Medium-scale reference tests and calculations of spontaneous ignition in wood pellets - the LUBA project

(1)

LUBA

project

Ida Larsson, Per Blomqvist, Anders Lönnermark

and Henry Persson

Fire Technology SP Report 2012:50

S

P

T

ec

hni

ca

l R

es

ear

ch I

ns

tit

ut

e of

S

w

ed

en

0 100 200 300 400 500 600 0 10 20 30 40 50 Tem per at ur e ( °C) Time (h) Test 1 - TC 10 Test 2 - TC 9 Test 3 - TC 14 Test 4 - TC 11

(2)

Medium-scale tests and calculations of

spontaneous ignition in wood pellets -

the LUBA project

Ida Larsson, Per Blomqvist, Anders Lönnermark and

Henry Persson

(3)

Abstract

Medium-scale tests and calculations of spontaneous

ignition in wood pellets - the LUBA project

The use of biomass pellets is increasing. As a consequence, large indoor storage facilities are needed along the transportation chain. The increased production volumes,

transportation, handling and storage of pellets result in increased risks. A number of fire incidents due to spontaneous ignition in wood pellets have been reported. Increased efforts concerning safety and quality assurance are, therefore, important. The aim of present work is to provide methods for estimating risks for self-heating from pellets stored in bulk quantities. This report compares medium scale tests of pellets in bulk with smaller screening test; micro calorimeter and crossing point. Two types of pellets; one “reactive” and one “less reactive” were compared. Kinetic parameters from the crossing point and micro calorimeter tests was used as input data for Frank-Kamenetskii

calculations and compared with the medium scale test results. Calculations of the critical ambient temperature and time to self-ignition have also been preformed on four different types of representative types of full scale storages that either exist today or might be possible in the future.

The results clearly reveal that results in medium scale can be predicted by using results from small scale screening methods like isothermal calorimetry or crossing point. The small scale test methods show the same indications as medium scale when comparing reactive and less active pellet types.

The medium-scale tests were effective in separating the self-heating activity of the two types of pellets investigated.

Key words: Self-heating, spontaneous ignition, wood pellets, emissions

SP Sveriges Tekniska Forskningsinstitut

SP Technical Research Institute of Sweden SP Report 2012:50

ISBN 978-91-87017-68-1 ISSN 0284-5172

(4)

Abstract

3

4 Acknowledgements

5 Sammanfattning (in Swedish)

6

1 Introduction

7

2 Medium scale tests

8

2.1 Experimental design 8

2.1.1 Pellets 8

2.1.2 Test configuration 8

2.1.3 Description of tests 14

2.2 Results 15

2.2.1 Pre-heating and temperatures in the enclosure 15

2.2.2 Bulk temperature 16

2.2.3 Gas emissions 19

2.2.3.1 On-line analysis of CO2, CO and O2 19

2.2.3.2 Analysis of organic species 21

3 Calculations of spontaneous ignition of wood pellets

24

3.1 Theory and calculation tool 24

3.2 Comparison of kinetic parameters 25

3.3 FK-calculations on the medium scale tests 27

3.4 Examples on FK-calculations for large scale storage 29

4 Discussion

32

5 References

33 Appendix 1 - Notes from reference tests

34 Appendix 2 - Temperature measurements during medium scale tests 38

(5)

Acknowledgements

The presented work was part of the research project ”Large Scale Utilization of Biopellets for Energy Applications - LUBA”, sponsored ForskEL (Kontrakt Projekt nr. 2010-1-10541), DONG Energy and Vattenfall, which is gratefully acknowledged. We would further like to thank our colleges at SP that have contributed to the experimental work reported here.

The authors would also like to thank Leif Fjällberg from CBI for their assistance in connection with the performance of experiments.

(6)

Sammanfattning (in Swedish)

LUBA-projektet – Referensförsök i mellanskala och beräkningar av självantändning i träpellets

Användningen av bio-pellets ökar. Detta medför att stora anläggningar för inomhuslagring behövs längs transportkedjan för denna typ av bränslen. Ökade

produktionsvolymer, transport, hantering och lagring av pellets leder till ökade risker. Ett antal brandtillbud på grund av självantändning i träpellets har rapporterats. Ökade ansträngningar för säkerhets- och kvalitetssäkring är därför viktigt. Syftet med föreliggande arbete är att tillhandahålla metoder för att uppskatta risker för

självuppvärmning av pellets som lagras i stora kvantiteter. I rapporten jämförs försök på pellets i bulk i mellanskala med mindre screeningförsök; mikrokalorimeter och crossing point. Två typer av pellets, en "reaktiv" och en "mindre reaktiv" jämfördes. Kinetiska parametrar från crossing point och mikrokalorimeter försök användes som indata till Frank-Kamenetskii beräkningar och jämförs med försöksresultaten från mellanskala. Beräkningar av kritisk omgivningstemperatur och tid till självantändning har också genomförts på fyra olika typer av representativa fullskaliga lager som antingen finns i dag eller kan vara möjliga i framtiden.

Resultaten visar tydligt att resultaten i mellanskala kan förutsägas med hjälp av resultat från små screeningmetoder som isotermisk kalorimetri (mikrokalorimeter) eller crossing point. De småskaliga testmetoderna visar samma indikationer som mellanskala när man jämför reaktiva och mindre aktiva pelletstyper.

Referensproven i mellanskala var effektiva i att separera självuppvärmande aktivitet för de två typerna av undersökta pellets.

(7)

1 Introduction

A series of medium-scale tests was conducted at SP in May 2011. The aim of the test series was to create strictly controlled conditions for self-heating in a comparatively large (1 m3_{) bulk of wood pellets that would result in spontaneous combustion. Two of the}

pellet types earlier studied in LUBA with various laboratory characterisation methods were selected for the medium-scale reference tests. The pellets selected were those denoted L and M. These pellets showed a significant difference in the characterisation tests [1] where L was regarded as active and M was considerable less active. The main objective of the medium-scale reference tests was to investigate if the characterisation of activity made in lab-scale also would prove to be true in a larger scale.

Additional objectives of the experiments were to study the processes in the bulk of the pellets and to collect data for later use as validation data for the mathematical prediction model investigated in LUBA. Detailed instrumentation in the tests gave data of the temperature development in the bulk of the pellets during the self-heating process. Further, the gas transport within the bulk and emissions from the pellet bulk were characterized.

Calculations of the risk for spontaneous ignition in various types of storages of wood pellets using the Frank-Kamenetskii theory have been made in the LUBA project. The medium-scale reference tests were used as validation for these calculations. Basic theory and the results of calculations and the validation are given in the report.

(8)

2 Medium scale tests

2.1 Experimental design

The medium-scale tests were conducted to make an assessment of the self-heating propensity of wood pellets in a relatively large scale. The scale was, however, not large enough to potentially allow significant heat production from the pellets at normal ambient temperatures. In order to give conditions allowing higher self-heating rates, the

experiments were designed as semi-adiabatic tests where an elevated ambient temperature were maintained and kept constant during the test.

2.1.1 Pellets

Two types of pellets were investigated in the tests. These where pellets that had earlier been investigated in LUBA using different small-scale test methods. The pellets selected were those denoted L and M. These pellets showed a significant difference in the characterisation tests [1] where L was regarded as ‘active’ and M was considerable less active.

