Sunteți pe pagina 1din 7

An Oven Explosion: Lessons

Learned on PSM Applications


Michael D. Cazabona and Kirby Ericksonb
a
FM Global, Heath, TX 75032; michael.cazabon@fmglobal.com (for correspondence)
b
GAF-Elk, Ennis, TX 75119

Published online 12 August 2009 in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/prs.10346

An explosion occurred at one of our customer’s ponents to Process Safety Management. We will see
locations in late 1998 involving unburned and par- how these come into play here, at a facility not gov-
tially burned natural gas while the unit was in high erned by 29 CFR 1910.119. This article suggests that
fire mode. Investigation of the incident led to many PSM ideas should be applied via the general duty
contributing causes, most of which can be catego- clause throughout all facilities by suppliers, contrac-
rized broadly as lack of process safety information, tors, and employees to assure profitable, reliable, and
failure of a fuel combustion control system, contrac- safe operation.
tor control and lack of management of change. A common thread through most industrial proper-
Combustion systems are found throughout all ties is the mundane process of burning hydrocarbons
industries and homes. Code-based safeguards are to generate controlled heat for the purpose of doing
thoroughly and adequately described in NFPA. But is work. The nature of fire has been with mankind for
the spirit and intent of the code always applied and millennia. We have many prescriptive codes with altru-
interpreted safely? Instrumentation application and istic intent—no explosions, and no ‘‘unfriendly’’ fires.
installation are similarly well described in manufac- We are familiar, generally, with both fire and the codes.
turer’s literature and the ISA standards. But are these And, as we know, familiarity breeds contempt. We
systems installed and arranged properly? Is it impor- think we can out-smart the code, bend the laws of
tant to manage change in something as mundane as physics and chemistry and make equipment do fan-
fuel fired equipment? Is explosion venting on indus- tastic things. As for the codes themselves, blind appli-
trial ovens a sound way to limit damage? Is it even cation of the letter may not fulfill the intent. Or
possible in all instances? Are there alternatives? worse, it may bring on undesirable affects.
This article highlights the investigation activities, This article will describe, in general terms: the pro-
answers these questions, presents conclusions and cess, critical changes in design, key components
suggests remedies while showing the applicability of involved, applicable codes, discoveries during the
PSM concepts to industries other than the CPI. investigation and solutions deployed during the
This article will describe, in general terms: the pro- rebuild to address these issues.
cess, critical changes in design, key components
involved, applicable codes, discoveries during the inves- THE PROCESS
tigation and solutions deployed during the rebuild to
address these issues. Ó 2009 American Institute of Making Glass Fiber Mat
Chemical Engineers Process Saf Prog 29: 87–93, 2010 This operation takes chopped glass fibers, blends
Keywords: explosion; fuel; combustion; safeguards; them with a water soluble binding agent, and distrib-
instrumentation utes the solution on a ‘‘wire’’ to form a sheet. This
sheet is then dried with hot air in a multizone oven.
The sheet is rolled and used in the making of roofing
THE PROMISE shingles. This process resembles the making of arti-
In many incidents, there is no single issue that cle. We will focus on the Zone 1 dryer section, as
causes the havoc associated with a sudden generation shown in Figure 1.
of heat, light, and sound. Many systemic failures
come into play, which is why there are so many com- The Dryer
The four zones, gas fired dryer removes moisture
Ó 2009 American Institute of Chemical Engineers through evaporation. Each of the four zones is direct

Process Safety Progress (Vol.29, No.1) March 2010 87


Figure 1. Process flow diagram. [Color figure can be viewed in the online issue, which is available at
www.interscience.wiley.com.]

