Ignition and lateral flame spread characteristics of certain composite materials

NISTIR

89-4030

Ignition

and Lateral

Flame

Spread

Characteristics of

Certain

Composite

Materials

T. Ohlemiller

S. Dolan

U.S.

DEPARTMENT

COMMERCE

National Institute of Standards and Technology

(Formerly

National Bureau of Standards)

National Engineering

Laboratory

Center for Fire Research

Gaithersburg, MD 20899

January 1989

Sponsored

by:

David Taylor Research Center

United

States

Navy

Annapolis, MD

U.S.

DEPARTMENT OF COMMERCE

William Verity, Secretary

Ernest Ambler, Acting Undersecretary

for Technology

NATIONAL INSTITUTE

OF STANDARDS

AND TECHNOLOGY

Raymond

Kammer, Acting Director

TABLE OF CONTENTS

List

of Tables

List

of Figures .............................

Abstract ............................ 1

Introduction ....................... 1

Experimental

............................

Materials .............................

Measurement

Techniques

................

Results and

Discussion

.......................

3.1.

Ignitability

...................

Sample

Behavior

................

Heat Flux

Dependence of Ignition Delay 5

Correlation of Ignition Data ...........

3.2.

Lateral

Flame

Spread

................

Sample Behavior

..........

Lateral Flame Spread Parameters.

........

4. Summary

and Conclusions 11

5. References

.......................

iii

LIST OF TABLES

Page

Table 1.

Parameters Inferred from

LIFT Ignition

Data

Table

Parameters Inferred from LIFT Flame

Spread Data

LIST

FIGURES

Page

Figure 1.

Ignition delay

time vs. incident

heat

flux

for CG Camel honeycomb

panel. Equation

shown is least-squares

fit and

R is the

correlation coefficient.

............

Figure

Ignition

delay

vs.

incident heat

flux for CG

Camel and CG Yellow

honeycomb panels

............. 15

Figure 3. Ignition

delay

time vs.

incident heat

flux

for CG Yellow honeycomb

panel; face and

edge ignition data points

are distinguished

......... 16

Figure 4. Ignition delay time vs.

incident heat flux

for CG Green Vinyl

honeycomb panel.

Equation

shown

is least- squares fit

and R is the

correlation coefficient

................... 16

Figure 5. Ignition delay time vs.

incident heat flux

for OC Armor composite armor

panel. Face and

edge data points

are distinguished. Equations

shown

are leas t- squares fits

and R

values

are

correlation

coefficients

. ........

Figure 6. Ignition model

correlation for CG

Camel and

CG Yellow

treated as a single

material. Data

points

for face ignition only.

........ 18

Figure

7. Ignition

model correlation

for CG Green

Vinyl.

Data points

for face ignition only

Figure

8. Ignition model

correlation for face ignition

OC Armor

............

Figure

9. Ignition

model correlation

for

edge ignition of

OC Armor

................. 19

Figure

10. Experimental data

for lateral flame

spread

OC Armor.

Edge and

face

data are distinguished.

The value

is the peak incident flux on

the sample

face

...........

Figure

11.

Experimental

data

for lateral flame spread on

CG Camel honeycomb panel.

Data

shown

are for

sample edges.

Each test

was

run at a

different

value

of the peak incident heat flux,

Figure

12.

Experimental data for lateral

flame

spread

CG Camel honeycomb panel. Data are for sample

face. Results for two different values of the

peak incident heat flux are shown

Figure 13. Model

correlation

lateral

flame

spread on

Green Vinyl honeycomb panel. Line

shown

has

been

forced to

pass

through

value

0.87 W/cm

on the horizontal axis

Figure 14.

Distribution of

heat flux

incident on

sample

surface,

normalized

peak value

Ignition and

Lateral Flame Spread Characteristics

of Certain

Composite Materials

Abstract

The LIFT

apparatus

was used to obtain information on the ignition

and lateral

flame

spread

characteristics of

two types of composite materials. The

first

type

was

honeycomb

sandwich panel; three different facings

were tested

with

this

material.

The

second type of material

was

a composite armor.

There

was

substantial

variation in

the ignitability of the

various material

combinations

with

vinyl

-faced honeycomb panel being the

most ignitable

and

the

composite

armor being the

least ignitable.

The ignition

behavior of the

facings of all

materials was correlated by

a simple

predictive model. Only

the vinyl-

faced honeycomb

panel showed significant

normal flame

spread under

the conditions

examined

though some flame

"advancement"

was

seen

with the

others. All

of the

materials exhibited

worse flammability

properties at their

edges

as compared

to their facings.

Introduction

This report

summarizes the results

of ignitability and

lateral flame

spread

tests

performed on

two

substantially

different types

of composite

materials,

Nomex

sandwich panels and a

composite ballistic armor.

The tests

were

performed using the Lateral

Ignition and Flame

Spread (LIFT) apparatus

at the

NIST Center for Fire Research. The

objective is to begin

to establish

database

the flammability of materials of

potential

interest to the

Navy

for shipboard

applications such as the

Composite Deckhouse.

The general approach

flammability assessment

followed here is discussed in

detail

in Reference

Briefly, the

approach consists

of determining

sets of

effective parameter values which can be

utilized in the

context of

simplified

models of

ignition or flame

spread to predict this

behavior

variety

different

contexts. Examples of such

parameters

are the

minimum surface

temperature

for ignition and the

minimum incident

heat flux

on a material

surface necessary

to support

continued flame

spread.