Table 1 Information on the types of pellets investigated.

Label of

pellets Pellet type Bulk density (kg/m3₎ Pellet diameter _(mm)

L Agro Energi, Sweden

Sampled relatively fresh from production

715 8

M Verdoe, Scotland

Transported and stored for 3 months before sampled

719 8

2.1.2 Test configuration

The test set-up included a cylindrical metal container in which the pellets were placed. The set-up further included equipment to introduce pre-heated air into the pellet bulk and into the insulated enclosure where the pellet container was placed. The test set-up is schematically outlined in Figure 1 and Figure 2.

The cylindrical test container was made of 2 mm steel, had a radius of 1100 mm and a total height of 1920 mm. A 160 mm Ø inlet were located centrally in the bottom of the container and a circular 200 mm diameter flow distributor plate was positioned 100 mm up from the bottom inlet. A coarse grid-iron was mounted 420 mm up from the bottom. This is shown in Figure 1. A much finer metal net was laid on top of the coarser one to hold the pellets. The finer net had a mesh size of 5 mm and was made of 1.25 mm metal wire. The relative opening area of the finer grid was 60 %. Wood pellets were filled in the container to a height of 1100 mm above the grid.

(9)

Figure 1 Side view of the test configuration. V2 and V3 identifies bi-directional probes (o). The positions of single thermocouples inside the enclosure are marked with (x).

Y X Z 400 1100 420 360 3600 2495 1100 Ø 600 95+10 2 Pre-heating inflow 95+10 TC_54 V3 160 Ø ₁₀₀ 3000 V2

(10)

Figure 2 Top view of the test configuration. V1 identifies a bi-directional probe (o). The positions of single thermocouples are marked with (x). The positions of the two thermocouple trees inside the enclosure are marked with (x-circle). A, B and C identifies valves in the heated air distribution system.

Additionally, during a test a metal lid with a centrally located 160 mm Ø opening was placed on the top of the container. The test container was further equipped with a hatch located at the front side for removing pellets after completion of a test. The hatch can be seen in Figure 3-c. 2500 X Z Fan Heater Pellets 160 Ø 25 0 Ø 2500 600 10 16 24 25 40 Outflow A ll du cts in su la te d A. B. C. 100 Ø TC 52 V1 600 250 TC 46-48 TC 49-51 TC 53 V2

(11)

(a) (b)

(c) (d)

Figure 3 Photos of the test set-up. (a) The test enclosure with front wall sealed closed. (b) Rear view of the test enclosure with the insulated duct system. (c) Enclosure with the front wall removed and pellet container visible. (d) The surface of the pellet bulk with TC trees and sampling lines visible.

To introduce pre-heated air into the test enclosure a fan was connected to a duct heater (22 kW) and the heated air was transported in an insulated 250 mm duct with valves for distributing the heated air to the pellet container (valve B) or alternatively directly into the enclosure (valve A). The velocity of the heated air was measured with bi-directional probes in the main duct (V1) and in the duct leading to the pellet container (V2), see Figure 2. A bi-directional probe (V3) was also positioned in the inlet to the pellet container, see Figure 1.

The walls of the enclosure were made of 10 mm non-combustible (Promatect®- H) boards insulated at the outside with 95 mm glass wool. Additionally 6 mm Plywood was covering the back wall and parts of the removable front wall (see Figure 3). The ceiling was constructed as the walls. Note that 30 mm mineral wool was covering the floor of the enclosure during tests to reduce the heat losses.

(12)

In order to be able to extinguish the wood pellets in tests where spontaneous ignition occurred, the injection of nitrogen gas (N2) was prepared from the bottom of the test

container (can be seen in Figure 3-c).

There were thermocouples (TCs) positioned at different heights (2.9 m, 1.925 m and 0.75 m) at two locations within the enclosure for measuring the ambient temperature seen by the pellet bulk, TCs 46-48 and TCs 49-51 (see Figure 2).

A large number of thermocouples were further positioned within the pellet bulk for a precise mapping of the temperature distribution during a test. The positions of the 0.5 mm type-K thermocouples placed in the pellet bulk are shown in Figure 4- Figure 6, where P1-P9 denotes TC trees and where the locations of individual TCs are identified with numbers.

Figure 4 Top view of the pellet container with positions of vertical measurement trees shown (P1-P9).

There were also gas samples extracted for analysis of oxygen (O2), carbon monoxide

(CO) and carbon dioxide (CO2), intermittently (every 60 minutes) from two selected

locations in the pellet bulk (no. 8 and 14) and continuously from the void over the pellets surface (no. 4). These measurements were carried out using a NDIR CO2/CO analyser

and a paramagnetic oxygen analyser. Additionally, analysis of the emissions of organic compounds were conducted in one test with each type of pellets. Air was sampled from the void over the pellet surface (no. 4) for analysis of VOC and low-molecular aldehydes. The sampling devices used for VOC-species were Perkin Elmer sampling tubes packed with Tenax®_{adsorbent material. The range of compounds sampled includes non-polar}

and moderately polar organic compounds, corresponding to the hydrocarbon range C6 to

C20. Examples of such species include aliphatic and aromatic hydrocarbons, phenols,

alcohols, ketones, some amines, larger aldehydes etc. Two Tenax®_{tubes in series were}

used in order to collect the VOCs efficiently. Further, the smoke gases were sampled

Rear side

Front side (with hatch)

P1 P2 _P8 P9 P3 P7 P6 P4 P5

Z

X

(13)

through two lines of tubes in parallel, with sampling flows of 50 ml/min and 100 ml/min, respectively, to get duplicate samples. The sampling was conducted for 5 minutes. GC-MS/FID analyses of the various adsorbed VOC-species were conducted after completion of the experiments.

Reactive low-molecular aldehydes were sampled on Sep-pack cartridges coated with 2,4-dinitrophenylhydrazine (DNPH). Due to their reactivity, these aldehydes have to be stabilized prior to analysis. During sampling, they are thus derivatized by reaction with DNPH in order to form hydrazones. The smoke gases were sampled with a sampling rate of 600 ml/min. The sampling was started simultaneously with the VOC-sampling but continued for 12 minutes. The cartridges were subsequently extracted with acetonitrile at the analytical laboratory. The collected hydrazones were separated by reversed-phase High Performance Liquid Chromatography (HPLC) and analysed by atmospheric pressure chemical ionization-mass spectrometry (APCI-MS).

Figure 5 Locations of temperature measurements and gas sampling in the pellet container (P1-P5 orientation).

1500 1100

1100

550

Thermocouple (0.5 mm type K) Thermocouple + gas sampling

100 50 17. 16. 15. 14. 13. 12. 10. 9. 8. 7. 6. 5. 18. 21. 20. 23. 30. 27. 24. 26. 32. 33. 29. 11. 25. 28. 31. 19. 22. 200 150 4.

P1

P2

P3

P4

P5

200 200 200 150 270

(14)

Figure 6 Locations of temperature measurements and gas sampling in the pellet container (P6-P9 orientation).

2.1.3 Description of tests

The test container was filled with a fresh load of pellets directly before a new test was started. The tests started with a pre-heating period. During this period a hot air flow was heating up the pellets to a predefined temperature through inlet B. After that, the heated air flow was redirected into the enclosure (inlet A) and there was no forced flow through the pellet bed. However, valve C was opened at this point in time during a test to allow a convective flow through the bed. The criteria used for ending the pre-heating period was that thermocouple no. 7 had reached the predefined temperature. Temperatures further down in the pellet bulk could thus increase over the predefined temperature from self-heating reactions.

