a r t i c l e i n f o a b s t r a c t
Article history: A fast response thermocouple was developed for measuring surface temperatures of aluminum compo-
Received 11 June 2009 nents in ICE combustion chambers. The key features of the design are the use of the aluminum substrate
Received in revised form 23 September as one of the thermocouple metals and the use of a thick copper layer as the hot junction at the surface.
2009
The copper equalizes the hot junction temperature with the surrounding aluminum to correct for the dif-
Accepted 13 October 2009
ferences in thermal properties between the two materials. FEA determined the optimum thickness of the
copper layer to be between 100 and 125 lm. Under typical SI engine heat ux conditions, the thermocou-
ple should be able to measure average surface temperatures within 0.19 C and the magnitude of temper-
Keywords:
Internal combustion engine
ature swings within 6% of true values.
Surface thermocouple Following the FEA, the optimized thermocouple was tested in a SI engine. Experimental results dis-
Fast response thermocouple played the same trends as the FEA at measuring average temperatures and temperature swings, suggest-
Heat ux ing the thermocouple was performing as predicted.
2009 Elsevier Inc. All rights reserved.
0894-1777/$ - see front matter 2009 Elsevier Inc. All rights reserved.
doi:10.1016/j.expthermusci.2009.10.008
184 M.A. Marr et al. / Experimental Thermal and Fluid Science 34 (2010) 183189
1500 RPM. The lower boundary condition was set at 100 C, and
the aluminum side wall was modeled as thermally insulated.
The geometry was meshed by COMSOL with variable spacing
with the highest density around the hot junction. The time step
was set at 0.1 ms. Trials found that further renement of the mesh
or time step did not substantially alter the FEA output.
Fig. 5 shows FEA computed hot junction temperatures for a
range of copper thicknesses. The No Copper temperatures were
obtained by running the FEA model without the copper layer,
and the True temperatures were obtained by running the model
without the copper layer and with all other materials changed to
aluminum. The True results represent the surface temperatures
that would be measured by a perfectly accurate thermocouple.
Most striking about Fig. 5 is the No Copper curve, which has an
average temperature and temperature swing signicantly greater
than True values. This demonstrates the importance of the copper
layer for removing heat from the hot junction to equalize its tem-
perature with the substrate.
The thermocouple is highly accurate at measuring cycle aver-
aged temperatures over a wide range of copper layer thicknesses.
The range of thicknesses shown in Fig. 5, 75400 lm, all measure
the average temperature within 0.23 C of the True average tem-
Fig. 3. FEA model geometry and boundary conditions (all dimensions in mm, not to perature. If copper thickness is reduced to 10 lm, the thermocou-
scale). ple would still measure an average temperature within 0.8 C of
the True value.
When measuring the magnitude of surface temperature swings,
Table 1
Material properties.
the specic thickness of the copper layer is more important. In
Fig. 5, the 75 lm case overestimates the True temperature swing
Material k (W/m K) qkg=m3 C (J/kg K) by 11% while the 400 lm case underestimates it by 44%. The opti-
Constantan 19.5 8900 390 mal thickness of the copper layer was found to be between 100 and
Teon 0.23 2190 1010 125 lm, as shown in Fig. 6. These thicknesses will measure the
Ceramic cement 1.2 2260 1000
True temperature swing within 2% and 6% accuracy respectively.
Aluminum 155 2750 915
Copper 400 8700 385 Furthermore, in this thickness range the delay in peak temperature
occurrence from the True value is less than 5 Crank Angle (CA).
This delay is on the order of 10% of combustion duration, but is a
necessary trade-off for improved temperature measurement accu-
In the analysis, the thermocouple output was assumed to be the
racy. If desired, temperature data could be shifted by a processing
temperature at the hot junction marked in Fig. 3. This assumption
algorithm to offset the delay. Table 2 provides a numerical sum-
therefore ignores the temperature difference across the copper be-
mary of FEA results shown in Figs. 5 and 6.
tween the hot junction and the aluminumcement interface. This
To conrm the thermocouples accuracy over a range of heat
is a reasonable assumption because the temperature difference
transfer conditions, the analysis was repeated twice using different
was generally small, and copper and aluminum have similar ther-
convective heat transfer boundary conditions. First, the model was
moelectric powers which are considerably smaller in magnitude
run using twice the Alkidas and Myers convective heat ux, the
than constantans. For example, at 100 C the thermoelectric volt-
results of which are shown in Fig. 7. Second, the analysis was per-
ages of copper, aluminum and constantan are approximately 1.0,
formed with heat ux calculated from in-cylinder gas tempera-
0.5 and 4.0 mV respectively when paired with platinum with
tures and a heat transfer coefcient. Gas temperatures were
the reference junction at 0 C [7].
taken from previous SI engine work by Enomoto and Furuhama
The time varying boundary condition for the upper surface was
[9], and the heat transfer coefcient was calculated with the
convective heat ux as shown in Fig. 4. These data were taken from
Woschni equation [10] using temperature and pressure data from
measurements by Alkidas and Myers [8] in the cylinder head of a
Enomoto and Furuhama. Hot junction temperatures from the anal-
propane-fueled SI engine running slightly lean at part load and
ysis with the Woschni equation are shown in Fig. 8.