The

principal measure of

ignitability is

the delay time

between

the

onset of a

constant

heat flux to the surface of

a material

and the first

appearance

flammable

mixture of gases issuing

from the

surface of the

material.

The

presence of a flammable gas mixture

is sensed by

placing a

small pilot

flame

in the gas

stream issuing from

the surface

(positioned so

that it

does not add

to the

external heat flux on the

sample surface)

The delay

time is dependent

on the

incident

flux

level; a

complete

ignitability

characterization thus uses

a range

of heat fluxes from 7 or 8

W/cm

down

the minimum flux

necessary

for

ignition. This minimum is

dependent

the

physical and chemical

characteristics

the

material as

well

as the

conditions of

heat exposure.

These

measured data are fitted to a simple

model of the

ignition process to

infer the

values of two parameters in the model:

the effective

thermal inertia

of the

sample (product of thermal conductivity, density

and heat

capacity) and

the minimum

surface temperature

for ignition. The model equation

can then

used to

predict the ignition delay for the sample

material

other

conditions

such

for

larger scale vertical surfaces.

Flame

spread

rates on

vertical

flat

surfaces

are approximately equal

in both

the lateral

and downward directions; spread rates

on horizontal flat

surfaces

are

also comparable

[2,

3].

Thus only one of these

configurations (all of

which involve opposed flow flame spread; see Ref.

needs to be measured

obtain

the expected behavior

for all

three.

The LIFT apparatus measures

lateral spread on a vertical surface

This rate is dependent on

the

temperature of the surface ahead of the spreading flame. This temperature

could be

increased

a fire by an external heat flux (from some other burning

object, for

example) incident for a

varying

amount of time. Generally such an

external flux can increase the sample surface temperature only up to an

equilibrium

value

dependent on that flux

level and

on the rate of heat loss

from the surface. The LIFT apparatus takes this preheating effect into

account

first allowing the sample surface to equilibrate locally

with

external flux. This local flux

varies

monotonically along the

surface of

the

sample so that one can obtain a measure of the heat flux

dependence of

the

lateral flame spread velocity. Furthermore, one can obtain the minimum flux

at which sustained lateral spread is possible. By fitting these data to a

simplified model of opposed

flow

flame spread, one can obtain

another

effective parameter for the material, a measure of the flame heat transfer to

the sample surface

during flame spread. The

fitted model

can then be

used to

predict the opposed flow flame spread rate of the material in any of the

configurations

mentioned

above

for larger-sized samples.

These are the procedures applied to the

materials of

interest

here.

They

met

with

varying degrees

of success due to peculiarities of sample

behavior,

discussed below.

Experimental

Materials

All

samples

were

obtained from the U. S.

Navy David

Taylor

Research Center;

they are the

first in a series

of composite materials to be

subjected

to an extensive flammability

characterization in

aid of their

evaluation for

shipboard use. The first

type of

composite was a Nomex

honeycomb sandwich

panel

Three different panels were tested

having

different finishes on

the heated surface. Two of the finishes

appeared

to be

substantially similar,

differing only

color. One

was described as "camel"

In color; the

other

was yellow.

For both of these the color

layer had the

appearance

of a

low

gloss paint intimately bonded to the layers

below.

The

third surface

finish on the honeycomb panels was apparently a

textured vinyl

sheet, light green-yellow

in color, that appeared to be

glued to the outermost

Product

names are

provided

here only for purposes of

clarity; they do

not

imply any

endorsement by NIST.

layer

the sandwich panel; this vinyl layer

was

readily

strippable from the

sandwich

panel. All of

these

panels

were

nominally 15 h mm

(5/8

in.)

thick

and the

honeycomb

core accounted for approximately

8 mm

(1/2

in.) of this

thickness.

The

sandwich faces were

multilayer in character including

one

plies of

woven glass fibers. In the results that follow, these

panels

are

designated as GG

Camel, CG Yellow and CG Green Vinyl.

The

second

type of

material was

composite armor based

on S-2

glass

and a

phenolic

resin;

this was a ballistic armor designed for stopping the

penetration

of small,

high speed metal fragments. The glass is in the form of

a large

number of

woven roving

plies

with

the

resin

dispersed throughout.

The

high glass

content

(80

weight)

necessitated that this material be cut to

size

for the tests by

means of a diamond

saw.

The nominal thickness

was

12 h

(1/2

in.). In

the results

below

this material is designated

as OC Armor.

Measurement

Techniques: The LIFT apparatus was used for both the ignitability

and

lateral flame

spread measurements. For the ignition tests the sample

was

156 mm

1/8

in.) square

and it was placed in the end of the LIFT sample

holder

which

provides

nearly uniform heat flux from a gas -fired

panel. The

procedures

followed with regard to calibration

and

test

measurements

were

those currently

under consideration by ASTM

[4].

Briefly,

a dummy specimen,

fabricated to support a

heat flux gage, is mounted in the

apparatus. The

radiant

panel and the pilot flame

(acetylene

air)

positioned

above

the

top

edge

of the sample, are

stabilized and the

flux

to the specimen

surface is

recorded. The dummy

specimen holder is quickly

removed and the actual

specimen holder rapidly inserted. At full insertion a

stopwatch is started.

The

test ends when a flame is attached

somewhere

on the

sample and the

requisite time for this is noted from

the

stopwatch.

