
Abstract
A
sizable fraction of ignitions of structures are due to electrical faults
associated with wiring or with wiring devices. Surprisingly, the modes
in which electrical faults progress to ignitions of structure have not
been extensively studied. This paper reviews the known, published information
on this topic and then points out areas where further research is needed.
The focus is solely on single-phase, 120/240 V distribution systems. It
is concluded that systematic research has been inordinately scarce on
this topic, and that much of the research that does exists is only available
in Japanese.
Background
The latest
statistics of the National Fire Protection Association [ [1] ], available for 1993 –
1997 are that 41,200 home structure fires per year are attributed to ‘electrical
distribution.’ These electrical distribution fires account for 336 civilian
deaths, 1446 civilian injuries, and $643.9 million in direct property
damage per year. These figures include a proportional distribution of
fires with unknown equipment involved in ignition, but do not include
power cords or plugs which are attributed to specific appliances. The
41,200 structure fires account for 9.7% of total home structure fires
in the period, placing electrical distribution 5th out of 12
major causes. The $643.9 million in property damage represents 14.4% of
total damage, putting electrical distribution in second place (behind
incendiary or suspicious causes). Earlier statistics compiled for 1985
– 1994 by FEMA [ [2] ]
showed very similar results: electrical distribution was the fifth-ranked
cause of fires, the fourth-ranked cause of fire fatalities, and the second-ranked
cause of property loss. The electrical distribution causes [1]
are itemized in Table 1.
Table 1
Causes of US residential fires due to
electrical distribution
Cause of fire
|
Percent
|
fixed
wiring
|
34.7
|
cords
and plugs
|
17.2
|
light
fixtures
|
12.4
|
switches,
receptacles, and outlets
|
11.4
|
lamps
and light bulbs
|
8.3
|
fuses,
circuit breakers
|
5.6
|
meters
and meter boxes
|
2.2
|
transformers
|
1.0
|
unclassified
or unknown electrical distribution equipment
|
7.3
|
The high
losses sustained due to electrical distribution fires do not imply that
the systems are unreliable. There are about 270 million people in the
US, occupying about 100 million housing units, with the average housing
unit having 5.4 rooms [ [3] ]. This
means there are 2.7 persons per housing unit, or 2 rooms per person. If
there are 4 outlets per room, then the number of receptacles is 4´2´270´106 = 2.16 billion. A certain percentage should be subtracted
for receptacles not in use. It may be estimated that half the receptacles
have a device plugged in. Of the remaining half, it will be assumed that
half are “daisy-chained” to another outlet, and that the other outlet
is in use. Thus, the actual number of receptacles carrying current is
estimated as ¾ of 2.16 billion, or 1.62 billion. NFPA statistics indicate
that 4700 fires originate at “switches, receptacles and outlets,” but
CPSC [ [4] ] further breaks down the statistics
for switches, indicating that these account for 30% of the above figure.
Subtracting out the switch fires, 3290 fires per year are due to receptacles/outlets.
The failure rate is then estimated as 3290 / 1.62´109, or 2´10-6/yr. The very low
failure rate indicates that electrical receptacles are highly reliable.
The problem lies not with a high probability of failure, per device, per
annum. Instead, the issue is that electrical distribution involves an
extraordinarily high number of devices distributed throughout the built
environment. Each one supplies energy, and each one can potentially fail
and cause a fire.
Modes of
ignition
Given that
the electrical distribution ranks second in the dollar loss due to fires,
one might conclude that there has been a large body of work examining
the failure mechanisms that lead to ignition of fires. This proves not
to be correct, and, in fact, the research has been fragmentary at best.
The examination of failures can be approached in several different ways:
(a)
identifying the act(s) or omission(s) leading to failure
(b)
classifying failures by the functional nature of the device or part thereof
that failed
(c)
studying the basic physics of failures.
Both (a)
and (b) are essential in reconstruction of accidents, but the focus of
this paper will be on (c). This is especially important since several
authors [ [5] ][ [6] ] have already reported
studies along the lines of (a) or (b).
A consideration
of the failure mechanisms reveals that there are only a few main ways
that electrical insulation, or combustibles close by to electric distribution
components, can be ignited, although there are diverse aspects to each:
(1)
arcing
(2)
excessive ohmic heating, without arcing
(3)
external heating.
Some ignition
types involve a combination of mechanisms, so they must not be viewed
as mutually- exclusive causes of fire.
Arcing
Topologically,
an arc can be either a series arc (Figure 1) or a parallel arc (Figure 2).
Some authors consider a third form of arc—line-to-ground—is
possible when the circuit contains a ground in addition to a neutral.
But the topological arrangement is identical to that of the parallel arc,
since the load is not in series with the arc. The distinction between
the two basic forms of arcs is essential. In the case of the series arc,
the occurrence of the arc decreases the current flow in the circuit.
Thus, an over-current protection device cannot be expected to respond.
The causes of arcs can be many, but the primary ones are:
(1)
carbonization of insulation (arc tracking)
(2)
externally induced ionization of air (created by flames or an
earlier arc)
(3)
short circuits
Carbonization
of insulation
In 120 VAC circuits, it is not difficult to cause sustained
arcing if there is a carbonized conductive path. This is sometimes called
‘arcing-across-char.’ The mechanism has been known in electrical engineering
for a very long time [ [7] ]. How
a carbonized path gets established across an insulating material is not
a trivial question. There turn out to be more than one way of creating
such a path. The simplest way, used in some standard test methods [ [8] ], is to create an arc directly at the
surface of the insulation, for example, by placing two electrodes on the
insulator and applying a high voltage across them. Another mechanism involves
the combined effects of moisture and pollutants on the surface. This process
is sometimes called ‘wet tracking’ and has been a particular problem for
aircraft wiring with aromatic polyimide insulation [ [9] ]. The combined effects
of surface moisture and pollutants cause leakage currents across the surface
of the insulator, which, in time, can lead to formation of carbonized
tracks [ [10] ].
