www.nature.

com/scientificreports

OPEN Deep neural networks for crack

detection inside structures
Fatahlla Moreh 1, Hao Lyu 1,2*, Zarghaam Haider Rizvi 1,3 & Frank Wuttke 1

Crack detection is a long-standing topic in structural health monitoring. Conventional damage

detection techniques rely on intensive, time-consuming, resource-intensive intervention. The current
trend of crack detection emphasizes using deep neural networks to build an automated pipeline
from measured signals to damaged areas. This work focuses on the seismic-wave-based technique of
crack detection for plate structures. Previous work proposed an encoder–decoder network to extract
crack-related wave patterns from measured wave signals and predict crack existence on the plate.
We extend previous work with extensive experiments on different network components and a data
preprocessing strategy. The proposed methods are tested on an expanded crack detection dataset. We
found that a robust backbone network, such as Densely Connected Convolutional Network (DenseNet)
can effectively extract the features characterizing cracks of wave signals, and by using the reference
wave field for normalization, the accuracy of detecting small cracks can be further improved.

Keywords Neural network, Deep learning, Crack detection, Wavefield

Crack detection is a long-standing structural health monitoring (SHM) research topic. Cracks can change mate-
rial behaviour and weaken structural strength. Even tiny cracks can evolve into severe damage to structures and
cause heavy loss of property and life. The ultimate goal of SHM is to estimate the structure’s state from sensor
data and use the information to make decisions regarding the maintenance and decommissioning of the struc-
ture. Rytter pointed out that the SHM task is a hierarchy of levels, including detection, localization, assessment,
and ­prediction1. Recent developments in SHM show increasing demand for more sophisticated data processing
and analyzing techniques for the four levels of SHM ­tasks2. Deep learning (DL) approaches are getting particu-
lar ­attention3. Because DL approaches enable machines to automatically learn task-specified features without
relying on hand-crafted features from experts and ­researchers4. Deep neural networks (DNN) take raw data
as input and develop hierarchical representations through stacked neural network layers with nonlinear map-
pings. Automated image acquisition platforms and deep learning image processing algorithms are extensively
used in current research for surface crack detection, replacing manual visual ­inspection5–7. These algorithms
use images as input and estimate damage from different levels, ranging from categorizing damage/non-damage
classes to identifying different severity levels or damage types. Attempts were also made to use DNNs in wave
propagation-based damage detection. The forms of input data greatly influence the selection of DNN variations.
Dense Networks, 1D-Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), and Long
Short-Term Memory (LSTM) are used to learn vibration signatures from the time d ­ omain8–10. 2D CNNs are
usually employed to learn damage-relevant features from the frequency ­domain11 or an array of ­sensors12. The
propagation properties of waves depend on the properties of the media they travel through and the parameters
of the excited wave. The interaction between waves and cracks modifies the wavefield. The wavefront reflects
and refracts when it encounters the cracked interface. When assuming that sensors are used to capture this
modification, a crack detection method aims to establish the connection between the observed wave data and the
damaged region. In previous work, we developed a spatially asymmetric encoder-decoder model to approach the
­connection13,14. The detection of cracks on plates is addressed as a semantic segmentation problem, where each
point in the material domain is classified into damaged/intact classes. The “encoder network” receives observed
data as input and extracts crack-related features. The “decoder network” masks the damaged region according
to the extracted features. For the specific task, the developed model is tailored to the specific challenges of the
problem because it takes the wavefield measurements from a sparsely distributed sensor array and makes dense
predictions for the whole material domain. The backbone network of the encoder, e.g., Visual Geometry Group
(VGG)15 or Residual ­Networks16, is modified to a 1D version according to the format of wavefield data. Upsam-
pling layers rescale the encoded features into a desired spatial resolution.

