PCBs AND ASSOCIATED AROMATICS 919
temperature, a high molecular weight paraffinic hydrocarbon
oil produced a smaller fireball, lower fireball temperature and
shorter fire residence time than any of the other fluids tested,
including silicone dielectric oils. Catastrophic failure tests
methods such as that described above are difficult to stan-
dardize because of the many factors which contribute to the
results. Nevertheless, the scenario used is substantially more
realistic than many other fire tests applied to less flammable
fluids. For example, ignitability tests to determine whether a
fluid would contribute to a fire showed that, under all the test
conditions applied, a high temperature hydrocarbon fluid
exercised a considerable advantage over other fluids with
respect to time of ignition. The advantage was lost in badly
conceived experiments where gasoline was added to the
fluid as a fire starter to the hydrocarbon oil. Tests to deter-
mine the ignitability of fluids are difficult to standardize and
have been abandoned because of it. However, the ignition of
a combustible material is the first step in any fire scenario
and therefore is important to fire prevention.
Since arcing is the most common failure mode of
devices incorporating insulating fluids, R. Hemstreet^90 made
an attempt to assess spray flammability. His experimental
arrangement starts a flow through a nozzle and the liquid tem-
perature is gradually increased until the spray sustains ignition
at a measured distance of the igniter flame from the nozzle.
The burning rate in a gas is proportional to the square of
the gas pressure. Therefore, for two fluids of different vis-
cosity, or the same fluid at different temperatures, the pres-
sure in the aerosol cloud will be greater for the thinner fluid
because more of it can be pumped through the nozzle in a
given time. One would therefore expect that the minimum
distance of the igniter flame to cause combustion would be
larger in the case of a non-viscous fluid than in a viscous
one or, equivalently, that the minimum igniter distance will
increase for a given fluid as the temperature increases. This
is shown to be the case in Hemstreet’s results. Clearly, the
experimental factors in an apparently simple test to rank
fluids can be extremely complicated. For example, it was
found that an air velocity of greater than about two feet per
second had a considerable effect on the fluid temperature
required for ignition as the nozzle distance was increased.
The overall shape of a temperature–distance curve in
experiments of this sort is likely to be strongly influenced
by the presence of decomposition products, either originally
present in the fluid or induced by the igniter flame. The min-
imum temperature at which a liquid ignites depends on such
factors as the degradation produced in the fluid by the flame
and prior thermal and electrical stresses.
Heat release rate was measured in Hemstreet’s work
after removal of the ignition source and measurements are
sometimes significantly different from results obtained
using a sustained ignition source. Heat release raw data
derived from measurements done on a quiescent pool of
fluid provide information which has been used to calculate
suggested clearances between the burning pool, presumably
around the transformer, and the walls and ceilings of the
room in which the unit is housed. The test does not consider
the interference of silica crust produced over a quiescentpool of silicone oil. In the unlikely event that the conditions
necessary to maintain a quiescent condition were to exist in
a real building fire, the wicking action of the crust causes
the silicone oil to burn longer than high temperature hydro-
carbon oils under the same conditions (Webber^89 ).Thermal Decomposition of PCB Replacement Fluids
The suitability of fluids as replacement dielectrics for PCBs
depends not only on the electrical properties of the liquids
but also their decomposition products. Part of the reason is
that the buildup of oxidation products in in-service fluids
degrades its electrical properties and also because the
decomposition products may be toxic. In particular, combus-
tion products derived from PCB replacement fluids should
not be toxic.
Several fluids have been promoted as replacement dielec-
tric fluids in capacitors and transformers, of which, some are
listed in Table 42.
The capacitor fluids are usually based on a single major
constituent whereas the transformer oil are typically derived
from oils or are polyalkylated siloxane oils. Decomposition
products derived from single or, at least well characterized,
compounds are more easily analyzed than those obtained
from complex mixtures such as refined oils. Dielectric min-
eral oils have been in use in transformers for many years
and are not usually considered to represent a high health
risk even though they contain many aromatic compounds.
Indeed, part of the usefulness of naphthenic oils which gives
them an advantage over their paraffinic counterparts, is their
ability to solubilize the sludge of oxidation products formed
during in-service aging. Table 43 lists the percentage of dif-
ferent classes of hydrocarbons found in naphthenic trans-
former oils to illustrate the starting complexity with which
oxidation studies are faced.
RTEmp transformer oil was developed as a non-toxic
PCB replacement fluid and was derived from high molecu-
lar weight paraffinic material. The major thermal decom-
position products at 600°C in air include benzene, which is
known carcinogen, and traces of other compounds, as shown
in Tables 44 and 45. Benzene is also formed, in greater yield,
when RTEmp is heated^158 at 700°C under nitrogen. It is very
difficult to assess the health risks associated with the pres-
ence of trace quantities of pyrolysis products. Since the for-
mation of monoalkylated benzenes is two to three orders of
magnitude less than the amount which occurs naturally in
nontoxic naphthenic transformer oils, it is not likely that the
decomposition products formed from high molecular weight
paraffinic oils would cause much concern.
Thermal decomposition products derived from PCB
replacement capacitor fluids are much more easily character-
ized and are shown in Table 46. While the thermal decom-
position products of PCBs yield PCDF and PCDD products
which are toxic at the μ g/Kg level, the toxicities of products
derived from the alkylaromatic capacitor fluids in the absence
of chlorine sources, are toxic at the mg/Kg level. In the pres-
ence of chlorine sources, however, it would be possible to
generate PCBs as well as polychlorinated methylanthracenes
and, conceivably, PCDFs as well.C016_003_r03.indd 919C016_003_r03.indd 919 11/18/2005 1:12:41 PM11/18/2005 1:12:41 PM