Insulating materials vary widely in their susceptibility
to arc tracking. A large fraction of wiring in 120/240 V circuits is insulated
with PVC, but unfortunately PVC is one of the less-satisfactory polymers
with regards to arc tracking [10].
Noto and Kawamura [ [11] ] have
reported on extensive wet-tracking experiments with PVC-insulated cables.
Using the standard IEC 60112 test method [ [12]
], they documented a number of specimen types that led to flaming
ignition of the cable.
When PVC is exposed to temperatures of 200 – 300ºC, it
chars and the char is a semiconductor. Not surprisingly, this can lead
to leakage currents and arcing. But Nagata and Yokoi [ [13]
] found that if virgin PVC was heated to the rather low temperature
of 160ºC, impressing 100 V across 1 mm of insulator thickness was sufficient
to cause ignition of the insulation. Furthermore, if the insulation had
previously been preheated to 200 – 300ºC, then ignitions occurred when
the preheated insulation was raised to only very mild temperatures during
the voltage test—from room temperature to 40ºC—were found sufficient (Figure
3).
Figure 3
Effect of preheat temperature and test
temperature on ignition of PVC wire insulation when subjected to 100 VAC
across 1 mm insulation thickness
Hagimoto
et al. [ [14]
] conducted laboratory studies on arcing faults (parallel arcing)
of electrical cords. They identified that the process typically proceeds
in a repetitive, but irregular fashion. They identified the following
sequence of steps:
- initial
current flow occurs due a carbonized layer.
- the current
flow increases and results in local arcing
- the arcing
causes melting of metal and expulsion of the molten pieces.
- once
the molten pieces are expelled, current flow drops
- continued
current flow through carbonized material eventually leads again to a
sizeable current flow.
The process
repeats indefinitely. The authors also measured the current waveforms
during the process, and found peaks up to 250 A, but these peaks were
rare, and the waveform typically showed peaks no greater than 50 A. Consequently,
a long time would take before a circuit breaker would be expected to open.
(Note, of course, that the actual current values will depend on the resistance
of the particular circuit tested).
Externally-induced
ionization of air
The intrinsic
dielectric strength of air is high (roughly 3 MV m-1, for all
except very small gaps), but breakdown can occur at much lower values
if the air space is ionized by some means. Two such means are flames and
pre-existing arcs. If a serious arc-fault occurs in a distribution bus,
a large amount of ionized gases will be ejected. These can travel a certain
distance, and if they encounter another circuit, they can readily cause
a breakdown and new arcing at the second location
[ [15] ]. The decreased breakdown strength
of air due to presence of flames has been documented in laboratory studies
by Mesina [ [16]
], who showed that the dielectric strength of air drops to
ca. 0.11 MV m-1 in flames. Mesina’s study, however, only encompassed
conditions at 1600 V and higher.
Fire-induced
arcing is considered that the most common situation for arcing damage
to be encountered in fire scenes [ [17] ]. It can involve either
carbonization of insulation, externally-induced ionization of air, or
both, but for the case of 120 V branch circuits, only a few limited empirical
studies are available that do not give general guidance.
Short
circuits
The term
short circuit is commonly applied in the situation where a low-resistance,
high-current fault suddenly develops in a circuit. This can take two forms:
(1) a bolted short where a good metal-to-metal contact is made across
a full-thickness section of metal; (2) an arcing short, where initial
metal-to-metal contact is not sustained and current flows through an arc.
In a bolted short, heating is not localized at the fault but distributed
over the entire length of the circuit. A bolted short can readily be created
by mis-wiring a circuit and then turning on the circuit breaker. The circuit
breaker then typically trips before anything ignites. It is, in fact,
exceedingly hard to create a fire in branch-circuit wiring from a bolted
short [ [18] ][ [19]
].
An arcing
short results from a momentary contact of two conductors. This causes
melting of the material around the contact area. Magnetic forces tend
to push the conductors apart, and the liquid bridge between them then
gets broken. Sparking may be observed as the conductors come apart. After
an arcing short, large-diameter conductors can often be seen with a notch
on the surface; smaller-diameter wires may be severed entirely; both
results are illustrated in NFPA 921 [ [20] ].
It is also
hard to ignite combustibles from arcing shorts in normal branch circuits
protected by 20 A or smaller circuit breakers or fuses. For example, Béland
[17] hammered cables, armored
cables, and conduits until the circuit breaker opened; these produced minimal mechanical sparks and could
never ignite wood, although in some cases loose fibers from wood fiberboard
insulation did ignite. On the other hand, Kinoshita et al. [ [21] ] successfully
ignited cotton gauze when creating bolted shorts with wires having 1.6
mm diameter solid conductors and also with ones using 1.25 mm2
stranded conductors. In their experiments, this required a thermal-mode-only,
20 A circuit breaker; when using a 20 A thermal/magnetic breaker, ignitions
were not observed.
