Few experimental data sets exist in the literature to support the development and evaluation of digital twins 
predicting structural degradation. The literature is especially sparse for system tests where multiple failures 
occur and interact. In this work, a laboratory-level experiment is conducted to mimic many of the properties of 
larger and more complex marine structures with redundant load paths, failure interaction, and component-to-system 
level integration. In the experiment, such properties are reflected by a hexagon tension specimen with four 
propagating fatigue cracks tested under displacement-controlled loading. The hexagonshaped specimen contains 
four pre-cracks at each corner to boost the initiation of cracks and restrict the crack propagation in desired areas.
The test is conducted on MTS 810 material testing system. The applied loading cycles and corresponding crack lengths 
are recorded as the major time-varying data of degradation, with the resisting force at maximum extension used as 
the system capacity. A novel computer vision method is used to measure the crack length. Strain gauges are also 
used to monitor the structure's status. The experimental data is presented in this dataset. The resulting data 
sets can be used to evaluate the performance of different digital twin updating approaches.

Five hexagon specimens have been manufacutred and tested. The first hexagon specimen has horizontal pre-cracks 
while the rest four hexagon specimens have pre-cracks with 15 degree orientation which helps the propagated 
crack be straight. The five hexagon specimens have the same dimensions and materials. A grip is designed to mount 
the specimen onto the MTS 810. The details of the design of hexagon specimen and grip can be found in the file 
named "5_Design of hexagon specimen and grip".

The data gathered from the five experimens is the applied loading cycles and corresponding crack lengths 
as the major time-varying data of degradation, with the resisting force at maximum extension used as 
the system capacity. The loading cycles and the resisting force at maximum extension are readed from MTS 810 
material testing system directly. To gather crack lengths from the experiment, three methods were used in this study: 
visually using a machinist scale, a novel computer vision method and a digital image correlation (DIC) method. For visually 
measuring the crack length, a machinist scale with marking of 0.5mm is used. The applied cycles, visually measured crack length,
and the max reaction force for the five hexagon experiments are stored in the file names "1_crack length & max reaction force.xlsx".
The crack lengths are also measured by computer vision method and a digital image correlation (DIC) method, which requires taking 
pictures of the crack area. The details of these two methods are listed below:


Computer vision method

The computer vision method takes a digital image of a specimen with an embedded crack, and attempts to identify and 
measure the crack length on the level of individual pixels in the image. A GoPro HERO4 camera with 12.0 Mega Pixel (MP) 
resolution is used to take JPEG pictures at a 4000*3000 pixel resolution. A NEEWER 12.5X macro lens is mounted on the 
GoPro camera to focus on the small region around a growing crack. The camera is mounted on a flexible tripod that can 
be easily moved around a specimen to take images of all active cracks. In order to increase the contrast between the 
structure surface and the crack, the crack area is painted with a fluid fluorescent dye showing a bright maize color 
under ultraviolet (UV) lights. The UV dye is painted on the surface of the structure when the crack is closed by 
removing the loading on the structure (e.g. pausing the fatigue test apparatus at zero or a low load value). After 
the crack is re-opened by re-applying load to the specimen, the dye makes the crack opening clearer in the image. The 
computer vision method is applied in the 2nd, 4th, and 5th hexagon experiments. The captured images are inside 
"The Second Hexagon - CV data", "The Fourth Hexagon test - CV data", and "The Fifth Hexagon - CV data" in the folder named
"2_Crack figures-CV method". For example, the folder "2_Crack figures-CV method\The Second Hexagon - CV data\300000\left top" stores 
the images of the lefttop crack of the second hexagon experiment at 300000 cycles. All the images are JPG files.