Flame

attachment means

that a gas phase flame has its base on the sample so

that its own heat flux

assures a continuing flow of fuel gases from the

sample to feed the flame. It

is not necessary that the foot of the flame

cover

large portion of the

sample

face or the sample edge to satisfy this

criterion and, in fact, it

frequently did not.

Prior to

flame attachment, flashes of

flame were often

seen; these would

propagate rapidly from the

pilot flame to some region of the

sample surface

and then disappear. Such a

phenomenon does not represent true

ignition.

These procedures were repeated

with new samples at

several

heat

flux levels

to establish the variation of ignition

delay time

with

heat flux.

Establishment of the

minimum heat flux

for ignition generally

requires

several

tests

which

one attempts to bracket the limit.

The longest delay time one

waits

for ignition

has been arbitrarily set

at 22 minutes. It is

apparent

from the

resulting data that this is adequate.

The lateral

flame spread tests were run in

the same apparatus but

with longer

samples

(794

mm or 31

in.). As

noted

above, the sample

temperature affects

the flame

spread rate.

The normal

procedure

calls for pre-heating

the sample

until

reaches equilibrium with the

local incident flux

(recall this

varies

along

the length

of the sample)

This

procedure

was followed except in the

case

of the

CG Green Vinyl material.

Waiting for full thermal

equilibrium

here

allowed

the

vinyl

surface coating to

fully degrade

to the point

where

was

not

ignitable by the usual ignition method.

The length of time

required

for

thermal

equilibrium

with

the incident heat

flux is determined

from the

ignitability

results by correlating them

in a

manner which is

illustrated

below.

After

pre -heating (typically with the maximum flux just above

the

minimum for

ignition) the gases

issuing

from

the

surface are

ignited with

pilot

flame

and flame

spread on the pre-heated surface

is initiated. Progress

of the flame front

across the sample face (toward the direction

of lower

incident heat flux) is

followed visually with

optical

arrangement which

allows determination of the times at

which

the flame reaches

pre - determined

positions (typically 25 mm apart). The flame front is not necessarily

flat;

one normally

focuses

on the progress

of the flame

across the mid-height of

the

sample. The actual flux incident at each position is precisely calibrated

beforehand. Thus

when

the flame reaches a position on the sample face beyond

which

will

not propagate, one

knows the corresponding minimum heat

flux for

lateral flame

spread (and

the

corresponding minimum sample surface temperature

for

continued spread)

This same procedure is repeated

for

three separate

samples of each material.

Results and

Discussion

3.1)

Ignitability

Sample Behavior.

All of the samples

exhibited some form of idiosyncratic

behavior

when

heated. CG Camel and CG

Yellow behaved alike. Both underwent

explosive delamination of the outermost layers relatively early in the

heating process. This appeared to be due to the

volatility of some component

binding these layers together;

when

it began to vaporize, the

vapors

could not

escape

through the non-porous "paint" layer.

Bubbles formed between

layers,

expanded and then blew loose all

layers

down to the outermost glass ply.

Sometimes these bubbles

turned

the

sample

face into

irregular surface with

several "flaps" of loosened material opening out randomly; other times most of

the sample surface was occupied

by one

large "flap"

disrupted material.

all cases

these flaps undoubtedly sharply altered the

flow

of heat into the

sample

interior. Soon after the explosive delamination the

various

protruding

fragments

of the outer layers charred rapidly with a considerable efflux of

white smoke.

Only at high incident fluxes did this smoke

ignite and

only then

did the

actual face of the sample ignite. When a sample

was

exposed to

lower

fluxes this smoke emission ended

and many

minutes passed before

visually

smaller amounts of smoke began to be emitted once again. In the

interim the

glass

plies detached from the Nomex core, also as a result

of internal

pressure,

but they remained intact, merely bulging

outward about

1/2

cm in the

sample center;

this too must

have

affected the

heat flow in the sample. This

internal

pressure

was

evidently relieved out through the

edge of the sample;

this appeared to establish a pathway to the outside used

preferentially by all

subsequent

gases generated in the sample interior.

As a

consequence, the

ignition and subsequent burning of the sample at

these

lower

fluxes

was

at one

edges

exclusively.

The CG Green Vinyl samples might potentially have

behaved in a similar manner

since it appears to have a structure similar to the

other CG materials.

However,

the surface layer of (presumed) vinyl material

was so much more

flammable

than the surface layers

the

other two CG materials

that it

dominated

the

observed behavior.

This layer

rapidly softened

and bubbled,

even at

very low fluxes. The bubbles

swelled considerably,

sagged and

charred

with

emission of visible smoke. The

bubbles, some

which were several

centimeters in

diameter,

undoubtedly have

substantial

effect on the

heat

transfer

in the

sample. The evolved smoke was readily ignited

the pilot

flame

with

subsequent flaming

over

most of the sample face. The burning

this vinyl

facing

lasted long enough (up to

two

minutes) for it to be

considered the

primary hazard. The

subsequent ignition

and burning of

the

deeper layers

of the

composite, analogous

the

behavior

noted for the

other

two

materials,

was noted in a

few tests but

was

not followed in detail.

The OC Armor,

with its drastically different

structure, behaved quite

differently

from the

above materials in some

respects.

The geometry

of the

material

never

changed at any time in an

ignition test. The surface

simply

turned

black

slowly with

very

little

visible smoke emission. The high

mass

per unit

of facial

area of the material

slowed the heat-up and subsequent

ignition

process.