fired with natural gas as the fuel. These zones qualify corresponds to the pressure found at 70–90% of the
as Class B ‘‘furnaces’’ in NFPA [1]. Each zone exhausts rated flow yielding the possibility of two flow values
to a regenerative thermal oxidizer for environmental for the same pressure.
compliance. The exhaust zone is also provided with
a direct fired heater to assure the stream is above the About Combustion
dew point at all times. From here, the exhaust is sent To get complete combustion of natural gas, one
to a regenerative thermal oxidizer for environmental needs to have about 10 parts air per 1 part fuel by
compliance. volume. In the case of natural gas, the flammable
Zone 1 is by far the largest of the four in terms of range is 5–15%, so near-stoichiometric conditions
BTU input and air flow rates. The single burner for exist at a 10:1 mixture. Standard engineering contin-
this zone had been rerated to 40 MM BTU/hour from gency estimates of 130% in design apply here. It is
the design specification of 30 MM. The recirculation typical to specify the fan capacity at 130% of the
air flow was rated at a maximum of 196,000 scfm. demand created by maximum gas flow.
Make up air is provided via combustion air and Natural gas has a requirement to supply 1000
inward leakage around the dryer hood. This unit BTU/scf. Therefore, a 30 MM BTU/hr burner would
operates nearly balanced, but with a slight negative require 500 scfm of gas. Thus, the rated air flow
pressure (20.03 to 20.3 in. water column (WC)). should be about 6500 scfm.
System control is achieved via a distributed control
system (DCS) of proprietary design. The DCS func-
About Burners
tion is to provide an overview of all devices and
provide general commands to field mounted pro- A manufacturer developed a burner that would
grammable logic controllers (PLC). The PLC provides have a 15:1 turndown ratio, pushing the envelope of
actual commands to motor control centers. The com- stable flame range. To achieve a huge turndown ratio,
bustion system has one more layer of control: a hard- a creative design was developed. The main ring burner
wired combustion safeguard interlock system. This was supplemented with a ‘‘low fire’’ burner which
system is completely independent of the PLC and would take over when the demand for heat was scaled
DCS system control input. The combustion safeguards back. This low fire burner amounted to a 3=4 in. line
only communicate status to the PLC. The PLC also from the main gas supply piped independent from the
communicates status to (in addition to taking orders throttle assembly. This burner provided about 7% of
from) the DCS system. the total heat input (per NFPA 85 [2], a Class 2 igniter).
NFPA 86 [1] shows many burner arrangements and
specifically states that pilots within the cone of the
About Fans main burner flame need no separate flame detection:
Combustion air fans are often centrifugal fans with (a) A burner in which the pilot fuel is issued from
forward curve blades. They develop peak static pres- the same ports as the main flame or merges with the
sure at some flow rate, typically between 30 and 50% main flame to form a common flame envelope with a
rated flow. The static, or zero flow pressure, also common flame base and (b) each pilot and main

