Experimental validation exploited manually, but since the number of configuration 409
is great and the mesh is very condensed, this kind of man- 410
PR
359 Experimental validation tests are performed using the induction ual work of input and result extraction will take a very long 411
360 machine located at the École de Technologie Supérieure time [10]. To optimize this time and effort problem, all 81 412
361 (Montreal, Canada). This machine contains solid state con-
D simulations were automatized by writing a MATLAB code 413
362 verters (10 kHz) and thyristor radio frequency generator coupled with a COMSOL built-in functions. This automat- 414
363 415
TE
(200 kHz). The first one provides a maximum power of ed method solution reduced the calculation and results ex-
364 550 kW and the second generator delivers a maximum power traction into just 6 h even with a very condensed mesh size. 416
365 of 450 kW. This machine is capable to control both frequencies Results were the temperature distribution alongside the 417
EC
366 using the sequential double frequency heating concept and it is edge and the medium plan of the heated disc. Statistical 418
367 a numerically commanded using two numerical axes (Fig. 8). analysis was performed to analyze the effect of these inde- 419
368 The first test was done under the following machine param- pendent geometrical and machine parameters on tempera- 420
R
369 eters and geometry configuration. This tuning data (Table 2) is ture distribution and hardness profile. 421
R
1000
373
Temperature ( C)
form across the part section. Due to the edge effect, the mag-
375 netic field is concentrated more at the edges than in the mid- 950
U
376 dle. This fact generate a deep layer with important concentrat- 925
377 ed currents that heat more the edges than the middle. After 900
378 been transformed to austenite, the regions are transformed to 70.0 72.5 75.0 0.40 0.45 0.50 2.4 2.5 2.6 6.3 6.5 6.7
379 thin martensite. Due to lack temperature diffusion and skin PM (kW) TH (s) GS (mm) WP (mm)
380 effect, the non-affected zone toward the center of the disc
381 remains with its initial hardness value and no structural and (a)
382 phase change occurs in this region.
383 As illustrated in Fig. 10, the hardness curve measured at the
384 edge and the medium has a typical hardness profile as expect-
385 ed with significant difference on critical depth for both posi-
386 tions, due to edge effect. It can be shown that a conformity in
387 the shape exists between simulation and experimental tests for
388 both edge and medium profile. One can remark that the hard-
389 ness at the surface are slightly the same at the middle and the
390 edges. It is also interesting to observe that the low hardness is (b)
391 about 370 HV and the over-tempered zones have the same Fig. 12 Effect of geometry factors and machine parameters on
392 width in both cases. temperature, b middle and a edge
AUTHOR'S PROOF! JrnlID 170_ArtID 2892_Proof# 1 - 23/10/2018
(a) (b)
F
Fig. 13 Response surface methodology of surface temperature in the a middle and at the b edge
O
422 4.1 Simulation data 4.3 Simulated versus predicted temperatures 446
O
423 The following data presented in Tables 4 and 5 are extracted One goal of this study is to find a model that gives the most 447
PR
424 from results simulations and summarize the temperature valid prediction of the edge and surface temperature given 448
425 values at medium and edge of the part. It can be shown that the geometrical and machine parameters. The ANOVA sta- 449
426 final process temperatures are above the austenitization tem- tistical study gives equation that describes the predicted 450
427
D
perature Ac3 as desired and respects the non-melting condition relationship between temperatures in function with all oth- 451
428 452
TE
for medium region but it can be reached at the following er parameters for the edge and the medium point. The fit-
429 condition at the edge (75 kW, 0.5 s, GS = 2.4 mm and WP = ness of the model should be examined and evaluated. 453
430 6.3 mm), and the following configuration is the extreme case Experimentally, it is a difficult task to measure instantly 454
EC
431 and temperature is at its highest values. the temperature distribution across the part section as the 455
process happens very quickly. Figure 11 presents the scat- 456
432 4.2 Contributions ter plot for TE and TM. It is clear that for each simulated 457
R