The experiments continued until significant self- heating and temperature runaway was observed in the bed, or until it was quite clear that the elevated ambient temperature selected for the experiment was below critical. In case of spontaneous ignition and temperature runaway the combustion was quenched using nitrogen gas.

In total, four experiments were conducted and general information on the tests are summarized in Table 2.

1500 1100

1100

550

Thermocouple (0.5 mm type K) Thermocouple + gas sampling

100 50 17. 16. 15. 14. 13. 12. 10. 9. 8. 7. 6. 5. 34. 37. 36. 39. 43. 40. 42. 45. 11. 41. 44. 35. 38. 200 150 4.

P6

P7

P3

P8

P9

200 200 200 150 270

(15)

Tests with elevated ambient air temperatures of 90 ºC and 105 ºC were conducted with both types of pellets. The differences seen in the time for pre-heating the pellet bulk to the predefined temperature was due to restrictions in the amount of energy available from the duct heater. The higher air temperature was only possible to reach by lowering the flow through the heater. Further was manual adjustments of the flow rate through the heater necessary, which contributed to the differences in pre-heating times.

Table 2 Description of experiments.

Test Pellet

type Air temperature in enclosure (ºC) Pre-heating time (h:min) Total time for experiment (h:min) Notes 1 L 90 6:05 49:37 -

2 L 105 7:05 41:18* * Test terminated for

extinguishment. Analysis of VOC and aldehydes.

3 M 90 7:45 48:00 -

4 M 105 8:22 49:01 Analysis of VOC and

aldehydes.

2.2 Results

2.2.1 Pre-heating and temperatures in the enclosure

The tests started with a pre-heating period. For the first hour of the test the hot air flow was directed into the enclosure to heat the interior of the test chamber; the flow was then switched to heat up the pellets to the predefined temperature. After that, the heated air flow was redirected into the enclosure for the continuation of the test. Notes from the tests with the time for different events in the tests are given in Appendix 1.

The temperature in the enclosure can be seen in Figure 7 a-d for the four tests conducted. The redirection of the heated airflow from the enclosure to the pellet container and back directly into the enclosure can be clearly seen around the first hour in respective test. Note that it was possible to keep a relatively steady air temperature in the test enclosure during the tests. The temperatures aimed for were 90 °C in Test 1 and Test 3 and 105 °C in Test 2 and Test 4.

The pre-heating period with a heated air-flow through the pellet bulk was judged as necessary to limit the time extension of the tests. However, a possible negative consequence of the pre-heating was that oxidation of the pellets took place during this period and that the ‘reactivity’ of pellets was reduced somewhat for the continuation of the test.

(16)

(a) (b)

(c) (c)

Figure 7 Temperatures in the enclosure measured with two separate thermocouple trees (TCs 46-48 and TCs 49-51, see Figure 2). (a) Test 1, (b) Test 2, (c) Test 3, and (d) Test 4.

2.2.2 Bulk temperature

The results of the temperature measurements with the TCs placed in the pellet bulk is summarised in Table 3. The table shows the maximum reading and identification of the TC that measured the highest temperature for each experiments. The table further shows the average temperature of the five TCs in each experiment that measured the highest temperatures (top 5 TCs). Time resolved data for the complete experiments is shown in Figure 8 - Figure 11, and data for all individual thermocouples are given in Appendix 2. One can conclude that all experiments showed some degree of self-heating as the maximum temperature in all cases was significantly higher than the constant ambient air temperature maintained by the heated air flow through the enclosure. Further, the temperature increase was not localised to single points but was distributed to a larger volume as can be seen from the relative small differences seen between the maximum temperatures measured and the average temperature from the five TCs that measured the highest temperatures.

It is clear from the results shown in Table 3 that there is a significant difference in reactivity, i.e. self-heating rate, between pellet type L (Agro) and pellet type M (Verdoe). For both levels of elevated air temperature, 90 ºC and 105 ºC, pellet type L showed the highest temperatures and Test 2 resulted in temperature runaway and spontaneous-ignition. This was the only experiment which had to be extinguished with N2-injection.

(17)

Table 3 Maximum bulk temperatures measured.

Test Pellet

type Air temperature in enclosure (ºC) Maximum temperature (ºC) TC no. Average of the top 5TCs (ºC) TC no. 1 L 90 121 10 119 9, 10, 11, 12, 41 2 L 105 508 9 416 7, 8, 9, 10, 40 3 M 90 104 14 103 12, 13, 14, 15, 42 4 M 105 131 11 130 9, 10, 11, 12, 13 Table 4 shows the temperature differences between the measured temperatures at selected times in the experiments and the elevated ambient temperature maintained during the experiment. The time scale used in the table is the time in the experiment starting from the time when the pre-heating period ended. The temperatures data used is the average of the five top TCs.

The temperature differences seen for pellet type L show a clearly increased temperature in the 90 ºC test but the temperature difference does not grow significantly with time after 30 h. In the 105 ºC test, however, the temperature difference increases considerably already after 20 h and increases almost exponentially between 30 h and 35 h. The thermal runaway that resulted can be seen in Figure 8 and Figure 10.

Pellet type M showed considerably lower activity. In the 90 ºC test there is a measurable positive temperature difference but the difference is actually decreasing between 30 h and 35 h. It is unlikely that this experiment would have reached spontaneous ignition however long time that it would have been allowed to continue. The 105 ºC test with pellet M showed higher activity and there was no obvious decreasing trend in the end of this experiment. However, the temperature increase was relatively low and the increase rate was not growing when this experiment was ended, as can be seen in Figure 11.

Table 4 Temperature increases measured during the tests (average of the top 5 TCs).

Test Pellet

type Air temperature in enclosure (ºC) Pre-heating + 5h + 10h + 20h + 30h + 35h 1 L 90 16.1 18.9 23.4 26.5 27.8 2 L 105 20.8 24.7 34.4 55.4 304.6 3 M 90 4.2 7.2 11.6 12.6 12.3 4 M 105 5.8 10.2 17.1 21.4 23.4

(18)

Figure 8 The graph shows the thermocouple (TC) that gave the maximum temperature in each test.

Figure 9 Close up of Figure 8 showing more detailed temperatures.

0 100 200 300 400 500 600 0 10 20 30 40 50 Tem per at ur e ( °C) Time (h) Test 1 - TC 10 Test 2 - TC 9 Test 3 - TC 14 Test 4 - TC 11 0 20 40 60 80 100 120 140 0 10 20 30 40 50 Tem per at ur e ( °C) Time (h) Test 1 - TC 10 Test 2 - TC 9 Test 3 - TC 14 Test 4 - TC 11

(19)

Figure 10 The graph shows the average temperature of the five maximum temperatures in each test.

Figure 11 Close up of Figure 10 showing more detailed temperatures.

2.2.3 Gas emissions

2.2.3.1 On-line analysis of CO

2

, CO and O

2

The on-line measured gas concentrations at the top of the pellet container (Location 4 – at TC_4) and within the pellet bulk (Locations 8 – at TC_8 and Location 14 – at TC_14) are shown in the graphs below. Figure 12 shows the results for O2, Figure 13 the results for

CO2 and Figure 14 shows the results for CO.

