Harold SchockGiles Brereton

Department of Mechanical Engineering,

Michigan State University,

East Lansing, MI 48824

Eldon CaseDepartment of Chemical Engineering

and Materials Science,

Michigan State University,

East Lansing, MI 48824

Jonathan D’AngeloTim Hogan

Department of Electrical Engineering

and Computer Science,

Michigan State University,

East Lansing, MI 48824

Matt LyleDepartment of Mechanical Engineering,

Michigan State University,

East Lansing, MI 48824

Ryan MaloneyDepartment of Chemical Engineering

and Materials Science,

Michigan State University,

East Lansing, MI 48824

Kevin MoranJames Novak

Department of Mechanical Engineering,

Michigan State University,

East Lansing, MI 48824

Christopher NelsonCummins Inc.,

Columbus, IN 47201

Andreas PanayiTrevor Ruckle

Department of Mechanical Engineering,

Michigan State University,

East Lansing, MI 48824

Jeffery SakamotoDepartment of Chemical Engineering

and Materials Science,

Michigan State University,

East Lansing, MI 48824

Tom ShihSchool of Aeronautics and Astronautics,

Purdue University,

West Lafayette, IN 47907

Ed TimmDepartment of Mechanical Engineering,

Michigan State University,

East Lansing, MI 48824

Long ZhangDepartment of Chemical Engineering

and Materials Science,

Michigan State University,

East Lansing, MI 48824

George ZhuDepartment of Mechanical Engineering,

Michigan State University,

East Lansing, MI 48824

Prospects for Implementationof Thermoelectric Generators asWaste Heat Recovery Systemsin Class 8 Truck ApplicationsWith the rising cost of fuel and increasing demand for clean energy, solid-state thermo-electric (TE) devices are an attractive option for reducing fuel consumption and CO2

emissions. Although they are reliable energy converters, there are several barriers thathave limited their implementation into wide market acceptance for automotive applica-tions. These barriers include: the unsuitability of conventional thermoelectric materialsfor the automotive waste heat recovery temperature range; the rarity and toxicity of someotherwise suitable materials; and the limited ability to mass-manufacture thermoelectricdevices from certain materials. One class of material that has demonstrated significantpromise in the waste heat recovery temperature range is skutterudites. These materialshave little toxicity, are relatively abundant, and have been investigated by NASA-JPL forthe past twenty years as possible thermoelectric materials for space applications. In arecent collaboration between Michigan State University (MSU) and NASA-JPL, the firstskutterudite-based 100 W thermoelectric generator (TEG) was constructed. In this paper,we will describe the efforts that have been directed towards: (a) enhancing thetechnology-readiness level of skutterudites to facilitate mass manufacturing similar tothat of Bi2Te3, (b) optimizing skutterudites to improve thermal-to-electric conversionefficiencies for class 8 truck applications, and (c) describing how temperature cycling,oxidation, sublimation, and other barriers to wide market acceptance must be managed.To obtain the maximum performance from these devices, effective heat transfer systemsneed to be developed for integration of thermoelectric modules into practical generators.[DOI: 10.1115/1.4023097]

Contributed by the Advanced Energy Systems Division of ASME for publicationin the JOURNAL OF ENERGY RESOURCES TECHNOLOGY. Manuscript received July 15,2011; final manuscript received August 8, 2012; published online January 25, 2013.Assoc. Editor: Gunnar Tamm.

Journal of Energy Resources Technology JUNE 2013, Vol. 135 / 022001-1Copyright VC 2013 by ASME

Downloaded From: http://energyresources.asmedigitalcollection.asme.org/ on 04/02/2015 Terms of Use: http://asme.org/terms

Implementation of Solid-State Thermoelectric Energy

Conversion Devices

Thermoelectric devices are typically used for heating, cooling,and power generation. The heart of the TEG is a module, whichproduces power and is shown schematically in Fig. 1. The coupleoperates by converting heat to electrical energy and in the reverse,when applying a voltage to the device it creates a temperaturedifference on each side. This effect is often called the Peltier–Seebeck effect and a detailed discussion of these principles isfound extensively in the literature [1]. Two major issues toaddress when implementing a thermoelectric generator in an auto-motive system are heat transfer considerations and thermoelectricmaterial selection for the generator. The Hendricks [2] work eluci-dates the requirement for carefully designed heat exchangers toget maximum power output from the thermoelectric generator.Crane and Bell [3] describe the issues related to temperaturevariation by conducting an analysis using a three-section thermo-electric system, where the materials in each of the sections areselected to produce highest performance at the expected tempera-ture gradients. Although multimaterial thermoelectric generatorsindeed provide the best performance, issues related to cost andsystem thermal management must be addressed. Other systemshave been used to promote energy conservation in what could beconsidered traditional single IC engine powered applications.They include combined power cycles [4], organic Rankine cycles[5,6], and dual fuel mode systems [7] sometimes combined withcomplex optimization schemes for decisions of mode of operation[8]. The thermoelectric system directly converts heat to electric-ity, enabling precious power to be distributed where it is neededon a vehicle for operating a motor, powering sensors, or charginga battery for later energy usage.