88 March 2010 Published on behalf of the AIChE DOI 10.1002/prs Process Safety Progress (Vol.29, No.1)
burner flame shall be equipped with flame supervi- ects, economic predictions changed resulting in a re-
sion in one of the following ways: vised specification of 1.5X feet/minute. To remove
the water from the mat, the heat needed to be
a. Main and pilot flames supervised with independ-
increased. This resulted in greater demands on the
ent flame sensors
gas fired burners and the installation of an additional
b. Main and interrupted pilot flames supervised with
drying zone. The original 30 MM btu/hr burner was
a single flame sensor
rerated by the burner manufacturer to 40 MM btu/hr
c. Self-piloted burner supervised with a single flame
by merely increasing the gas and air pressure and
sensor.
thus throughput. As seen from above, this has a
The intent of the burner operation was to have the significant impact on the combustion fan size require-
igniter light the pilot and the pilot to light the ‘‘low ment. The change should have resulted in a require-
fire’’ burner. Once the proof of ignition took place, ment of about 8500 scfm. Lacking a formal MOC
the high fire burners began to flow gas and ignited process, this critical item was overlooked. The origi-
from the ‘‘low fire’’ burner. Because this is a ‘‘self- nal fan specification remained at about 6500 scfm. To
piloted’’ burner, only one flame sensor—focused at compound this oversight, the burner design itself was
the low fire burner—was provided. As this burner a bit suspect.
ramped up to 40 MM btu/hr, the flame moved further When operating by itself during periods of low
from the spokes and cone. This flame was reportedly heat requirements, the low fire burner did not pro-
22 feet down the duct and about 4 feet away from duce a stable flame as the input from the combustion
the cone! Yet the code allows a single sensor. air fan kept blowing it out. The first attempt to cor-
Intuitively, the advice in NFPA 85 on boilers seems rect this was the installation of a weighted relief valve
to make more sense: on the combustion air line to truncate the fan pres-
sure curve. This appeared to meet the following
a. The main burner shall be equipped with a perma- items from NFPA 86:
nently installed igniter.
b. Where a Class 2 igniter is used, it shall not be a. All burners shall maintain the stability of the
used to ignite main fuel under uncontrolled or designed flame shape, without flashback or blow-
abnormal conditions. off, over the entire range of turndown that is
c. The burner shall be operated under controlled encountered during operation. . .
conditions to limit the potential for abnormal oper- b. Burners that cannot be ignited at all firing rates
ation as well as to limit the charge of fuel to the shall have provision to adjust the burner firing rate
furnace in the event that ignition does not occur during light-off to a level that ensures ignition of
during light-off. the main flame without flashback or blow-off.
d. If the Class 2 igniter is not shut down once the Unfortunately, this was a noisy way to meet the
main flame sequence is successfully completed,
aforementioned conditions and yet another change
then the main burner flame shall be proven by a
occurred. To eliminate the noise associated with the
flame scanner independently of the igniter.
venting of several hundred (or thousands) of cubic
feet of air, a butterfly valve was installed in the com-
About Venting bustion air duct and controlled by a pressure trans-
NFPA 86 and FM Global Data Sheet 6–9 [3] recom- mitter. This too appeared to be allowed and encour-
mend venting for ovens, regardless of the presence of aged by NFPA 86. The code is not clear regarding
volatiles generated by the process: (a) Fuel-fired fur- where the control needs to reside for this ‘‘provision.’’
naces, and furnaces that contain flammable liquids, In this case, the control loop for the pressure trans-
gases, or combustible dusts, shall be equipped with mitter and valve actuator resided in the basic process
unobstructed explosion relief for freely relieving inter- control system instead of within the combustion safe-
nal explosion pressures. . ., (b) the minimum design guard system.
shall be at least 1 ft2 (0.093 m2) of relief area for each
15 ft3 (0.424 m3) of furnace volume. (NFPA) [1], and The Venting Issue
(c) provide explosion vents for ovens where there are No explosion vents were provided and the struc-
fuel or vapor explosion hazards (FM global) [2]. ture was not designed to support their inclusion nor
their operation. This occurred despite the presence of
published and recognized engineering standards.
Uncontrolled or Managed Change?
In any project, if we let change occur without man- The Instrumentation Issue
agement systems, unknown outcomes will result. Since As many of us have witnessed in our careers, ‘‘as-
this site is not governed by PSM regulations, formal built’’ and ‘‘as-designed’’ often have some differences.
management of change (MOC) is not practiced or Fortunately, these are usually minor and do not play
established. A series of unmanaged changes occurred, into the hands of Murphy’s Law. In this case, of
resulting predictably, in an undesirable event. course, the differences did.
Many recommendations and standards [4–6] have
The Combustion Issue established certain commandments which the wise
The original machine specification was to produce should heed. One of these says that instrument taps
X feet/minute of glass fiber mat. As with many proj- should engage the fluid which they are measuring.

Process Safety Progress (Vol.29, No.1) Published on behalf of the AIChE DOI 10.1002/prs March 2010 89
Figure 2. Damaged oven enclosure. [Color figure can Figure 3. Damaged supply duct. [Color figure can be
be viewed in the online issue, which is available at viewed in the online issue, which is available at
www.interscience.wiley.com.] www.interscience.wiley.com.]

