Developing a Phenomenological Model of Infrared Emissions from Detonation Fireballs
for Explosives Identification
Kevin C. Gross, Glen P. Perram, Ronald F. Tuttle
Air Force Institute of TechnologyRiverside Research Institute
Ninth Biennial HITRAN Conference
Harvard-Smithsonian Center for Astrophysics
Cambridge MA
26 June 2006
3
Introduction• Traditional battle space characterization• Classification of transient, infrared events
• Bomb detonations, muzzle flashes, rocket and missile plumes
• Classifying explosives is difficult• No simple model exists for describing emissions from
detonation fireballs
• High-explosive detonations are non-reproducible
• Inherent irreproducibility (age, mixture tolerances, casing design, impact angle, etc.)
• Environmental interaction (soil type, atmospheric conditions, etc.)
• Cost and safety concerns lead to small-scale tests with limited reproducibility
• Broadband absolute radiometric signatures not apparently useful for classification
• Roughly, variance within explosive class same size as variance between classes
4
Introduction
Framework for solving the explosives classification problem
• Collect data using spectrometers, radiometers, and several banded imagers
• Develop a low-dimensional phenomenological model for fireball emissions
• Spectrometers: Chemistry
• Imagers: Fluid dynamics
• Extract key features (fit model to data)
• Reproducible within the same explosive class (small within-class scatter)
• Distinguishing for different explosive classes (large variance between classes)
• Invariant to uncontrollable factors
• Constrained by physics
• Quantify classification potential of extracted features using pattern-recognition codes
5
Field Tests• Radiant Brass III: Conventional Bomb
• Brilliant Flash II: Enhanced Novel Explosives (ENEs)
• Bronze Scorpio: IEDs
• ABB/Bomem MR Series FTS• RB3: 16 cm
-1 / 21 Hz (InSb: 1800–7100 cm-1, MCT 500-6000 cm-1)
• BF2: 4 cm-1 / 8 Hz (InSb: 1800–7100 cm-1, MCT 500-6000 cm-1)
• BS: 4 cm-1 / 38 Hz (InSb: 1800–7100 cm-1,InGaAs 6000-11000 cm-1)
• Radiometers (4 MWIR bands)
• Banded Imagers (Vis, NIR, MWIR)
RB3 BF2 BS
56 Events 44 Events 58 Events
3 distinct compositions
5 distinct compositions
3 distinct compositions
4 sizes 4 sizes 2 sizes
Half delivered by aircraft
Uncased charges
Cased artillerary shell
RIVERSIDE RESEARCH INSTITUTE Temporal Profile
7
˜ 2100 cm 1
˜ 5500 cm 1
t (s)
I obs˜,t(W/sr-cm
1 )
0 0.5 1 1.5 2 2.5 30
500
1000
1500
2000
1.5 N2 2.5 H2O3.5 CO 3.5 C HD
C7H5N3O6
Detonation (µs)
3.5 CO 3.5 C 5.25 O27 CO2 HAB
Afterburn (s)
1.0
0.0
0.5
1.5 2000
3000
4000
5000
6000
0.0
1.0
2.0
3.0
0.5
I obs˜,t
Wk/srcm
1
˜ cm 1
t s
Fast-scanning FTS collects time-resolved spectra
Temporal profiles reveal detonation and afterburn timescales
Spectral re
solutio
n degraded
2000 2500 3000 3500 4000 4500 5000 5500 6000 6500 7000
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
2000 2500 3000 3500 4000 4500 5000 5500 6000 6500 7000!0.04!0.02
00.020.04
Uncased Explosive, 4 cm!1, 8Hz
I obs"#,t$[nounits]
t ! 0.000 s
Im!I obs"
# #cm!1$
Typical Spectra
8
Typical Spectra
2000 2500 3000 3500 4000 4500 5000 5500 6000 6500 7000
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
2000 2500 3000 3500 4000 4500 5000 5500 6000 6500 7000!0.02
0
0.02
Uncased Explosive, 4 cm!1, 8Hz
I obs"#,t$[nounits]
t ! 0.362 s
Im!I obs"
# #cm!1$ 9
Scene-Change Artifacts
10
At each frequency, assume spectrum’s temporal evolution is quadratic over the scan time of the interferometer
2000 2500 3000 3500 4000 4500 5000 5500 6000 6500 70000
0.2
0.4
0.6
0.8
1
1.2
2000 2500 3000 3500 4000 4500 5000 5500 6000 6500 7000
!505
x 10!3
2000 2500 3000 3500 4000 4500 5000 5500 6000 6500 7000!2
0
2 x 10!4
Re !I obs
"[nounits]
Simulation - Cooling Planckian
IobsI1/!I2/!I3/!