1
Geomechanics and Geotechnics, Kiel University, Kiel 24118, Germany. 2Competence Centre for Geo‑Energy,
Kiel University, Kiel 24118, Germany. 3GeoAnalysis Engineering GmbH, Kiel 24118, Germany. *email: hao.lyu@
ifg.uni-kiel.de

Scientific Reports | (2024) 14:4439 | https://doi.org/10.1038/s41598-024-54494-y 1

Vol.:(0123456789)
 www.nature.com/scientificreports/

In this work, we extend the previous work, set up a baseline with extensive experiments on different compo-
nents of the asymmetric encoder-decoder structure, and propose a new normalization strategy for input wave
signals. The rest of the paper is organized as follows: Sect. ”Related work” briefly introduces recent crack detec-
tion achievements and highlights the deep learning approaches. Section Problem definition presents the formal
definition of our crack detection problem and indicates the difference between the crack detection problem and
the image semantic segmentation problem. The crack detection dataset is reported in the section “The crack
detection dataset”. In “Deep neural networks for crack detection” and “Benchmarking experiments”, we have
described the DNN models and reported the results of benchmarking experiments. Finally, in the “Conclusion”,
we discussed the paper with final concerns and remarks.

Related work
Damage detection in shell structures
Active lamb wave-based methods are one of the most promising techniques for SHM of shell structures. Lamb
waves can travel long distances without significant loss in intensity. Thus, large areas of shell structures can be
examined with a sparse array of transducers. Yuan et al. developed an active monitoring system for honeycomb
sandwich and carbon fibre composite structures based on wide-band Lamb ­waves17. They employed a two-layer
neural network to extract the wave signature of two types of damage: delamination and impact damage. Golato
et al. developed the SHM method based on the sparse reconstruction technique to identify the size and shape of
the ­defect18. The proposed method produces a heat map of damage probability for the region of interest through
the fundamental A0 wave mode propagation captured by multiple sensors. Their method was validated using
simulated data for a thin aluminium plate. The interaction of Lamb wave modes with damages can be strongly
dispersed, which leads to complex signals being interpreted. Numerical simulations are often needed to offer
some possibilities to interpret measured signals. Gravenkamp and his colleagues applied the scaled boundary
finite element method (SBFEM) to simulate Lamb waves in cracked plates with more efficiency and accuracy
compared to ­FEM19. Besides sparsely distributed sensors, complete wavefield data is also an effective damage
detection, localization, and visualization tool. Kudela et al. applied a set of signal processing techniques to extract
crack-relevant features without using baseline ­signals20. A high-resolution scanning laser Doppler vibrometer
acquires the complete wavefield data. However, applying the methods to high-resolution wavefield data is still
challenging, requiring an extended processing time and much storage space.
Another category of SHM systems passively receives the sound waves from a cracking site in a structure,
namely the acoustic emission (AE). AEs are caused by sudden changes of strain in fracturing material that lead
to a burst of energy released by high-frequency sound waves. For the general application of AE in SHM, we
refer to the review p­ aper21. Here, we only name several AE-based SHM approaches applied to shell structures.
Farhidzadeh and his colleagues analyzed the b-value of AEs to monitor the fracture process of a large-scale
reinforced concrete shear ­wall22. The b-value is the log-linear slope of the frequency-magnitude distribution
of acoustic emissions. The fracturing events are distinguished from baseline data through a statistical outlier
analysis. The authors have identified the pattern of fracturing in AE signals. These patterns, though not used
to indicate the location of the crack, can be used as an early warning of structural failure. Van Driessche et al.
investigated the AE patterns in a fracturing wall of textile-reinforced cement for two types of damages: mode-I
cracking and shear-dominated debonding and ­delamination23. Argus et al. developed an AE-based SHM system
for thin-walled carbon fibre reinforced plastics (CFRP) s­ tructure24. The AEs were simulated through an artificial
source and received by piezoelectric sensors. They employed an artificial neural network to localize the sound
source from the monitored acoustic signals with high precision.

