Thermal combustion and oxygen chemisorption of wood exposed to low temperature long term heating 4

Chapter Four: Spontaneous Heating and Cone Calorimeter Ignition Tests and Experiments This chapter presents the methodology and experiments for measuring ignition and combustion of wood

Chapter Four: Spontaneous Heating and Cone Calorimeter Ignition Tests and Experiments

This chapter presents the methodology and experiments for measuring ignition and combustion of wood in bench-scale tests: spontaneous heating of wood cubes and ignition testing in Cone Calorimeter For spontaneous heating, the mathematical considerations of thermal explosion were first considered as it laid the groundwork for the experimental design For ignition testing in Cone Calorimeter, the preparation of both green wood and preburn were discussed The determination of thermophysical properties for green wood, preburn wood, and moisture was also highlighted An analytical method was proposed for correlation of ignition data in both green and preburn wood

In the assessment on self-heating propensity in wood, two main methods have been used: heat-based method and the non-heat based method (Wang, Dlugogorski and Kennedy, 2006) The measurement of thermal runaway of wood cubes in preheated isothermal oven

is a heat-based method Heat-based method measures heat generation and hence the critical temperature for autoignition as a basis for evaluating the tendency of the material

to self-heat The non-heat based method on the other hand assesses oxygen adsorption, or the capacity of the material to consume oxygen, as a measure of the propensity to self-ignition The chemical aspect of self-heating in wood by oxygen adsorption is discussed separately in Chapter 5

4.1 Spontaneous Heating

According to Frank-Kamenetskii theory, spontaneous ignition involves a size effect Ignition occurs when heat generated by exothermic processes exceeds heat dissipated to the surrounding; in this case, heat generation is related to the cube of dimension while heat dissipation is related to either dimension or square of dimension (Walker 1967, Cuzzillo 1997) This section discusses the principle underlying the critical F-K parameterδ in determining the critical size of wood cube and the experiment cdesign used to study spontaneous ignition

4.1.1 Method to determine the size for thermal runaway

This method was developed by Bowes and Cameron, based on the Kamenetskii model for thermal ignition of packed solids (Bowes and Cameron 1971, Bowes 1984) The mathematical considerations that underlie the experimental design of spontaneous heating of wood slab are shown below

where it is subjected to the following boundary conditions:

where r represents the radius or half-thickness of the slab, h denotes the convective heat

transfer coefficient at the solid surface, and T∞ is the temperature of the ambient

At steady-state, the non-dimensional steady state form of the general energy conservation

ρδ

The boundary conditions for Equation (4.6), as deduced from the definition of z and that

δ is known as the critical Frank-Kamenetskii parameter, which is primarily a function of

geometric shape of the solid and the average surface Biot number Beever (1995) gives

an excellent review of critical parameters for various geometries and the corrections required for proper application

predict critical temperatures T∞ as a function of size r of the wood slab

However, this method can be cumbersome because of the lengthy experimental iteration process to determine the kinetic parameters, involving first the evaluation of a series of

critical ignition temperature T∞ for different sample sizes A number of trial and error experiments have to be carried out to “bracket” the right critical ignition temperature for

a given sample size As a result, a large number of experiments are required just to get the kinetic parameters

Chen and Chong (1995, 1998) introduced the “crossing-point” method which reduces the number of experiments required to determine the kinetic parameters For a sample undergoing self-heating, there will be some point in time where the temperature profile at

the centreline becomes locally flat At the unique time where

2 2 0

where T is the centre-plane temperature at which the conduction term vanishes, i.e the ctr

centre region of the slab is adiabatic

Taking logarithm of both sides of Equation (4.9) yields

c , the log of the theoretical maximum adiabatic rate of temperature rise

In terms of experimental work, only the T needs to be determined, not the critical ctr ignition temperature T∞ T is the centre plane temperature where the conduction term ctr

vanishes In practice, two thermocouples are placed at a short distance away from the centre line of the sample, and the temperature is determined when the temperatures of the two thermocouples cross each other In Chen’s method, every experiment is able to yield

a T ; in Frank-Kamenetskii method, many experiments are however required just to ctr produce one T∞

In experimental determination of the critical size relating to spontaneous combustion for

a given wood slab, and/or its associated kinetic parameters, the crossing point method (Chen and Chong 1995, Chen and Chong 1998) or simply known as Chen’s method has clear advantages over Frank-Kamenetskii model First, it reduces the number of experiments required to determine the kinetic parameters Since the shape of the material and the Biot number hr k/ constitute the most important parameters affecting δ (Bowes c