433 After performing statistical analysis on simulation results, the due to the low value of the residuals errors. Consequently, 459
O
437 shown that TE and TM are highly influenced by the power and
438 the heating time. Both contribute in more than 70% on final
U
t8:2 Parameter dH dL dI
(a) (b)
Fig. 15 Response surface methodology of case depth in the a middle and at the b edge
F
O
460 for each value of temperatures, the predicted and simulated 4.4 Parameters effects 479
O
461 curves are nearly identical, which explains the good agree-
462 ment between the predicted and the measured values. The The average effect of geometrical and machine parameters on 480
PR
463 linear fitting criteria that the prediction equation is able to the final edge and machine temperature are presented in Fig. 481
464 assume a satisfactory level of accuracy with simulation 12a, b. Each factor contribute by its own degree on the tem- 482
465 results. perature. The power and heating time have the biggest influ- 483
466
D
After eliminating the non-significant terms, the empir- ence. An increase of 2.5 kW in machine power contribute to 484
467 ical relationship between the temperatures at th edge and an increase of 80 °C in edge temperature and 40 °C on medi- 485
TE
468 in the middle of the heated part by Eqs. 22 and 23. The um temperature. A short increase on the heating time (0.05 s) 486
469 equations present an approximate model based on linear leads to an important temperature rise for both points, at about 487
EC
470 regression and allows the evaluation of the temperature 50 °C increase on the edge temperature and 40 °C on medium 488
471 according the machine parameters and the geometrical temperature. Geometrically, the temperature decrease signifi- 489
472 factors. cantly when increasing the gap. A 0.1 mm of gap variation 490
R
about 40 °C at the medium. The part width (WP) has no sig- 492
þ 101:8xPM xT H ð22Þ nificant influence of the final temperature distribution. A 493
O
478 þ 2402xGS þ 114:5xPM xT H −37:76xPM xGS ð23Þ medium temperature than the edge temperature. 497
U
477
476
(a) (b)
Fig. 16 Simulated case depths versus predicted case depth in the middle (a) and at the edges (b)
AUTHOR'S PROOF! JrnlID 170_ArtID 2892_Proof# 1 - 23/10/2018
t9:1 Table 9 Experimental planning also demonstrates that the geometrical factors contribute much 526
t9:2 Parameters PM TH WP GS Fr more in the middle than at the edge. 527
t9:3 Test 2 42.9 1 6.5 2.5 200
t9:4 Test 3 56.1 1 6.5 2.5 200
5.1 Parameters effects 528
t9:5 Unit kW s mm mm kHz
Figure 14 represents respectively the case depth in the edge 529
and at the middle layer after heating and assuming a perfect 530
498 The response surface methodology (RSM) given by quenching process in function of machine power and heating 531
499 ANOVA analysis is a fast and powerful way to predict the time, maintaining the part width and gap in their centered 532
500 final temperature in function with the process parameters. value. It is clear that case depth has a linear profile and in- 533
501 Edge and medium temperature are presented according to crease in the same rate with time and power. As is known, due 534
502 the power and heating time values. The gap and the part width to edge effect, the case depth has bigger values on edge than in 535
503 are maintained in typical centered values of 2.5 mm and the medium layer. The case depth value is up to 1.62 mm in 536
504 6.5 mm. It can be shown that for both points, temperature the edge and 1.04 mm in the medium. 537
F
505 has a parabolic profile and increase proportionally with the Figure 15 describes the average effect of geometrical and 538
machine parameters on the case depth. The response of case 539
O
506 power and heating time. Figure 13 illustrates the RSM obtain-
507 ed at the edge and in the middle of the heated part. depth is analogous to temperature distribution and the effects 540
O
of variant parameters on it are similar. A small increase in the 541
machine parameters increase the case depth rapidly while in- 542
PR
creasing the geometrical parameters values have the opposite 543
508 5 Sensitivity study of case depth effects. The case depth is higher under these conditions 544
D (75 kW, 0.5 s, GS = 2.4 mm and WS = 6.3 mm), and lower 545
509 The final temperature distribution is converted to hardness under the following configuration (70 kW, 0.4 s, GS = 546
547
TE
510 curve using Matlab code. In fact, the temperatures according 2.6 mm and WS = 6.7 mm).
