tools for evaluation of steel truss bridge gusset plates christopher higgins quang nguyen o. tugrul...
TRANSCRIPT
Tools for Evaluation of Steel Truss Bridge Gusset Plates
Christopher HigginsQuang NguyenO. Tugrul Turan
School of Civil and Construction EngineeringOregon State University
OUTLINE
Introduction & Background
Digital Imaging Tools
CAD- FE Analysis
Summary and Future Work
INTRODUCTION & BACKGROUND
Introduction & Background
• Gusset plate connection performance of interest.
•Structural evaluations require accurate as-built data.
•Collection of field data is time consuming and expensive. Results likely to be highly variable.
•Rating methods rely on specification-based on design provisions. Need for improved tools.
Collapsed I35W Bridge in Minneapolis, MN
Overall Approach
•Capture images in field with digital cameras (snooper, climbing, from ground level)
•Process images into scaled orthophotos (MatLab)
•Query and extract geometric data (MatLab)
•Directly produce CAD drawings (AutoCAD)
•Directly produce FE models (ABAQUS)
•Run nonlinear finite element analyses to rate connections (ABAQUS)
•No high-level training required (scripted CAD drawing and FE modeling)
•Integration from field inspection to rating and archiving
Collecting Field Data is a Challenge
Inspectors Get Creative
Is there a better way?
Methodology
• Gusset plates lie in plane : 2-D correspondences simplify transformation
• Point correspondences used to map points based on central projection.
• Need to establish control points in the real world image.
X’ is point in photographic image plane (pixel)
X is point in “real world” imageHartley and Zisserman, Multiple View Geometry in Computer
Vision, 2004
Methodology
• Relate image points (pixels) to real world control points via transformation matrix
• Remove image perspective (rectification)
•Direct Linear Transform (DLT) algorithm
•H is 3x3 transformation matrix
• Need minimum of 4 control points (8 DOF and 1 for scale)
• If over constrained then can estimate errors.
X’ and Y’are point coordinates onphotographic image plane (pixles)
'
'
1 1
i i
i i
X X
H Y Y
X and Y are point coordinates inreal world image
Methodology
8” and 24” Reference Targets
• Ultra Corrosion-Resistant Pure Titanium (Grade 2) bars
• Nine aluminum control points
• Safety chain and vise-grip clamp
Reference Targets
Base magnetically attaches and sits 2” off of the face of the plate
Methodology
Methodology
Methodology
Root-Mean-Square Error:
where θ’ is the predicted value, θ is the true value, and n is the sample size.
Standard deviation:
where e is the location error, ē is the average location error, and n is the sample size.
2
1
1( )
n
ii
e en
2
1
1( ' )
n
i ii
RMSEn
Error Estimation
Methodology
Location Uncertainty:
Measurement Error for Length Measurements:
where RMSE is the predicted error value, 3σ provides a 99.7% confidence interval for a normal distribution, and the given pixel size.
Measurements of length require two points thus double the location uncertainty.
2 3measurementE RMSE pixel
int 3poE RMSE pixel
Error Estimation
Methodology
Standoff correction•Targets are offset from plates
•Image dimensions are shorter than true dimensions due to standoff
• Correction may be performed:
actual imageoffset
DL L
D D
Actual dimension onplate
Distance to plate, D
Standoff distance of targets, Doffset
Image dimension
Implementation
Consumer grade camera• Single-lens reflex (SLR)
digital camera (Nikon D200)
• Nikon AF NIKKOR 50mm 1:1.8D lens
• Resolution: 3872x2592 Pixels (10 Megapixel)
Fill the image frame with the gusset plate as much as possible.
Take a well focused image. Take the best available image
with the camera as orthogonal to the gusset plate surface as possible.
Deploy flat-field lenses.Take at least one physical length
measurement on the gusset plate (and mark on the plate so it is visible in the image).
Take a measurement of the plate thickness, as this cannot be determined directly.
Map corrosion damage, if any.
Implementation
MATLAB w/ Image Processing Toolbox - Gusset Plate Rectification Tool
• Loads gusset plate image
• Calculates H matrix
• Calculates the measurement error and standoff correction
• Creates and saves rectified scaled orthographic photograph
• Allows measurements
• Length
• Perimeter
• Fastener locations
• x, y coordinates
• Angles
• Outputs all data into an Microsoft Excel file
Software
Implementation
Microsoft Excel – LSP file creator
• Take data from Gusset Plate Rectification Tool and create a LSP file to be used by AutoCAD to create a scale drawing of the Gusset Plate
• Also creates LSP file for ABAQUS FE analysis
Software
Implementation
Mock Gusset Plate:• Connection U3• Southbound Booth
Ranch Bridge on I-5 Oregon.
• As-constructed drawings from March 1967
Example
Example
Design Drawing Typical View of Connection
Example
Image Rectification Tool
MATLAB program
Example
Example
Camera 3 – 15ft @ 45° Off-center
Example
Camera 2 – 250ft @ 0° Off-center
Typical View of Connection
Compare Design with As-Built
Extra bolts!
Work point shifted due to member alignment!
Plate clipped differently
Remember I35W?
ODOT Field Trials X Y
-17.733230.1962
6
-18.06 27.0375
-18.386823.5519
6
-15.119130.1962
6
-15.554826.9285
7
-15.663723.6608
9
-12.39630.3051
9
-12.7228 27.0375
-12.831723.7698
1
-9.6729230.0873
4
-9.6729226.9285
7
-9.9996823.6608
9
-6.9498430.1962
6
-7.0587626.8196
5
-7.1676923.7698
1
-4.2267730.1962
6
-4.22677 27.0375
-4.3356923.6608
9
-1.6126129.9784
2
-1.61261 27.0375
-1.6126123.6608
9
1.219383 29.8695
1.328306
26.71073
ODOT Field Trials (hidden edges)
ODOT Field Trials
Outside Agencies and Consultants
US Army Corp Unknown Consultant
Close-Range Requirements (climbing)
+
Combine Multiple Images
=Also can get 3D
Mosaic from Multiple Images
Rectified Images Become Models
* FE Model created from panoramic image
Structure, Member, Connection, Fastener
FEA
Nonlinear FE Modeling and Analysis
X,Y coordinates (digital image),
Thickness of plate (field measured), Material
(drawings),and Member forces (stress
sheet)
Automatic Output Generation (Strength)Multiple Members : von Mises Stresses
Validating Methods with Full-Scale Experiments at OSU July 2010
SummaryMethod for rapid and accurate collection of
field geometry of gusset platesImages are rectified and scaled to permit
metrificationData query export via Excel to CAD and FEACompare as-built connections with design
drawingsAnalyze and rate connections directly using
FE methodsQuantitative imaging enriches field
inspection and evaluation
Acknowledgements
Bruce Johnson – Chief Bridge EngineerSteven Soltesz – Research Coordinator
Oregon Transportation Research and Education Consortium
Questions?