0 50 100 150 200 250 300 350 400 450 0 10 20 30 40 50 Tem per at ur e ( °C) Time (h) Test 1 Test 2 Test 3 Test 4 0 20 40 60 80 100 120 140 0 10 20 30 40 50 Tem per at ur e ( °C) Time (h) Test 1 Test 2 Test 3 Test 4

(20)

(a) (b)

(c)

Figure 12 Comparison of O2 concentrations. (a) At location 4, (b) at location 8, and (c) at

location 14, during test 1-4.

(a) (b)

(c)

Figure 13 Comparison of CO2 concentrations. (a) At location 4, (b) at location 8, and (c) at

location 14, during test 1-4. The CO2 measurement at location 4 was not working

during test 3. 17 17,5 18 18,5 19 19,5 20 20,5 21 21,5 0 10 20 30 40 50 O2 (v ol % ) Time (h) O2- Location 4 O2 - Test 1 O2 - Test 2 O2 - Test 3 O2 - Test 4 0 5 10 15 20 25 0 10 20 30 40 50 O2 (v ol % ) Time (h) O2- Location 8 O2 - Test 1 O2 - Test 2 O2 - Test 3 O2 - Test 4 0 5 10 15 20 25 0 10 20 30 40 50 O2 (v ol % ) Time (h) O2-Location14 O2 - Test 1 O2 - Test 2 O2 - Test 3 O2 - Test 4 0 0,5 1 1,5 2 0 10 20 30 40 50 CO2 (v ol % ) Time (h) CO2- Location 4 CO2 - Test 1 CO2 - Test 2 CO2 - Test 4 -2 0 2 4 6 8 10 12 0 10 20 30 40 50 CO 2 (v ol % ) Time (h) CO2- Location 8 CO2 - Test 1 CO2 - Test 2 CO2 - Test 3 CO2 - Test 4 -2 0 2 4 6 8 10 12 0 10 20 30 40 50 CO 2 (v ol % ) Time (h) CO2- Location 14 CO2 - Test 1 CO2 - Test 2 CO2 - Test 3 CO2 - Test 4

(21)

(a) (b)

(c)

Figure 14 Comparison of CO concentration at location 4, 8 and 14 during test 1-4. The CO measurement at location 4 was not working during test 3.

It is clear from the on-line data that the greatest gas production within the pellet bulk is taking place in the first part of the experiments. It is further obvious that the gas

production from pellet type L (Tests 1 and 2) is significantly higher than that from pellet type M (Tests 3 and 4). Test 2, that went to spontaneous ignition, is different from the other experiments as the gas production increases rapidly in the end of the experiment. The oxygen concentration in location 2 reaches zero and both CO2 and CO concentrations

are above the maximum span of the analyser. The high gas production in Test 2 is obviously a result of the thermal runaway and the resulting pyrolysis of the pellet material.

2.2.3.2 Analysis of organic species

The results of the analysis of emitted individual organic species are given below. The sampling for these compounds was conducted during the pre-heating time in Test 2 with pellet type L and in Test 4 with pellet type M. Although these experiments had different pre-defined elevated air temperatures (105 ºC respective 90 ºC), the sampling was made during comparable conditions as the sampling in both cases started when a temperature of 44 ºC was reached at TC no 6. This was at 5 h 3 min in Test 2 and at 5 h 19 min in Test 4. The results from the VOC analysis are given in Table 5 and Table 6. The VOCs found in highest amounts have been quantified and the concentrations are given in Table 5. All species that could be identified are listed in Table 6. It can be seen from Table 5 that aldehydes were emitted from the pellet bulk in the highest concentration. Pentanal and Hexanal are the dominant species. Further were terpenes found with α-pinene emitted in the highest concentration. Other groups of VOCs found included alcohols, ketones and carboxylic acids. 0 0,1 0,2 0,3 0,4 0,5 0 10 20 30 40 50 CO (v ol % ) Time (h) CO - Location 4 CO - Test 1 CO - Test 2 CO - Test 4 -0,5 0 0,5 1 1,5 2 2,5 3 3,5 0 10 20 30 40 50 CO (v ol % ) Time (h) CO - Location 8 CO - Test 1 CO - Test 2 CO - Test 3 CO - Test 4 -0,5 0 0,5 1 1,5 2 2,5 3 3,5 0 10 20 30 40 50 CO (v ol % ) Time (h) CO - Location 14 CO - Test 1 CO - Test 2 CO - Test 3 CO - Test 4

(22)

Table 5 Quantification of main VOCs. Test Pellets Test 2 Type L Test 4 Type M Sampling time: TC no. 6 temperature: 05:02:50 +5 min 44.5 °C 05:19:00 +5 min 44.6 °C VOC species (mg/m3_): Total VOCs 180 62 Pentanal 22.4 3.4 Hexanal 57.1 14.3 Nonanal 5.7 3.0 Pentanol 9.8 1.1 α-pinene 3.4 3.9

Table 6 Identified VOCs.

Type of compound Individual species identified in both tests

(listed in order of occurrence)

Aldehydes Hexanal Pentanal Heptanal Octanal Nonanal Butanal

Unsaturated aldehydes 2-methyl-2-propenal

Pentenal Hexenal 2-heptenal 2-oktenal 2-nonenal 2-decenal 2,4-decadienal Alcohols Pentanol Heptanol Ketones Methylvinylketone 2-heptanone

Carboxylic acids Hexanoic acid

Pentanoic acid Octanoic acid Heptanoic acid Butanoic acid Etanoic acid

Terpenes and sesqui-terpenes α-terpene

3-carene β-fellandrene Terpineol Longifolene α- and γ-muorlen

Sampling and analysis specifically for low-molecular reactive aldehydes was further made in Test 2 and Test 3. Unfortunately, the sampled amounts of aldehydes were to high

(23)

which resulted in total consumption of the reagent in the Sep-pack sampling cartridges used. This results in an underestimation of the actual concentrations of these aldehydes (which is clear if the results in Table 7 are compared to the results from the VOC-analysis in Table 6). The results presented in Table 7 can thus only be interpreted as

semi-quantitative. What can be concluded from these results is that the strong irritants

formaldehyde and acrolein was found in Test 2 (pellet L) and that only formaldehyde was found in Test 4 (pellet M).

Table 7 Results from specific aldehyde sampling and analysis using Sep-pack cartridges.

NOTE: the 2,4-dinitrophenylhydrazine (DNPH) reagent was consumed in both tests, which implicates that all results below are likely to be underestimated.

Test Pellets Test 2 Type L Test 4 Type M Sampling time: Pellet surface temp:

05:02:50 +12 min 44.5 °C 05:19:00 +12 min 44.6 °C Aldehydes (mg/m3_): Dimetylbenzaldehyde N.D. N.D. Toluolaldehyde N.D. N.D. Benzaldehyde 1.6 0.35 Hexanal 10.0 2.7 Pentanal 1.3 0.31 Butanal 0.54 0.14 Crotonalaldehyde N.D. N.D. Acetone 0.29 0.15 Propanal 0.33 0.16 Acrolein 1.0 N.D. Acetaldehyde 1.0 0.54 Formaldehyde 1.2 0.72 N.D. = Not detected.