A bit of
experimental ingenuity reveals that there are modes of parallel arcing
caused by short circuits that have a high probability for ignition. Franklin
[ [22] ] described that fires
were readily started in blankets and in paper, when a power cord was cut
with diagonal cutters. The fires ignited from the molten copper droplets
which are ejected. In such a situation, a bolted-short condition persists
only very briefly, since the magnetic forces induced by the short circuit
push the conductors apart, converting the bolted short into an arc. He
was able to create up to thirty such short circuits on a power cord before
a 20 A circuit breaker tripped. Nishida [ [23] ] found
that cotton and paper (but not PVC) could be ignited when a single 0.18
mm strand contacted a strand from the other leg of a stranded cable. But
he concluded that ignition was occurring due to the high temperature reached
by the strand and not through arc energy.
The cutting
of an energized electrical cord by an electric saw can result in ignition
of nearby combustibles having a low thermal inertia. UL has a ‘guillotine’
test which simulates a sawing accident [ [24] ]. Cheesecloth is placed
nearby as the ignition target.
Excessive
ohmic heating
The causes
of excessive ohmic heating can be subdivided into:
(1)
gross overloads
(2)
excessive thermal insulation
(3)
stray currents and ground faults
(4)
overvoltage
(5)
poor connections
Gross
overloads
It is easy
to start fires by creating a gross overload in an electric cable. But
the circumstances required for it do not tend to correspond to ways by
which electric wiring fires normally start. The smallest power cords or
extension cords in general use in the US are 18 AWG, and these are rated
for 10 A. Experimental studies on the gross-overload ignition mode are
meager, but they indicate that currents 3 – 7 times the rated load are
needed for ignition [ [25] ][ [26] ][ [27] ]. Since branch circuits
are normally protected by 15 or 20 A circuit breakers or fuses, a gross
overload must be considered a rare cause of fires in branch-circuit wiring.
Excessive
thermal insulation
There are
simple ways in which a fire can be created with an electric cord that
is neither damaged nor subjected to a current in excess of its rated capacity—loop
it up upon itself several times, or provide a high amount of external
insulation, or both. Laboratory demonstrations have verified that ignition
readily occurs [ [28] ];
in one case, simply coiling the cord three times and covering with a cloth
sufficed [ [29] ]. A special form of
this hazard occurs with the old knob-and-tube wiring, which was common
in the US prior to World War II. This type of wiring uses two separate
conductors which are not grouped into a cable, but are individually strung
on widely-spaced porcelain knobs. The current-carrying capacity is dependent
on there being unobstructed air cooling of the wires, and fires have occurred
when the wires were buried in thermal insulation [6].
Stray
currents and ground faults
Stray currents
occur when circumstances cause current to flow through paths not intended
to carry current. Ground faults are a well-known example [ [30] ][ [31] ]. They can occur if a conductor is
abraded or damaged and contacts metal siding, roofing, etc. Kinoshita
et al. [ [32] ] documented
that only 5 A was required for ignition when a 3-conductor, PVC-insulated
cable contacted a galvanized iron roof. An unusual mode of ignition from
a ground fault is where current flows through a gas line. The current
causes overheating of the metal and eventually a failure occurs [ [33] ].
In cold climates, it is not rare for individuals to thaw a frozen water
pipe by attaching a welding transformer and passing current through it.
Fires have resulted due the very large currents that are
involved [ [34] ]. Sanderson [ [35] ] studied a case where thawing activity
did not ignite the house that was being worked on, but caused ignition
in six neighboring houses fed from the same power utility connection.
Overvoltage
All indications
are that this is a rare form of ignition, as concerns branch-circuit wiring.
The materials used for wires and wiring devices are well able to withstand
the normal surges that are a regular event in a power distribution system.
To experience ignitions, one of three events are generally needed:
(a)
lightning strike
(b)
accidental delivery of high voltage into low voltage wiring
(c)
floating neutral
Lightning
strikes can result in massive ignitions, not just of wiring, but of all
sorts of combustibles. The problem has generally not been studied in connection
with 120/240 V wiring systems. Occasional fire reports are encountered
where, due to some malfunction in the power distribution network, high
voltage got applied to wiring intended to carry only 120/240 V. These
cases are rare enough that no systematic study exists. Floating neutral
problems are a bit less rare, but again, no systematic studies exist.
The basic problem is illustrated in Figure
4. A normal load, such as Rx, expects to see 120 V presented
to it. But if a break in the neutral occurs, it will be presented with
a voltage that can range from slightly above 0, up to almost 240 V; the
exact value is determined by the other loads on the system, R1
and R2. Ignitions are not surprising in such circumstances.
Figure 4
Floating neutral
Poor
connections
If a connection
is not mechanically tight and of low resistance, it can start to undergo
a progressive failure. The process often has the quality of an unstable,
positive-feedback loop. High resistance creates localized heating, heating
increases oxidation and creep, the connection becomes less tight, and
further heating occurs, until high temperatures are attained. At a certain
stage, a poor connection can become a glowing connection which shows very
high temperatures. At that point, nearby combustibles may be subject to
ignition. The process generally appears to be one of ohmic heating albeit
with a highly complex resistive element (but as indicted below, there
is some possibility that arcing also plays a role in glowing connections).
One of the
earliest efforts to study glowing connections dates to 1961 [ [36] ]. The primary results
are shown in Figure 5. The connection
acts as a non-linear circuit element. For currents over 10 A, drops of
around 2 V were found. But for small currents, voltage drops in the tens
of volts can be found. At a maximum current of 20 A, ca. 50 W is dissipated
in a copper/brass connection and around 35 W for copper/iron. The study
noted that the power dissipation depends only on the materials involved
and not on the nominal size of the contacts. It was also found that to
start the glowing process, a current of 4 – 6 A had to be supplied; glowing
of freshly-made connections could not be started with smaller currents.
