PB98-195698
Boilers Coficing Hazardous Waste
Effects of Hysteresis on
Performance Measurements
(U.S.) Environmental Protection Agency
Cincinnati, OH
Apr 88
J
-------
L PA/•=>'''•')/D-8 3 •
April 1953
BOILFRS COFIfclHG HAZARDOUS WASTE: ETFtCTS OF
HYSTERESIS ON PERFORMANCE MEA
Ivars J. Lids
Thermal Destruction Branch
Alternative Technologies Divi
Hazardous Waste Ergineerlng Reesearch Laboratory
and
Howard B. Mason
Acurei Corporation
555 Clyde Avenue
Mountain View, California
ETA Project Officer
Ivars .!. Llcls
H.-V!ARPOrS WASTE ENdXEERI* RCSLARCH LABORATORY
OFFICE OF KFSEARCH AND DEVELOPMENT
I'.S. E.NTIRON*MF.XTAL PROTECTION AGENCY
CIXCIXXATI. OH 45263
-------
fttil It
i'**' 9» /»<• rr*m* bmrt tv-
A««O futtr
Boilers Cofirlng Hazardous Naste: Effects of
Hysteresis on Performance Measurements
DA?t
A:-ril 198?
rtft*OftM*«C 0*GA*li.;ATtON CO DC
Q» !*<
Licis, Ivars J. and Mason, Howard B,
MAMC **«O AOO*ESS
IDS, AID, HWERL
Cincinnati, Ohio and
Acurex Corporation
Mountain View, California
'0
ny>Q .
CDNlBACT.
*»AMt
O*
Offke of Research and Development
Hazardous Haste Engineering Research Laboratory
U.S. Environmental Protection Agency
'Cincinnati, Ohio 45268
ACf MC<
pA /(t,r»n/j 2
'5 SUp*ltMf NTAMV
The Haiardous Waste Engineering Research Laboratory (HWp%i_) ha$
conducted full scale and pilot scale boiler testing to determine hazardous
waste destruction and removal efficiencies (ORE's) and other associated boiler
performance parameters during the last five years. This effort investigated
several anomalous boiler performance areas dealing «lth non-steady and off-
design boiler operation and Included measurements of volatile and seal-volatile
organics in the flue gas as well as the fate of some hM»y »etals Introduced
into the system. The tests were performed at pilot scale §nd investigated
hysteresis effects, phenomena detected during earlier full-scale testing These
hysteresis effects are defined as the propensity for a boiler to retain certain
organic components of the feed material (beyond the duration to a typical samplina
protocol}t to bt released over a longer time frawe In an unspeciflc Banner.
Additionally, a long time effect has been observed that aanifests itself as a
gradual buildup of background concentration during successive testing of initially
clean hardware. This effort attempted to Identify the presence of these effects
beyond random variation, backgnjynd concentrations and sanpHng problems such is
sample contamination. These tests were partially successful *t characterizing
hysteresis phenomena as we',1 «$ establishing • qutllutlve significance of these
effects on the accuracy of ORE §nd rther performance •eisurewents.
11 r won OS AMO 0OCUMCWT **t*i.r«
-------
N01ICI
This document has been reviewed in accordance with
U.S. Lnvironroental Protection Agency policy and
approved for prblication. Mention of trade na-es
or commercial products does not constitute endorse-
ment or *ecom»endation for use.
-------
BOILERS COF1RINS HAZAP.DO-JS WASTE: EFFECTS Of HYSTERESIS OS
' PEPFOPH&NCE MEASUREMENTS
Ivars J. Lids*, HVERL, USEFt; Howard E. Masci, Acurex Corporation
ABSTRACT
The Hazardous Waste Engineering Research Laboratory [HWERL] has
conducted full scale and pilot scale boiler testing to determine hazardous
waste destruction and removal efficiencies [DPE's} and other associated
boiler performance parameters during the last five /ears. This effort inves-
tigated several anomalous boiler performance areas dealing witM ncn-steady
and off-design boiler operation a^d Included measurements of volatile and
semi-volatile organics 1n the flue gas as well as the fate of so^e heavy
metals Introduced Into the system. The tests were performed at pilot scale
and Investigated hysteresis effects, phenomena detected during earlier full-
scale testing. These hysteresis effects are defined as the propensity for a
boiler to retain certain organic components of the feed material [beyond tie
duration of a typical sampling protocol], to be released over a longer
time frame In an uspecified manner. Additionally, a long time effect has
been observed that manifests Itself as a gradual buildup of background con-
centration during successive testing of Initially clean hardware. This effort
attempted to Identify the presence of these effects beyond random variation,
background concentrations and sampling problems such as sample contamin-
ation. These test were partially successful! at characterizing hysteresis
phenomena as well as establishing a qualitative significance of tiese effects
on the accuracy of ORE and other performance measurements.
