Studies of flashback in the TARS burner FM seminar 2016.02.17

Studies of flashback in the TARS burner
FM seminar 2016.02.17
SZASZ R.Z., SUBASH A.A., LANTZ A., COLLIN, R., FUCHS L. , GUTMARK E.
Background
Wish list
Action
Consequence
Flame
Ignite it!
Pollutants
Low pollutant
emissions
Lean
(Partially)
Premixed
Lean Blow
Out,
Flashback
Stability
Bluff body,
swirl,
active/passive
control, ...
Cost,
performance,
may not work
Fuel flexibility
Flashback
• When
–S>U
• Flashback ’types’ [Sommerer 2004]
– Core flow
– Boundary layer
S
U
– Combustion Induced
Vortex Breakdown (CIVB)
– Combustion Instabilities
– Autoignition
• Can be a combination
Flasback in a wall boundary layer
• Often quantified in terms of
the critical velocity gradient
y
– g=du/dy
• Critical velocity gradient
– mostly estimated based on
Re, Ub
Quenching
distance
u
– recently measured using
micro-PIV
– Difficulties with topology
changes
Flashback due to CIVB
• Combustion Induced Vortex
Breakdown
• Vortex Breakdown?
• Combustion induced?
𝐷𝜔
1
= 𝜔 ⋅ 𝛻 𝑈 − 𝜔 𝛻 ⋅ 𝑈 + 2 (𝛻𝜌 × 𝛻𝑝)
𝐷𝑡
𝜌
Stretching
Tilting
Dilatation
Baroclinic
Dominating
[Kiesewetter et al. 2007]
• Adverse pressure gradient
helps VB
• Flame generates adverse
pressure gradient
Goals
• Flashback
• Flash-forward
• When?
• How?
• Why?
• Sensitivity: Re, f, geometry, T,
...
Experimental rig
• TARS 45/45/0 (low swirl)
• CH4/NG – air
Experimental set-up
• Flame: HS Chemiluminescence
• Flow: HS PIV
• Not simultaneously
• 1000 frames pre-triggered, 1200-2000 collected, 1 kHz
Method
• Keep constant Re
• Change phi by adjusting fuel
flow
• Base case – intermediate
points as well
• Sensitivity study – fcrit
• 260 cases
Flashback hysteresys
• FB & FF both on the lean and
rich sides
• Critical phi approaches 1 with
increasing Re
Hysteresys
• Why?
– Changes in flow/flame
topology
– Adverse pressure
gradient
– Rig temperature
Average
and
RMS
flame intensity
Axial
Radial
Before flashback
Velocity field
After flashback
Post-processing
• Integrate intensity in
radial direction
• When?
• How fast?
Detect flame position
• Threshold options
– Constant across all datasets
– % Max for a dataset TI10, TI15, TI20
– Intensity @ plate height, PH
– % Max for a frame, LI10, LI20, LI30, LI40
Verification
Case
#Missed
TI10
12
TI15
2
TI20
14
PH
22
L10
19
L20
3
L30
1
L40
3
More verification
• 4 FB and 4 FF events
• Predictions compared to ’exact’ values
• Few cases, but:
– TI15, L30 best prediction
– LXX less sensitive to threshold limit
Event duration
• FB & FF duration
– In average, FB on the
lean side is significantly
slower
Influence of fuel
• NG includes
higher
hydrocarbons
• Larger SL
Confinement effect
• No significant
effect
Protrusion effect
• No significant
effect
Preheating effect
• Preheating
• Sl increases
• Re decreases
• [Kalantari 2015] used
Da to quantify
flashback propensity
• Here: Le, Ttip
neglected
Conclusions and future work
• There is a hysteresys for flashback/flash forward
• Automatic quantification of FB/FF events
• Threshold based on instantaneous maximum intensity
most robust
• Lean FB slower than FF or rich FB
• Sensitivity study
– Important: Re, phi, Tin
– Less important: Geometry
• Future work
– Why? Why not (How to predict/prevent)?
Thank you!
QUESTIONS?