structural evaluation of flexible pavements using non destructive
TRANSCRIPT
Structural Evaluation of Flexible Pavements Using Non Destructive Tests
Simona Fontul
Young Researchers Seminar 2009Torino, Italy, 3 to 5 June 2009
Outline of the paper
Structural Evaluation of Pavements using NDT
Simona Fontul
1. Introduction 2. Improved Method for Pavement Evaluation
2.1. Non Destructive Tests 2.2. Interpretation of Results2.3. Artificial Neural Networks
3. Case Study
4. Final Remarks
3.1. General Presentation3.2. Data Analysis
Introduction
Requirements for existing roads are more demanding
Efficient methods pavement monitoring
structural evaluation
Pavement maintenance important issue.
Non Destructive Tests (NDT)
Structural Evaluation of Pavements using NDT
Simona Fontul
Introduction
Pavement structural design process
h1
E4, ν4
h2
h3
E1, ν1
E2, ν2
E3, ν3
Asphalt Layers
Granular Materials
Subgrade
“Rigid” layer∞
p0
aP0 pavement structural model
Residual life estimationResidual life estimation
load testing (FWD) deflections layer thickness (GPR)
layer moduli
Structural Evaluation of Pavements using NDT
Simona Fontul
Outline of the paper
Structural Evaluation of Pavements using NDT
Simona Fontul
1. Introduction 2. Improved Method for Pavement Evaluation
2.1. Non Destructive Tests 2.2. Interpretation of Results2.3. Artificial Neural Networks
3. Case Study
4. Final Remarks
3.1. General Presentation3.2. Data Analysis
Improved Method for Pavement Evaluation
Non Destructive Tests (NDT) Falling Weight Deflectometer (FWD)
Structural Evaluation of Pavements using NDT
Simona Fontul
Improved Method for Pavement Evaluation
Non Destructive Tests (NDT) Falling Weight Deflectometer (FWD)
0
50
100
150
200
250
300
350
400
450
500
0 1 2
D0D0D1D2D3D4D5D6
k
Structural Evaluation of Pavements using NDT
Simona Fontul
Improved Method for Pavement Evaluation
Non Destructive Tests (NDT) Ground Penetrating Radar (GPR)
AntennasPulse
generator Transmitting Receiving
Structural Evaluation of Pavements using NDT
Simona Fontul
Improved Method for Pavement Evaluation
Interpretation of results
h1
E4, ν4
h2
h3
E1, ν1
E2, ν2
E3, ν3
∞
p0
aP0
GPR
FWD
E1
E2
pavement structural model
Structural Evaluation of Pavements using NDT
Simona Fontul
Improved Method for Pavement Evaluation
Artificial Neural Networks (ANN) General presentation
ANN structureartificial neuron
biological neurons
y=F(I)
w1 w2 wn
xx11 xx22 …… xxnn
yy
y=F(I)
w1 w2 wn
xx11 xx22 …… xxnn
yy
∑==
n
iii xwI
1
Structural Evaluation of Pavements using NDT
Simona Fontul
InputsDeflections: D0 to D6 (FWD)Layer thickness: h1, h2 (GPR)
OutputsLayer Elastic Modulus: E1, E2 and E3
Layer thickness: h3
Improved Method for Pavement Evaluation
Artificial Neural Networks (ANN) Application of ANN for Interpretation of NDT Results
Asphalt Layers
E1 ; ν1
GranularBase
E2 ; ν2
SubgradeE3 ; ν3
h1
h2
h3
‘Rigid’ layerE4 ; ν4
h1 h2D6D0
h3E1 E2 E3
Structural Evaluation of Pavements using NDT
Simona Fontul
Outline of the paper
Structural Evaluation of Pavements using NDT
Simona Fontul
1. Introduction 2. Improved Method for Pavement Evaluation
2.1. Non Destructive Tests 2.2. Interpretation of Results2.3. Artificial Neural Networks
3. Case Study
4. Final Remarks
3.1. General Presentation3.2. Data Analysis
Case StudyGeneral Presentation
Tests performed
Airport pavement bearing capacity evaluation
• built in 1965, then extended and reinforced in 1980 and 1990• 2400 m long• 45 m wide
Pavement structure
h1: 180 - 230 mmh2: ~100 mm
h3: ~300 mm
h4
Asphalt layers
Bituminous macadam
Unbound aggregate
Subgrade
h1: 180 - 230 mmh2: ~100 mm
h3: ~300 mm
h4
Asphalt layers
Bituminous macadam
Unbound aggregate
Subgrade