Introduction
The USEPA has been Investigating the performance of boilers and other
industrial furnaces and their ability to destroy hazardous waste cofired with
a convential fuel such as natural gac, oil, or coal. The Investigations
Included the testing of over 14 full-scale boilers, and a number of labora-
tory and pilot scale tests. The tests typically determined ORE, the main
organic constituents In t!ie flje gas, characterized effluents and residues
and recorded the associated boiler operating parameters and continuous
emissions monitoring parameters such as 02* CO, COz, SOX and NO, and total
unburned hydrocarbons (TUHC).
The results of these Investigations showed that high levels oT DREs
could be expected (a mass weighted average for the 14 full-scale tests of
99.998 was calculated)(l) over a broad range of boiler designs, waste
characteristics and operating conditions.(2)
The ORE values during sub-stoichiometric air settings, malfunctioning
burner atomizer tests, and waste flow transients were found to be uniformly
high and very similar to those during normal operation.
Three full-scale boiler tests were conducted at non-steady off-design
conditions tc map the ORE window and identify operating conditions which
could cause ORE values to fall below the acceptable four nines limit. No ORE
limits were found within the constraints of allowable plant operation during
the three tests.
A number of phenomena Identified during the full-scale boiler testing
were "eft unexplained however, and certain questions pertinent to boiler per-
formance 1n destruction of hazardous waste were left unanswered.
During several of the full-:cale tests, chlorinated organics were ob-
served in the baseline data at the same order of magnitude, a day after
I
-------
cofiring of these substances had ceased. Between the earlier tests and the
baseline sample the following day the boiler had been operated on natural gas
only. Instrument error and contamination were ruled out.
Soot blowing 's a routine method by which the heat transfer surfaces are
prtened to maintain heat transfer efficiency. limited data were obtained
during soot blowing cycles indicating that higher organic concentrations were
detectable in the gis phase. However only the gas phase was analyzed.
Organlcs adsorbed or otherwise present on the part: -late were not assessed.
• The hazardous organlcs present In the flue g- • •£ typically separated
Into two groups. The principal organic hazardous Constituents (POHCs) are
used for determining ORE. The remaining hazardous oryanics are sometimes
referred to as Products of Incomplete Combustion or PiCs. The PIC's, as
defined for the full scale boiler testing, were the semlvolatlle and volatile
organic (other than POHC's) detected via the modified method 5 (««5) and the
volatile organlcs sampling train (YOST) respectively. For the full-scale
boiler tests, It was found that, while the average DSEs were uniformly high,
the concentration of PICs relative to POHCs were at ratios ranging from less
than one (1) to 700. This phenomenon was considered significant 1n that
available data showed PIC to POHC ratios were generally one (1) for hazardous
waste incinerators. A ten-fold Increase 1n PICs could have the sane effect
on stack emissions as a reduction of t*e ORE by one nine (i.e. a starting
99.99 would become 99.9).
Limited data (one test) were obtained at full-scale to determine
the fate of heavy metals co-f1red 1n boilers as part of hazardous waste.
Because the metals are not destroyed by the thermal process, data regarding
the fate, form and particulat* size distribution are Important for evalatlng
boiler performance.
The above topics were Incorporated Into a pilot-scale test program to
complete the data base Initiated by the full-scale test program. This paper
focuses on the pilot-scale effort and, more specifically, on the hysteresis
effects Investigated and their significance to the measurement of boiler
performance.
Defining The Problem
The protocols tor assessing boiler performance Include the determination
of their ORE. The calculation of ORE requires the sampling and analysis of
the boiler flue gas for hazardous organlcs Identified as POHCs. In addition,
it has been of Interest to determine the remaining organic constituents
because the full-scale boiler tests have Indicted that. In relation to In-
cinerators, boilers can produce larger amounts of PIC concentrations.
The methods for sampling organlcs have evolved over the duration of
hazardous waste Incineration testing and at present are basically a volatile
organlcs sampling train (VOST) and a Modified Method S train (MM5) for semi-
volatlles (3). Both trains arc usually necessary to characterize the organic
destruction efficiency of a boiler. For many cf the boiler tests a simpli-
fied VOST train called a m1n1-VOST has been used 1n addition to the VOST.
The three organlcs sampling trains concentrate the organlcs In the flue gas
by adsorbing them onto polymeric resins over a period of time. The t1«e
required for each sampling train Is a function of expected concentrations in
the flue gas and detection limits of the organic species to be analyzed.