Im!I obs"
I 2!Re !I obs
"
" #cm!1$
!t " 0.1 s, !# " 8 cm$1T ! !T
H" T
L" e"k t # T
L TH! 2000K, T
L! 300K, k ! 1 s"1
Iobs!"# $ I
2%
1
4&2L2'2 !2I
1% 4I
2( 2I
3"
'"2
!i1
2"L# !I
3$ I
1"
#%
I3! Iobs"#, x ! L$I
2! Iobs"#, x ! L/2$I
1! Iobs"#, x ! 0$
Y YzpdYzpdYY
kT/kI
Error
Conventional (Cased) Munition
12
Transmittance
0
0.2
0.4
0.6
0.8
1
H2OCO2O3N2OCOCH4O2
˜ (cm 1)
I obs˜,t(W/sr-cm
1 )
2000 2500 3000 3500 4000 4500 5000 5500 60000
500
1000
1500
2000 t 0.000 st 0.247 st 0.493 s
t 0.740 st 0.986 st 1.233 st 1.479 s
I obs!",t#$arb. %
t 0 st 0.25 st 0.5 s
˜ ( cm 1)
I obs˜,t(W/sr-cm
1 )
2500 2550 2600 2650 2700 2750 2800 2850 29001200
1400
1600
1800
2000
2200
2400
Iobs ! Iobs"#i, t j$/ Iobs"#i % k&#, t j$
Atmospheric Compensation
13
Iobs!", t# $ %!"# Isrc!", t#
! " c/cold
Iobs ! "#
m "r Isrc
!"#$ % e!i !i"#$ ci l
Iobs!", t# $ %i!"#&% j'i!"# Isrc!", t#
Find single set of absorber concentrations for entire data cube
Weighted linear regression to estimate δ
Atmospheric Compensation
log!"m#
log!I obs/"rIsrc#
!2 !1.5 !1 !0.5 0 0.5 1 1.5 2
!2
!1.5
!1
!0.5
0
0.5
1
1.5
2
14
Iobs!", t# $ %!"# Isrc!", t#
! " c/cold
Iobs ! "#
m "r Isrc
log ! Iobs!r Isrc
" " # log$!m%
!"#$ % e!i !i"#$ ci l
Iobs!", t# $ %i!"#&% j'i!"# Isrc!", t#
Iteratively recompute τm with new concentration until δ = 1
Estimate of log(τm)which varies with time
Beer’s Law not strictly appropriate for moderate resolution spectra
Find single set of absorber concentrations for entire data cube
!" !"#$ ILS"# % #&$ d #&#' ( " !"#$' ILS"# % #&$ d #&
Atmospheric Compensation
15
Test Number
H2O
(ppm)
5 10 15
2500
3000
3500
4000
4500
Test Number
CO2(ppm)
5 10 15
360
370
380
Test Number
N2O
(ppm)
5 10 150.295
0.3
0.305
0.31
0.315
0.32
Test Number
CH4(ppm)
5 10 15
1.55
1.6
1.65
371±9 ppm366 ppm
1.59±0.04 ppm1.70 ppm
305±7 ppb310 ppb
Rad
iant
Bra
ss II
I Fie
ld T
est
Radiative Transfer
17
(Over-) Simplified RT for fireball
No sources except fireball
No gradients (uniform T, ρ)
Local thermodynamic equilibrium
No scattering
Fireball parameters: ρ(H2O, CO2, CO,Tg), Tc
Rough approximation to full RT solutionIgnore geometry
Include continuum emitters additively
Iap ! ta !Ac B"Tc# $ Ag "1 % tg#B "Tg#$ $ "1 % ta#B "Ta#tg ! tg !Tg, "H2O# , "CO2# , "CO#$tg ! exp
"####$%L &!