(24)

3 Calculations of spontaneous ignition of

wood pellets

3.1 Theory and calculation tool

A calculation tool, in Microsoft Excel, for calculations with Frank-Kamenetskii theory has been developed as a part of the project. In this section there is a brief presentation of the theory used and a summary of which limitations the Frank-Kamenetskii-theory has when it comes to predicting spontaneous ignition in storages of wood pellets.

The assumptions underlying Frank-Kamenetskii´s original theory are the following [2]: 1. It is assumed that only one heat generating reaction is present and that the speed

of this reaction complies with the Arrhenius equation. This means that the rate of reaction is a function of temperature in such a way that it is proportional to e-E/RT_.

This means that the reaction rate at a given temperature is constant over time. 2. The activation energy (E) is assumed to be high.

3. Heat transfer through the body is assumed to be from conduction only.

4. Heat transfer, by radiation and convection, from the body to its surroundings is assumed to be high enough for the surface of the body to have the same temperature as the surroundings (high Biot number).

5. The material is assumed to be isotropic and homogenous and to have physical properties that do not depend on temperature.

Deviations from these assumptions can be corrected for to different degrees which are described by Beever [2].

The main concept for calculations of critical conditions is to calculate the value of a dimensionless parameter called the Frank-Kamenetskii parameter δ from:

δ₌ρ𝑄𝐴_λ ⋅ 𝐸𝑟2

𝑅𝑇₀2⋅ 𝑒−𝐸 𝑅𝑇⁄ 0 ,where

δ = the Frank-Kamenetskii parameter (-) ρ = bulk density (kg m-3₎

Q = heat of reaction (J kg-1₎

A = the pre-exponential factor in the Arrhenius expression for heat production in a body (s-1₎

λ = thermal conductivity (W m-1 _K-1₎

E = activation energy (J mol-1₎

r = characteristic length for the storage configuration (m) R = the universal gas constant (R=8. 314 J mol-1_K-1₎

(25)

The calculated value is to be compared to tabulated critical values, δc, which depend on

the geometry of the body. Spontaneous ignition can occur if δ>δc.

Provided that spontaneous ignition can occur, that is δ>δc, it is of great value to be able to

determine how long time it will take before it comes about. Several models for

calculation of time to ignition, often abbreviated TTI, exists. Two of these are ”The Zinn and Mader-model” and “The Buddigton-model”[3]. Both these models are available in the calculation tool developed in Microsoft Excel.

There are a fairly large number of factors that restrict the Frank-Kamenetskii-theory´s suitability for use on storages of wood pellets. One of them is that the effect of moisture is not included in the calculations. Another is that there may be more than one reaction contributing to heat production, if so the assumption of the theory saying that only one heat generating reaction is present does not match reality. As long as the pellets are handled and stored properly it is, however, unlikely that biological activity contribute to self-heating. Yet another factor that limits the suitability of the theory for use on storages of wood pellets is that these stockpiles, primarily due to zones with fine particles and the gaps formed between individual pellets, does not constitute a homogeneous material and this cause deviation from one of the basic assumptions of the theory. A further limitation of the model is that the thermal conductivity is fixed and is not dependent on the

temperature of the bulk material.

The developed calculation tool for estimation of self-ignition with Frank-Kamenetskii-theory is easy to use. The results from calculations with the tool should be interpreted with the Frank-Kamenetskii theory’s assumptions in mind. A preliminary validation of the calculation tool showed that results from the tool can be useful to make a first assessment of the risk of spontaneous ignition [3].

The calculation tool provides two models for calculation of time to ignition. During the process of validation, it turned out that one of them overestimated the time while the other understated the time. A rough estimate of time to ignition seems to be feasible by looking at the results from both models. The sensitivity analysis for the parameters thermal conductivity (λ), heat of reaction multiplied by the pre-exponential factor in the Arrhenius expression for heat production (Q×A), the activation energy (E) and density (ρ) carried out with the calculation tool shows that, within the studied range of values,

variations in the value of the activation energy affects the result most. Q×A also has strong effect on the results but only when it comes to low values of the parameter. Varying the values of λ and ρ, on the other hand, changes the results in a comparatively small extent. The tool could be improved by addition of more corrections.

3.2 Comparison of kinetic parameters

Using the crossing point method (CP) and the micro calorimeter (µ-cal), values for activation energy (E) and heat of reaction times the frequency factor (Q×A) can be determined from the experimental data. These data are given in Table 8. Also see [1] for more details. Data of the bulk density (ρ) and specific heat capacity (Cp) has been

(26)

Table 8 Bulk properties and kinetic parameters from experimental data.

Wood pellet ρ, bulk

(kg/m3₎ Cp, bulk _(J/kg.K) E (kJ/mol) Q×A

Data Crossing Point (CP)

M 719 1650 74 3.9 × 109

L 715 1600 82 5.1 × 1010

Data Micro calorimeter (µ-cal)

M 719 1650 76 4.8× 1010

L 715 1600 52 8.6 × 107

We know from the crossing point measurements, the micro calorimeter and the results from the medium scale tests that L is the most active pellet type. However, if just

comparing the kinetic parameters E and Q×A individually, it is difficult to conceive this. A better method is to compare the heat production rate for each pellet type, where all kinetic parameters are included. As mentioned in Section 3.1, it is assumed in Frank-Kamenetskii theory that the heat production rate follows an Arrhenius function:

𝑞̇′′′_{= 𝜌𝑄𝐴𝑒}−𝐸�_𝑅𝑇

If using the equation above and the kinetic parameters for each pellet type, the heat production rate can be plotted against temperature, see Figure 15. Both the crossing point test and the micro calorimeter test results show that pellet type L is the more active pellet compared to pellet type M and is thereby more likely to experience self-heating. As mentioned above, this was also shown with the medium scale tests where pellet type L caused self-ignition in one case. However, if looking at the low level of heat production which is given in Figure 16, it can be seen that Pellet M actually shows a higher heat production up to about 100 °C from the crossing-point (CP) tests compared to Pellet L.

(27)

Figure 15 The heat production rates for pellet M and L.

Figure 16 Close-up of the low region of heat production from Figure 15.

Figure 15 further clearly shows that there is a difference in the heat production rate from the crossing point method and from the micro calorimeter for the same type of pellet. Tests conducted with the micro calorimeter gives a more rapid heat production rate compared to the crossing point method.

This indicates that the reaction rate differs in different temperature regions. The micro calorimeter tests are conducted in the temperature range 40-80 °C and the crossing-point tests are conducted in the much higher temperature range, 180-200 °C.

3.3 FK-calculations on the medium scale tests

The calculation tool for calculations with Frank-Kamenetskii-theory was also used for calculations on the medium scale tests. The kinetic parameters from the crossing point

0 100 200 300 400 500 600 700 800 900 1000 0 50 100 150 200 Hea t p ro du ct io n ra te J /m 3s Temperature (°C) Pellet M - CP Pellet L - CP Pellet M - µ-cal Pellet L - µ-cal 0 10 20 30 40 50 60 70 80 90 100 0 20 40 60 80 100 120 Hea t p ro du ct io n ra te J /m 3s Temperature (°C) Pellet M - CP Pellet L - CP Pellet M - µ-cal Pellet L - µ-cal

(28)

and micro calorimeter tests was used as indata for calculations of the critical ambient temperature and time to self-ignition of the medium scale tests. These results where then compared to the actual results in medium scale. The input data for these calculations are shown in Appendix 3, Fel! Hittar inte referenskälla. and the results are shown in Table 9.