Flaming ignition,

when

it finally

occurred, was either at

the sample edge or at

the sample face.

Face ignition

was always localized to

one

or a few

very small (few mm) jets of gas

which appeared to issue from

pinpoint

defects in the

sample face. Typically,

as time increased, the size

of each flame

increased

somewhat, as did the number of jet- like

flames.

However, the full facial area of the

sample was never

involved

flaming.

Edge

ignition was more

diffuse; it presumably resulted from

defects

which

happened to be near

the cut edges of

the sample. Unlike the

situation

with

the CG

samples, here edge

ignition or face ignition could

occur over the same

flux range (see

below)

Heat Flux Dependence of Ignition

Delay, Figure 1 shows the

ignition delay

time for CG Camel as a

function

of the

incident radiant heat

flux. The line

through

the

data points is a best fit polynomial,

not an ignition

model line;

the model fit is discussed

below.

Clearly,

the incident flux has

a drastic

effect on

the ignition delay time

especially as one gets

near the minimum

flux

for ignition (here about

2.2

W/cm

What

not apparent from the

smooth line

the

Figure is that there is a transition

in the range

from about 3

0 to

4.0 W/cm

from edge

ignition (at

lower fluxes) to

face ignition (at

higher

fluxes)

The implication

of this behavior is that

edge effects are

extending the flux

range over which

this material will ignite.

That is, it

appears that in the

absence

of edge effects, the minimum flux for

ignition might

be closer to 3.2

rather than

2.2 W/cm

This is only an

estimate of

the edge effect;

we will

pursuing

this issue more precisely in

the future

with

different sample

holder

which

should eliminate edge

ignition. It is

important to separate

out

edge

effects

for

two

reasons. First, they

are real

and probably

are

found

in full scale fires where they may

worsen the

flammability of a

material,

as here. Thus

it is

important to be

able to isolate

and study these

effects

separately.

Second,

these effects

probably will not

follow the

simplified

ignition model used to apply

face ignition

results from

the LIFT

apparatus

other situations.

There is reason

to believe

that the edge

effects

may be scale dependent, in contrast

to face

ignition

behavior. As a

result

one has to be cautious

using the

polynomial fit

shown in

each of the

Figures having

ignition data; the

polynomial describes

the data

shown quite

well but it may fail

to do so at

some larger or smaller

scale for

the heated

sample area.

was

noted

above

that

the

physical

behavior

of the CG Camel

and CG Yellow

samples

was

essentially the same. Figure 2 shows that

their ignition behavior

was indistinguishable within the variability of

the samples. Evidently

they

are virtually identical in composition except

for the pigment

in the face

"paint"

The reflectivity of the two surfaces (Camel

and Yellow)

for infrared

radiation from the gas -fired panel in the LIFT apparatus is apparently

not

significantly different. Figure

with an expanded vertical

scale,

shows

the

sharp separation, seen

with

the CG

Yellow

material, between edge ignition

and

face ignition.

Figure 4 shows the ignition behavior

of CG Green

Vinyl;

note the expanded

vertical

scale. The

vinyl

facing clearly has a deleterious effect on

ignitability

The minimum flux for ignition has moved down to about

W/cm

The

ignition

delay time at 2

W/cm

has

gone

from something greater than 20

minutes (perhaps infinity) for the CG Camel or CG Yellow down to about 40

seconds

Figure

shows the ignition results for the OC Armor composite; this is the

most ignition-resistant material of all those examined here. Note that

edge

effects once again

have extended

the

flammability of a material; here the

minimum flux for face ignition is about

W/cm

higher than it is for edge

ignition (approx. 4.5

vs.

3.5 W/cm

Now, however,

the edge

effects continue

throughout the

flux

range where ignition

was observed.

This implies that for

any given sample at any flux, ignition may occur unpredictably at an edge or

on the sample face. Since

ignition

appears to be a result of gases generated

or preferentially escaping from localized defects in the sample, the

ignition site apparently depends on

whether weaker

defects happen to be on the

sample

face

near an edge. For this material, in

contrast to the

materials above, edge

ignition

was

not

obviously

associated

with internal

delamination

creating

lower

resistance gas flow paths out through the sample

edges. However,

the

fact that edge ignition

always occurred

more readily than

face

ignition suggests that

gases could more easily escape to the edges

when

defect

was present

near an edge.

Correlation

of Ignition

Data. As discussed in Ref.

the

ignition delay time

data from

a variety of materials

can

be correlated in

simple manner from

which

one

can infer parameters

for

the ignition and flame spread

models. The

ignition model

predicts that

plot

(through

the origin) of

)

vs the

square root of the

ignition delay time will give a straight

line. Here

the minimum flux necessary

for ignition and

the incident flux needed to

yield a given

ignition delay

time.

The

model

underlying this

relationship is

one-dimensional.

Thus

does not anticipate the edge

effects seen with some

these materials.

For this reason

we have

applied

this correlation,

for the

most

part, only

to the face ignition data. The results

are shown in Figures 6

the correlation

is reasonable in all cases. Note

that the CG Camel

and

CG Yellow

materials are

treated as one for this

purpose in Fig. 6.

The

objective

seeking this correlation

is the

parameter values

one

can

infer

from it. The intercept of the correlation

line with

the unity value of

the flux

ratio

gives

one a working measure

of the time

needed for an

externally

heated sample of the given

material to come

to thermal equilibrium

with

an external

flux (regardless

of the

value

of the

flux)

This value is

used

as a pre-heat time in the lateral flame spread

tests discussed

below

(with some

exceptions as noted below)

A second

parameter is the effective

thermal

inertia (product of thermal conductivity,

density and heat capacity)

of the

sample.