In this work, a system describing how thermoelectric-baseddevices can lead to improvement in fuel economy by convertingwaste heat to electricity for class 8 truck applications is presented.We also discuss how additional savings can be obtained by power-ing the “hotel loads” associated with normal engine-off periodsusing a thermoelectric-based auxiliary power unit. Two thermo-electric systems are described; the first is termed an exhaust gasrecirculation thermoelectric generator (EGR TEG). A traditionalEGR system is replaced with one that uses a thermoelectric as-sembly to cool the exhaust gas while producing electricity. Thesecond system is termed an energy recovery system-auxiliarypower unit (ERS-APU). The ERS-APU will function as an auxil-iary power unit when the vehicle is stationary, providing electrical

energy for hotel loads, heat for the cabin as required, and produc-ing electrical energy while traveling on the highway. This energywill be used to operate electrical systems, thus reducing the loadon the alternator and returning the remainder of the electricalpower to the output shaft using a system such as a belt-integratedmotor generator. Throughout this paper, several “next steps” arediscussed to illustrate the tasks that must be undertaken to furtherdevelop TE technology for integration with class 8 truck engines.

The implementation of waste heat recovery devices reducesthe carbon footprint of a vehicle, reduces greenhouse gases, andlessens a countries dependence on foreign oil. Due to the rela-tively high fuel consumption and long life of a class 8 truckengine, often over 1,000,000 mile, the goal of a 2% reduction infuel consumption would result in a 4000 gallon savings per vehi-cle. When one adds the savings due to using the ERS-APU toeliminate engine idle fuel consumption while the vehicle isparked, an additional 10,000 gallons of fuel could be saved overthe life of the vehicle.

The MSU group, with the support of Cummins, conducted anextensive analysis of the potential gains associated with fitting anengine (2007 Cummins ISX) on a class 8 truck (minimum grossvehicle weight 33,001 lb or more) with thermoelectric generatorsfor waste heat recovery [9]. Figure 2 shows in schematic form, thehardware associated with a typical class 8 truck powertrain. Aone-dimensional wave dynamic engine simulation called WAVE,a code widely used in the transportation industry, was used toevaluate the potential gains with the thermoelectric generator forboth EGR and the ERS-APU energy recovery systems. TheWAVE simulation solves the unsteady partial differential equa-tions that describe the heat transfer implemented with models thatestimate thermoelectric generator performance. The MSU groupestimated that based on measured thermoelectric properties oflead–antimony–silver–tellurium (LAST(T)) and bismuth tellurideat a road load operating point, the system efficiency could beimproved 3–5% by the implementation of a TEG system, depend-ing on the assumptions employed and the efficiency of the ther-moelectrics. The economic aspect was evaluated, including theeconomics associated with the use of a waste heat energy recoverysystem that also provides an auxiliary power unit to the vehiclethat utilizes solid-state thermoelectric generators, instead of idlingthe engine to generate electrical power when the vehicle is station-ary. To facilitate the cost and pricing study, the ERS-APU understudy was divided into four major subsystems: electrical/electron-ics; thermoelectric generator; burner; and cooling.

The MSU and Cummins analysis used data generated from the2007 engine studies and assumed an operating condition termedB62 (62% of full load equal to 1500 rpm, 245 kW), typical of aroad load condition at that time. The benefits come from energyrecovery while the vehicle is in operation and by a savings duringidling periods, allowing one to turn the engine off. Assumptionswere a 4% system efficiency when operating as an energy conver-sion device and a 7% efficient thermoelectric generator duringAPU operation, as the APU can operate at a higher temperaturethan in the exhaust heat recovery mode, for 1 and 5 kW TEGsystems. The vehicle was assumed to operate 150,000 miles peryear in 300 days with 8 h of idling each day. This savings is likelythe upper limit that could be anticipated for this configuration, butis conservative in that it does not include the gains that would berealized by replacing electrical loads on the alternator, which hasan efficiency of about 50–60%, with electrical energy producedby the TEGs. In these applications, the 1 and 5 kW units wouldpay for themselves in 1 yr and 3 yr, respectively, with fuel costsestimated to be $4 per gallon. The 1 kW and 5 kW systems do notscale linearly, as a significant fraction of savings is during opera-tion of the 1 kW APU, which is assumed to operate at this powerlevel even with the 5 kW system. In developing a “retail” price forthe whole system, a price for each of the subsystems was calcu-lated by adding its components to establish the total price. Theprice includes such things as fully accounting for cost of labor,overhead (building cost, cost of capital, etc.), and general and

Fig. 1 Supply heat to hot side is engine exhaust; supply cool-ant is engine coolant or a separate fluid, there were 200 couplesin TEG demonstrated in this work

022001-2 / Vol. 135, JUNE 2013 Transactions of the ASME

Downloaded From: http://energyresources.asmedigitalcollection.asme.org/ on 04/02/2015 Terms of Use: http://asme.org/terms

administrative costs, at a production rate of 10,000 units per year.The estimate for the price that a manufacturer would have tocharge for a 1 kW system, to cover cost and provide a reasonablereturn, is $4773. These prices were developed by an engineerexperienced in assessing mass production costs for automotiveapplications under the direction of the first author of this work. Arecent breakthrough in construction of a 100 W nominal sizedthermoelectric generator at MSU caused us to reduce the esti-mated cost of the generators significantly as machine assemblyusing technology for building circuit boards is now possible and isnot reflected in the above estimate.