Typical sampling times encountered during boiler testing are:
VOST - 20 m1n.
mlnl-VOST - 10-15 mln.
MM5 -3-4 hrs.
-------
The resulting concentrations of organics calculated from tnese
ments are average values over the span of time set-pled. The duration of a
t/pical test point 1s therefore defined by the KM5 sampling Interval. One or
mere VOST or mlnt-YOST samples **y be taken during t^i* Interval.
In order to dero'strat? a L>E of 95.9?, the se^piing and an./.ysis
methods must be able to provide resolution to the tMrd ded*-- ^lace (I.e.
99.999) or detect the target substances at a concentration '.'iju.OOO of the
feed value. Man/ test series (for both boilers and Incinerators) have
relied on a method that enhances resolution of the target substances by in-
jectlng surrogate organic* species with the waste (or Instead of the waste).
The surrogates are Injected at feed concentrations to provide flue gas con-
centratlons above detection limits at the required levels of destruction.
The use of surrogates and sampling trains in the manner discussed pre-
sumes that conditions within the boiler are generally at equilibrium djrin;
the sampling period. Specifically:
1. The combustion products within the boiler system come to an equi-
librium within a short period of time (for liquid injection units this wojl d
be on the order of the gas phase retention time).
2. The combustion products within the boiler sj-steri generated by the
injected surrogates closely follow the injection schedule (I.e. t^cy start
and stop at approximately the same times).
3. For the purposes of sampling organlcs the boiler firing history is
Insignificant.
During full-scale boiler testing, data were obtained that appear to
conflict with the above presumptions.
Case Study: A boiler burning non-chlorinated waste was being tested for
POHC and PIC destruction by Injection of chlorinated organics with natural
gis used as the primary fuel. At the conclusion of testing, the boiler was
left running overnight burning pure natural gas. On the following morning, a
baseline run was conducted using oil as primary fuel. Although no chlorina-
ted components were present 1n the feed the flue gas was found to contain
chlorinated species on the order of magnitude of those recorded during con-
elusion of the previous day's testing. A repeat test confirmed the first
readings.
On the strength of this example and other similar data a pilot-scale
effort was launched to investigate the existence and impact of hysteresis
effects on boiler performance measurements.
Toward Solving the Problem
In order to provide a test platform for investigating hysteresis effects,
a pilot-scale facility was configured and operated in a manner simulating the
tine and temperature profile of an industrial boiler. Figure 1 illustrates
the operating envelope of the simulator and also the time/temperature profiles
of a typical industrial boiler and, for comparison, a coal-fired utility
boiler. The test facility Is discussed in more detail later.
A pilot-scale test facility was selected to enable exploration of the
boiler operating envelope at non-steady and off-design conditions beyond
those possible at a full-scale field Installation. Additionally, closely
controlled test conditions would have been difficult to maintain at an
operational full-scale facility.
-------
2.500
2,000
+-•
CO
fe
a
1,500
1,000
Coat-fired utility
Industrial package
watertube
Pilot-scale
5 cooling coils
Pilot-scale no cooling
1 2
Residence time (sec)
Figure 1* Boiler ti»e/t«iper«ture profiles.
-------
Test Objectives
The pilot-scale test program included a larger effort for characteri-
sing off-design and unsteady state boiler operations as well as a smaller
test series cofiring heavy metals. Urder the hysteresis portion of the
program two objectives were identified:
1. Determine if hysteresis effects are present at levels significant to
C?L and PJC measurements.
2. Characterize the hysteresis effects in teTs of maximum values and
time variations.
Test Description
The pilot-scale facility 1s a f iei
-------
Convective section
Hot sampling ports
Heat exchange -^
tube bundles V*
Cold sampling -,
station \
Tobaghouse
Ashpit
Figure 2% Pilot-scale boiler simulator.
-------
!• tlcct
I* Qhl
1. PI lot -«»r.tl «» hnttfr
lr .
-------
sect to*
Ports 8
Potmtlal heat
e»cha*ofr
locations
Figure 4. Sampling locations and potential conv«ctlve heat exchanger
locations In the pilot-scale test facility.
-------
recorded every 5 minutes. The bailer operating parameters such as terpera-
ture pressures and flow rates were a'iso recorded.
The Investigation of hysterp$«s effects involved t*iree test series.
The organic* data was acquired by V05T, tr.ir. i-VOT , e~d »*^5 sa^ir-g tra^s.
