Modelling of a vented
corn starch dust
explosion using an open
source code
-
Global Dust Saftey Conference 1-3, March 2021
Chen Huang and Andrei Lipatnikov
Outline
• Background • Method
• Experimental setup • Numerical setup
• Results and discussions • Conclusions
Background
3
• Dust explosion threats the industries which deal with combustible powders, e.g. pellets, food, metal and so on;
• National and global statistics, e.g. Swedish working environment authority, and Combustible Dust Incident Report by DustSafetyScience;
• “Dark figure” - unreported incidents;
• Once per week instead of once per month (Nessvi and Persson 2019); • One seventh incidents were reported in Germany from 1965 – 1985
(Eckhoff 2003, Yuan 2015);
Project status
• A model of dust explosion has been implemented into OpenFOAM • Implementation has been verified against analytical solutions
• Model and implementation have been validated against Leeds fan-stirred explosion vessel experiments for corn starch. (Huang et al. 2020)
• To apply the model and code to simulate large-scale industrial dust explosions
4
Method
• Dust explosion resembles a gas explosion for fine dust particles and high volatile content (Bradley et al. 1988, 1989).
• Flame Speed Closure model focusing on flame propagation in a turbulent premixed flame
• FSC model was quantitatively tested for laboratory gaseous turbulent premixed flames from different groups with different conditions
(Lipatnikov 2002).
Method
Combustion progress variable c
Image by Fox & Weinberg, Proc. R. Soc. London A268:222-239, 1962.
δ
L
<<δ
Unburned
t
c=0
Burned
c=1
Flame brush
0<c<1
Method
Flame Speed Closure (FSC) Model
7
𝜕 ҧ𝜌 ǁ𝑐
𝜕𝑡 + ∇. 𝜌ҧ𝐮 ǁ𝑐 = ∇. [ ҧ𝜌ሺ𝜅 + 𝐷𝑡)∇ ǁ𝑐] + 𝜌𝑢𝑈𝑡 ∇ ǁ𝑐 + 𝑄 + ҧ𝜌𝑊𝑖𝑔𝑛 transient
convection
flame structure (thickness)
burning velocity ignition 𝐷𝑡 = 𝐷𝑡,∞ 1 − exp −𝑡𝑓𝑑 𝜏𝐿 𝑈𝑡 = 𝑈𝑡,∞ 1 − 𝜏𝐿 𝑡𝑓𝑑 + 𝜏𝐿 𝑡𝑓𝑑exp − 𝑡𝑓𝑑 𝜏𝐿 Τ 1 2 𝑈𝑡,∞ = 𝐴𝑢′𝐷𝑎1 4Τ = 𝐴𝑢′3 4Τ 𝐿1 4Τ 𝑆𝐿1 4Τ 𝛿𝐿− Τ1 4 𝐷𝑡,∞ = 𝐶𝜇 𝑃𝑟𝑡 ෨𝑘2 ǁ𝜀 = 𝐶𝜇 𝑃𝑟𝑡 ෨𝑘1 2Τ 𝐿 𝐶𝑑
8
Thermophysical properties of corn starch dust
• Corn flour and corn starch
• Chemical formula,
C6H7.88O4.98• The standard heat of formation
• Specific heat capacity
9
Laminar burning velocity of corn starch dust
• Leeds data from Bradley
• Leeds data from Sattar
• Dahoe
10
Laminar burning velocity of corn starch dust
• Leeds data from Bradley
• Leeds data from Sattar
• Dahoe
11
Laminar burning velocity of corn starch dust
• Leeds data from Bradley
• Leeds data from Sattar
• Dahoe
• Gexcon
12
Laminar burning velocity of corn starch dust
• Leeds data from Bradley
• Leeds data from Sattar
• Dahoe
13
Laminar burning velocity of corn starch dust
• Leeds data from Bradley
• Leeds data from Sattar
• Dahoe
14
Laminar burning velocity of corn starch dust
Dust properties play role (Huang et al. 2019)
15
(a) Correlation between Pmax and D50 (b) Correlation between KStand D50. Fig. 6. Correlation between dust explosion characteristics and D50for 49 wood dust samples from literature [9-17] and this work. Symbol size corresponds to D50of dust samples, whereas symbol
color corresponds to the moisture content. White bubbles indicate missing information about moisture content. Current results are shown as diamond symbols.
Huang C., D. G. J., Nessvi K., Lönnermark A., Persson H., the 9th International Seminar on Fire and Explosion Hazards, St. Petersburg, 21-26, April; St. Petersburg, 2019; pp 366-375.
16
17
Sensitivity study of other parameters
18
Sensitivity coefficients on different input parameters on the computed flame arrival time at the vent opening based on ҧ𝑐=0.1 and
ҧ𝑐=0.5.𝑢′=0.5 m/s, 𝐿=0.1 m, 𝐶𝑜=0.1, a whole domain with 2.5 million
Comparison with experimental data
20 Time [ms] 2 4 6 8 Mean temper ature 280 1800 [K]
Conclusions
• Model for vented corn starch explosion was setup with the first
result of calculated explosion overpressure being in good agreement with measure one before the rupture of the vent panel.
Future work
• continue the simulation of vented explosion for Rembe explosion vessel with a focus on the period after the rupture of the vent panel. • One more industrial case will be simulated provided with the
availability of experimental data.
23
Acknowledgements
• Thank AFA-Försäkring for financial support of this project (grant number 180028).
• Thank Marius Bloching at Rembe Research and Technology Center for providing
experimental data.
• The computations were enabled by resources provided by the Swedish National
Infrastructure for Computing (SNIC) at HPC2N partially funded by the Swedish
24