The MSU group’s analysis of the potential gains from fitting a2007 Cummins ISX class 8 truck engine with thermoelectric gener-ators when operating at the B62 condition was based on a detailed,fully transient, cycle-resolved simulation of its performance andyielded improvements in brake specific fuel consumption, whichvaried between 3% and 5%, depending on the particular values cho-sen for thermal and thermoelectric characteristics.

Selection of TE Material

The process of down-selecting a TE material for generatortesting compared three key characteristics of the materials underconsideration: (a) optimal temperature range, (b) mechanical du-rability, and (c) availability of material. As the target applicationfor the ERS-APU and EGR TEG systems is a class 8 truck engine,a temperature range of 600–850 K was selected. For this tempera-ture range, two candidates were compared: LAST-LASTT andskutterudites (skutterudite is a cobalt arsenide mineral that hasvariable amounts of nickel and iron substituting for cobalt witha general formula: (Co,Ni,Fe)As3. Some references give thearsenic a variable formula subscript of 2–3). A broad rangeof LAST (AgaPbbSbcTed) compositions underwent mechanicaltesting, within the mole fraction ranges of 0.006< a< 0.043,0.417< b< 0.480, 0.011< c< 0.050, and 0.496< d< 0.517.These LAST compositions had hardness (H), values from 0.53 to0.92 GPa [10] and Young’s modulus (E), values from 24 GPa to68 GPa [11], with similar E and H values for LASTT [12]. In con-trast, the skutterudite specimens tested had H values of about6.0 GPa and E values of 140 GPa [13]. Although fracture tough-ness (KC), measurements were difficult to perform on LAST,LASTT, and skutterudite specimens, the decreased tendency ofthe skutterudites to chip and spall during cutting and handlingmay be indicative of a higher KC value for the skutterudites thanLAST and LASTT. Additionally, LAST and LASTT materials

contain tellurium, which has potential scarcity concerns, whereasskutterudite materials are relatively abundant. Thus, skutteruditeswere down-selected as the material for TEG testing.

Effort in Advanced TEG Design and Construction

There have been very few experiments performed that demon-strate the actual power generated by the skutterudite thermoelec-tric modules in a realistic environment. MSU, in collaborationwith JPL, has designed, fabricated, and tested the first earlyskutterudite-based thermoelectric generators (50 W and 100 W).The results of a 10 module generator test are shown in Fig. 3. At aDT of 550 �C, the generator produced a combined power output of50.1 W. The maximum power output for the modules ranged from3.9 to 5.9 W. Based on an energy balance and assuming similarenergy recovery by adding additional modules, we estimate thatthe technology demonstrated could be implemented in construc-tion of a thermoelectric generator, which would recover about 4%of the energy in an air (or exhaust) stream. MSU’s most recentgenerator, shown in Fig. 4, has a nominal 100–200 W poweroutput and is comprised of 200 couples. It is managed by an elec-tronic system that MSU terms couple bypass technology (CBT),allowing for generator designs that permit a large number ofcouples to be placed in series, thereby boosting voltage output tousable levels.

Critical TEG Subsystems Development

Our generator demonstration work has identified the followingchallenges to be overcome for a viable automotive/transportationTEG:

(1) High-efficiency TE materials for the appropriate tempera-ture range.

(2) Unavailability of low-cost, high-efficiency insulationtechnology.

(3) Poor heat transfer on hot-side surface.(4) Sublimation suppression.(5) Failure of a single couple causes complete failure of an

entire module.(6) Failure at interfaces due to mismatch in coefficients of

thermal expansion and/or poor bonding.

Challenge 1 was discussed above and 2–6 are discussed in thenext sections. Further challenges that are not discussed in thispaper include: scale up/mass production of TE materials,

Fig. 2 Efficiency improvements using an EGR TEG and ERS-APU

Journal of Energy Resources Technology JUNE 2013, Vol. 135 / 022001-3

Downloaded From: http://energyresources.asmedigitalcollection.asme.org/ on 04/02/2015 Terms of Use: http://asme.org/terms

packaging the system to enhance heat transfer while maintaining alow pressure drop, mitigating soot buildup that can result inreduced heat transfer to the hot side, identifying an efficientburner required for APU operation, and defining manufacturingrequirements.

TEG Module Insulation

Two thrusts concerning the aerogel-insulation activities havebeen carried out to date. The first generation aerogel formulationwas integrated into multiple couples, subassemblies, individualplates, and generator designs. The formulation was first devel-oped at JPL and has been modified by the Sakamoto group atMSU. Second, efforts to improve and characterize the next gen-eration aerogel were conducted to improve the consistency frombatch to batch. The key to reducing costs will require the abilityto dry the aerogel under ambient conditions rather than in anautoclave as is currently done. This can be accomplished by thedevelopment of cast-in-place aerogel-insulation (CPAI) methodsusing freeze drying and polar solvents instead of a supercriticalautoclave.

A multifoil insulation system is used in radioisotope thermo-electric generators (RTGs); however, terrestrial TEGs will requirethermal insulation that can reduce radiative, solid, and convectiveheat transfer. To limit the amount of material and the size ofthe generator, the thermoelectric couples should be relativelyshort (in the few mm range). This requires highly effective ther-mal insulation to maintain large gradients and limit parasitic

losses. Furthermore, the small-scale complex geometry and inti-mate contact required to improve thermal efficiency will requirethe thermal insulation to be cast into place.