Table 9 Calculations of critical ambient temperature and time to self-ignition in the medium scale tests. Kinetic parameters from micro-calorimeter (µ-cal) tests and crossing-point (CP) tests are compared.

Ambient

temperature Explanation Unit Pellet M (µ-cal) Pellet M (CP) Pellet L (µ-cal) Pellet L (CP)

Calculations of critical ambient temperature:

- Calculation with no

correction K (°C) (48) 321 (74) 347 (10) 283 (75) 348

- Calculation with

correction* K (°C) (42) 315 (66) 339 276 (3) (68) 341

Calculations of time to self-ignition:

90 Zinn- and Mader

model, with correction*

h 227 395 141 398

90 Buddington model,

with correction* h 3 58 <1 53

105 Zinn- and Mader

model, with correction*

h 161 303 95 299

105 Buddington model,

with correction* h <1 14 <1 11

* Correction for surface temperature ≠ ambient temperature.

Critical ambient temperature

Calculations based on crossing point data shows little difference in critical ambient temperature between pellet type M and type L; 66°C and 68 °C respectively. However, as seen in Figure 15, the heat production rate is quite similar for pellet type M and type L for temperatures below 80 °C.

Calculations based on micro calorimeter data shows very low critical ambient

temperatures for pellet type M and type L; 42 °C and 3 °C respectively. Pellet type L is clearly more reactive than pellet type M.

In the medium-scale tests the ambient temperature was set to either 90 °C or 105 °C. If the calculated values above were entirely correct predictions we would have experienced self-heating in all 4 medium scale tests, which was not the case. The temperature increase seen for pellet type L show a clearly increased temperature in the 90 ºC test but the temperature difference does not increase significantly with time after 30 h. In the 105 ºC test, however, the temperature difference increases considerably already after 20 h and increases almost exponentially between 30 h and 35 h. The thermal runaway that resulted can be seen in Figure 8 and Figure 10.

Pellet type M showed considerably lower activity. In the 90 ºC test there is a measurable positive temperature difference but the difference is actually decreasing between 30 h and 35 h. It is unlikely that this experiment would have reached spontaneous ignition however long time that it would have been allowed to continue. The 105 ºC test with pellet M

(29)

showed higher activity and there was no obvious decreasing trend in the end of this experiment. However, the temperature increase was relatively low and the increase rate was not growing when this experiment was ended as can be seen in Figure 11.

It seems thus that the F-K model is conservative in the predictions of critical ambient temperatures for the medium-scale reference tests. A reason for this might be that the water content of the pellet is disregarded in the model.

The prediction is more accurate if the kinetic parameters from the crossing-point tests are used as input data compared to if using the kinetic data from the micro calorimeter tests. However, the data from the micro calorimeter separate the degree of ‘reactivity’ seen from the two types of pellets better. It might be that different oxidation reactions takes place in different temperature regions. Easily oxidised compounds might produce heat at a high rate in the lower temperature region and these compounds might then have been depleted when the higher temperature region is reached.

Time to ignition

Calculations in Table 9 shows that the time to ignition is considerably shorter using micro calorimeter data compared to crossing point data. This is obviously due to the low values of critical ambient temperature for micro calorimeter data. In Table 9, two different models have been used to calculate time to ignition; ”The Zinn and Mader-model” and “The Buddigton-model”. The calculations have been corrected for Frank-Kamenetskii assumption no 4 (see Section 3.1), i.e. the surface temperature equals the ambient temperature.

The calculations clearly show that the Buddigton-model overestimates and the Zinn and Mader-model underestimates time to ignition. The answer is probably somewhere between. In the actual medium scale test thermal runway was observed for pellet type L after 30-35 h, and only when the ambient temperature was 105 °C. None of the medium scale tests were performed longer than 48 h and it is possible that thermal runway would have appeared eventually also for the test with pellet type M at 105 °C and the test with pellet type L at 90 °C. With pellet type M at 90 °C it is unlikely that this experiment would have reached spontaneous ignition even if allowed to continue for a long period of time.

As shown above, the limitation of the F-K model put restrains on the accuracy of

prediction for wood pellets. However, useful indications of critical ambient temperatures for wood pellet bulk storages seems to be possible.

3.4 Examples on FK-calculations for large scale

storage

The calculation tool was also used on four different types of representative types of full scale storages that either exist today or might be possible in the future, see Table 10. The kinetic parameters from the crossing point and micro calorimeter tests was used as input data for calculations of the critical ambient temperature and time to self-ignition. The input data for these calculations are shown in Appendix 3, Table 13 and the results are shown in Table 11.

(30)

Table 10 Input data for four representative large-scale storages. For definitions on characteristic length, see [2].

Storage 1:

Tower silo Storage 2: Normal storage silo

Storage 3: Big

storage silo Storage 4: Ground storage

Shape Cylinder Cylinder Cylinder Rectangular box

Dimensions Diameter: 8 m

Height: 40 m Diameter: 25 m Height: 25 m Diameter: 60 m Height: 40 m Length: 50 m Width: 30 m Height: 10 m

Wall material Concrete Sheet metal Concrete Assumption: no

walls Wall thickness 20 cm 5 mm 7 cm - R, characteristic length, depends on geometry 4 m 12.5 m 30 m 5 m L, length of one of the sides of a rectangular box - - - 25 m M, length of one of the sides of a rectangular box - - - 15 m

L, half the height

of a short cylinder - - 20 m -

Table 11 Calculations* of critical ambient temperature and time to self-ignition for four representative full-scale storages of wood pellets.

Storage

no Explanation Unit Pellet M (µ-cal) Pellet M (cross

point)

Pellet L

(µ-cal) Pellet L (cross point) 1 Critical ambient temperature K (°C) 276

(3) (21) 294 -** (26) 299 Time to ignition: Zinn- and

Mader model Dygn 708 1837 - 5760

Time to ignition: Buddington

model, with correction Dygn 36 983 - -***

2 Critical ambient temperature K (°C) -** 275

(2) -** 281 (8) Time to ignition: Zinn- and

Mader model Dygn - 6736 - 7960

model Dygn - 217 - 489

3 Critical ambient temperature K (°C) -** -** -** -**

Time to ignition: Zinn- and

Mader model Dygn - - - -

model Dygn - - - -

4 Critical ambient temperature K (°C) -** 282

(9) -** (15) 288 Time to ignition: Zinn- and

Mader model Dygn - 3147 - 4071

model Dygn - 377 - 908

* Correction for surface temperature ≠ ambient temperature.

** Predicted critical temperature < 274 K.

(31)

The results from the Frank-Kamenetskii predictions given in Table 11 again show that the model is conservative if applied to a bulk of wood pellet. When using the reaction

parameters from the crossing-point method, critical temperatures are found for normal ambient temperatures (~20-30 °C) or for low ambient temperatures (~0-20 °C) for storage 1, 2 and 4. The critical temperature for storage 3 is predicted as < 0 °C and the output from the model does not give any more precise information. These temperatures seems to be very low predictions of the critical ambient temperature in practice. The time to ignition, however, predicted by the two models used are in all cases long which seems reasonable.