Since some of

the samples undergo drastic physical changes

during

the ignition

process,

this parameter is quite

empirical, being an

average

over the changing sample structure.

A third parameter is the

effective surface

temperature of the sample

at ignition. This uses a

experimental

correlation

[4]

developed for materials whose physical

deterioration during

ignition was

much less than

some

of that seen here

so,

again, the

result must be

viewed

as an effective empirical value rather than

the real

surface temperature (the surface of

some

of the present materials

not at all

well-defined by the time of ignition)

With these caveats in mind, the inferred parameter

values

can be found in

Table 1. The

thermal inertia and

effective

ignition temperature are used in

the model

equation

for

lateral flame spread rate (see below)

The slope in

plots such as Figures

6-8,

denoted here as "b"

is used in the

following

expression for calculating ignition delay times.

l, t

’m

Here t is the

ignition

delay

time and t is the

thermal equilibrium time

discussed above.

Putting the appropriate

parameter values from Table 1 into

this

expression, one finds that it generally gives a fairly

accurate

prediction of

the observed ignition delay times for the sample

face,

especially

for engineering purposes; the accuracy of

the predictions is

comparable

to that seen for the correlation lines in

Figures 6

-8.

These

predictions

for face

ignition

delay time should be adequate for

these

materials

if the heated area is three to four times larger

than that used

the

LIFT apparatus

[1].

Beyond this range some checking

should be done.

The same data

reduction procedures and correlations can be

used for the edge

ignition results.

Figure

shows such a correlation for OC

Armor; the

correlation

looks quite good.

general,

however, the edge correlations do

not

look this

good. Furthermore, as noted

above, the model used

for both

ignition

and lateral

flame spread does not

consider such non one

-dimensional

effects

as edge

ignition and burning. Again,

there is good reason

to believe

that edge

effects

are scale dependent. This

whole issue needs further

study;

it appears

to be

of considerable importance in

the flammability of composites

which

exhibit

non- isotropic properties.

3.2)

Lateral Flame Spread

Sample

Behavior.

The physical behavior of the samples

was in all cases

similar to that described

above

for ignition. The impact of this

behavior on

flame spread varied with the sample

formulation.

The primary data

from

the flame spread tests are plots of

the most forward

flame

position as a function of

time. Examples of these are shown

in Figures

12.

On these plots the incident flux varies with

position as shown

Figure 14. There one can see that it is fairly constant for the first

150 mm

or so; it then decays monotonically in a nearly linear manner. At 500 mm

the

incident flux is

down

about

20%

the peak

value.

The slope

of the plots

Figures

is the inverse of the spread

velocity. Thus the nearly horizontal sets of points in Fig.

11,

for example,

imply a

very high spread rate. Actually flame spread in the normal sense

was

not seen

with

most of these samples

because

complications

their

physical

behavior, as discussed

below.

CG Camel delaminated the most

the high flux

end of the sample, as

would

expected. While

there

was

good deal of flaming,

which

could last up to 20

minutes

subsequent to ignition, there

was

real organized flame spread

process during this time. An organized

flame spread process is one in

which

heat transfer from the flames in the ignited region of the

sample causes a

smooth

movement

of the

flame toward regions of the sample that

are

cooler by

virtue of their receiving a lower incident heat flux

from the radiant panel.

was noted above,

the

delamination

process

left fuel gas exit paths at

random

locations around the edges of the

sample; here these paths led out

through the three edges around the high flux end of

the

sample.

Burning of

the gas streams emerging from various points along these edges did not

lead

smooth propagation of the

flames

along

the edge.

The appearance of flame

spread along

the edges

was

noted in some cases, i. e.,

after some time flames

did appear

at points along the edges which were in the

direction of

lower

incident radiant flux. However, this was always spatially

discontinuous; it

seemed to be due

delayed

piloted ignition of gas

streams that may

have been

present from the original delamination process.

the

delamination process

itself

was

propagating, it

was

not apparent. Some attempts

were made to

obtain

flame spread on the front

face

material

using a

pre-heat flux high

enough to cause

face ignition; these samples tended

to ignite spontaneously

during the attempted pre-heat interval (which was evidently too long)

and did

not

give

the desired

facial

flame

spread.

The explosive delamination

early

the heat-up

may preclude this since it breaks the

facial material

into random,

disconnected segments. CG Yellow was

not

tested for lateral

flame spread

since it behaved so

similarly to CG Camel in

the ignition

tests.

CG Green Vinyl

came closest to exhibiting a simple

flame

spread process across

the sample face. There were two complications.

First, as

noted above, the

vinyl

layer swelled with large scale (several cm)

bubbles;

these made the

spread process somewhat erratic. Second, as was also

noted

above, the

vinyl

surface

layer tended

to char

and become largely inert if

was subjected to

an extended period

of pre-heating.

For

this

material,

the normal

procedure

was

altered.

Instead of pre-heating the sample

to its

thermal

equilibrium

time, it

was pre -heated only for the normal ignition

delay time for

the

particular flux

chosen. This interval was assured

by keeping the

acetylene

air

pilot flame lit

during

the pre-heat

period;

it ignited

the

evolved

gases

soon as they

reached a flammable concentration.