CPAI can be used as it has the lowest thermal conductivity ofany known material. Another insulation alternative is a high-temperature aerospace insulation such as Microtherm

VR

or Min-KVR

and has been considered for NASA RTGs that operate undera cover gas. Prior efforts to integrate Microtherm

VR

or Min-KVR

insulation proved to be problematic in that the insulation was dif-ficult to handle when consistently machined and integrated intocomplex, precise shapes.

Recent MSU work, supported by DOE-EERE, demonstratedthe efficacy of CPAI in multiple thermoelectric generator proto-types [14]. One aspect of this work was to quantify the thermalefficiency of the generator technology. To clarify, thermoelectricefficiency typically refers to the fraction of heat flow passingthrough thermoelectric couples which is converted to electricalpower, i.e., not accounting for parasitic heat losses. At MSU, sev-eral skutterudite TEG prototypes were made between 2009 and2011, ranging from 25 to 100 W. For this work, the comparisonfocused on commercially available fiberglass and CPAI.

The performance of the two thermal insulations was comparedin a 50-couple TEG, which was square in shape and comprised offour plates. The top and bottom plates (1 and 3) were blanks, i.e.,did not have thermoelectric couples. Figure 5 shows that theother two plates each had 25 skutterudite couples, whereby one(plate 4) used packed commercial-fiberglass insulation and theother (plate 2) used CPAI. The system was configured to haveequal flow rates of hot gas impinging on the two plates that con-tained TE couples.

The temperatures of the gas entering and exiting the TEG, aswell as the increase in cooling fluid temperature for each plate,were used to determine the heat flow through each plate.

It was found that the heat flow through the plate containing thecommercial-fiberglass-insulated couples was 30% greater thanthe heat flow through the plate containing the CPAI-insulatedcouples. This resulted in steady-state thermal efficiencies of 3.2%and 4.2% for the commercial-fiberglass-insulated couples andthe CPAI-insulated couples, respectively. Thermal efficiency isdefined as electrical power out divided by heat transfer throughthe plate plus electrical power produced.

The next step will be to simplify processing and enable new,high thermal efficiency packaging. Instead of placing wet gels in asupercritical autoclave, the wet gels or TE modules encapsulatedin wet gel should be placed in a freeze drying chamber or soakedin nonpolar solvents and allowed to dry under ambient conditions,while controlling the heat transfer. If the thermal conductivity isfound to be comparable to supercritically dried aerogel, efforts tofreeze or ambient-dry aerogel-encapsulated TE modules can bemade. Another beneficial step would be to investigate how aerogel

Fig. 4 Completely assembled 100 W-TEG (left) and subassembly module for the TEG (right).This is the largest thermoelectric generator built using skutterudite material, with dimensionsof 6 in. diameter and 6 in. in length. All couples are in series facilitated by MSU-developed CBT.

Fig. 3 Results of 10 skutterudite modules operated at a DT of550 �C

022001-4 / Vol. 135, JUNE 2013 Transactions of the ASME

Downloaded From: http://energyresources.asmedigitalcollection.asme.org/ on 04/02/2015 Terms of Use: http://asme.org/terms

might be used to insulate the exhaust system upstream of theTEGs.

Fluid Flow and Heat Transfer Considerations

Purdue University, in collaboration with MSU, has conductednumerous multidimensional modeling efforts [15–18] to study fluidflow and heat transfer with the TEGs as well as within the modules.Aside from parasitic heat transfer losses, there are a number ofother challenges involving heat transfer. These other heat transferissues for TEGs can be divided into two interconnected areas. Thefirst is how to get the temperature gradient across the TE couple tobe as high as possible for given hot (Th) and cold (Tc) fluid tempera-tures so that the electric current generation is at a maximum. Thisrequires heat transfer enhancement techniques that can reduce thethermal boundary layer thickness on the two sides of the TE coupleto increase the heat transfer rate (e.g., restarting boundary layersand jet impingement; see Fig. 6). This must be accomplished withminimum loss in stagnation pressure of the fluids. When the hotfluid is a gas (e.g., waste heat from the combustion gas of an inter-nal combustion engine), heat transfer enhancement can be quite achallenge. This is because gases have much lower thermal conduc-tivities than liquids, and the heat transfer rate per unit area thatneeds to be achieved, depending on the length of the TE legs andthe temperature of the hot and cold fluids, can be as high as 5 to10 W/cm2. The second major challenge is how to extract a signifi-cant portion of the available energy in the hot fluid in a compact

fashion. Since the objective of a TEG is to extract energy from thehot gas, a significant portion of the available energy in the enteringhot gas must be removed by the TEG by the time the hot gas exits.Thus, one definition of the efficiency of the heat exchanger or heatsink is (Th,inlet� Th,outlet)/(Th,inlet� Tc), where Th,inlet and Th,outlet arethe temperature of the hot gas at the TEG inlet and outlet and Tc isthe temperature of the cold fluid. With this definition, the efficiencyis 1% or 100% if all of the available energy (Th,inlet�Tc) isextracted by the TEG. Otherwise, it is less than 100%. To increaseefficiency or usefulness of the TEG in extracting energy from thehot fluid, innovative designs are needed that could involve a com-plex system of microchannels (e.g., channels with height of about5 mm), where there is high surface-area-to-volume ratio. Thesemicrochannels must be linked with heat transfer enhancements toincrease the heat transfer rate with minimum pressure loss. Sincethe channel dimensions may be quite small, the flow may be lami-nar. Thus, the heat transfer enhancement techniques must inducestreamwise vorticity to entrain the “hotter” and “cooler” fluids nearthe center of the channel to the walls that sandwich the TE coupleto increase the heat transfer rate, which in turn increases the electriccurrent generation. This research should focus on both challengesof the heat exchanger/heat sink by exploring, developing, and eval-uating heat transfer enhancement and available energy extractionconcepts through computational fluid dynamics with validationthrough experimental measurements. These efforts must be coordi-nated and integrated with the mini/microchannel heat exchanger