FWD GPR Cores
0
100
200
300
400
500
600
700
800
0 200 400 600 800 1000 1200 1400 1600 1800 2000 2200 2400
μ
D0 D1 D2 D3 D4 D5 D6
Structural Evaluation of Pavements using NDT
Simona Fontul
Case Study
0
100
200
300
400
500
600
700
800
0 200 400 600 800 1000 1200 1400 1600 1800 2000 2200 2400
defle
xões
(m)
D0 D1 D2 D3 D4 D5 D6
FWD - deflections
0
10
20
30
40
50
60
70
80
0 200 400 600 800 1000 1200 1400 1600 1800 2000 2200 2400
profun
dida
de (c
m)
camadas betuminosas camadas granularesResultados obtidos com radar
GPR – layer interface depths cores
General Presentation
Structural Evaluation of Pavements using NDT
Simona Fontul
Case Study
BituminousMixturesE1 ; ν1
GranularMaterialsE2 ; ν2
Subgrade
E3 ; ν3
h1
h2
h3
BedrockE4 ; ν4
Elastic layer model
D0
D1
196 1890
156 1590
D2
D3
134 1370
116 1170
D4
D5
85 829
43 588
D6 16 333
D0
D1
196 1890
156 1590
D2
D3
134 1370
116 1170
D4
D5
85 829
43 588
D6 16 333
Data Analysis ANN training
h1
E1
0,22 0,38
1000 6000
h2
E2
0,28 0,32
200 500
h3
E3
0,80 2,40
80 240
Structural Evaluation of Pavements using NDT
Simona Fontul
Case StudyData Analysis ANN training
E1
E2
E3
Training process
Structural Evaluation of Pavements using NDT
Simona Fontul
Case StudyData Analysis ANN training
Structural Evaluation of Pavements using NDT
Simona Fontul
300
450
600
750
900
300 450 600 750 900
D0 meas ured (μm)
D0
calc
ulat
ed -
ANN
stru
ctur
e (μ
m)
a ve rage e rro r: 4%200
300
400
500
600
700
800
200 300 400 500 600 700 800
D1 meas ured (μm)
D1
calc
ulat
ed -
ANN
stru
ctur
e (μ
m)
a ve rage e rro r: 4%
ou tlie rs ou tlie rs
200
300
400
500
600
700
200 300 400 500 600 700
D2 meas ured (μm)
D2
calc
ulat
ed -
ANN
stru
ctur
e (μ
m)
a ve rage e rro r: 5%
ou tlie rs
100
200
300
400
500
600
100 200 300 400 500 600
D4 meas ured (μm)
D4
calc
ulat
ed -
ANN
stru
ctur
e (μ
m)
a ve rage e rro r: 8%
ou tlie rs
50
150
250
350
450
50 150 250 350 450
D5 meas ured (μm)
D5
calc
ulat
ed -
ANN
stru
ctur
e (μ
m)
a ve rage e rro r: 10%
ou tlie rs
100
200
300
400
500
600
100 200 300 400 500 600
D4 meas ured (μm)
D4
calc
ulat
ed -
ANN
stru
ctur
e (μ
m)
a ve rage e rro r: 8%
ou tlie rs
300
450
600
750
900
300 450 600 750 900
D0 meas ured (μm)
D0
calc
ulat
ed -
ANN
stru
ctur
e (μ
m)
a ve rage e rro r: 4%200
300
400
500
600
700
800
200 300 400 500 600 700 800
D1 meas ured (μm)
D1
calc
ulat
ed -
ANN
stru
ctur
e (μ
m)
a ve rage e rro r: 4%
ou tlie rs ou tlie rs
200
300
400
500
600
700
200 300 400 500 600 700
D2 meas ured (μm)
D2
calc
ulat
ed -
ANN
stru
ctur
e (μ
m)
a ve rage e rro r: 5%
ou tlie rs
100
200
300
400
500
600
100 200 300 400 500 600
D4 meas ured (μm)
D4
calc
ulat
ed -
ANN
stru
ctur
e (μ
m)
a ve rage e rro r: 8%
ou tlie rs
50
150
250
350
450
50 150 250 350 450
D5 meas ured (μm)
D5
calc
ulat
ed -
ANN
stru
ctur
e (μ
m)
a ve rage e rro r: 10%
ou tlie rs
100
200
300
400
500
600
100 200 300 400 500 600
D4 meas ured (μm)
D4
calc
ulat
ed -
ANN
stru
ctur
e (μ
m)
a ve rage e rro r: 8%
ou tlie rs
Case Study
Data Analysis
0
100
200
300
400
500
600
700
800
900
0 500 1000 1500 2000 2500Location (m)
Def
lect
ions
(m
)
D0 D1 D2 D3 D4 D5 D6FWD
0
100
200
300
400
500
600
700
800
900
1000
0 500 1000 1500 2000 2500
laye
r int
erfa
ce d
epth
(mm
)
asphalt / granular base
granular base / subgradeGPR
ANN results - asphalt modulus E1
0
1000
2000
3000
4000
5000
6000
7000
8000
50 200
350
500
650
800
950
1100
1250
1400
1550
1700
1850
2000
2150
2300
Location (m)
E1 a
spha
lt m
odul
us (M
Pa)
ANN results - granular base modulus E2
0
100
200
300
400
500
600
50 200
350
500
650
800
950
1100
1250
1400
1550
1700
1850
2000
2150
2300
Location (m)
E2 g
ranu
lar b
ase
mod
ulus
(MPa
)
ANN results - subgrade modulus E3
0
50
100
150
200