Fig. 4. Convective heat ux data taken from Alkidas and Myers [8]. Fig. 5. Hot junction temperature with varied copper thickness.
186 M.A. Marr et al. / Experimental Thermal and Fluid Science 34 (2010) 183189
Fig. 6. Hot junction temperature with 100 and 125 lm copper layers. Fig. 9. Hot junction temperature with 0.5 mm YSZ TBC.
4. Engine testing
Figs. 7 and 8 conrm that, with a 125 lm copper layer, the ther-
mocouple should produce highly accurate temperature
measurements.
Finally, as the authors intend to use the thermocouple to study
TBCs, the FEA was performed with the thermocouple behind a
layer of Yttria-Stabilized Zirconia (YSZ). The geometry shown in
Fig. 3 was again used, except a 0.5 mm thick layer of YSZ Fig. 10. Cylinder wall access port conguration. Adapted from [11].
M.A. Marr et al. / Experimental Thermal and Fluid Science 34 (2010) 183189 187
Table 3
Engine test results.
Fig. 13. Representative cycles of Thermocouple 3 in Port C (temperatures shifted to Fig. 15. Sample cycle of Thermocouple 2 with YSZ coating in Port B (temperatures
provide average temperature of 0 C). shifted to provide average temperature of 0 C).
M.A. Marr et al. / Experimental Thermal and Fluid Science 34 (2010) 183189 189
5. Conclusions References
The thermocouple developed by Heichal et al. was successfully [1] D. Assanis, E. Badillo, Evaluation of alternative thermocouple designs for
transient heat transfer measurements in metal and ceramic engines, SAE
modied to measure accurate surface temperatures in aluminum Technical Paper Series 890571.
combustion chamber components. The primary modication was [2] J. Chang, Z. Filipi, D. Assanis, T. Kuo, P. Najt, R. Rask, Characterizing the thermal
the addition of a thick copper layer at the surface to conduct heat sensitivity of a gasoline homogeneous charge compression ignition engine
with measurements of instantaneous wall temperature and heat ux,
away from the constantan wire. Using FEA, the optimal thickness International Journal of Engine Research 6 (2005) 289309.
of the copper layer was found to be between 100 and 125 lm for [3] T. Hendricks, J. Ghandhi, J. Brossman, Instantaneous local heat ux
accurately measuring the true temperature swing. The addition measurements in a small utility engine, in: Proceedings of the ASME Internal
Combustion Engine Division Spring Technical Conference, 2009.
of this layer will introduce a modest delay in peak temperature [4] S. Furuhama, Y. Enomoto, Heat transfer into ceramic combustion wall of
occurrence of less than 5 CA. The thermocouple is very accurate internal combustion engines, SAE Technical Paper Series 870153.
at measuring average temperatures, a task which is less sensitive [5] D. Buttsworth, Transient response of an erodable heat ux gauge using nite
element analysis, Proceedings of the Institution of Mechanical Engineers, Part
to copper thickness. If the thermocouple is used to measure tem-
D: Journal of Automobile Engineering 216 (8) (2002) 701706.
peratures under a TBC, a wide range of copper thicknesses can be [6] Y. Heichal, S. Chandra, E. Bordatchev, A fast-response thin lm thermocouple
tolerated while still producing accurate measurements. to measure rapid surface temperature changes, Experimental Thermal and
Fluid Science 30 (2) (2005) 153159.
Experimental engine testing of the thermocouple displayed the
[7] P. Kinzie, Thermocouple Temperature Measurement, John Wiley & Sons, Inc.,
trends predicted by the FEA. The results suggest the thermocouple New York, NY, 1973.
is accurate at measuring average temperatures and temperature [8] A. Alkidas, J. Myers, Transient heat-ux measurements in the combustion
swings. chamber of a spark-ignition engine, Journal of Heat Transfer, Transactions
of the American Society of Mechanical Engineers 104 (1) (1982) 6267.
[9] Y. Enomoto, S. Furuhama, Measurement of the instantaneous surface
Acknowledgments temperature heat loss of gasoline engine combustion chamber, Proceedings
SAE 1 (1984) 1.581.63.
[10] G. Woschni, A universally applicable equation for the instantaneous heat
The authors are grateful to Tiegang Li, a researcher at the Uni- transfer coefcient in the internal combustion engine, SAE Transactions 76
versity of Toronto Centre for Advanced Coating Technologies, for (1967) 670931.
all his assistance applying the copper and ceramic coatings. The [11] H.S. Imming, The effect of piston-head temperature on knock-limited power,
Technical Report, National Advisory Committee for Aeronautics, 1944.
authors would also like to thank the Natural Sciences and Engi- [12] S. Smith, The Scientist and Engineers Guide to Digital Signal Processing, rst
neering Research Council of Canada (NSERC) for their nancial sup- ed., California Digital Pub., San Diego, CA, 1997.
port of this study.