Fig. 5 Insulation test comparing commercially available insulation and CPAI

Fig. 6 Heat and flow schematic

Journal of Energy Resources Technology JUNE 2013, Vol. 135 / 022001-5

Downloaded From: http://energyresources.asmedigitalcollection.asme.org/ on 04/02/2015 Terms of Use: http://asme.org/terms

designs to achieve optimum hot-side and cold-side heat exchangeconfigurations and designs for the ERS-APU and EGR TEGsystems.

TEG System Durability

Three major issues related to durability have been identified asimportant in past efforts: (a) Develop sublimation suppressiontechniques, (b) develop electronic controls for each couple of thehigh-temperature generator, and (c) manage stresses at interfaces.These are discussed in the next sections.

Thermal Packaging Technology to Improve Durability

Sublimation is one of the primary modes of degradation inTEGs. As volatile Sb sublimes away, the leg cross-section isreduced, causing the internal resistance to rise (Fig. 7). NASA’sSiGe-based RTGs employed Si-based coatings to suppress thesublimation of Ge and PbTe-TAGS TEGs. They use a 1 atm inertcover gas to suppress the sublimation of Te.

If skutterudite-based TEG technology is to mature, measuresmust be taken to suppress the sublimation of Sb. If operated above600 �C in a vacuum or argon, the sublimation rate of Sb fromSb-based skutterudite is above 10�3 gSb/cm2h, which could causea significant degradation in performance in a few 100 h of opera-tion. Previous work at JPL investigated various coating techniquesinvolving sol–gel, sputtering, and electrochemical processes.However, it was found that Sb vapor is highly corrosive, thusreacting with most coatings. Additionally, skutterudite has a coef-ficient of thermal expansion (CTE) of 9–11 ppm/K. There are fewceramic materials with CTEs in this range, thus making it difficultto create a stable bond. Using a 1 atm argon cover gas, as is usedin PbTe-TAGS RTGs, was also considered as a method for sup-pressing Sb sublimation from skutterudite. However, it was foundthat 1 atm argon does not significantly slow Sb sublimation inskutterudite. One of the most promising methods for suppressingSb sublimation in skutterudite happens to be the same materialtechnology that shows promise as thermal insulation: aerogel.Work conducted by JPL demonstrated that the sublimation rate ofSb can be reduced by a factor of approximately 1000 through theapplication of aerogel coatings, which are stable against sublima-tion for 2500 h at 873 K. As is the case with the aerogel-basedthermal insulation, the ability to cast the aerogel into placecreates intimate contact between the TE element and the coating.Although the aerogel is porous, the pores are nanometer to sub-nanometer range, thus creating a tortuous path for Sb vapor to per-meate, as what holds true for gas conduction also holds true formetal vapor diffusion.

A beneficial next step to improve couple durability would be astudy of the sublimation rate of gradient-density aerogel coatingsdried using various techniques. The gradients in this study wouldconsist of varying concentrations of opacifier (titanium powder)and varying silica aerogel porosity for optimum radiation scatter-ing at high temperatures and sublimation suppression, respec-

tively. Since aerogel cannot form an airtight seal, the skutteruditemodules should be sealed with an inert cover gas such as argon,either as individual modules or as a complete system.

dc–dc Boosting, System Diagnostics, and Fault

Management

The thermoelectric generator system architecture is similar to abattery system. Each TE element provides relatively high currentbut low voltage, which makes it necessary to arrange the TE ele-ments in both series and parallel connections [19,20]. Figure 8shows the high-level architecture of the proposed TEG system,where the entire TEG system consists of multiple TEG modulesand each TEG module is formed by serial connection of the TEelements (devices) to increase the TEG module output voltage.There are two main functions for the TEG control system: one isto maximize the power output of each TEG module by controllingthe output voltage of the corresponding dc–dc convertor used tosum the TEG module powers, and other is to communicate withthe TEG module controller for diagnostics and electrical powermanagement. The purpose of the dc–dc conversion is to convertthe electrical power provided by these TEG modules to a voltagesuitable for charging the battery and usage by electrical devices.By adjusting the dc–dc convertor output voltage at the summation,the power output of each TEG module can be adjusted (ormaximized).

The fault management system is capable of communicatingwith each TEG module. The fault information of each TEG mod-ule will be used to manage the maximum power output of eachTEG module, along with other system parameters such as exhaustgas temperature, flow rate, etc.