Subsequently,
a number of research projects delved into further details of glowing connections,
especially following the popularization of aluminum wiring in residential
and mobile home construction in the 1970s. Hotta [ [37] ] identified a number of fire cases
attributable to this cause and conducted studies where he found that approximately
15 W was dissipated in a glowing copper-copper connection drawing 1 A,
and about 25 W at 2.5 A. By means of X-ray analysis, Hotta identified
that the high resistance in a copper-copper connection is due to progressive
formation of Cu2O at the junction. Kawase [ [38] ] further studied the glowing process
with copper-copper connections. Using an AC source of less than 100 V
and 0.5 to 1.0 A currents, he noted the following sequence of events
when an intermittent copper-to-copper connection is made. Initially, when
the contact is made and broken, blue sparks are generated. After a number
of make/break cycles, the sparks become red, instead of blue. If after
this time, contact is made continuously, a “Cu2O breeding process”
begins to take place. Layers of Cu2O begin to grow on both
contacts. Along the layer of Cu2O, a single bright filament
emerges. Molten metal is located along this thin filament, which meanders
“like a worm.” Kawase measured the voltage-current relationship of the
glowing connection and found that it cycles between high- and low-conductivity
states. He interpreted the cycling as a recurring breakdown of the interface
between Cu and Cu2O. Hagimoto et al. [ [39] ] found that in AC circuits,
for 1 mm wires, the minimum current necessary for glow to be sustained
was 0.3 – 2 A , while for 2 mm wires, it was 1 – 2.5 A.
Figure 5
Power dissipation and voltage drop across
glowing connections of two types
Sletbak et
al. [ [40]
] studied additional details of the Cu2O breeding
process and found that the filament glows at 1200 – 1300ºC. The process
is able to sustain itself, since copper continues to be oxidized underneath.
The high temperatures attained can readily lead to ignition. With a current
of 1 A, values of 200 – 350ºC were recorded at a 10 mm distance from the
glowing point. If a temperature of ca. 1250ºC is taken to be as typical
for the hot part of a glowing Cu-Cu connection, it can be noted that it
is very close to 1230ºC, the melting point of Cu2O. Hagimoto
et al. [39]
explain that the pulsing waveform found for glowing connections is accounted
for by spatter (mechanical sparks) that is emitted from the connection.
The spatter ejects material and this causes a momentary fluctuation in
current.
Meese and
Beausoleil [
[41] ] conducted a series of experiments specifically
focusing on glowing at the screw terminals of an AC duplex outlet. Glowing
connections readily occurred when the screw was not tightly tightened.
Visible glow occurred for currents carrying as little as 0.3 A in a 120
V circuit and also in low-voltage (3 – 4 V) circuits carrying less than
1 A. In low voltage applications, glowing connections could be established
in circuits with a voltage of less than 10V. A poor connection which is
glowing can re-establish the glow if the current is cut off and later
turned back on. There does not appear to be any time limit for glows;
in one experiment Meese and Beausoleil saw a connection glow for 129 h.
In a circuit carrying 20 A, a glowing connection was seen to dissipate
20 – 40 W; this is contrasted with 0.08 – 0.2 W for a good connection
at 20 A. A glowing connection in a typical residential duplex outlet may
be dropping only about 1 – 2 V across it—this is why the problem may not
be noticed at an early stage. Meese and Beausoleil also found that steel
screws are much more likely than brass screws to produce a glowing connection.
An interesting
question is whether some pairs of metals might be immune to glowing. Several
research groups have made claims that a particular pairing cannot lead
to glow. But a different research group typically succeed in eliciting
a glowing connection with the selfsame pairing of metals. At the moment,
there does not appear to be any confirmed non-glowing pairs of contact
metals.
Complicating
matters somewhat, UL [24]
has proposed, on the basis of unpublished experimental work, that a phenomenon
identified as ‘micro arcing’ is involved in a glowing connection. When
two metals are separated by a metal-oxide layer, conduction is essentially
nil across the layer, which is a dielectric. But the applied voltage can
cause a breakdown of the oxide layer. This discharge can cause a fine
metal bridge to be created across the dielectric. Substantive current
will flow through the metal bridge, but because of its limited current
carrying capacity, it shortly overheats, melts, and breaks apart. The
process then continues, but because of the high temperatures being created
locally, oxide layers are further built up. Other researchers have not
attempted to prove or disprove this UL hypothesis.
IEC [ [42] ] and
Sandia National Laboratories [ [43] ] both
developed different test methods intended to simulate a glowing connection
as a means of testing the ignitability of electric wires and cables from
this source, but neither method has been validated for ignition of building
components.
CPSC has
found that, in a flagrant violation of both regulations and good sense,
a number of fires which were caused by amateurs who made connections to
building wires by simply twisting two wires together, and neither soldering
nor using a wire nut on the connection [6]. Similarly, individuals sometimes repair
electric cords simply by twisting the wires together and insulating them
with electrical tape. This leads to a poor connection, and Hijikata and
Ogawara [ [44]
] measured the characteristics of some joints of this type
at currents of 10 – 20 A. They found that the temperature of the joint
increased linearly with current, typically being 50 – 95ºC for 10 A, and
going up to 130 – 300ºC at 20 A.
Long-term
failures of twist-on connectors was studied by Béland [ [45] ]. When two copper wires were joined
by a twist-on connector without adequate tightening, he found that failures
commonly occur due to metal loss, but this always occurred “several inches”
away from the connector, not at the connector itself. This was discovered
to be a corrosion problem. Overheating of the connector liberates HCl
gas from the PVC wire; the gas is corrosive and attacks copper. Over long
periods, metal loss occurs to the point that a connection can be completely
severed.