1984, Cuzzillo 1997), the exclusion of critical Frank-Kamenetskii parameter eliminates the direct Biot number effects on data interpretation Heat transfer coefficient h in the oven and conductivity k of the wood slab need not therefore be known during the experiments Chen’s method certainly offers a much simpler and neat experimental method

4.1.2 Self-heating experiment

Cuzzillo (1997) has used Chen’s method to derive a critical size of 89mm wood cube corresponding to a critical ignition temperature of 200ºC In this study, the same critical size and critical ignition temperature is used: Kapur wood was sized into 89mm cubes and placed in an isothermal oven maintained at 200°C to investigate the effects of self-heating The purpose of adopting the same sample size and critical temperature is to enable comparison of results to be made between these two studies directed at investigating self-heating in wood cubes

The 89 mm (3.5”) cube was to be heated symmetrically in the isothermal oven at 200oC until thermal runaway was observed To study spontaneous heating, thermocouples type-

T were inserted at the centre of the cube and connected to a data logger Yokogawa DAQ station DX230 A total of 18 thermocouples were used Holes were drilled to reach half the depth of wood cube i.e 45mm and sixteen thermocouples were inserted The details

on the location of thermocouples are discussed in Section 4.1.2.2 If the centre temperature exhibited thermal runaway, the temperature would show a “peak” on the thermocouple readings The wood cube experimental design was discussed below

4.1.2.1 Specimen preparation and orientation

The cube was oriented so that the two side-grain surfaces (i.e grain of wood is perpendicular to heat flux) facing two heating modules of the oven The wood cube was

bottom faces with calcium silicate wrapped with rock wool and aluminium foil to prevent escape of volatiles through these faces Wood cube preparation was shown

diagrammatically as in Figures 4.1 and 4.2

The purpose of insulation was two-fold First, it prevented the escape of volatiles, mainly through the end-grain faces in order to minimise the non-isotropic effects on the spontaneous combustibility of wood Secondly, the insulation enabled the wood domain

to be treated as one-dimensional which was the case for consideration in this study

Radiative heat loss

Figure 4.1: The three principal axes of

flow in wood cube

Figure 4.2: Schematic domain on plan view

4.1.2.2 Location of thermocouples

Eighteen type-T thermocouples (operating temperature: -200°C to 350°C) were used to measure the temperature history of the wood cube throughout the heating regimes The thermocouples in use were grouped into 3 groups:

Group 1: thermocouples were positioned along transverse axis of the cube (01, 02, 03, 04,

05, 06, 07, 08, 09) Transverse axis is the line connecting two center points of any two opposite side-grain surfaces as shown in Figure 4.3 Table 4.1 shows the locations of Group 1 thermocouples with respect to the side grain face

Group 2: thermocouples were positioned along longitudinal axis of the cube (09, 10, 11,

12, 13, 14, 15, 16, 17) Longitudinal axis (or axial axis) is the line connecting two center points of two end-grain surfaces, as shown in Figure 4.3 Table 4.2 shows the locations of Group 1 thermocouples with respect to the side grain face

Group 3: thermocouple no.18 hung in the oven to measure the oven air temperature The

location and distribution of thermocouples are shown in Figure 4.3 and Figure 4.4 respectively

Table 4.1 Group 1 Thermocouples

01, 08 0 mm (on side-grain surface)

Table 4.2 Group 2 Thermocouples

Side Grain

Side Grain Face

Heat

Heat

Transverse Centerline Longitudinal Centerline

Figure 4.4 – Location of thermocouples

(Plan View)

4.1.3 Concluding remarks for self-heating experiments

The self-heating experiment described the heat-based experimental method to evaluate the propensity of wood to combust spontaneously The experimental temperature field data from wood cube heating served to validate the porous model and the underlying phenomenological models on low-temperature heating and evaporation of moisture in wood The non-heat based test method was also employed to evaluate self-heating in wood, but it entailed a chemical principle of oxygen chemisorption and is discussed separately in Chapter 5 The methodology framework for assessing self-heating in wood and the use of temperature field development to evaluate self-heating and the effects of evaporation was shown diagrammatically below