As the proposed computer vision method was new, it was validated by two standard eccentrically-loaded single edge crack 
tension specimens, which were tested on an MTS 810 testing system. The standard specimens were preloaded at 8 kN to 
guarantee a slack-free connection between the specimen and the fixture bolts. The tests were conducted under 
displacement control with a maximum displacement amplitude of 0.14 mm resulting to a maximum reaction force of 23.6 kN. 
The images of crack are captured using the same way as hexagon experiment. Take folder "2_Crack figures-CV method\CV validation 1st\150000" 
as an example, it contains the images of the crack of the 1st validation experiment at 150000 cycles. All the images are JPG files.

digital image correlation (DIC) method

DIC is a post-processing method to acquire the displacement and strain field of a structure under deformation.  By looking for the 
characteristic strain field around a crack tip, DIC can be used for detecting the crack tip and calculating the crack length. 
A Blackfly BFLY-PGE-31S4M GigE Charge Coupled Device (CCD) camera manufactured by FLIR was used to take pictures for DIC method. 
The reason for choosing CCD camera is that its sensors can create high-quality and low-noise images. A Sony IMX265 sensor is used 
in the Blackfly CCD camera, capturing monochrome images at 35 Frame per Second(FPS) with a resolution 2048*1536. A Tamron 23FM25SP lens 
is combined with the CCD camera, this is a 25mm focal length lens, and on the hexagon specimens it can capture a figure around 10cm*8cm area 
at a 26cm objective distance. The camera was connected to a GigE host adapter with a RJ45 connector and controlled by a related piece of 
software named FlyCapture. In addition to the CCD camera system, a LimoStudio 700W photography lighting system is chosen to provide 
high-quality and uniform light on the specimen. A white back drop hung behind the specimen for a clean background to help in the image analysis. 

The DIC images are gathered in the 2nd, 4th, and 5th hexagon experiments and stored in the folder named as "3_Crack figures-DIC method". Take 
"3_Crack figures-DIC method\The Second Hexagon - DIC data\300000\Left-Bot" as an example, it contains the images of the left bottom crack of the 
second hexagon experiment at 300000 cycles. There are two images files in this folder named "min.png" and "max.png". "min.png" is the crack image
before applying displacement to the specimen which is usually used as the reference image in DIC method. "max.png" is the crack image after applying
the max displacement. All the images are PNG files.


Besides measuring crack length with a machinist scale, a computer vision method and a digital image correlation (DIC) method, strain gauges are also 
used to monitor the structure's status in the 3rd, 4th, and 5th experiments. OMEGA uni-axial strain gauges with 350 Omega resistance were employed 
in the hexagon tests. A quarter-bridge Wheatstone bridge setup was used for the gauges in this case, and a compensating strain gauge for the effect 
of temperature was not used since both the room temperature and specimen temperatures were stable during the 3-5 day time of these experiments.  
A National Instrument measurement system containing NI 9236 and cDAQ-9181 were used to capture the strain data for these tests. NI 9236 is a 350ohm, 
quarter bridge input module with 8 channels. cDAQ-9181 is a chassis with one slot capable to connect one NI 9236 modules to a computer via Ethernet. 
The system measures dynamic strain on all channels simultaneously allowing for synchronized, high-speed measurements. The data recording is controlled 
by a LabView algorithmm, storing the strain by taking a 1 second sample every 20 seconds with 1000 Hz sample rate. The test frequency for this test 
was normally between 5 Hz and 7 Hz, so the 1 second sample time captured several complete load cycles. The strain data is stored in lvm text files inside
the folder named "4_Strain data". For example, "4_Strain data\Third Hexagon Strain Gauge Data\data-100000to200000.lvm" stores the strain data of the third
hexagon from 100000 cycles to 200000 cycles.


To summary, the data was collected from the hexagon experiments listed as follows,

(1) "1_crack length & max reaction force.xlsx": Crack length recorded by machinist scale and maximum reaction force from MTS testing machine

(2) "2_Crack figures-CV method": Crack figures captured for computer vision method (the 2nd, 4th, and 5th hexagon experiments), including two validation tests

(3) "3_Crack figures-DIC method": Crack figures captured for Digital Image Correlation method (the 2nd, 4th, and 5th hexagon experiments)

(4) "4_Strain data": Strain data measured by strain gauges (the 3rd-5th hexagon experiment)

(5) "5_Design of hexagon specimen and grip": Design of hexagon specimen and grip

If you are interested in more details of the study, please visit http://hdl.handle.net/2027.42/155231 for the thesis.