Ignitions
from poor connections
A glowing
connection might typically be found in a wall cavity, where the closest
combustibles—thermosetting plastics used as case materials for outlets
or switches, along with wood studs—are high-thermal-inertia substances
unlikely to be easily ignited. Thus, the question arises as to what exactly
a glowing connection in one of these electrical devices can ignite. On
this crucial question, only three very limited, unpublished studies can
be found. Aronstein [ [46] ] states that he successfully ignited:
- low
thermal inertia furnishings (bedding, drapes, upholstery) placed directly
against the face of an outlet, from a connection dissipating 28 W.
- thermosetting-plastic
receptacle cover plates, from a connection dissipating 30 W (thermoplastic
cover plates, however, were prone to melt away rather than to ignite).
- wood
studs, from a connection dissipating 35 – 50 W.
Unfortunately,
Aronstein gave few details of his experimental work. He did note that
the burning of the thermosetting cover plates was a flameless, glowing
combustion, and that a lightweight material (e.g., drapes) would have
to be contacting the cover plate for further propagation to take place.
In the case of ignitions of studs, again he found that initial ignition
was of a glowing type or smoldering type, but that this turned into flaming
when it broke to the other side of the stud, or, in the case of wood paneling,
when it broke through the front face of the paneling. Aronstein also reports
that glowing connections were able to ignite:
-
male plugs, cords, and small transformers plugged into the outlet,
- vapor
barriers inside the wall cavity,
but he gave
no details about the conditions needed for these ignitions to occur. He
also found that a connection glowing at a 45 – 50 W level was able to
melt aluminum wiring, and the gobs of molten aluminum could ignite a cardboard
box filled with papers.
Ontario Hydro
[ [47] ] conducted
a series of tests using duplex outlets wired with aluminum wire and previously
exposed to a modest overload of 27 A. The outlets were cycled using a
15 A load applied for 3.5 h, then off for 0.5 h. A mockup up stud space
was built, including thermal insulation inside the cavity and combustibles
placed at the face. No male plug was used, the current being drawn by
a daisy-chain connection. The results are summarized in Table
2.
Table 2
Results from Ontario Hydro testing of
duplex outlets with poor connections
Wall paneling
|
Insulation
|
Covering over outlet plate
|
Wiring method
|
Results
|
wood
|
cellulose
|
cotton
blanket
|
screw
|
scorching
only
|
wood
|
cellulose
|
cotton
blanket
|
back-wired
|
wood
paneling and cellulose insulation ignited after 4 cycles
|
wood
|
cellulose
|
cotton
blanket
|
back-wired
|
fuse
blew in 4th cycle; paneling and insulation ignited after
current flow had ceased
|
gypsum
wallboard
|
fiber
glass
|
cotton
drapery
|
back-wired
|
no
ignitions after 42 cycles; plastic outlet parts charred
|
Hagimoto
[ [48] ] reports
one unpublished experiment where a glowing connection was made in a knife
switch. The switch was housed in a molded plastic case and attached to
a wood board. After about 5 h of carrying 10 A, the wood had partly carbonized
and a piece of plastic from the case melted, dropped onto the hot conductor,
and ignited. This, in turn, caused the wood board to ignite.
Combined
effects
A number
of fire scenarios can involve a sequence of two steps: overheating first,
followed by arcing and ignition. For example, a wire may become heated
either due to excessive current or due to a poor connection. This may
soften the insulation sufficiently, so that a short circuit occurs at
a place where the wire is bent or passes a metal edge.
The most
important of the combined-effects situations is perhaps the last-strand
problem. A number of fires occur either at the junction between a cord
and the male plug, or at another place along the cord where repeated bending
has taken place. This has been studied by several groups of researchers.
Typically, it has been found that ignition of a plug or cord is associated
specifically with the breakage of the last strand. Mitsuhashi et al. [ [49] ] created failures of PVC-insulated
cords so that only one strand remained. Using test cords of 30×0.18 mm
strands (rated 7 A) or 50×0.18 mm strands (rated 12 A), they found that
for ignition to occur, the load current had to be within a relatively
narrow range. Ignition of the PVC insulation occurred only if the current
was between 10 and 20 A. Currents smaller than 10 A equilibrated to steady-state
temperatures of 100ºC or less and did not lead to fusing of the last strand
and ignition of the PVC. Conversely, currents over 20 A caused a rapid
fusion of the strand, and consequently did not deliver sufficient energy
into the already-broken strands to raise their temperature sufficiently
to ignite the insulation. PVC used for electric cords is moderately resistant
to ignition—applying local flames or hot temperatures normally does not
lead to a propagating fire of the polymer. But this changes if the material
has been preheated. Thus, Mitsuhashi discovered that the overheating must
not be too rapid. The sequence of events, then, is: overheating →
fusing → arcing → stopping of current flow. The ignition occurs
initially at only a tiny spot, but because a certain portion of the cord
has been preheated to over 100ºC, rapid flame spread can occur away from
the ignition location. The authors did further heat transfer modeling
and concluded that a gap of at least 1 mm is needed for ignition to occur.
Nagata [ [50] ] also conducted experiments and theoretical
modeling and came to broadly similar conclusions. UL uses the last-strand
problem in their ‘rotational flexing’ test [24],
where a stranded electrical cord is rotated enough times that one of the
conductors suffers a break, and it is then examined whether cheesecloth,
used as an ignition target, will ignite.
A PVC male
plug can be ignited by repeated ‘hot plugging’ while carrying a heavy
load. Blades [ [51]
] demonstrated this using a 1500 W space heater as the load.