Thus, for gasses and vapors, taps should be at the The low combustion air interlock—a differential
top of the pipe or duct while for liquids, they should pressure switch—had its tap located about 30 feet
be located at the bottom. Another commandment above the switch and at the 6:00 position on the
states that condensate should be considered where duct. The ‘‘elegant’’ solution to the noisy relief valve
present and the tap should be arranged to drain had its pressure transmitter taps located a foot or so
away from the instrument via gravity. These bits of above the instrument. Considering that the fluid in
wisdom got lost along the way between the design the duct was air saturated with water at 4708F, con-
prints and the installation. densate was guaranteed in the instrumentation lines.

The Undesirable Event


The following events are based on the DCS time mark (CDT; 9/15/98):
21:15 The line is shut down due to poor quality mat. In this process, shutdowns require rethreading
from a low speed and thus the oven heat load resets to a very low demand. Zone 1 would
have been operating at or near low fire.
20:30 to 21:15 Two failed attempts to thread up the machine are made.
21:33:47 Incinerator temperature is 15048F (normal operating temperature)
21:34:27 Machine is successfully rethreaded and operator orders a 12 min ramp-up from minimum
speed to 8X min speed. Temperature set point is 4708F; oven supply pressure is 0.7 in. water;
exhaust is 0.73–1.0 in. water.
21:39:11 Zone 1 leakage high alarm. This is a calculated number, which assumes normal combustion
air based on measured gas flow rate (shown at 667 scfm) and other parameters. If all non-
measured parameters are as assumed, this would mean air is leaking out of the oven and into
the building. Operators noticed no fumes (formaldehyde smell would have been strong).
21:39:16 Incinerator temperature excursion begins; data from RTO.
21:41:36 RTO high temperature shutdown (18008F).
21:41:37 RTO alarm initiates machine shutdown and bypasses all zone exhaust to atmosphere.
21:41:55 Incinerator bypass valve in position for full vent to atmosphere.
21:41:58 Sheet break alarm indicating residual product has exited the ovens.
21:45:33 Operator ‘‘clears’’ sheet break alarm (acknowledged alarm).
21:46:04 Zone 1 pressure control disturbance (onset of explosion).
21:46:06 No information; all computer components have been knocked out of the cabinet.

The aforementioned events are generally con- Despite the magnitude of the event, there were no
firmed by the operators. One operator (in control injuries.
room) observed a ‘‘bright flash’’ in the vicinity of the
elbow in the supply duct near the west wall of the The Evidence
building. The rewinder operator only recalls getting A detailed investigation followed the incident im-
up from the floor, but not how he got there. Other mediately. Instruments were sequestered, valve posi-
operators heard an explosion and were knocked tions were noted and operability tests were per-
down. Indeed, neighbors reportedly heard the explo- formed. Instrumentation tubing was disassembled
sion 6 miles away! As shown in Figures 2 and 3, the and conditions noted. Damage was photographed
consequences of this accident were significant. and assessed.

