Journal of the Institute of Energy, June 1996,69' pp 80'86

Combustion measurements in an industrialglass-melting furnaceM COSTA, U VOUNAO, J BALTASAR ANd M G CARVALHO*

This paper describes the results of an experimental study undertaken in an industrial regenerative 'horseshoe'-ttow-iype oil{ired furnace for the production of glass containers. The data reported include in-flame gas speciesconcJntrations of 02, CO, CO2 and NO,, and gis temperatures,.for one of the furnace operating conditions. Theresults reveal: a) in-iiame peaft temperdtures of about 1690 'C; b) exhaust-gas temperatures of about 1520 'C;aM c) exhaust lbvels of Cb at 0.1%voland NO, at 580 ppm. The data presented are especially usefulforevaluation and development of 3-D mathematical models.

1 lntroductionThe increased interest in mathematical modelling by man-ufacturers of full-scale combustion equipment as a meansof determining design improvements cannot be over-emphasised. Mathematical models demand reliable anddetailed data for their validation. Unfortunatelr'. at presentthere are relatively little data from full-scale plants, and toa large extent thii is because of difficulties in controllingand rieasuring such flames.This is especially the case in theglass industry.as a consequence, validation of mathemat!cal-modelling proposals against full-scale data is virtuallyimpossible for glass-melting furnaces. For some years oneof

-the present group's main areas of research has been the

modeliing of industliat glass furnacesi.The group haslongsuffered in insufficiency of good full-scale data, and themain objective of the present work is to help redress theproblem.

*Mechanical Engineering Department, Instituto Superior T6cnico,Av Rovisco Pais" 1096 Lisboa Codex, Portugal.

This paper reports new combustion data obtained in anindustrial glass-melting furnace, the property of the SantosBarosa Company, which is in Marinha Grande, Portugal'Measurements have been made for in-flame mean gasspecies concentrations of 02, CO, CO2 and NO*, and meangas temperature. These data supplement the limitedamount from the other related study, notably Cassiano etaP, who reported results for a regenerative 'horseshoe'-flow-type oil-fired furnace similar to that used in the pre-sent study. Data from other sources for furnaces of thistype are scarce. Barklage-Hilgefort & Siegerr have studiedthe effectiveness of flue-gas recirculation and air-staging incontrolling NO* emissions. That study took place in anindustrial glass-melting furnace but it involved only flue-gas measurements. Nakamura et alahave also investigatedNO*-reduction methodg and have released detailed in-flanie data-but for a scaled-down compartment of a full-scale glass-melting furnace.

The remainder of this paper is in five sections: describingthe industrial glass-melting furnace and instrumentation;presenting the experimental conditions, results and relateddiscussions; and summarising the most important conclu-sions.

Left firingcycle

Chequerworkbricks

Melting zone

Fig.1 Schematic view of furnace.

Regenerators

Chimney -...-.-

80

COS]'A: COMBUSI-lON IN A GLASS-Mlll-l-lNG ITURNACE

Sample out

,Cotton wool

8l

The rear wall arrangement.

Zerogas

Waterout

Condenser

OiaphragmpumP

The combustion air, which is preheated in the regenera-tors, enters the furnace through'the port above the burnerat a downward angle of 14"; the flame forms a.loop withinthe combustion chamber, and exits through the paltnerport. The furnace is fired from each port alternately, toenhance the uniformity of heat-transfer to the glass sur-faces, and to allow regeneration; the regenerators areswitched every 20 minutes. In the combustion chamber arethree square inspection ports (L35 x 135 mm) for probingand viewing, as shown in Fig.l. In Tiiers can be found adetailed description of a furnace of this type, and its oper-ation.

-:-ilMll

g3mm

l_T-

Fig.3 The water-cooled probeand the gas-species analysis system.

Spangas

2 The indirstrial glass'melting furnace

Fig.1 is a schematic view of the glass furnace used in thepresent work; it is a regenerative 'horseshoe'-flow-typeiurnace firing heavy fuel oil. The combustion chamber isabove the tank in which the melting and refining of theglass take ptace; the melting area is 48 m2, and the refiningirea is 24 n:?. Fig.2 illustrates the arrangement of the twocombustion-air/combustion-products ports and of the twoburners located at the rear wall of the combustion cham-ber; all the relevant dimensions are shown in mm. Eachburner comprises three air-assisted nozzles, inclinedupwards at an angle of 9 ".

(co, co2)

oooao.tieo

82 COSTA: COMBUSTION IN A GLASS-MELTING FURNACE

.9S60-CCt3tc>o>()*o

3 e o.zs=o-E P o.zoo:o o O-1Sc>Oro :L o-lo69o o.o5

0.00

goG:!o-E6595b>o-fltlo-o

E=oL=agr=qoEOJ.q3-, e.og.