Using the reaction parameters from the micro calorimeter, in all cases but one, predicted critical ambient temperatures are < 0°C. The predicted ignition times are not presented in these cases.

If comparing the relative order of predicted critical temperatures for the storage cases investigated, the results in order of decreasing criticality are: storage 3 (big silo) > storage 2 (normal silo) > storage 4 (ground storage) > storage 1 (tower silo).

(32)

4 Discussion

The medium-scale tests were effective in separating the self-heating activity of the two types of pellets investigated. The experiments clearly showed that pellet type M had a very moderate activity which resulted in limited temperature increases and less gas emissions. Pellet type L showed very clearly a higher activity which resulted in spontaneous ignition in the test with the highest elevated ambient temperature. The gas production was further significantly higher for this type of pellet both during the pre-heating phase and during the subsequent self-pre-heating phase. The concentrations of

emittants measured during the thermal runaway and spontaneous ignition of pellet L were very high compared to the self-heating phase.

A possible disadvantage of the test procedure used here is the pre-heating period where the pellet bulk is flushed with heated air to reach the pre-determined “test temperature”. This procedure could potentially consume a significant amount of reactive material in the pellet which could lead to less reactivity in the subsequent self-heating phase. If this is a problem is unclear but the alternative to omit the pre-heating phase would most probably for most types of pellets lead to an exceedingly long experimental time.

Calculations of the medium-scale tests using the Frank-Kamenetskii theory based on crossing point data showed little difference in critical ambient temperature between pellet type M and type L; 66 °C and 68 °C respectively. However, the heat production rate calculated from the crossing point tests was quite similar for pellet type M and type L for temperatures below 80 °C. The calculations based on micro calorimeter data showed significantly lower critical ambient temperatures; 42 °C for pellet type M and 3 °C for pellet type L.

It was shown that there is a difference in the heat production rate from the crossing point method and from the micro calorimeter for the same type of pellet. Tests conducted with the micro calorimeter gave a more rapid heat production rate compared to the crossing point method. This indicates that the reaction rate differs in different temperature regions. The micro calorimeter tests are conducted in the temperature range 40-80 °C and the crossing-point tests are conducted in the much higher temperature range, 180-200 °C. The results from Frank-Kamenetskii predictions of some typical storage cases showed that the model is conservative if applied to a bulk of wood pellet. When using the reaction parameters from the crossing-point method, critical temperatures are found for low to normal ambient temperatures (0-30 °C) The time to ignition, however, predicted by the two models used are in all cases long which seems reasonable. Using the reaction parameters from the micro calorimeter, in all cases but one, predicted critical ambient temperatures were < 0 °C.

If comparing the relative order of predicted critical temperatures for the storage cases investigated, the results in order of decreasing criticality are: storage 3 (big silo) > storage 2 (normal silo) > storage 4 (ground storage) > storage 1 (tower silo).

(33)

5 References

1. Lönnermark, A., Blomqvist, P., Persson, H., Rahm, M., and Larsson, I., "Use of small-scale methods for assessment of risk for self-heating of biomass pellets", SP Technical Research Institute of Sweden, SP Report 2012:49, Borås, Sweden, 2012.

2. Beever, P. F., "Self-heating and Spontaneous Combustion". In The SFPE Handbook of Fire Protection Engineering National Fire Protection Association, 1995.

3. Jonsson, M., "Beräkningar av självantändning i lager av träpellets", Department of Fire Safety Engineering and Systems Safety, Lund University, Sweden, 5339, ISSN: 1402-3504, Lund, 2010.

(34)

Appendix 1 Notes from reference tests

Test 1: Pellet type L – 90 °C ambient air

Time (h:min:s) Event

0:00:00 Start of data acquisition

0:10:00 Start of heated air flow to room (valve A open, valve B closed) 0:30:00 Temperature of air in main channel (at V1): TC_52 = 105 °C 1:10:00 Heated air flow opened to pellets

(valve A partly closed, valve B opened)

Temperature of air to pellets (at V3): TC_54 = 53 °C (TC_52 = 91 °C)

1:20:00 Temperature of air to pellets has stabilized TC_54 = 90 °C (TC_52 = 93 °C)

6:01:00 TC_7 reaches 90 °C

6:05:30 Air flow to pellets closed (valve B closed, valve A fully opened)

6:06:30 Valve C is opened

48:04:00 The heater is turned off (the air flow is still on) 48:05:45 Valve C is closed

48:07:00 Nitrogen injection into the pellet bulk is started 49:22:13 The air flow is turned off

49:27:30 The hatch of the pellet container is opened Notes from Test 1:

• Weight of pellets used for the test was 744.8 kg.

• Temperature in test hall was 22.4 °C at the start of the test. • Air moisture content was 20.4 %.

(35)

Test 2: Pellet type L – 105 °C ambient air

(valve A still open, valve B opened)

1:20:00 Temperature of air to pellets has stabilized TC_54 = 110 °C (TC_52 = 116 °C)

1:22:20 Valve A closed

7:05:30 TC_7 reaches 105 °C

7:05:30 Air flow to pellets closed (valve B closed, valve A opened)

41:18:00 The heater is turned off (the air flow is still on) 41:20:00 The air flow is turned off

41:20:00 Valve C is closed

41:22:00 Nitrogen injection into the pellet bulk is started 41:26:30 The front wall of the test enclosure is removed 41:44:00 The air flow is turned off

45:46:30 The hatch of the pellet container is opened Notes from Test 2:

• Weight of pellets used for the test was 749.6 kg.

• Temperature in test hall was 25.3 °C at the start of the test. • Air moisture content was 37.4 %.

(36)

Test 3: Pellet type M – 90 °C ambient air

(valve A closed, valve B opened)

1:20:00 Temperature of air to pellets has stabilized TC_54 = 90°C (TC_52 = 93 °C)

7:45:00 TC_7 is 88 °C

48:00:00 The test is terminated Notes from Test 3:

• Weight of pellets used for the test was 752 kg.

• No notes of the time for turning of the fan or opening of the test enclosure. But all of this took place after that the test was terminated.

(37)

Test 4: Pellet type M – 105 °C ambient air

(valve A still open, valve B opened)

1:12:00 Valve A is closed

1:20:00 Temperature of air to pellets has stabilized TC_54 = 111°C (TC_52 = 113 °C)

8:22:30 TC_7 reaches 105 °C

48:00:00 The test is terminated Notes from Test 4:

• Weight of pellets used for the test was 752 kg.

• No notes of the time for turning of the fan or opening of the test enclosure. But all of this took place after that the test was terminated.