Armor also

did not exhibit an organized flame spread

process. As noted

above,

flaming

was always localized to small jets emerging

from the sample

face

or to

gases emerging from the sample

edges

There was essentially no

spread of

the edge flames. There

was some semblance

spread with

the face

jets in that

newly flaming jets

appeared

after

some

time

interval

cooler

regions

of the sample face.

The new jets

were

never contiguous

jets,

however.

Thus they were not the result of localized heat transfer

causing

a given

jet to become larger; they

could have

been the result of

heat

transfer

through the

sample from the total assembly of flaming jets on

the

sample face.

Lateral

Flame Spread

Parameters.

light of the

above

discussion,

there is

correlation to be

made of

the

data for any of the materials except CG

Green

Vinyl because only it

appeared to yield a true flame spread

process. Before

turning

to this last material,

however, one can estimate lower bounds for

pair

of useful parameters

for the other materials. These

parameters are the

minimum incident heat flux

necessary to support lateral

flame spread and the

minimum

temperature of the sample surface

(ahead of any flame heating

effects)

necessary to support lateral

spread. These two parameters are

interrelated;

the

temperature is that

achieved

the sample surface at

thermal equilibrium

with

the

minimum incident heat flux. This

flux is estimated

from the position

of the

forward-most flames in the LIFT

spread tests coupled

with the pre-

calibrated flux versus position.

The results are

shown on Table

the

flux incident on the

location of most forward

flame appearance and T

s min

the

surface

temperature achievable at

thermal equilibrium

when this flux

incident

For CG Camel the most forward flames

were on the edges at

a position

corresponding to the

flux

shown; the large

scatter in this

flux (from three

tests)

is a result of the random

emergence of the

gas streams

along the sample

edges.

The corresponding minimum

sample surface

temperature for

flame spread

shown

parentheses because it is higher

than the

minimum ignition

temperature for this material

shown

Table 1. In

the context of

the

simplified model

for ignition

and lateral

flame spread

described in

Reference

this result

contradictory. The

highest

sample surface

temperature

allowed

equal

to the

minimum ignition

temperature

itself

because

achieving

this

temperature

pre-heat would lead

to an

infinite flame

spread

rate upon

ignition

(ignition

everywhere

once). The

source of

this

discrepancy

for

this material

is probably the

dominance of

edge effects

not

accounted for in

the model

(plus scatter in

the experimental

data

)

is not

uncommon, even for

materials

which

behave better than

those

studied

here, for the minimum

ignition

temperature and

minimum

sample surface

temperature

for flame spread to

appear to

be inverted as

they do

here. Data

scatter coupled with

simplifications in

both the

ignition and

flame

spread

models probably

account for this

Green Vinyl also

shows

a discrepancy between

the estimate

of the

minimum

pre-heat temperature for flame spread and the

ignition temperature

of the

material in Table

This cannot be attributed

to edge effects

since the

behavior under consideration for

this material is

strictly facial

(both

ignition

and lateral flame spread)

Recall that this

material was

unique in

that the full pre-heat time could

not be

allowed;

it caused

complete

charring

of the thin vinyl layer and a subsequent

lack of flame spread.

It is probably

this same

phenomenon which

halts flame spread (or at least

contributes

to the

halt) since the

equilibrium

pre-heat time

(268 s

from Table

which

causes

complete

charring

of the vinyl,

is comparable to the absolute heat flux

exposure times at which flame spread stops in some of the tests.

This is a

reactant

depletion effect which

not included

the models

of ignition and

flame spread

used for correlating the data.

It is

particularly pertinent

thin

flammable layers atop a less flammable substrate.

Oc Armor did not exhibit normal lateral flame spread, as discussed above, but

the

behavior of the minimum pre-heat temperature in Table

2 is

at least

consistent with the model expectations; it is

°C less than the minimum face

ignition temperature in Table

The flame on the sample edges typically

reached slightly greater distances than that on the sample face; the minimum

heat flux for

the

edge

flames is thus about 10

15% lower than the value

shown

in Table 2. The high minimum pre-heat temperature

and

the corresponding

high value for the minimum flux for flame spread indicate the

relatively

high

stability of this material. It is relatively difficult to ignite and it

resists lateral flame spread.

As noted

above,

CG Green Vinyl came closest to exhibiting a normal

flame

spread process though this had to be examined at a

lesser than usual extent

pre-heating.

For

this reason an attempt

was

made to correlate the

spread

data

in the manner used to obtain a parameter for the flame spread model

equation.

The model equation

is the

following:

(kpC)(T

)

Here

is the lateral flame spread velocity, $ is the parameter

whose value we

seek;

it is a measure of the heat transferred from

the

flame

to the sample

surface

ahead of the flame, (kpC) is the thermal

inertia of the

sample

(its

value was

inferred from the ignition data: see Table

1),

the sample

surface temperature

at ignition (again see

Table

and T

the temperature

of the sample ahead of

the flame

front (likely to be

elevated by

pre-heating

from an external

flux; this effect is

calculable; Ref.

1).

Thus

the only

unknown

in this equation for flame spread

velocity is $. A

plot of the flame

spread data such

that shown

Figure 13

allows one to infer

the value of

;

it is related to the slope. The

data in Fig. 13 are

problematical in that

regard, however.

The scatter is

such

that

the true

slope is quite

hard to

discern.