250
300
50 200
350
500
650
800
950
1100
1250
1400
1550
1700
1850
2000
2150
2300
Location (m)
E3 s
ubgr
ade
mod
ulus
(MPa
)
Zone 1 Zone 2
Zone 1 Zone 20
100
200
300
400
500
600
700
800
900
0 500 1000 1500 2000 2500Location (m)
Def
lect
ions
(m
)
D0 D1 D2 D3 D4 D5 D6FWD
0
100
200
300
400
500
600
700
800
900
1000
0 500 1000 1500 2000 2500
laye
r int
erfa
ce d
epth
(mm
)
asphalt / granular base
granular base / subgradeGPR
ANN results - asphalt modulus E1
0
1000
2000
3000
4000
5000
6000
7000
8000
50 200
350
500
650
800
950
1100
1250
1400
1550
1700
1850
2000
2150
2300
Location (m)
E1 a
spha
lt m
odul
us (M
Pa)
ANN results - granular base modulus E2
0
100
200
300
400
500
600
50 200
350
500
650
800
950
1100
1250
1400
1550
1700
1850
2000
2150
2300
Location (m)
E2 g
ranu
lar b
ase
mod
ulus
(MPa
)
ANN results - subgrade modulus E3
0
50
100
150
200
250
300
50 200
350
500
650
800
950
1100
1250
1400
1550
1700
1850
2000
2150
2300
Location (m)
E3 s
ubgr
ade
mod
ulus
(MPa
)
ANN results - asphalt modulus E1
0
1000
2000
3000
4000
5000
6000
7000
8000
50 200
350
500
650
800
950
1100
1250
1400
1550
1700
1850
2000
2150
2300
Location (m)
E1 a
spha
lt m
odul
us (M
Pa)
ANN results - granular base modulus E2
0
100
200
300
400
500
600
50 200
350
500
650
800
950
1100
1250
1400
1550
1700
1850
2000
2150
2300
Location (m)
E2 g
ranu
lar b
ase
mod
ulus
(MPa
)
ANN results - subgrade modulus E3
0
50
100
150
200
250
300
50 200
350
500
650
800
950
1100
1250
1400
1550
1700
1850
2000
2150
2300
Location (m)
E3 s
ubgr
ade
mod
ulus
(MPa
)
Zone 1 Zone 2
Zone 1 Zone 2
Inputs Outputs
Structural Evaluation of Pavements using NDT
Simona Fontul
Case Study
Data AnalysisANN results vs classic approach - asphalt modulus E1
0
2000
4000
6000
8000
50 200
350
500
650
800
950
1100
1250
1400
1550
1700
1850
2000
2150
2300
Location (m)
E1 (M
Pa)
ANN classic approach
ANN results vs classic approach - granular base modulus E2
0
200
400
600
50 200
350
500
650
800
950
1100
1250
1400
1550
1700
1850
2000
2150
2300
Location (m)
E2 (M
Pa)
ANN classic approach
ANN results vs classic approach -subgrade modulus E3
0
100
200
300
50 200
350
500
650
800
950
1100
1250
1400
1550
1700
1850
2000
2150
2300
Location (m)
E3 (M
Pa)
ANN classic approach
ANN results - asphalt modulus E1
0
1000
2000
3000
4000
5000
6000
7000
800050 200
350
500
650
800
950
1100
1250
1400
1550
1700
1850
2000
2150
2300
Location (m)
E1 a
spha
lt m
odul
us (M
Pa)
ANN results - granular base modulus E2
0
100
200
300
400
500
600
50 200
350
500
650
800
950
1100
1250
1400
1550
1700
1850
2000
2150
2300
Location (m)
E2 g
ranu
lar b
ase
mod
ulus
(MPa
)
ANN results - subgrade modulus E3
0
50
100
150
200
250
300
50 200
350
500
650
800
950
1100
1250
1400
1550
1700
1850
2000
2150
2300
Location (m)
E3 s
ubgr
ade
mod
ulus
(MPa
)
classic approach
Structural Evaluation of Pavements using NDT
Simona Fontul
Final Remarks
GPR provides continuous information on layer thickness, for the bound and also the unbound layers. Consequently all the FWD test points can be used for backcalculation.
The results obtained using ANN are realistic and similar to the one obtained through classic approach.
It is possible to reduce the computation time and to include some engineering judgment in the process by limiting the range of pavement structure parameters used in the training process.
Future developments
Use ANN for pavement residual life estimation directly from deflections.
Application to other type of pavements (rigid or composite).
Use of more sophisticated pavement models (visco-elastic, non-linear response).
Structural Evaluation of Pavements using NDT
Simona Fontul
Structural Evaluation of Pavements using NDT
Simona Fontul