The TE elements (devices) are serially connected to increasethe output voltage of each TEG module without using dc–dc con-vertors. The disadvantage of this architecture is that if one of theTE elements fails (open circuit), the entire TEG fails since thewhole generator becomes an open circuit. To mitigate this risk, asystem of bypassing failed couples, termed CBT, has been devel-oped. The performance of each TE element is monitored (patentdisclosure filed) for its output voltage and the measured voltage iscompared against the average voltage of the TE elements to diag-nose if the corresponding element is operating normally or hasfailed (open circuit). If a TE element fails (open circuit) duringoperation, the TEG controller will short the corresponding TE ele-ment, therefore, enabling the TEG module to continue providingthe maximum power output. The TEG module will also communi-cate with the power summation dc–dc convertor control system toreport the failure information, such that the dc–dc convertor canbe optimized.

Due to the high number of TE elements contained in a TEGmodule, the TE elements can be grouped into arrays (such as 10elements per array) where each array is diagnosed and managedby a microcontroller. A supervisory microcontroller communi-cates with all TE array microcontrollers through control area

Fig. 7 High-level TEG architectureFig. 8 As Sb sublimes from the hot side of a skutteruditeelement, depletion or “necking” occurs on the hot side

022001-6 / Vol. 135, JUNE 2013 Transactions of the ASME

Downloaded From: http://energyresources.asmedigitalcollection.asme.org/ on 04/02/2015 Terms of Use: http://asme.org/terms

network communications. Figure 9 shows the architecture of aTEG module with diagnostic and fault management capabilities.The array microcontroller diagnoses each TE element (device) bysampling its output voltage and comparing it with the referenceand average TE element voltages and making a judgment if theTE element has failed or not. Nothing will be done if the TEelement is healthy. If the TE element has failed (open circuit), themicrocontroller turns on the corresponding switch (see Fig. 9) toshort the failed TE element so that the current can flow throughthe rest of the TE elements. With the help of the array diagnosticsand fault management microcontroller, the entire TEG systemfailure rate then equals the individual TE element failure rate, andhence, maximizes the entire TEG system efficiency. Using theMSU developed system, the couple monitoring and bypassingtake place at very low power consumption levels compared toother methods.

This electronic solution has significant positive implications forTE technology as it allows multiple couples to be connected inseries, thereby increasing the generator output voltage without therisk of a failed couple negating the power production of the entiregenerator.

MSU has developed and demonstrated a beta system, shown inFig. 4, which enables the following [21]:

• Integration of the TE array microcontroller with the TE arrayelements.

• Power summation dc-dc convertor development.• TEG module diagnostics and fault management system

development at low power consumption.

Structural Consideration at Interfaces

Structural integrity research of the TEG must focus on twokey elements: (a) understanding failure modes that have beenobserved at interfaces of thermoelectric material and matingsurfaces, and (b) developing designs using simulations and experi-ments to enable robust interfaces during heating and cycling [22].This will assist in developing an understanding of the nature offailures that are experienced at the interfaces between the thermo-electric material and its contacts. This usually involves differen-tials in the coefficient of thermal expansion during heating andcycling. The broad impact will be the development of a new class

of interfaces such as graded interfaces and the development ofnew machining methods or other unique interfaces that can main-tain their integrity during transients.

During recent testing of our generators (over 100–5 and 10 cou-ple modules), many of the TE couples have failed during testing,thus decreasing the efficiency of the thermoelectric generator. Fail-ures tended to occur in the P-type leg, at the skutterudite– material2 interface. It is hypothesized that thermal expansion is the drivingforce for these failures, as the different material layers comprisingthe TE legs possess different coefficients of thermal expansion.

A schematic of the TE couple is shown in Fig. 10. For the analy-sis, the symmetry of the couple is utilized to reduce mesh size andspeed up computation time. The couple is sectioned to account forthe different materials. This procedure assumes a clean boundarybetween material layers, i.e., there is no intermediate layer of mixed(unknown) material properties between two successive materialslayers in the TE legs. Such a layer would exist in the actual legs, asthey are manufactured by hot-pressing material powder. It is alsoassumed that the material properties are temperature independent;thus, we examine trends as we know these properties are not tem-perature independent. For this simulation, these properties weretaken from data with temperatures of 20–100 �C.

The finite element analysis is performed in two stages. A ther-mal analysis is performed to obtain a temperature differential ofabout 450 �C from the hot to the cold side of the couple, assumingconvective heat transfer. This approximates the temperature distri-bution of the TE couple under operating conditions. Second, astructural analysis is performed in four steps as shown. Table 1 tosimulate the cyclic thermal loading the TE module experiencesduring operation. Here, it is assumed that the TE legs are stressfree at the beginning of the analysis, i.e., there are no residualstresses from the hot-pressing process and cutting of the legs.

Figures 11 and 12 show the von Mises stresses at the legs at theend of process 1 (temperature distribution as shown in Table 1)and cycle 2 (TE module at 30 �C), respectively. Examination ofthe second principal stresses at the end of processes 2 and 4 whenthe TE module is cooled to 30 �C predicted stress levels thatwould cause mechanical failures to occur in the P-type legs at theskutterudite–material 2 interface.

Simulations attribute the causes of these failures to the exces-sive stress fields predicted during thermal loading of the TEmodule. As material 2 expands more, it pulls the neighboringmaterials along their interfaces. At the material 1 interface, largerstresses are observed. At the skutterudite interface, lower stressesare observed as the pulling action of material 2 yields the skutteru-dite close to the interface (Fig. 11).