The effect appears to be somewhat similar to the last-strand problem,
in the repeated hot-plugging erodes the contact material, creates a poor
connection, and heats up the PVC locally. Finally, arcing is able to cause
ignition. The general problem of ignitions due to poor connection at the
plug/receptacle interface has only been studied to a limited extent [ [52] ], and systematic studies
are not available in the English-language literature to describe conditions
needed for a structure ignition. Several studies have been conducted in
Japan, however. Based on laboratory studies, Ashizawa et al. [ [53]
] concluded that the steps leading to ignition are:
(a)
overcurrent and poor connection
(b)
thermal degradation of PVC
(c)
release of HCl gas from PVC
(d)
absorption of moisture by hygroscopic action of calcium carbonate
filler
(e)
initiation of surface and internal scintillations
(f)
formation of carbonized paths in the PVC, both on the surface
and internally
(g)
arcing
(h)
ignition.
Okamoto et
al. [ [54]
] also conducted laboratory studies and came to a roughly similar
conclusion. Uchida et al. [ [55] ] studied the problem
of failures in plugs where attachment of the wire is by means of a screw
connection.
A poor connection,
followed by arcing and ignition can be created when a nail or staple splits
apart a conductor in a nonmetallic cable. This, again, is a low probability
event, but at least two researchers have documented it in laboratory experiments.
Roberts [ [56] ] demonstrated this by splitting 12
and 14 AWG nonmetallic cables with a nail. Brugger [ [57] ] used a staple to split
a nonmetallic cable and reported that a glow occurs first.
External
heating
Most cases
of external heating involve the wire or wiring device as ‘victim’ of fire
and not as the initiator of fire. But some situations do exist where external
heating of wiring serves as the initiating event. In many cases, arcing
occurs after sufficient overheating. Chavez [ [58] ] examined the electrical failure of
two cables as a function of oven heating. Electrical failure was considered
to be a short circuit or a low-resistance condition developed across the
line; experiments were not conducted to actually elicit ignitions. A
cable with cross-linked polyethylene insulation failed at 270ºC, while
a cable with polyethylene/PVC wire insulation and PVC jacket failed at
250ºC. A NIST study on lighting fixtures [ [59] ] examined
the effect of over-temperatures on 60ºC-rated normal building wiring.
When overlamping of a fixture created 202 – 205ºC temperatures in the
electrical junction box, failure occurred in less than 65 h. The wire
insulation became brittle, cracked, fell away from the conductors, and
this led to a short circuit.
Conclusions
In 1974 the
author of a textbook on electrical insulation [ [60] ] wrote: “The fundamental breakdown
processes are not understood; not for lack of experimental observations
but because our background knowledge is too crude.” Unfortunately, even
today this statement remains true, as concerns wiring and wiring devices
in buildings.
Claims are
sometimes made that a significant fraction of fires assigned to electrical
causes has been mis-investigated [ [61] ]. But, despite the recent efforts
in NFPA 921 to make high quality information available to the investigator,
this is very hard to do in the absence of adequate published research
studies.
It is surprising
how little systematic research has been done to elucidate and quantify
the mechanisms whereby electric wiring faults lead to structure ignitions.
Almost all of the experimental papers that could be found studied problems
only of a very narrow scope. In addition, a number of them (mostly not
reviewed here) have approached the topic by attempting to prove that certain
modes of ignition “cannot happen.” This, of course, is hardly good scientific
methodology, but is an easy trap to fall into, when it is realized that
failures of highly reliable devices are involved.
Not a single
paper from a US university was found on the topic, nor is there any agency
or research institute in the US that has carried on long-term research
on these problems. It might be noted that in Japan, elucidating the nature
of electrical ignitions has been considered to be a problem of national
priority, and several institutes and universities have done considerable
long-term research, but these studies are generally available only in
Japanese.
Without adequate
laboratory studies documenting and quantifying electric-wiring-related
fire ignition scenarios, little progress can be expected either in improving
fire investigations or in reducing fire losses of this origin.
In the US,
the safety of wiring and wiring devices is generally assessed according
to UL standards, but there exists almost no published material from UL
that would document their studies of ignition mechanisms, nor to provide
a basis for judging whether their test procedures have a traceable connection
to field failure modes.
Aging of
plastic materials can lead to increased failures. This has been studied
in other electrotechnical areas (e.g., aircraft wiring), but no studies
exist for building wiring.
Acknowledgement
This paper
was originally presented at the Fire and Materials 2001 conference and
the permission of Interscience Communications Ltd. to reprint it in the
Fire and Arson Investigator is gratefully
acknowledged.
Presented
by: Dr V Babrauskas at the 7th international Fire & Materials
conference, 2001, San Francisco, USA, pp39-50. Proceedings are available
from the publisher, Interscience Communications, UK. intercomm@dial.pipex.com

References
[1] . Rohr,
K. D., The U.S. Home Product Report (Appliances and Equipment Involved
in Fires), Fire Analysis & Research Div., National Fire Protection
Assn., Quincy MA (2000).
[2] . Fire in the United States 1985-1994, 9th
ed., US Fire Administration, Emmitsburg MD (1997).
[3] . Statistical
Abstract of the United States 1999, Tables 2, 1212, and 1213, Government
Printing Office (1999).
[4] . Ault,
K., Singh, H., and Smith, L., 1996 Residential Fire Loss Estimates,
Consumer Product Safety Commission, Washington.
[5] . Hall, J. R., Jr., Bukowski, R. W., and Gomberg,
A., Analysis of Electrical Fire Investigations in Ten Cities (NBSIR
83-2803), [U.S.] Natl. Bur. Stand., Gaithersburg MD (1983).