Figure 4.5: Methodology framework for analysing wood cube data

HEAT-HEAT-BASED TEST METHOD

NON-WOOD CUBE HEATING

-temperature field data -self-heating development

POROUS MODEL WITH SURFACE EVAPORATION

POROUS MODEL WITH INTERNAL EVAPORATION

OXYGEN ADSORPTION

- Elovich parameters

- Ignition temperatures

SEMENOV MODEL

- to analyse temperatures of ignition of wood chars

4.2 Ignition testing in Cone Calorimeter

This section discusses ignition tests in the Cone Calorimeter in which wood slabs were exposed to variable heat fluxes in a vertical orientation to determine both piloted and spontaneous ignition Times to flaming ignition and glowing ignition were measured Flaming ignition is defined as the initiation of flaming combustion; glowing ignition involves solid phase combustion where the surface glows but no flaming is observed (Janssens 1991a) Minimum heat fluxes for each piloted and spontaneous ignition were determined experimentally The samples were heated in descending heat fluxes, starting from high incident heat flux of 50kW/m2 to subsequently lower heat fluxes Minimum heat flux is determined at the heat flux level at which no ignition in Cone Calorimeter can

be observed

When tested in Cone Calorimeter, the specimen was to be oriented such that heat flux was perpendicular to the direction of the wood grain The predominant direction of the annual rings was perpendicular to the sample surface Therefore, the direction of the heat flux was tangential to the annual rings The specimen orientation with respect to the irradiance arrangement is illustrated in Figure 4.6

The purpose of this investigation was to provide a rational framework to assess ignition

of wood in relation to moisture and its thermophysical properties To examine moisture effects on ignition, pre-burn wood samples were created and tested in addition to green wood These pre-burn wood samples were heated at 250º until a constant weight was

achieved These pre-burn samples thus contained reduced moisture content The experimental preparation of pre-burn samples was described in Section 4.2.2 Green wood samples were tested as fresh, oven-dried wood The use of green, oven-dried samples and pre-burn wood samples provide a basis of comparison on the effects of moisture as well as thermophysical properties on ignition and burning characteristics

Figure 4.6 Specimen orientation in Cone Calorimeter

Pilot flame For piloted ignition

For spontaneous ignition, the igniter would be removed

(1 − ε ) q e′′ + h Tc( s − T∞) + εσ ( Ts + T∞)

e q′′

T Ambient Temperature∞

4.2.1 Specimen conditioning

Specimens measuring 80mm x 80mm x 25mm were used for Cone Calorimeter piloted ignition experiments The intention to keep to a smaller sample size, in lieu of the standard 100mm x 100mm x (up to 50mm) sample, was to maintain the end-use condition of this type of hardwood that was available for commercial purposes Sizing up the sample would require gluing the smaller pieces together; however, complications may arise as the sample may crack prematurely along the adhesion lines, thus compounding the test results Fresh wood samples were tested as oven-dried samples These green wood were first oven dried at 105°C to a constant mass and then stored in a desiccators at

RH 50 + 5 % and 23°C according to ASTM E1354-93 (ASTM, 1993) The green dried wood has an average moisture content of 13.83% The moisture content measurement was shown in Appendix 3

oven-4.2.2 Preburn samples preparation

For pre-burning of Nyatoh wood, wood slabs of the size 100mm x 85mm x 20mm were heated in an isothermal oven at 250°C fitted with a natural inflow of atmospheric air,

as shown in Figure 4.7 These wood slabs were allowed to burn until they achieved an average of 50% reduction in weight The degree of preburn β in wood slabs is expressed

asβ =(m o−m) /m o , where m o is the initial sample weight and m the final sample weight

at the end of the oven test The extent of pre-burn is related to the unconverted mass fraction of wood μ asµ=1−β The collection of pre-burn data was collated and is shown

These pre-burn samples (as shown in Figure 4.8) were stored in a dessicator prior

to ignition test in Cone Calorimeter For pre-burn samples, no oven drying was required, though the pre-burn samples were conditioned to the relative humidity and temperature as fresh wood prior to Cone Calorimeter testing Pre-burn wood had been conditioned to achieve an average moisture content of 6.6%

Figure 4.7 Pre-burning of green wood Figure 4.8 Preburn sample

4.2.3 Determination of minimum heat flux

To determine minimum heat flux for both oven-dried green wood and pre-burn wood, slabs were exposed to successively lower incident heat fluxes until ignition no longer occurred Surface temperatures pertaining to ignition were measured using thermocouples attached to the specimen surface While Babrauskas (2001) questioned the reliability of judg ing ignition criterion based solely on thermocouple readings, Thomson, Drysdale and Beyler (1988) however pointed out that satisfactory performance of the thermocouple in measuring a true surface temperature could be achieved if data from

tests in which thermocouples detached or sank into the material during tests are rejected These two cases produce thermocouple data that are characterised by high noise or low response to ignition, and should be excluded The findings by Thomson et al were made

in reference to thermoplastics; it remains notoriously difficult to measure surface temperatures on wood using thermocouples The thermocouples were staple stitched to the sample surface, and only the thermocouple data that did not detached from the surface were recorded in this study