90 March 2010 Published on behalf of the AIChE DOI 10.1002/prs Process Safety Progress (Vol.29, No.1)
Instrumentation The Answers
Calibration tests were conducted on switches and • Because the changes to the base design require-
transmitters. One significant anomaly noted was the ments were not managed, the burner was
low combustion air pressure switch. This device rerated without a commensurate increase in the
showed that adequate differential pressure existed combustion air blower capacity. Under best case
when none was applied. Subsequent destructive conditions, there would be only a slight oxygen
examination revealed the switch had corroded in the excess.
‘‘made’’ position (e.g., contacts corroded closed). • However, the best case was eroded first through
Thus, one defense against too little combustion air the pressure relief valve and subsequently the
had been removed from the combustion safeguard combustion trim damper, both of which allowed
system. We suspected as much when we removed adjustments to the flow/pressure of combustion
the switch from the ¼ in. instrument line and many air downstream and thus were unmonitored by
tens of milliliters of water drained out. Whether the the pressure safety switch. This certainly
contacts had corroded together or not, the conden- reduced the actual flow through friction losses.
sate WC would have provided enough pressure to • Because the fan curve was truncated, we had a
close the contacts. continuum of flow rates at a given pressure differ-
We had noted that the butterfly valve in the com- ential, from near zero to near rated. Because the
bustion air duct was mostly closed. This position was instrumentation lines were improperly arranged,
intuitively incorrect; the data from the DCS indicated condensate at least partially filled them. This had
the unit was on high fire. This valve should have the affect of blinding the low combustion air
been wide open. Examination of the wiring in the interlock switch and distorting the controller for
operator and stroke testing confirmed this valve oper- the combustion air butterfly control valve.
ator was a ‘‘fail-last position’’ device. On disassembly, • Because of the ‘‘self-piloted burner’’ design, we
we again noted several milliliters of water in the only needed to monitor one flame—the low fire
instrument line. We took careful measurement to flame which is burning regardless of the main
assure proper orientation for testing purposes. We flame throttle position. The main flame base,
also ran a full calibration on the pressure transmitter, according to reports from numerous sources, was
finding its read out to be true and accurate. We then several feet away from the burner nozzle at high
conducted tests on the effect of water in the instru- fire and extended 201 feet down the combustion
mentation line. chamber. The flame monitoring system never saw
The line connecting the low pressure tap to the that flame. So, when the main flame went out, the
DPT had about 10–15 ml of water in it. This line was flame scanner never noticed it missing.
removed, and two calibrators were provided to simu- • Because of the specific design of the burner, the
late the input ranges this device would normally see combustion air, though inadequate for the main
(8.5 in. WC on the high side, 4 in. WC on the low flame, was ample to support the low fire burner.
side). Water was injected into the sensing line in vari- This flame stayed lit and the annulus of combus-
ous amounts. We observed that the variations were tion air prevented the flame from contacting the
due to the capillary effect taking place in the 1/4 in. increasing concentration of unburned or partially
diameter instrumentation lines. burned gas in the dryer duct work—for a few
A 50% error can occur resulting in a false high read- minutes. Eventually, the diluting affect of the
ing. This would then drive the butterfly valve toward combustion air was overwhelmed by high con-
closure (about 30% more closed than it should have centration of gas and ignition occurred.
been). A DCS air pressure calculation assumed a 14:1
air/gas volumetric ratio. About 13:1 is the minimum
prudent engineering design guideline. About 10:1 is CORRECTING THE PROBLEMS
theoretically necessary to support combustion with
essentially no excess oxygen. Since the combustion air Is Venting the Answer?
fan was not properly sized for the rerated burner, only NFPA requires explosion venting for applications
a 10:1 air/gas mixture was available with no restriction! that do not use refractory or tubes (e.g., high temper-
With the condensate-induced error, the volumetric ratio ature furnaces and boilers). Yet these devices are
is depleted to 7:1 or 14% natural gas in air. This is the exposed to the same fuel explosion exposure as low
upper flammable limit as suggested in NFPA, though temperature ovens. These low temperature ovens are
other sources indicate the UFL is as high as 17%. typically located inside buildings. It seems a bit dan-
Regardless of the actual UFL, it is clear there was gerous to vent any overpressure into a building. FM
very inefficient combustion and a buildup of Global Data Sheet 6–9 [3] offers some solace by
unburned material was occurring. This is confirmed showing that only minimal pressures would be
by the RTO temperature excursion and shut down, observed if the vents were ever called on to deploy.
moments before the explosion. Still, minimal pressure would lose its meaning if you
The flame detector was proven functional, so there were standing in front of the vent when it deployed.
must have been a flame. But where was it? Why was Requiring venting as a needed layer of protection
the gas not consumed? Why did an explosion hap- suggests that prescribed combustion safeguards are
pen? A flow chart, developed during the investiga- insufficient protection in low temperature applica-
tion, is attached. tions. Exempting refractory lined equipment and