The line

drawn there was forced to go

through the

point on the

horizontal

axis equal to 0.87

W/cm

This is

the

minimum heat

flux for

ignition determined

for the CG

Green

Vinyl

material

(see

Table

1);

the model

says the

data should converge on this point.

The line

in the

Figure is

plausible

but

not

fully convincing.

The

value

<£

one

infers from its slope

710 in

units appropriate for use with the

other parameters

Table

Using

this to predict

flame spread

velocities as a

function of

(and thus as

function of

equivalent

incident heat fluxes, assuming

thermal

equilibrium)

gives a set

of numbers

which

behave plausibly but which

cannot

be checked

directly

since they call

for

incident

fluxes below

the

minimum for

the LIFT

apparatus

(approximately 1

W/cm

)

Again, there is a

phenomenon occurring

with

this

material,

reactant depletion, which

is not included

in the theory

which gives

rise

to the

above

equation. Thus one cannot

expect

to give

perfect

correlation of the data or to produce a perfect

predictive result.

The

preceding does

illustrate the

process

whereby

one goes

from the

raw

data

predictive equation; this process has been shown elsewhere

[2,5]

to yield

useful

results for a

variety of materials, including

composites for aircraft

interiors

Summary and

Conclusions

The

two

types

of composite material

tested here were

distinctly

different

physical structure; the first type

was a honeycomb sandwich panel and the

second was a high density

composite armor. Three different facings were

examined on the

honeycomb panels. Two apparently differed only in

color;

the

ignitability

behavior

(ignition delay time

versus incident heat flux level) of

these

two was

so similar that separate

lateral

flame

spread tests were deemed

unnecessary. The third honeycomb panel had a facing that appeared to be a

thin vinyl sheet.

This

sheet proved to be much more flammable than the

underlying

panel

structure;

its

ignition delay time at any given flux

was

less

than that of

the

other two honeycomb panels,

was its minimum

incident

flux

for lateral

flame spread. The composite armor material had a

higher mass per

unit

of exposed facial area than the honeycomb panels, potentially

providing a

greater

fuel load, but it also was

80%

percent glass by

mass. This material

was

the most

ignition resistant of those tested here.

Only the honeycomb

panel

with

the vinyl facing

exhibited a simple

flame spread

process; even

this

was

made

erratic

by large bubbles

and

was also evidently

affected

by charring

of the vinyl ahead

of the flame front at long

exposure

times. The other

materials allowed some slight

advancement

flames on their

heated

faces but this was

not flame spread in the

normal sense

Here again

flames

progressed

the least on the composite armor.

The

ignition behavior of

these materials

was correlated

with reasonable

success

by the

simplified ignition model

described

Reference

An attempt

was

made to

apply the flame spread model described

there only to the

vinyl

faced honeycomb

panel;

this

was

partial success

apparently limited by

reactant

consumption

effects not included in the

model.

All of the above

materials

exhibited

greater flammability

around the sample

edges

than

on the sample face.

ignition this

meant lesser

ignition delays

the

edges or

preferential ignition at

the edges in part of

the incident

heat

flux range

In lateral flame spread this meant that

the flames

progressed

further along the edges than along

the face. The edge

effects

appear

to be tied to a tendency for the composite to delaminate

and produce

preferential paths

for gas flow out

the edges rather than out through

the face

even though this latter path may be much shorter. These edge effects

deserve

much

closer study to determine how they vary

with

sample size and

to ascertain

what extent they can be predicted and controlled.

References

Ohlemiller,

"Assessing the Flammability of Composite Materials",

National

Institute of Standards and Technology Interim Report, in

review.

Quintiere, J., Harkleroad, M. and Walton, D.,

"Measurement

Material

Flame

Spread

Properties", Combustion Science

and

Technology 32

(1983),

Atreya, A.,

Carpentier, C. and

Harkleroad, M.

"Effect of Sample

Orientation on

Piloted Ignition and Flame Spread"

Proceedings of the

First

IAFSS

International Symposium

Hemisphere Publishing Co.,

New York,

(1986), p.

Quintiere, J.

and Harkleroad, M.

"New Concepts for Measuring

Flame

Spread

Properties",

Fire Safety

Science and Engineering

ASTM

Special

Technical

Testing

Publication

882,

Philadelphia,

(1985), p.

239

Harkleroad, M.

"Ignition and

Flame Spread

Measurements of

Aircraft Lining

Materials",

National

Bureau of Standards

NBSIR

88-3773,

May,

1988

Table

Parameters Inferred from

LIFT Ignition Data

Material d, ('W/cm

T. (°C) b(s~M

(kpc) m

CG Camel

Yellow 3.50 450 0.100

100 0 88

CG Green Vinyl 0.87 275 0.061 268

OC Armor 4.70 630 0.033 930

7 20

*above are for face ignition, not edge ignition

Table

Parameters Inferred from LIFT Flame Spread Data

Material

XW,

(°

CG Camel

3.1

(530)

CG Green Vinyl

1.2 ±. 2

(330)

OC Armor

3.8 ± 1

585

edge

face

CG Camel

Ignition

Figure 1

Ignition delay time

vs. incident heat flux for CG

Camel honeycomb

panel. Equation

shown is

least-squares fit and R

the

correlation

coefficient.

Camel

& CG

Yellow

Ignition

Camel

Ignition

Yellow

Ignition

Flux (W/cm2)

Figure

Ignition

delay vs. incident

heat flux for CG

Camel

and

Yellow

honeycomb

panels.