On the cool down process, material 1 suffers little plasticity andcontracts back to its original shape. The skutterudite, however,has suffered unrecoverable expansion (plastic deformation) andtends to stay deformed. This creates a bulge (due to skutteruditeplasticity) at the skutterudite–material 2 interface (Fig. 12). Thisbulge is also observed in experiments. These opposing tendenciesof the two materials result in high stresses at the interface.

Results show the second principal stress in the negative direc-tion is the highest at the skutterudite–material 2 interface, withhigh concentrations around the corners. These high stresses canresult in fatigue after multicycle loading and can lead to separa-tion at interfaces. The stress distributions, after subsequent ther-mal loading cycles, show little difference. Sources of failure canbe due to crack initiation and growth. Imperfections in the surfaceintroduced during the cutting process of the legs or chemical reac-tions at interfaces, combined with the high stress concentrations,can promote crack growth and separation. Analysis indicates thatthe stresses are high at the skutterudite–material 2 interface.Hence, it is essential to explore paths to reduce those stresses. Anew design, which is under review, shows where the surfaceintroduction of a graded material layer (one that has mechanicalproperties between those of mating) can be used at theskutterudite–material 2 interface. Such a layer can provide asmooth transition of material properties between the skudderudite

Fig. 9 TEG module architecture

Journal of Energy Resources Technology JUNE 2013, Vol. 135 / 022001-7

Downloaded From: http://energyresources.asmedigitalcollection.asme.org/ on 04/02/2015 Terms of Use: http://asme.org/terms

and material 2, thus increasing robustness. This can be investi-gated numerically using models from the functionally gradedmaterial theory.

Summary

A study has been conducted that shows with the appropriatesupport, successful demonstration and potential commercializa-tion of TEG technology for waste heat recovery in a class 8 truckapplication are possible. As discussed, three factors that wouldsignificantly improve the energy harvesting capabilities of TEGsare performance improvements in thermoelectric materials,exhaust system/combustion chamber insulation and overall systemdesign, including required electronics. During the past five years,the efficiency of skutterudites has improved by about 50%.Options for waste heat recovery are neither limited to transporta-tion systems nor are the auxiliary power units proposed for theclass 8 truck application. To be a viable system, the mechanicalstability of hot-side interfaces must be improved. Although smallIC-engine-driven generators, fuel cell systems and bottomingcycles may currently be more efficient alternatives to TEGs, it isonly the thermoelectric system that has the capability to recoverwaste heat while the vehicle is in operation and also serve as anAPU. It is likely that the first application of TEGs for power gen-eration may be expanded versions of how they are currently used:in specialty applications such as to drive low-wattage electricaltransducers or small fans. Liquid fossil fuel, supplying energy toIC engines, will likely dominate transportation needs for the nextquarter century. Given the performance of currently available seg-mented materials and assuming modest improvements in TE andheat exchanger technology, one could expect a 2–4% improve-

ment in fuel economy for a class 8 truck, depending on the operat-ing conditions, TE material efficiency, and insulation technologyemployed.

Acknowledgment

The work described in this paper has been supported by the USDepartment of Energy, Energy Efficiency and Renewable Energy,DE-FC26-04NT42281, Samual Taylor and John Fairbanks, DOEProject managers.

References[1] Pollock, D. D., 1995, CRC Handbook of Thermoelectrics, D. M. Rowe, ed.,

CRC Press, Boca Raton, FL, pp. 7–17.[2] Hendricks, T. J., 2007, “Thermal System Interactions in Optimizing Advanced

Thermoelectric Energy Recovery Systems,” ASME J. Energy Resour. Technol.,129(3), pp. 223–231.

[3] Crane, D. T., and Bell, L. E., 2009, “Design to Maximize Performance of aThermoelectric Power Generator With a Dynamic Power Source,” ASME J.Energy Resour. Technol., 131(1), p. 012401.

[4] Demirkaya, G., Besarati, S., Padilla, R., Archibold, A., Goswami, D. Y.,Rahman, M., and Stefanakos, E., 2012, “Multi-Objective Optimization of a

Fig. 11 Process 1—stresses at operating temperature. Defor-mation magnification factor of 175 and units of MPa.

Fig. 12 Process 2—stresses after cooling to 30 �C. Evidence ofplasticity is seen in both legs. Deformation magnification factorof 175 and units of MPa.

Fig. 10 Thermoelectric couple section and materials

Table 1 Steps for structural analysis

Step Temperature field

Process 1 30 �C to 450 �C, operating temperatureProcess 2 450 �C to cool down at 30 �CProcess 3 30 �C to 450 �C, operating temperatureProcess 4 450 �C to 30 �C, cool down at 30 �C

022001-8 / Vol. 135, JUNE 2013 Transactions of the ASME

Downloaded From: http://energyresources.asmedigitalcollection.asme.org/ on 04/02/2015 Terms of Use: http://asme.org/terms

Combined Power and Cooling Cycle for Low-Grade and Midgrade HeatSources,” ASME J. Energy Resour. Technol., 134(3), p. 032002.

[5] Srinivasan, K. K., Mago, P. J., Zdaniuk, G. J., Chamra, L. M., and Midkiff,K. C., 2008, “Improving the Efficiency of the Advanced Injection Low PilotIgnited Natural Gas Engine Using Organic Rankine Cycles,” ASME J. EnergyResour. Technol., 130(2), p. 022201.