Because the overall test program was one of screening for the in.poct of
various phenomena on OPE and PiC concentrations, only a few data points were
taken In each category and emphasis was placed on the volatile orgar.ics and
the rrini-V05T technique. The rationale was that beside this being a much
more economical screening method -over the MV5), the volatile organics we'-e
considered highly significant in this area and the ^hort sampling ti.-ne (10-15
minutes), on-site data analysis capability (ideally 20 min. cycle) provided a
feed-back loop that was very valuable in steering tie direction of the test-
\r-:.
A baseline test series provided information en background levels cf
riCs and POHCs when firing fuel unly (gas, oil and coal). A second series
was conducted cofiring waste with oil and gas (coal was cofired dying other
test series not involved in ireasuHng hysteresis).
The boiler was tested by
cofiring waste during the day, taking a hysteresis sample shortly after
cessation of cotiring (< 2 hours) and left overnight while burning natural
gas only. A hysteresis sample was taken the successive morning prior to
cofiring. On so^e occasions the unit was left on natural gas ove»- the
weekend and a hysteresis samnle was taken Monday morning, prior to start of
cofiring tests.
Another test series was comprised of taking numerous minl-YOST samples
during steady-state operation to establish the width of the noise band (i.e.
the amplitude variations in organics concentrations with the feed rate and
other operating parameters held as constant as practical).
Additionally, two soot blowing cycles were simulated. Soot blowing was
a manual operation performed with a compressed air lance. The concentrations
of organics 1n the gas phase were measured before, dur;ng and after each
cycle. The first soot blowing cycle was performed with the boiler cofiring
*6 oil with the surrogate waste. The second was performed later 1n the
test series with the boilers burning natural gas only. The timing of the
second test came after several test series had been conducted that were
expected to produce significant sooting and deposits on the boiler Interior.
Results
The main objective of the hysteresis testing for this series was to de-
termine If significant hazardous organics were present In the flue gas after
cessation of hazardous waste cofiring. Data were taken at three general
Intervals. These were < 2 hours, 14 to 24 hours (overnight) and an "extended
period' (after a weekend of burning natural gas only).
Table 1 presents ratios cf average hysteresis values for all PICs and
the individual POHCs measured < 2 hours after cessation of cofiring divided
by the average concentration values obtained from the gas and oil baseline
runs, taken before any_ cofiring had taken place. The table illustrates the
presence of hysteresis. Lack of residual effects would place these values
near unity.
Table 2 presents similar results by comparing the same average values
of < 2 hours hysteresis with average values obtained during cofiring of
-------
gas and oil. The PIC's measured during the pilot scale program are oe; n,=u
as volatile, chlorinated organics (other than t^ose identifies as POH.C's)
delected via the VOST and tne m'ni-vOST. The h/steresis effect is illustrated
here because without it the values In this case should be \ery small ard
definitely less than unity.
The values of Tables 1 ar..j 2 were normalized to take into account fuel
waste flow and airflow rates.
Table 1. HTSHRISIS EMISSION RATIOS
Average Hysteresis Concentrations < 2 hours After Cofiring
vs. Avenge Baseline Concentrations
CoT-pound Hysteresis Hysteresis
Oil Baseline Gas Basel ine
PIC 0.1 10.38
CC14 2.0 12.11
TC£ 0.7 17. S
PC5 149.0 NA (No EL)
where: CCl4«carbon tetrachloride; TCE«trichloroethylene; MC
PIC'S « volatile, chlorinated organics detected vi* VOST or miri-vCST
Table 2. HYSTERESIS EMISSION RATIOS
Average Hysteresis Concentrations <2 Hours After Ccfiring
vs. Average Cofiring Concentration For Oil ar.d Gas
Compound Hysteresis Co"c. Hysteresis Co'*-: .
Oi I Cofiring tone. Gas Cofiring Cone.
PIC 0.45 9.?3
CC14 2.0 1.79
TCE 0.93 7.19
KCB 7.0 44.0
Table 3 represents the levels of organic concentrations found during the
natural gas firing portions of this test series for which the most abundant
data points were taken. Based on this data, 1t can be seen that the total
Table 3. AVERAGE CONCENTRATIONS OF CHLORINATED COMPOUNDS
FOR GAS/WASTE COFIRING AND HYSTERESIS EFFECTS (ng/1)
Total Chlorinated €^4 TCE MCB
Species
Gas/Waste Cofiring 4.25 0.47 0.04 0.24
Hysteresis Concentrations
(< 2 hr after cofiring) 3.93 0.57 0.25 0.93
Average Hysteresis Concentrations
(all points for gas firing) 3.16 0.28 0.21 0.10
chlorinated species average concentrations during cofiring drops to about 921
of the original value In the < 2 hour timeframe and to about 741 when con-
sidering the overall average taken at three time Intervals from < 2 hours to
several days. It Is Interesting to note that the average concentration of
10
-------
CC1« at < 2 hours after coflrlng Increased to 1211 of the coflrlng value
while monochlorobenzene appears at 3881.