21.020.520.019.519.018.5

18.0

17.517.0

700

600

500

400

300

200

100

0o 20 40 80 80 100.120 140 160 180 200

Distance to the Wall (cm)

Fig.4 Measured gas-species profiles along port A, taken ontwo separate days.

oaaoo

o--oaaaaa

'6.rt*aI . **ul

3 Measuring techniques and experimentaluncertaintiesThe sampling of gases for the measurement of Ot, CO, COtand NO* was achieved by means qf a water-cooled stain-less-steei probe, whose design and dimensions are shownin Fig.3. It comprised a central tube of 3 mm i/d throughwhich quenched samples were evactnted; this was sur-rounded by two concentric tubes for cooling.

Fig.3 also shows a schematic of the gas-analysis system.Thesample was drawn through the probe and part of the sys-tem by a 100% oil-free diaphragm pump, and a condenserremoved the main particulate burden and condensate. Afilter and drier removed any residual moisture and parti-cles, so that a constant supply of clean, dry combustiongases was delivered to each instrument through a manifoldto give species concentrations on a dry basis. The probewas cleaned frequently by being blown back with high-pressure air to maintain a constant suction flow rate. Theanalytical instrumentation included a magnetic pressureanalyser (Horiba Model CFA-321A) for O, measure-ments, and non-dispersive infra-red gas analysers (HoribaModel CFA-311A) for measurements of CO, COr andNO,. Zero and span calibrations with standard mixtureswere performed before and after each measurement ses-sion.

The major sources of uncertainties in the concentrationmeasurements were associated with the quenching ofchemical reactions, sample handling and analysis, andaerodynamic disturbance of the flow. Fast quenching of thechemical reactions was achieved, and our best estimatesindicated that uncertainty due to the solubility of CO2,UHC and NO, in water was negligible.The error due to thedissolution was estimated by operating the sampling

3.5

3.02.5

2.O

1.5

1.0

o.5

0.00.30

r: I

tilr]t+--i-l

/-l:-=E\

3fi) pm wlrasPt/ Pt: 13% Rh

ThcrrnocoupleJunctlon

4 nm o.d.twln-boro

alumlna tubc

a a

o

aa!ot:

A 4mm

Fig.5 The temperalure-measurement syslem

COS-IA: COMBUSl'lON lN A GLASS-MELfING ITUIINACE

Thble I Furnace operating conditions

83

system with the probe supplied with samples of standardmixtures with compositions similar to those in the flamesstudied. No attempt was made to quantify the probe flowdisturbances. The repeatability of the data was checkedregularly during each experimental session, and on sepa-rate days. Fig.4 shows data obtained on separate days. Areasonable agreement is demonstrated, without which theusefulness of the data for the validation of mathematicalmodels would be questionable.

Measurements of gas temperature were obtained bymeans of uncoated Pt/Pt13o/oRh thermocouples of 300 pmwire diameter.The wires ran in a twin-bore alumina sheathwith an external diameter of 4 mm, placed inside a stain-less-steel water-cooled probe (Fig.5). As radiation lossesrepresent the major source of uncertainty in the measure-ments of mean temperature, an attempt was made to quan-tify them on the basis of a theoretical expressiondeveloped by Dee. The equation requires that tempera-tures be measured with three thermocouples of the samematerial but of different diameters (in the present work 40,80 and 300 pm). The calculation indicates that in theregions of highest temperature the 'true' temperatureexceeds the measured one by about 10%.As in the speciesmeasurements, the repeatability of the data was also eval-uated for the temperature measurements. Fig.6 shows atypical result where acceptable repeatability is demon-strated.

Both probes were mounted on a traverse mechanism thatallowed for movements along a line normal to the furnacewall through each inspection port, up to 2 m from the wall.The positioning of the probes in the furnace was accurateto +-1,0 mm. The analogue outputs of the analysers and ofthe thermocouple were transmitted via A./D boards to acornputer where the signals were processed and the rneanvalues were computed.

4 Experimental conditionsA No.6 heavy fuel oil was used during the present mea-surements. Its main properties, in 7owt, were: 88.4 carbon'9.3 hydrogen,2.0 sulphur, <0.2 nitrogen,0.04 ash; the HCVwas 43.95 MJ kg-t.1u6le 1 summarises the furnace operat-ing conditions for which the results reported herewithwere obtained. The values are averaged over the one-month duration of the experimental programme.

t700

Colour of glassOutput, t day-tSpecific melting rate, t m-2 day-lSpecific energy consumption, GJ t-l

Arnbert66.3z.J4.9

Fuel: flow rate, kg h-lpreheated temperature, oCpressure, bar

827.41132.8

Atomising air: flow rate, kg h-lpreheated temperature,oCpressure, bar

105950)A

Air: flow rate, kg h-lpreheated temperature, oC

10 5921400

Air:fuel ratio I4

0 20 40 60 80 100 120 140

Dlstance to the Wall (cm)

Fig.6 Measuredgastemperatureprofilesalong port A, taken on two separate days.