[6] . Smith,
L. E., and McCoskrie, D., What Causes Wiring Fires in Residences, Fire
J. 84, 19-24, 69 (Jan/Feb 1990).
[7] . Olyphant,
M. jr., Arc Resistance. I. Tracking Processes in Thermosetting Insulating
Materials, ASTM Bull. No. 181, 60-67. II. Effect of Testing
Conditions on Tracking Properties of Thermosetting Insulating Materials,
No. 185, 41-38 (1952).
[8] . Test Method
for High-Voltage, Low-Current Dry Arc Resistance of Solid Electrical
Insulation (ASTM D 495), American Society for Testing and Materials,
West Conshohocken PA.
[9] . Electrical
Arcing of Aged Aircraft Wire (Report N191-RPT4AU99), Report to NTSB
under Order No. NTSB18-99-SP0127, Lectromechanical Design Co., Sterling
VA (1999).
[10] . Billings,
M. J., Smith, A., and Wilkins, R., Tracking in Polymeric Insulation,
IEEE Trans. Elec. Insul. IE-2, 131-137 (Dec. 1967).
[11] . Noto,
F., and Kawamura, K., Tracking and Ignition Phenomena of Polyvinyl Chloride
Resin Under Wet Polluted Conditions, IEEE Trans. Elec. Insul.
EI-13, 418-425 (1978).
[12] . Method
for Determining the Comparative and the Proof Tracking Indices of Solid
Insulating Materials under Moist Conditions (IEC 60112), International
Electrotechnical Commission, Geneva (1979).
[13] . Nagata,
M., and Yokoi, Y., Deterioration and Firing Properties of Polyvinyl
Chloride Covering Cords at Elevated Temperatures, Bull. Japan Assn.
of Fire Science and Engineering 33:2, 25-29 (1983).
[14] . Hagimoto, Y., Watanabe, N., and Okamoto, Arcing
Faults on PVC-covered Electrical Cords, pp. 221-224 in Proc. 1st
Conf. of the Assn. of Korean-Japanese Safety Engineering Society,
Kyongju, Korea (1999).
[15] . Dunki-Jacobs,
J. R., The Escalating Arcing Ground-Fault Phenomenon, IEEE Trans.
Ind. Appl. IA-22, 1156-1161 (1986).
[16] . Mesina, J. G., Determination of Electrical
Clearances for Permissible Equipment Operating in Gassy Mines and Tunnels,
IEEE Trans. Ind. Appl. IA-30, 1339-1350 (1994).
[17] . Béland,
B., Electrical Damages—Cause or Consequence? J. Forensic Sciences
29, 747-761 (1984).
[18] . Zimmerer, C. W., and Neumer, F., Aluminum Building
Wires and Connectors (Bull. of Research No. 48), Underwriters Laboratories
Inc., Chicago (1954).
[19] . Ettling, B. V., Ignitability of PVC Electrical
Insulation by Arcing, IAAI-Oregon Chapter Newsletter, 6 (Mar.
1997).
[20] . Guide
for Fire and Explosion Investigations (NFPA 921), National Fire Protection
Assn., Quincy, MA (2001).
[21] . Kinoshita,
K., Hagimoto, Y., and Watanabe, N., Investigation Reports and Igniting
Experiments on the Electrical Causes of Many Fires Started after the
Big Earthquake in Kobe Area in 1995, published in Urgent Study Reports
on the Hanshin-Awaji Big Earthquake, Science and Technology Agency of
Japan, Tokyo (1995).
[22] . Franklin,
F. F., Circuit Breakers: The Myth of Safety, Fire and Arson Investigator
41, 42-45 (June 1991).
[23] . Nishida,
Y., Ignition Hazard by Short Circuit between Element Wires of a Stranded
Cord, Reports of the National Research Institute of Police Science
45:4, 57 (Nov. 1992).
[24] . Wagner,
R. V., Boden, P. J., Skuggevig, W., and Davidson, R. J., Technology
for Detecting and Monitoring Conditions That Could Cause Electrical
Wiring System Fires (UL Project NC233, 94ME78760), Underwriters Laboratories
Inc., Northbrook IL (1995).
[25] . Yamamoto,
T., et al., A Test on Cables on Fire from Over Electric Current, Mitsubishi
Cable Industries Review, No. 92, 41-47 (June 1997).
[26] . Lawson, D. I., McGuire, J. H., Fires Due to
Electric Cables (FR Note 55), Fire Research Station, Borehamwood, England
(1953).
[27] . Béland,
B., Considerations on Arcing as a Fire Cause, Fire Technology
18, 188-202 (1982).
[28] . Popular Extension Cord Reels Can Be Real Dangerous,
Fire Findings 2:1, 13 (Winter 1994).
[29] . Hagimoto,
Y., Kinoshita, K., and Watanabe, N., Fire Hazard of a Coiled or Bundled
Cord, Summary of 1994 Annual Meeting of JAFSE (1994).
[30] . Béland,
B., Fires of Electrical Origin, Fire and Arson Investigator 43,
35-41 (Dec. 1992).
[31] . Béland,
B., Ground Fault in Flexible Exhaust Duct, Fire and Arson Investigator
44, 44-45 (June 1994).
[32] . Kinoshita,
K., Hagiwara, T., and Kinbara, J., Ignitability of VVF Cable in Contact
with Grounded Object, J. Japanese Assn. Fire Science & Engrg.
28, No. 3, 30-37 (1978).
[33] . Goodson, M., Sneed, D., and Keller, M., Electrically
Induced Fuel Gas Fires, Fire and Arson Investigator 49:4,
10-12 (1999).