The longer ter» effects were analyzed by comparing hysteresis levels
early In the test series (the first hysteresis sample taken after the first
coflrlng waste using gas at 'normal* operating conditions) with subsequent
hysteresis test • <./•»!$. The test series was started using clean hardware.
The hysteresis c'..«cntrat1on data were compiled for the total chlorinated
species and the three Individual chlorinated volatile POHCs over the duration
of the test series (see Figure S). While hysteresis effects for the Individual
POHC remain approximately the sane, the total chlorinated species (PICs and
POHCs) seem to Increase with Increasing fuel/waste firing.
Data from the two soot blowing cycles were also analyzed. Generally
concentrations of POHC and PICs were highest during soot blowing and lowest
after the cycle was completed, while values before soot blowing were between
the two extremes. These concentrations were for the gas phase only and did
not consider concentrations of organlcs contained on the paniculate matter.
An attempt was made to correlate furnace temperatures with hysteresis
but no consistent trend could be found In the range of 1000*C (1B33*F) to
1150°C (210TF).
Conclusions
Data presented Indicate that target organic compounds were present In
significant quantities after the cessation of coflrlng. Their presence was
detectable shortly after coflrlng and at reduced concentrations a number of
days later. The overall average hysteresis effect based on 13 runs coflrlng
gas and three (3) coflrlng oil showed that around SOI of the original concen-
trations measured are still being emitted at a mean value of 43 hours (after
cessation of coflrlng. Because this phenomenon may significantly effect the
accuracy of the ORE and organlcs measurements quantitative and parametric
data need to be obtained In future studies.
The second objective was to characterize the time/amplitude characteris-
tics of these effects. Although the study was terminated before complete
characterization of these effects, some trends were dlscemable. Starting
with clean hardware. It appears that a boiler system accumulates organic
compounds with time and that over all, neither burning with natural gas nor
soot blowing return the system to the clean state. This effect can be
considered as "boiler aging" (4). An eventual "steady-state". 1f It exists,
was not reached within the framework of the 59 days of this series. While
the various organic compounds seem to exhibit different behavior patterns
within the hysteresis effects, there were not sufficient data for generalizing
about these differences.
The effects Investigated here were for volatile organlcs only. Due to
lack of developed short sampling time methods, semi-volatile effects are, more
difficult to obtain and provide relatively poor resolution (I.e. 3-4 hr
average readings). A rough time vs. concentration relationship data-set
could be produced as part of a follow-up test.
The organlcs adsorbed onto sooty partlculate need to be Investigated.
There Is a good chance that a significant aaount of hysteresis and a signifi-
cant amount of the mass of organlcs emitted may be on the solid phase.
Efforts have already been underway to provide more resolution In the
area of volatile organic quantities that are present In the flue gas after
11
-------
e
£
L.
e
u
u
9
r
L,
i
^
X
e
o
t*
e
u
e
u
"»
t*
w
•
X
12 •
U
10
9
e
7
1
6
5
it
3
2
1
0
D
O
Q
D
O 0
O
*3
0 10 20 30 *0 50 60
Day of ffeet Prflfrw
fatal
ANCB
•*B"U'a 9« n^v»m»w*v *>wu«.V.HUT«t,A1
(Data polata ar* ratios to aTatarcaia
first iey of tast)
12
-------
coflrlng ceases. Further studies are planned In determining maxl
hysteresis Impact that might be expected In full-scale boilers.
References
TIM. Wool, et.al., December, 1987 'Engineering Assessment Report: Hazard-
ous Waste Coflrlng In Industrial B'-'ler Under Nonsteadjr Operating Con-
ditions* draft report, prepared t»» **urex Corporation for U.S. CPA under
Contract 68-03-3241. Work Assignment 2-9.
2. H.f>. Mason, et.al., August 1987 'Pilot-Scale Testing of Nonsteady Boiler
Haste Coflrlng* draft report, prepared by Acurex Corporation for U.S. CPA
under Contract 68-03-3241, Work Assignment 2-14.
3. U.S. CPA, November 1986 'Test Methods for Evaluating Solid Vastes*.
SU-846, 3rd edition. Office of Solid Haste and Emergency Response,
Washington, D.C. 20460
4. C.D. Uolbach, March 1988 "Some Technical Aspects of Emission Testing
During Coflrlng of Hazardous Haste In Industrial Boilers" a paper
presented at the Second Annual National Symposium, Incineration of
Industrial Hastes, San Diego CA presented by Toxcon.
13
------- |