Crown temperature Zr, "CCrown temperature Ir, oC

Crown temperature Ir, oC

Bottom temperature Zr, oC

Fig.7 shows the location of the thermocouples at the crownand bottom of the furnace. Note that after the reversal ofthe regenerators a period of about 2 minutes was allowedfor the furnace to reach steady-state conditions, afterwhich data-collection began.

5 Results and discussionAs mentioned earlier, the measurements reported in thissection were made along ports A, B and C (Fig.7). Notethat ports A and B are in the melting zone whereas port Cis in the refining zone. Figs.8 to 12 show the profiles of Or,CO, CO2 and NO* concentrations and of gas temperatures,respectively, for the flame studied obtained at the threeinspection ports available.The meaning of left and right inthese figures can be discerned from Fig.7: the left profllescorrespond to the visible flame zone, and the right profilescorrespond to the'exhaust zone.

At port A the data for the left-hand firing cycle reveal aninner region, 100-200 cm, of intense combustion-as indi-cated by the low O, concentrations and relatively high COand CO, concentrations. The high CO concentrations(>6%vol) in the region between 140 and 200 cm indicatean intermediate stage in the reaction, because of insuffi-cient mixing within the fuel jet. The temperature peak ofabout 1690 "C,at120 cm, indicates the location of the flameboundarv where mixing is intense as the result of strongshear between the fuel spray and the combustion air. Notethat the temperature profile at this port is similar to that ofNO* concentration, which suggests that NO* is formedmainly via the thermal mechanism, as expected at thesehigh temperatures.

Further downstream, at ports B irnd C, for the left-hand fir-ing cycle the Or, CO, and NO* concentrations and the tem-perature profiles flatten progressively, and the COconcentration diminishes as the result of conversion toCOr. The conversion of CO to CO, appears to be slow,given the significant distance between ports; specifically,the CO concentration falls from a maximum value of7.7"/"vol at port A to a maximumof.2.3 at port B, with bothof these maxima located at a distance of 180 cm from thechamber wall. At port C the CO concentration still pre-sents a maximum of about 1%vol located at a distance of140 cm from the wall.

1531r50213951249

"o t65o

o 1600

E 1550o+ lsooEoF 1450

't400

loaaaaao'6 o. o o o o o o o

160 180

BC

84 COSTA: COMBUSTION IN A GLASS-MELTING FURNACE

, 3650 , 3300 | 27OO Ir--r-r__*1

A

16111210

I__1_ 6lrsl S 2bl e olE1613 14

a) Crown

Fig.7 The glass furnace, showing the measuring portsand the location of thermocouples.

ll I 'a

0 20 40 60 80 100 120 140 160 180 200

Dlstance to the Wall (cm)

Fig.9 Measured CO, concentration profilesalong ports A, B and C.

Port B

aI

1oaoo1.lr-

1

040 60 80 100 120 140 160 180 200

Dlstance to the Wall (cm)

Fig.8 Measured 02 concentration profilesalong ports A, B and C.

oaPort A

b) Bottom

brzE-10co8=g654(r^czoo0oo

5

4

3

2

0oEzsfE>20egtsc210Gc5oo5ood25oo20

I7654J2

SlIo!89t$o

5c€4!ec^oao61oO8B

5

0

Port C

rtrit

7654321

0 llllttlllro 20 40 60 80 100 120 140 160 180 200

Distance to lhe Wall (cm)

Fig.10 Measured CO concentration profilesalong ports A, B and C.

Port A

a

arataltlll

Port C

t:rrr:lro

lttltlll

aaloa

Port A

raslrtlrtt'.lloaaa

Port B aot::.I

COSTA: COMBUSTION

Thble 2 Comparison of present resultsand those of Cassiano er cPMeasured amount Cassiano Present work

IN A GLASS-MEUI'ING ITURNACE

500

400300

o 20 40 60 80 100 120 140 160 180 200

Distance to the Wall (cm)

Fig.11 Measured NO" concentration profilesalong ports A, B and C.

ti5

Maximum flame temperature, oC

Temperature of exhaust gas,'CMaximum concentration of CO, %volConcentration of exhaust CO, %volConcentration of exhaust NO*, Tovol

1690t520

7.70.1580

ln the refining zone, at port C all the data profiles for theright-hand firing cycle, particularly those of temperature,are similar to those for the left-hand firing cycle (as itshould be), revealing the good performance of the presentturnace. Moving towards the chamber exit from port C toport A, the data for the right-hand firing cycle show: a) arise in the Or concentration from 0.5 to about 1.8%volbecause of the mixing with gases from the inner regions ofthe flow; b) a fall in the CO concentrations because of theoxidation to CO2; c) a marginal rise in the NO* concentra-tion, which indicates formation of thermal NO* yet in theexhaust region as the result of the persisting high temper-atures. Near the exit of the combustion chamber, at port A,the levels of CO and NO* are about 0.1%vol and 580 ppmrespectively.