(38)

Appendix 2 Temperature measurements during

medium scale tests

Test 1 0 20 40 60 80 100 120 140 0 10 20 30 40 50 Tem per at ur e ( °C ) Time (h)

Temperature - vertical centrum line of silo

TC 1 TC 5 TC 6 TC 7 TC 8 TC 9 TC 10 TC 11 TC 12 TC 13 TC 14 TC 15 TC 16 TC 17 0 20 40 60 80 100 120 0 10 20 30 40 50 Tem per at ur e ( °C ) Time (h)

Temperature - horizontal, 15 cm below surface

TC 7 TC 34 TC 43 0 20 40 60 80 100 120 0 10 20 30 40 50 Tem per at ur e ( °C ) Time (min)

TC 7 TC 18 TC 27 0 20 40 60 80 100 120 140 0 10 20 30 40 50 Tem per at ur e ( °C ) Time (h)

TC 11 TC 35 TC 38 TC 41 TC 44 0 20 40 60 80 100 120 140 0 10 20 30 40 50 Tem per at ur e ( °C ) Time (h)

TC 11 TC 19 TC 22 TC 25 TC 28 TC 30 TC 31 TC 32 TC 33

(39)

0 20 40 60 80 100 120 140 0 10 20 30 40 50 Tem per at ur e ( °C ) Time (min)

Temperature - horizontal, 75 mm below surface

TC 13 TC 23 TC 26 0 20 40 60 80 100 120 0 10 20 30 40 50 Tem per at ur e ( °C ) Time (h)

TC 15 TC 20 TC 29

(40)

Test 2 0 100 200 300 400 500 600 0 10 20 30 40 50 Tem per at ur e ( °C ) Time (h)

TC 1 TC 5 TC 6 TC 7 TC 8 TC 9 TC 10 TC 11 TC 12 TC 13 TC 14 TC 15 TC 16 TC 17 0 50 100 150 200 250 300 350 400 450 500 0 10 20 30 40 50 Tem per at ur e ( °C ) Time (h)

TC 7 TC 34 TC 43 0 50 100 150 200 250 300 350 400 450 500 0 10 20 30 40 50 Tem per at ur e ( °C ) Time (min)

TC 9 TC 21 TC 24 -300 -200 -100 0 100 200 300 400 500 0 10 20 30 40 50 Tem per at ur e ( °C ) Time (h)

TC 11 TC 35 TC 38 TC 41 TC 44 -200 -100 0 100 200 300 400 500 0 10 20 30 40 50 Tem per at ur e ( °C ) Time (h)

(41)

0 50 100 150 200 250 0 10 20 30 40 50 Tem per at ur e ( °C ) Time (min)

TC 13 TC 39 TC 42 -250 -200 -150 -100 -50 0 50 100 150 200 250 300 0 10 20 30 40 50 Tem per at ur e ( °C ) Time (h)

TC 13 TC 23 TC 26 0 20 40 60 80 100 120 140 160 180 200 0 10 20 30 40 50 Tem per at ur e ( °C ) Time (h)

TC 15 TC 36 TC 45 -250 -200 -150 -100 -50 0 50 100 150 200 0 10 20 30 40 50 Tem per at ur e ( °C ) Time (h)

TC 15 TC 20 TC 29

(42)

Test 3 0 20 40 60 80 100 120 0 10 20 30 40 50 Tem per at ur e ( °C ) Time (h)

TC 1 TC 5 TC 6 TC 7 TC 8 TC 9 TC 10 TC 11 TC 12 TC 13 TC 14 TC 15 TC 16 TC 17 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 Tem per at ur e ( °C ) Time (h)

TC 7 TC 34 TC 43 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 Tem per at ur e ( °C ) Time (min)

TC 9 TC 37 TC 40 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 Tem per at ur e ( °C ) Time (h)

TC 11 TC 35 TC 38 TC 41 TC 44 0 20 40 60 80 100 120 0 10 20 30 40 50 Tem per at ur e ( °C ) Time (h)

(43)

0 20 40 60 80 100 120 0 10 20 30 40 50 Tem per at ur e ( °C ) Time (min)

TC 15 TC 20 TC 29

(44)

Test 4 0 20 40 60 80 100 120 140 0 10 20 30 40 50 Tem per at ur e ( °C ) Time (h)

TC 1 TC 5 TC 6 TC 7 TC 8 TC 9 TC 10 TC 11 TC 12 TC 13 TC 14 TC 15 TC 16 TC 17 0 20 40 60 80 100 120 140 0 10 20 30 40 50 Tem per at ur e ( °C ) Time (h)

TC 7 TC 34 TC 43 0 20 40 60 80 100 120 140 0 10 20 30 40 50 Tem per at ur e ( °C ) Time (min)

Temperature - Horizontal, 35 cm below surface

TC 11 TC 35 TC 38 TC 41 TC 44 0 20 40 60 80 100 120 140 0 10 20 30 40 50 Tem per at ur e ( °C ) Time (h)

(45)

0 20 40 60 80 100 120 140 0 10 20 30 40 50 Tem per at ur e ( °C ) Time (min)

TC 15 TC 20 TC 29

(46)

Appendix 3 Input data for FK-calculations

Table 12 Input data for calculations of critical ambient temperature and time to self-ignition in medium scale.

Parameter Explanation Unit Pellet M

(µ-cal) Pellet M (CP) Pellet L (µ-cal) Pellet L (CP)

ρ Bulk density Kg m-3 ₇₁₉ ₇₁₉ ₇₁₅ ₇₁₅ Q × A Activation energy times pre-exponential factor in the Arrhenius expression for heat production in a body J kg-1_s-1 _4.8×1010 _3.9×109 _8.6×107 _5.1×1010 λ Thermal conductivity (bulk) W m-1 _K-1 _0.17 _0.17 _0.16 _0.16

E Activation energy kJ mol-1 ₇₆ ₇₄ ₅₂ ₈₂

R The universal gas

constant J mol -1_K-1 _8.314 _8.314 _8.314 _8.314 r Characteristic length, depends on geometry m 0.55 0.55 0.55 0.55 C Specific heat capacity (bulk) J Kg -1_K-1 ₁₆₅₀ ₁₆₅₀ ₁₆₀₀ ₁₆₀₀ h* Effective heat transfer coefficient W m-2_K-1 ₁ ₁ ₁ ₁ T0 Ambient temperature K -** -** -** -**

* In a previous report [3] the effective heat transfer coefficient was calculated to 0.9262 and 0.9278 W m-2

K-1_{. However, in these calculation the effective heat transfer coefficient has been estimated to h=1W m}-2

K-1_{. A minor change of the effective heat transfer coefficient influence the results negligibly.}

** Only applicable when calculating time to ignition. 90 or 105 °C, depending on the calculations. See results in Table 9.

Medium-scale reference tests and calculations of spontaneous ignition in wood pellets - the LUBA project

LUBA

project

Ida Larsson, Per Blomqvist, Anders Lönnermark

and Henry Persson

S

P

T

ec

hni

ca

l R

es

ear

ch I

ns

tit

ut

e of

S

w

ed

en

Medium-scale tests and calculations of

spontaneous ignition in wood pellets -

the LUBA project

Ida Larsson, Per Blomqvist, Anders Lönnermark and

Henry Persson

Abstract

Medium-scale tests and calculations of spontaneous

ignition in wood pellets - the LUBA project

Contents

Abstract

3

Contents

4

Acknowledgements

5

Sammanfattning (in Swedish)

6

1

Introduction

7

2

Medium scale tests

8

3

Calculations of spontaneous ignition of wood pellets

24

4

Discussion

32

5

References

33

Appendix 1 - Notes from reference tests

34

Appendix 2 - Temperature measurements during medium scale tests 38

Acknowledgements

Sammanfattning (in Swedish)

1

Introduction

2

Medium scale tests

2.1

Experimental design

2.1.1

Pellets

2.1.2

Test configuration

Rear side

Front side (with hatch)

Z

X

P1

P2

P3

P4

P5

2.1.3