[34] . Béland, B., Some Fires of Electrical Origin,
Fire and Arson Investigator 37:2, 37-38 (Dec. 1986).
[35] . Sanderson, J. L., private communication (2000).
[36] . Tests of Insulating Materials for Resistance
to Heat and Fire, Report of CEE Working Group “Hot Mandrel Test,” CEE
(031) D126/61, Deutsches Komitee der CEE beim Verband Deutscher Elektrotechniker,
Frankfurt am Main (1961).
[37] . Hotta, E., On the Phenomenon of Glowing Connections,
J. Japan Assn. for Fire Science 24:1, 52-58 (1974).
[38] . Kawase, T., The Breeding Process of Cu2O,
IAEI News 47, 24-25 (July/Aug. 1975); Second Report, 49,
45-46 (Nov./Dec. 1977).
[39] . Hagimoto,
Y., Kinoshita, K., and Hagiwara, T., Phenomenon of Glow at the Electrical
Contacts of Copper Wires, Natl. Res. Inst. of Police Science Reports—Research
on Forensic Science 41, 30-37 (Aug. 1988).
[40] . Sletback, J., Kristensen, R., Sundklakk, H.,
Nåvik, G., and Munde, R., Glowing Contact Areas in Loose Copper Wire
Connections, pp. 244-248 in Proc. 37th IEEE Holm Conf.
on Electrical Contacts, IEEE (1991).
[41] . Meese,
W. J., and Beausoliel, R. W., Exploratory study of Glowing Electrical
Connections (NBS BSS 103), [U.S.] Natl. Bur. Stand., Gaithersburg, MD
(1977).
[42] . Fire
Hazard Testing. Part 2: Test methods. Bad-connection Test with Heaters
(IEC 60695-2-3), International Electrotechnical Commission, Geneva (1984).
[43] . Spletzer, B. L., and Horine, F., Description
and Testing of an Apparatus for Electrically Initiating Fires Through
Simulation of a Faulty Connection (NUREG/CR-4570; SAND86-0299), Sandia
Natl. Labs., Albuquerque NM (1986).
[44] . Hijikata, T., and Ogawara, A., Research on
Thermal Phenomena of Twist Joint Point of PVC Insulated Flexible Cords,
Summary of 1992 Annual Mtg. of Japan Assn. of Fire Science and Engineering
204-205 (1992).
[45] . Béland,
B., Behaviour of Electrical Contacts under Fire Conditions, Fire
and Arson Investigator 38, 38-41 (Sept. 1987).
[46] . Aronstein,
J., Fire Due to Overheating Aluminum-Wired Branch Circuit Connections,
Wright Malta Corp., Ballston Spa NY (1983).
[47] . Oda, S. J., Progress Report—Fire Initiation
Potential of Failing Electrical Receptacles (Report 78-92-K), Ontario
Hydro, Toronto (1978).
[48] . Hagimoto, Y., private communication, National
Research Institute of Police Science (2000).
[49] . Mitsuhashi,
N., Yokoi, Y., Nagata, M., and Isaka, K., Concerning the History of
Deterioration in Insulated Electric Wires and Fire Hazards, J. Japanese
Assn. for Fire Science & Engrg. 31, No. 1, 11-19 (1981).
[50] . Nagata, M., Firing Current and Energy Input
of Polyvinyl Chloride Covered Cords Having Disconnected Element Wires,
Bull. Japanese Assn. Fire Science and Engrg. 33:1, 1-7
(1983).
[51] . Blades,
R. “Arc Primer” video tape, RB Laboratories (1993).
[52] . Aronstein, J., Evaluation of Receptacle Connections
and Contacts, pp. 253-260 in Proc. 39th IEEE Holm Conf.
on Electric Contacts, IEEE (1993).
[53] . Ashizawa, K., and Omata, K., Property of Ignition
Mechanism Caused by Thermal Degradation on Plug, Abstracts of Annual
Meeting of JAFSE, Paper C-22, 386-389 (1997).
[54] . Okamoto, K., Watanabe, N., and Hagimoto, Y.,
Deterioration of Tracking Resistance of Dielectric Materials Caused
by Thermal Degradation, pp. 231-234 in Proc. 1st Conf.
of the Assn. of Korean-Japanese Safety Engineering Society, Kyongju,
Korea (1999).
[55] . Uchida,
M., Saito, H., Watanabe, S., and Sassa, Y., Experiments on the Heat
due to Poor Connections inside a Male Plug and on the Breaking of Wires
due to Repeated Pulling Out, Report of the Fire Science Laboratories
of the Tokyo Fire Department, No. 18, 8-16 (1981).
[56] . Roberts,
E. W., The Ground-Fault Circuit Interrupter and Fire Prevention, Electrical
Fires Conference, Univ. Wisconsin, Madison (April 3, 1986).
[57] . Brugger,
R. D., Glowing Electrical Connections, J. Natl. Academy of Forensic
Engineers 9, 21-33 (1992).
[58] . Chavez,
J. M., Steady-State Environment Cable Damage Testing (Quick Look Test
Report), Sandia Natl. Labs., Albuquerque NM (1984).
[59] . Fulcomer,
P. M., Temperature Measurement on Operating Surface Mounted Lighting
Fixtures (NBSIR 79-1912), [U.S.] Natl. Bur. Stand., Gaithersburg MD
(1979).
[60] . Sillars,
R. W., Electrical Insulating Materials and their Application,
Peter Peregrinus, London (1974).
[61] . Béland,
B., Electricity…The Main Fire Cause? Fire and Arson Investigator
32, 18-22 (Jan./Mar. 1982).
|