According to ASTM E1354-93 (ASTM, 1993), tests are terminated at 20 minutes if no ignition is observed A number of studies have shown that ignition may not occur until anything between several tens of minutes and up to one and half hours have elapsed (Babrauskas and Parker 1987, Spearpoint and Quintiere 2001, Mikkola and Wichman

1990, Janssens 1991b) Therefore, for piloted ignition, the green oven-dry hardwood was observed for ignition up to six hours duration, while the pre-burn wood was measured up

to one and half hour test duration, since the latter was less likely to re-ignite due to reduced calorific content

4.3 Analytical model: heat transfer model without moisture

For heat transfer through a solid slab, Fourier’s law postulates that a heat flux

n

q′′ in a direction of n, which is normal to an isothermal surface, is proportional to the

temperature gradient in the solid,

T

A general heat diffusion equation is derived by considering the conservation of energy over a control volume of size ∆x by y ∆ by z∆ as shown in Figure 4.9

Figure 4.9: Control volume of wood in Cartesian coordinates

(Control volume representation reproduced from Yuen, 1998)

The conservation of energy over the control volume is given as

& & & & &

where q&is the rate of accumulation of energy per unit volume; q , x q yand q are the z

normal heat fluxes through the control surfaces due to heat conduction and & is the rate q p

of internal energy production

For an infinitesimal control volume, Equation (4.12) can be expressed as

y

p q

where ρ is density; c is specific heat of the material and T is the absolute temperature

Writing q′′ ,x q′′yand q′′ in the form by Fourier’s law of heat conduction as shown in z

Equation (4.11), and expressing q′′ according to Equation (4.14), the entire expression of heat conduction into the control volume in Equation (4.13) can be written as

In a pure thermal problem, q p= 0; it is assumed that there is no energy release, whether endothermic or exothermic within (such as pyrolysis) or on the surface (such as surface oxidation or charring) of the solid In a simple thermal model, the energy balance in Equation (4.15) can be simplified to

4.3.1 Assumptions for heat transfer for solid slab

The study of wood combustion, whether piloted or spontaneous can be made simple by assuming the problem to be a thermal case The assumption is validated when the wood is semi-infinitely thick, and is considered as inert, opaque and totally absorbing, irradiated on one face by intensity q′′ , and losing heat from that face by Newtonian ecooling (Simms and Law 1967) The modelling of wood combustion as a thermal case has achieved great success, and a large number of correlation models for ignition in wood has been produced based on thermal model (Janssens, 1991a) For a problem to be considered as purely thermal, the mathematical description of the phenomenon inevitably

involves certain assumptions and simplifications For the heating of a solid wood slab, the following assumptions are imposed: -

1 Since the incident heat flux is perpendicular to the exposed surface, the problem can be formulated as a one-dimensional heat conduction problem

2 The solid slab is treated as inert with negligible chemical reactions prior to ignition

3 Convective heat transfer within the wood slab is negligible Experimentally, the effect of water vaporisation is minimised by using only the oven-dry wood slab

4 The heat losses from the surface are partly radiative and partly convective

5 The wood slabs behave as a semi-infinite solid so long as the thickness of slab is more than 16mm and can therefore be considered as thermally thick (Mikkola and Wichman 1990)

Under these assumptions, the heating of wood slab become a purely thermal problem Further assumptions were made with regard to diathermancy to simplify the mathematical model (see next section 4.3.2)

4.3.2 Mathematical model for heat transfer in solid slab

The one-dimensional heat transfer in the solid slab is schematically represented in Figure 4.10 below

Figure 4.10: One-dimensional heat transfer model

Equations for solid phase

Even in a pure thermal model, when a solid is considered as diathermanous, heat transfer within the solid is affected by the in-depth radiation and absorption Wood is diathermanous for small wavelength The one-dimensional energy balance for wood needs to account for the diathermancy,

where α ε= = −1 r, i.e Kirchoff’s law holds for the total α ,εand r For wood slab

heated in Cone Calorimeter, h assumes a value of 13.5 Wm c -2K-1 for vertical orientation (Janssens 1991b) An average value of ε =0.88 is adopted as emissivity approaches unity with increasing heating as the wood surface darkens (Melinek 1968)