Finally,Thble 2 presents a comparison between the presentsignificant results and those of Cassiano e/ 4P. Note thatihe data from the two studies are significantly consistent,with the exception of the NO* exhaust levels.The in-flamemeasurements of Cassiano et al appear to indicate eithersome limitations in fuel breakdown or lack of correct mix-ing of the combustion gases in the furnace. Undoubtedly',these are the reasons for the higher CO exhaust levelsobserved in that work. But these reasons may also play arole in the higher NO* emissions that the same workershave reported. In any case, the absence of idformation on'for example, fuel-nitrogen content, fuel atomisation andmixing efficiency prevents the interpretation of this dis-crepancy, which could not be traced to failings in the pre-sent measurements.

6 ConclusionsMeasurements have been obtained in an industrial glass-melting furnace. The results include in-flame gas speciesconcentrations of 02, CO, CO2 and NO*, and gas tempera-tures for one furnace operating condition. The main con-clusions are:

1 The CO and NO* exhaust levels were 0.17o vol and 580ppm, respectively; the temperature of the exhaust gaseswas approximately 1520'C.

2 Overall, the in-flame temperature profiles revealed aremarkable uniformity, especially in the refining zone; thisis evidence of the high performance of the present glass-melting furnace.

3 The data presented are of special importance to thecritical evaluation and development of 3-D mathematicalmodels, because of their detail and of the care exercised toensure reliability.

1660154050.54000

^mOER 10o

9o5 zooEu** soo

looo€ goo6: 200cE 1ooco0or 7ffi9 eoo

500400300200100

0

ooot(loo.EoF

1700

1650

1600

1550

1500

1450

14001700

1650

1600

1550

1 500

1450

14001700

'1650

1600

1550

1500

1450

1400

tttrflttto

Port C

0 20 40 60 80 100 120 140 160

Dlstanc€ to the Wall (cm)

Fig.12 Measured gas{emperature profilesalong ports A, B and C.

ala)ntrt-

aoaaa'

f r::t;t:2

r:::::lr:

O1

aaa

rlrr

a

llrll

Port A

-a'aloero'trlrrtlrl

Port B

180 200

86 COSTA: COMBUSTION IN A GLASS-MELTING FURNACE

7 AcknowledgementsFinancial support for this work was provided by the glassplant of Santos Barosa Vidros SA, (Marinha Grande.Portugal) and is acknowledged with gratitude.The authorswould like to thank the personnel of the glass plant fortheir valuable assistance during all the stages of the exper-iments, particularly Mr Victor Domingues. The authorswish also to thank undergraduate student Paulo Silva andtechnician Manuel Pratas who assisted them in the workpresented here.Thanks,are also due toTito Silva, who vol-unteered useful advice. and constructive criticism duringthe work; and to Rita Maia and Jorge Coelho for their co-operation during the preparation of the manuscript andthe figures.

I References1 CARVALHO M G, SEMIAO V S and COELHO P J. Modelling andoptimization of the NO formation in an industrial glass furnace. IraruASM E J Eng for Industry.!992, 114, pp 514-523.2 CASSIANO I HEITOR M V and SILV.A T F. Combustion tests of anindustrial glass-melting furnace. Fuel, 1994,73, pp 1638-1642.3 BARKLAGE-HILGEFORT H and SIEGER W. Primary measuresfor the NO* reduction on glass melting furnaces. Glasstech 8er,1989,67.pp 151-157.4 NAKAMURAI SMART J P and VAN DE KAMP W L.The effectsof fuel air mixing on NO, reduction and heat transfer in high tempera-ture, gas fired, giass melting furnaces. Institute of Energy l,st Int Conf onCombustion and Emissions Control, Cardiff, September 1993.5 TRIER W. Glass Furnaces-Design. Construction and Operation.(English trans.) Society of Glass Technology, Sheffield. 1987.6 DE D S. Measurement of flame temperature with a multi-elementthermocouple../ Inst Ene r gy, 1981, 54, pp 1 13-1 i6.

(Paper received November 1995.)