511 to the depth at the middle and at the edge are converted to
512 hardness curve as explained using Fig. 7. After interpolating 5.2 Simulated versus predicted case depths 548
EC
513 the critical depth values dH, dL, and dI corresponding respec-
514 tively to critical temperatures Ac3, Ac1, and Tr, statistical anal- ANOVA analysis gives the following equations describing 549
515 ysis was done to deduce the contribution of geometrical and critical case depth dH, dL, and dI in function of geometrical 550
R
516 machine parameters on these simulated critical depth. and machine parameters. It can be shown from Fig. 16 that the 551
R
517 According to statistical analysis (Tables 7 and 8), the most predicted and simulated depth values are strongly correlated. 552
The next part will examine the correctness of the predicted 553
O
518 important factor on the case depth is the heating time with
519 about 50% contribution percentage. The machine power con- model equations with experimental results. Figure 19 presents 554Q9
C
520 tributes significantly on the final case depth with about 35.4% the scatter plot for case depth at edges and in middle locations. 555
556
N
521 on the edge and 27% on medium. The influence of the part It is clear that for each measured response value, the predicted
522 width is insignificant with about 0.74% on edge and 2.5% on value is close to the diagonal line, due to the low value of the 557
U
523 medium. However, the gap has an important effect on the case residuals of the predicted values. For each value of case depth, 558
524 depth with about 13% contribution percentage on the edge and the predicted and measured curves are nearly identical, which 559
525 17% contribution percentage in the middle. The analysis result explains the good agreement between the predicted and the 560
measured values. 561
(a) (b)
Fig. 18 Predicted and measured hardness curves during test 2 at edge (a) and in the middle (b)
F
O
572 between the three curves is relatively low and is about 5 up to References 604Q11
573 19%. The marge of error is smaller for experimental tests
O
574 where input power and heating time values are close to those 1. Rudnev V, Loveless D, Cook R (2017) Manufacturing engineering 605
and materials processing. In: Handbook of Induction Heating, 2nd 606
PR
575 used in the simulation. Still, the mathematical model is reli-
edn. Boca Raton, CRC Press, Taylor & Francis Group xxi, 749 607
576 able for predicting the effect of input parameters on critical pages 608
577 depths and to predict the over tempering zone profile. 2. Guerrier P et al (2015) Three-dimensional numerical modeling of 609
D an induction heated injection molding tool with flow visualization. 610
Int J Adv Manuf Technol 85(1–4):643–660 611
3. Guo X et al (2015) Numerical simulations and experiments on 612
TE
fabricating bend pipes by push bending with local induction- 613
578 7 Conclusion heating process. Int J Adv Manuf Technol 84(9–12):2689–2695 614
4. Han Y, Yu E-L, Zhao T-X (2016) Three-dimensional analysis of 615
EC
579 The main feature of this study is the great number of simulation medium-frequency induction heating of steel pipes subject to mo- 616
tion factor. Int J Heat Mass Transf 101:452–460 617
580 done in order to create a reliable and acceptable model that 5. Hömberg D, Liu Q, Montalvo-Urquizo J, Nadolski D, Petzold T, 618
581
R
predicts the effect of geometrical and machine parameters on Schmidt A, Schulz A (2016) Simulation of multi-frequency- 619
582 the hardness depth of an AISI 4340 low-alloy steel disc heated induction-hardening including phase transitions and mechanical ef- 620
R
583 by induction. First, a 2D axis-symmetric model was developed fects. Finite Elem Anal Des 121:86–100 621
6. Achraf S et al (2016) Temperature history modelling and validation 622
O
587 convergence study was carried out to identify and select the of induction hardening of gear wheels. J Comput Appl Math 270: 626
588 best and typical machine, geometry configuration for the sim- 231–240 627
U
649 15. Bae K-Y, Yang YS, Hyun CM, Cho SH (2008) Derivation of sim- 19. Dmytro R et al (2012) Investigation of the surface residual stresses 663
650 plified formulas to predict deformations of plate in steel forming in spray cooled induction hardened gearwheels. Int J Mater Res 664
651 process with induction heating. Int J Mach Tools Manuf 48(15): 103(1):73–79 665
652 1646–1652 20. Kochure PG, Nandurkar KN (2012) Application of taguchi meth- 666
653 16. Barka N, Chebak A, el Ouafi A, Jahazi M, Menou A (2014) A new odology in selection of process parameters for induction hardening 667
654 approach in optimizing the induction heating process using flux of EN8 D Steel. International Journal of Modern Engineering 668
655 concentrators: application to 4340 steel spur gear. J Mater Eng Research (IJMER) 2(5):3736–3742 669
656 Perform 23(9):3092–3099 21. Jin J (2002) The finite element method in electromagnetics. John 670
657 17. Kristoffersen H, Vomacka P (2001) Influence of process parameters Wiley & Sons Inc., New York 671
658 for induction hardening on residual stresses. Mater Des 22:637–644 22. Barka N, el Ouafi A, Bocher P, Brousseau J (2013) Explorative 672
659 18. Besserer H-B, Dalinger A, Rodman D, Nürnberger F, Hildenbrand study and prediction of overtempering region of disc heated by 673
660 P, Merklein M, Maier HJ (2016) Induction heat treatment of sheet- induction process using 2D axisymmetric model and experimental 674
661 bulk metal-formed parts assisted by water-air spray cooling. Steel tests. Adv Mater Res 658:259–265 675
662 Res Int 87(9):1220–1227
676
F
O
O
PR
D
TE
EC
R
R
O
C
N
U
AUTHOR'S PROOF!
AUTHOR QUERIES
F
Q8. A citation for table 6 has been inserted after the sentence "Power has more effect on the edge
O
(45.62%) than the medium temperature (32.92%)...." Please check if appropriate.
O
Q9. Figure 19 was mentioned in the text but no figure image provided in the manuscript. Please
PR
check.
Q10. A citation for figure 17 has been inserted after the sentence "The obtained results demonstrate that
simulation combined to practical tests...." Please check if appropriate.
D
Q11. Please supply/verify the standard abbreviation of the journal name in References [11, 20].
TE
EC
R
R
O
C
N
U