[6] Boretti, A., 2012, “Energy Recovery in Passenger Cars,” ASME J. EnergyResour. Technol., 134(2), p. 022203.

[7] Bose, P. K., and Banerjee, R., 2012, “An Experimental Investigation on the Roleof Hydrogen in the Emission Reduction and Performance Trade-Off Studies inan Existing Diesel Engine Operating in Dual Fuel Mode Under ExhaustGas Recirculation,” ASME J. Energy Resour. Technol., 134(1), p. 012601.

[8] Yun, K., Luck, R., Mago, P., and Smith, A., 2012, “Analytic Solutions forOptimal Power Generation Unit Operation in Combined Heating and PowerSystems,” ASME J. Energy Resour. Technol., 134(1), p. 011301.

[9] Hartsig, A. T., 2008, “Thermoelectric Conversion of Waste Heat to Electricity in anIC Engine Powered Vehicle: An Engine Modeling Approach,” M.S. thesis, Depart-ment of Mechanical Engineering, Michigan State University, East Lansing, MI.

[10] Ren, F., Case, E. D., Timm, E. J., and Schock, H. J., 2008, “Hardness as a Func-tion of Composition for n-type LAST Thermoelectric Material,” J. AlloysCompds., 455, pp. 340–345.

[11] Ren, F., Case, E. D., Timm, E. J., and Schock, H. J., 2007, “Young’s Modulusas a Function of Composition for an n-Type Lead-Antimony-Silver-Telluride(LAST) Thermoelectric Material,” Philos. Mag., 87, pp. 4907–4934.

[12] Ren, F., Hall, B. D., Ni, J. E., Case, E. D., Timm, E. J., Schock, H. J., Wu, C.-I.,D’Angelo, J. J., Hogan, T. P., Trejo, R. M., and Lara-Curzio, E., 2008,“Mechanical Characterization of PbTe-based Thermoelectric Materials,” Ther-moelectric Power Generation, Materials Research Society Proceedings, T. P.Hogan, J. Yang, R. Funahashi, and T. Tritt, eds., Materials Research Society,Warrendale, PA, Vol. 1044, pp. 121–126.

[13] Schmidt, R. D., Ni, J. E., Case, E. D., Sakamoto, J. S., Kleinow, D. C., Wing,B. L., Stewart, R. C., and Timm, E. J., 2010, “Room Temperature Young’sModulus, Shear Modulus, Poisson’s Ratio and Hardness of Ce0.9Fe

3.5Co0.5Sb12 and Co0.95Pd0.05Te0.05Sb3 Skutterudite Materials,” J. AlloysCompds., 504, pp. 303–309.

[14] Schock, H., Hogan, T., Case, E., Kanatzidis, M., Peng, F., Ren, F., Sakamoto,J., Timm, E., and Shih, T., 2008, “Thermoelectric Conversion of Waste Heat toElectricity in an IC Engine Powered Vehicle,” U.S. Department of Energy,Washington, DC, Quarterly Report, October 10 (submitted).

[15] Hogan, T., and Shih, T. I.-P., 2005, “Modeling and Characterization of PowerGeneration Modules Based on Bulk Materials,” Thermoelectric Handbook:Macro to Nano-Structured Materials, M. Rowe, ed., Taylor and Francis Books,London.

[16] Zhu, B., Schock, H. J., Hogan, T., and Shih, T. I.-P., 2006, “NaturalConvection and Radiation Heat Transfer in High-Temperature ThermoelectricCouples,” Aerospace Sciences Meeting, Reno, NV, AIAA Paper No.2006–0574.

[17] Harris, R., Schock, H. J., Hogan, T., and Shih, T. I. -P., 2006, “Heat Transferand Electric Current Flow in a Thermoelectric Couple,” Aerospace SciencesMeeting, Reno, NV, AIAA Paper No. 2006–0575.

[18] Hu, K., Chi, X., Shih, T. I.-P., and Schock, H. J., 2009, “Heat TransferEnhancement in Thermoelectric Power Generation,” AIAA Paper No. 2009-1210.

[19] Cao, D., and Peng, F. Z., 2009, “Zero-Current-Switching Multilevel ModularSwitched-Capacitor dc-dc Converter,” 2009 IEEE Energy Conversion Congressand Exposition (ECCE), San Jose, CA.

[20] Cao, D., and Peng, F. Z., 2009, “A Family of Z-Source and Quasi-Z-Sourcedc-dc Converters,” 2009 Applied Power Electronics Conference (APEC),Washington, DC.

[21] Schock, H., Lyle, M., Maloney, R., Moran, K., Panayi, A., Ruckle, T.,Sakamoto, J., Timm, E., Zhang, L., and Zhu, G., 2010, “ThermoelectricConversion of Waste Heat to Electricity in an IC Engine Powered Vehicle,”U.S. Department of Energy, Washington, DC, Quarterly Report, October 29.

[22] Case, E. D., Schmidt, R., Schock, H., Sakamoto, J., and Timm, E., 2011,“Mechanical Characterization of Skutterudite Thermoelectric Materials,” U.S.Department of Energy, Washington, DC, Final Report, March 11.

Journal of Energy Resources Technology JUNE 2013, Vol. 135 / 022001-9

Downloaded From: http://energyresources.asmedigitalcollection.asme.org/ on 04/02/2015 Terms of Use: http://asme.org/terms