Process Safety Progress (Vol.29, No.1) Published on behalf of the AIChE DOI 10.1002/prs March 2010 91
boilers suggests that this risk may be acceptable as the various switches (position, pressure, etc.) be
the operating temperatures in these applications are improper before the purge starts, the DCS will lock
usually above the AIT of the fuel. out the start up command. This feature assures that a
false positive signal from the combustion air differen-
An Alternate Path? tial pressure switch will result in an alarm and no
ASME Code Case 2211 [4] suggests instrumented start up. Assurance for purging the oven prior to light
systems may be used to replace the need for physical off is controlled by the combustion safeguard system
venting in reaction systems. We reasoned that com- with the DCS backing up this system. Should the
bustion is an oxidation reaction and that there should purge timer fail in the safeguard system, the DCS will
be a way to achieve suitably reliable, controlled oxi- also lock out the start up command.
dation to avert the need to provide vents. We turned Regarding the reliability of the flame detection sys-
to NFPA 86 to see if such an approach might be tem to detect high fire flame, we replaced the ‘‘self-
allowed. Lo and behold: Nothing in this standard is piloting burner’’ with a conventional burner that is
intended to prevent the use of systems, methods, or provided with an interrupted pilot and no integrated
devices of equivalent or superior quality, strength, ‘‘low fire’’ burner. We found that the original 16 in.
fire resistance, effectiveness, durability, and safety flame rods were too short to detect flame at high fire
over those prescribed by this standard. and replaced these with 24 in. flame rods.
Would you instrument an oxidation reactor as All instrumentation lines were arranged to assure
poorly as current codes suggest we do for the oxida- condensate would flow back into the oven duct work
tion of fuel gas in an oven, furnace or boiler? The and we upsized these from ¼ to ½ in. diameter to
need to provide venting suggests that the prescribed assure no capillary affect would occur.
methods for controlling oxidation are insufficient. In Finally, detailed quarterly testing of all combustion
certain cases, it is deemed impractical. We believe safeguard functions as well as annual calibration of
this gap can be closed, making fuel oxidation safer in all pressure switches and tightness testing of all gas
boilers and furnaces while offering an alternative to valves was established. Management reviews these
venting explosions into a building. completed documents and retains them for annual
auditing by a third party.
Gas Supply
We know that with fuel in the proper flow range, Are We There Yet?
and air in the proper flow rate, a stable flame will be Although we addressed some areas that had no con-
maintained. So let’s look at a typical fuel train as a trol safeguards, we still rely on a single flame detection
safety instrumented system. system and mechanical linkage between the air and
A double block and bleed system is used in the fuel supplies. We are still vulnerable to gas build up
gas feed line. To control oxidation, we need to have under high fire conditions should the burner fall out of
the gas at the correct flow rate. We measure this ‘‘tune’’ due to linkage slip or combustion air obstruc-
indirectly through pressure constraints (hi/lo limit tion downstream of the pressure switch tap. Flame can
switches). Then, all we need is to assure a flame is be present and can satisfy the flame detection system
present and the proper amount of air. If the gas pres- even with partial combustion. We do not have separa-
sure is too high or low or the flame goes out, the sys- tion of the control system and its supervisory functions
tem shuts down by closing both block valves. This to the degree necessary to count as a layer of protec-
provides two safety functions. To protect against low tion. The combustion safeguard system only qualifies
pressure gas causing flame problems the PSL and as a SIL 1 system. We are not yet measuring what we
flame sensor are voted 1oo2 to shut the block valves. want to measure: the ratio of fuel to air.
To protect against the regulator or control valve fail-
ing open and the resulting high pressure gas flame Can We Get There?
problems, the PSH and flame sensor are voted 1oo2 There are many opportunities available to get us
to shut the block valves. to a much safer place. We can:

Air Supply • Measure fuel/air ratio by using flow measure-


To assure we have enough air, we again rely on ment devices (mass or volumetric), make these
pressure. A combustion air low differential pressure redundant and arrange their output to trip the
interlock is provided. As described earlier, we have fuel valves closed when dangerous out of ratio
additional detection for loss of flame. All are arranged conditions exist.
so that failure of one will trip the system (1oo2). The • Measure products of incomplete combustion
output from the wiring terminal is arranged to (CO) and/or excess oxygen. These can also be
de-energize solenoids holding open two gas valves. arranged to trip the fuel valves if key parameter
Thus, we have two outputs. are exceeded.
• Employ an additional means of flame detection.
• Make the logic solvers more robust and/or
Control redundant.
We modified the DCS to conduct a few additional
checks to cover the start up conditions and provide We can also accept that explosions happen and
a layer of protection we deemed necessary. Should that providing enough venting on the enclosure will

92 March 2010 Published on behalf of the AIChE DOI 10.1002/prs Process Safety Progress (Vol.29, No.1)
allow us to conform to codes and standards. How- 2. NFPA 85 Boiler and Combustion Systems Hazards
ever, we find this gives us little comfort. Code, 2007 Edition.
3. FM Global Property Loss Prevention Data Sheets
6–9 ‘‘Industrial Ovens and Dryers’’ (2003).
CONCLUSIONS
4. ‘‘Pressure vessels with overpressure protection by
• MOC, process hazard analysis, prestartup safety
system design,’’ Section VIII, Divisions 1 and 2,
review, and process safety information are
ASME Code Case 2211, 1995 Boiler Pressure Ves-
essential for safe and reliable operation in all
sel Code, American Society of Mechanical Engi-
industry sectors.
neers, 1995.
• There is no substitute for adequate engineering.
5. ANSI/ISA S84.00.01–2004, Functional Safety:
• There is no substitute for adequate installation.
Safety Instrumented Systems for the Process
• There is no substitute for adequate testing.
Industry Sector.
• We believe that the NFPA allowance of ‘‘self-
6. IEC 61511, Functional Safety: Safety Instrumented
piloting burners’’ in Chapter 86 is a contradiction
Systems for the Process Industry Sector, 2004.
to a requirement to assure flame is present at all
7. M. Cazabon and K. Erickson, ‘‘An oven explo-
burners as spelled out in Chapter 85. Addition-
sion—Lessons learned on PSM application,’’
ally, neither of these resources considers partial
Loss Prevention Symposium Proceedings, AIChE,
flame failure due to inadequate combustion air
New York (2009), pp. 663–680.
at high fire conditions.
8. J. Kauffman, T. Oakey, K. Price, and R.W. John-
• Given a typical fan curve, a low combustion air
son ‘‘Combustion safeguards test intervals—Risk
pressure switch may not provide useful informa-
study and industrial survey,’’ American Institute
tion—downstream obstructions could result in
of Chemical Engineers, Paper LPS-5e, LPS 2001
no flow, but adequate combustion air pressure.
Proceedings of the 35th Annual Loss Prevention
• We believe that measuring what you want to
Symposium, AIChE, New York (2001).
measure is critical. Measure mass flow or volu-
9. A. Summers, ‘‘Using instrumented systems for
metric flow of reagents and use this to assure
overpressure protection,’’ American Institute of
proper fuel to air ratios are met throughout the
Chemical Engineers, Paper No. 4A, LPS 2000
full firing range.
Proceedings of the 34th Annual Loss Prevention
Additional details of this article are in LPS Proceed- Symposium, AIChE, New York (2001).
ings [7]. Also, there are relevant teachings in other 10. J. Kauffman, T. Oakey, K. Price, and R.W. John-
AIChE publications [8–11]. son, Combustion safeguards test intervals—Risk
study and industry survey, Process Saf Prog 20
LITERATURE CITED (2001), 257–267.
1. NFPA 86 Standard for Ovens and Furnaces, 2007 11. A. Summers, Software-implemented safety logic,
Edition. Process Saf Prog 21 (2002), 161–163.

Process Safety Progress (Vol.29, No.1) Published on behalf of the AIChE DOI 10.1002/prs March 2010 93

S-ar putea să vă placă și