Ignition

Time

(s)

Face vs Edge Ignition of

Yellow

800

600

400

200

Edge Ignition

Face

Ignition

£3

0 123456789 10

Flux (W/cm

Figure

Ignition delay time

vs. incident heat flux

for

Yellow honeycomb panel;

face and

edge

ignition

data points are distinguished.

CG Green

Vinyl Ignition

Flux

(W/

Figure

Ignition

delay time vs.

incident heat flux for

Green Vinyl honeycomb

panel.

Equation

shown

is least-squares fit and R is the correlation coefficient.

Ignition

Time

(s)

Armor

Face vs Edge

Ignition

4.133e+4

-2.4127

0.98

Face Ignition

1.152e+4

-1.9744

0.98

Edge

Ignition

Figure

Ignition

delay time vs. incident

heat

flux for

Armor composite

armor panel.

Face

and

edge

data

points are

distinguished.

Equations shown

are

least-squares

fits and R

values

are correlation coefficients.

CG (Camel

Yellow) Ignition Correlation

Camel

Yellow

Figure

Ignition

model

correlation for

Camel and

Yellow

treated

as a single material.

Data

points

for

face ignition only.

CG Green

Vinyl Ignition Correlation

Figure

Ignition

model

correlation for

Green

Vinyl. Data points for face

ignition only.

Armor Face

Ignition

Correlation

Figure

Ignition model correlation for

face

ignition of

Armor.

Armor

Edge

Ignition

Correlation

(

1/2

)

Figure

Ignition

model correlation for

edge

ignition of

Armor.

Armor Flame

Spread

Figure 10

Experimental

data

for lateral

flame spread on OC

Armor.

Edge

and face data are

distinguished. The value

is the

peak incident flux on the sample face.

Camel Flame Spread

1500

1200

900

600

300

0 100 200

300

400 500

Flame Front Position (mm)

Figure

1 1

Experimental

data for lateral flame spread on

Camel honeycomb

panel. Data

shown

are for sample

edges. Each test was run at

different

value of

the peak

incident heat flux, qe.

=4.58 W/cm

q'=3.57

W/cm

q^=

5.05

W/cm

(?)

Time

(s)

Green Vinyl Flame Spread

Figure 12

Experimental

data for lateral flame spread on

Camel

honeycomb panel. Data are

for sample

face. Results for two different values

of the peak incident heat flux are

shown.

Green Vinyl Flame Spread

F(t)

(

W/cm“

)

Figure

Model correlation of lateral flame

spread on CG

Green

Vinyl honeycomb panel.

Line shown

has been forced to

pass

through a

value of 0.87

W/cm2 on the

horizontal

axis.

Figure

Distribution

heat

flux

incident

sample

surface, normalized

peak

value.

NBS-114A

irev.

2-80

U.S.

DEPT. OF

COMM.

BIBLIOGRAPHIC

DATA

SHEET

(See

instructions)

PUBLICATION

REPORT

NO.

NISTIR-89/4030

Performing Organ. Report

NoJ

TITLE AND

SUBTITLE

Publ

ication Date

January

1989

Ignition

and Lateral

Flame

Spread Characteristics

5. AUTHOR(S)

Ohlemiller

and S.

Dolan

6. PERFORMING

ORGANIZATION

(If

joint or

other than

N BS.

see instructions)

Composite Materials

Contract/Grant

No.

National Institute

of Standards

and Technology

U.S.

Department

of Commerce

Gal thersbura

MD 20899

8 .

Type

of Report

Period

Covered

SPONSORING

ORGANIZATION

NAME AND COMPLETE

ADDRESS (Street, City

State, ZIP)

David Taylor

Research Center

U.S.

Navy

Annapolis,

10.

SUPPLEMENTARY NOTES

Document

describes a

computer

program;

SF-185,

FlPS

Software

Summary, is attached.

11. ABSTRACT

(A 200-word

less

factual

summary

most significant in formation

. If

document includes a

gnificant

bibliography

literature survey,

mention it

here)

The

Lateral Ignition

and Flame

Spread

(LIFT) apparatus was

used to obtain

information

on the

ignition and

lateral

flame spread characteristics

two

types

composite materials.

The

first type was a

honeycomb sandwich panel;

three

different

facings

were

tested with

this material.

The second type of

material was

a composite

armor.

There was

a substantial variation

in the

ignitability

the various

material

combinations with

vinyl-

faced honeycomb

panel

being

the most

ignitable

and

the composite

armor being the least

ignitable.

The

ignition

behavior

of the

facings of all materials was

correlated

a simple

predictive

model. Only the vinyl

-faced honeycomb panel

showed

significant

normal

flame

spread under

the conditions examined

though

some

flame

"advancement"

was

seen with

the others. All of

the materials

exhibited

worse

flammability

properties

at their edges

as compared to their

facings

12. KE

WORDS (Six

twelve

entries; alphabetical order; capitalize

only

proper names;

and

separate key words

semicolon s)

composite

materials; flame spread;

flammability;

ignition

13.

AVAILABILITY

j y |

Unl irmted

For Official

Distribution.

Not Release

to NTIS

~| Order

From Superintendent of

Documents,

U.S.

Government Printing

Office,

Washington,

D.C.

20402.

Order From National

Technical Information

Service

(NTIS),

Springfield, VA. 22161

14.

NO. OF

PRINTED

PAGES

15.

Price

11.95

use

OMM-

D C

6043-P80

wTfm.