i
Ni 'i(),Tg*+,,,,-
H2O & CO: HITEMP (HITRAN) databaseCO2: CDSD
!""" !#"" $""" $#"" %""" %#"" #""" ##"" &""" &#"" '"""
"
"()
"(!
"($
"(%
"(#
"(&
"('
"(*
"(+
)
!""" !#"" $""" $#"" %""" %#"" #""" ##"" &""" &#"" '"""!"("!
"
"("!
Uncased Explosive, 4 cm!1, 8Hz
I obs"#,t$[nounits]
t ! ".$&! ,
Im!I obs"
# #cm!1$
I!"si# $ I!"si%1# e% ! si
si%1&!"s
'# ds'( " s
i
si%1
&!"s'#B! #T"s'#$ e% ! s'si%1 &!"s''# ds'' ds'
AtmosphereFireball FTS
t! " e#$s
i#s
i#1%&!t! I!"si#1$ % "1 # t!$B !Ti"
2000 2500 3000 3500 4000 4500 5000 5500 6000 6500 70000
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
2000 2500 3000 3500 4000 4500 5000 5500 6000 6500 7000!0.1
!0.05
0
0.05
p = [1.96E-01 2.01E+00 1.31E+03 1.90E+04 7.04E+04 9.16E+02]I obs!t," #$arb. %
SE = 4.20, RMS Err = 14.4%, Median |Rel Err| = 7.2%
" !cm&1"
I obs&I mdl
164DAT40-06
Model (FBM1)
2550 2600 2650 2700 2750 2800 2850
!5
0
5
10
15
20
x 10!3
Modeling Results
18
TNT (H2O/CO2 ~ 0.4)
2000 2500 3000 3500 4000 4500 5000 5500 6000 6500 70000
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
2000 2500 3000 3500 4000 4500 5000 5500 6000 6500 7000
!0.1
0
0.1
p = [3.20E-01 1.73E+00 1.91E+03 3.41E+04 3.24E+04 3.49E+03]I obs!t," #$arb. %
SE = 9.25, RMS Err = 10.0%, Median |Rel Err| = 6.5%
" !cm&1"
I obs&I mdl
164DAT43-02
Model (FBM1)ENE (H2O/CO2 ~ 9.5)
Modeling Results
19
Feature Extraction
20
TNT (L) vs ENE (R)
TK
Conc.[arb.]
H2O
/CO2
PNR(6592cm
1 )
t s t s
1000
1500
2000
2500
TK
0
1
2
3
4
Conc.[arb.]
1
1.5
2
2.5
3
3.5
H2O
/CO2
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
0
5
10
15
PNR(6592cm
1 )
CO2
H2O
CO2
H2O
TNT (H2O/CO2 ~ 0.4)
ENE (H2O/CO2 ~ 9.5)
Conclusions
21
• Conventional munitions
• Fireball emission well represented by a single-temperature Planckian distribution over most of the MWIR
• Non-Planckian emission observed in 2000-2200 cm-1 is likely due to hot CO2
• Accurate atmospheric correction key to connecting this residual to fireball phenomenology
• Temperature decays exponentially (some fireballs exhibit secondary maxima)
• Area dynamics can be determined without imagery (awaiting confirmation from MWIR camera)
• Enhanced novel explosives
• Substantial non-Planckian component is a function of H2O and CO2 concentrations
• Extracted concentration ratio [H2O]/[CO2] connected to explosive stoichiometry
• Simple model enables the study of fireball kinetics
• Explosives classification from optical signatures promising