Deep learning approaches for crack detection

The deep learning approaches for crack detection can be classified into two types according to their input data,
i.e., the vision-based approach and the wave-based approach.
The vision-based approach uses images and videos as input and adopts the advanced DNNs, which were
initially designed for computer vision tasks for surface inspection. Depending on the granularity of the crack
detection task, vision-based methods can be categorized into three types: (1). image classification, i.e., categoriz-
ing image patches into “damaged” or “intact” classes or into different severity ­levels5,25,26, (2). object detection
or region-proposal methods, identifying the crack’s existence and predicting its bounding box on the image
­patch7,27,28, and (3). semantic segmentation methods, classifying each pixel of an image patch into “damage” or
“non-damage” ­classes6,29–34. The semantic segmentation approaches have been widely applied as an alternative
to visual inspection for monitoring civil structures, such as ­roads35, ­tunnels36, and ­bridges37.
The wave-based approach is based on the idea that the crack’s existence changes the physical properties of
the structure (e.g., mass, damping, and stiffness) and thus leads to the change of modal parameters (e.g., natural
frequencies, modal shapes, modal damping)38. Usually, wave propagation can be induced by natural sources (e.g.,
seismic) or human-made sources (e.g., guided waves). The measured responses, such as displacement, velocity,
and acceleration, can be compared with the responses of a predefined numeric model or the benchmark responses
from the non-damage structure or analyzed statistically without any benchmark model.
Recent studies demonstrate that DNN can be a powerful tool to extract wave patterns from wave response
signals even with high noise ­levels39,40. Abdeljaber and his colleagues presented a 1D-CNN model to identify and
localize structural damage from raw vibration s­ ignals41. A network of accelerometers acquires these signals, and
the damage is simulated as a loose bolt in the structure. For each damage case, a separate CNN is optimized to
estimate the probability of damage (PoD). The model is evaluated for undamaged, single, and multiple damage
cases with high accuracy (0.54% average error). In the following paper, they applied 1D-CNN to address the
binary damage detection problem, distinguishing between undamaged and damaged ­cases8. Based on simulated

Scientific Reports | (2024) 14:4439 | https://doi.org/10.1038/s41598-024-54494-y 2

Vol:.(1234567890)
www.nature.com/scientificreports/

vibration responses, Lin et al. demonstrate that a 1D-CNN model performs better than the signal processing
method that analyses the wavelet packet transform (WPT) component energy to identify a local change in the
system ­parameters42. The Neural Network (NN) model is robust to white Gaussian noise and can detect the
location of single/multiple damages with corresponding damage degrees. To generate vibration responses, the
authors build a supported beam model consisting of 10 identical Euler-Bernoulli beam elements and calculate the
vertical accelerations of each beam element under different damage locations, damage levels, and burst random
excitation. Besides 1D-CNN, Long-Short Term Memory (LSTM) network is also commonly used as the pattern
extractor for damage detection. Hung et al. build an LSTM-based single end-to-end network that directly takes
raw acceleration time-series without requiring any signal preprocessing ­step10. The proposed approach is evalu-
ated on the vibration data collected from a laboratory-scale three-story frame and produces accurate damage
detection results. Considering the difficulty of collecting training data with precisely labelled damages from the
in-service infrastructures, Wang and Cha proposed an unsupervised method that combines CNN and support
vector machine (SVM) to learn signal patterns from a known undamaged scenario and identify damages of
unknown ­scenarios43. Though response data can be easily generated from numeric simulations, collecting data
for every structure part is impractical.
Goh et al. represent a study on predicting unmeasured mode shape magnitudes based on a limited number of
measured data with the NN ­model44. Compared with the Cubic Spline interpolation (CS) method, the NN model
provides more reliable results with limited measurement points. Ding et al. present a sparse deep belief network
(DBN) for structural damage detection with a noisy and limited amount of ­data45. The proposed model takes
vibration characteristics in the modal domain, i.e., natural frequencies and mode shapes, as input to the network
and predicts the structure’s damage locations and levels. To consider the sensors’ geometric location, Sajedi and
Liang developed a grid-based methodology for damage segmentation in large-scale civil ­infrastructures46. Their
fully convolutional encoder-decoder neural network takes wave response from an array of sensors and predicts
damage status for each sensor location. Cumulative intensity measures train the model as the input and damage
states of nodes as output. Their proposed approach yielded global accuracies of 96.3% and 93.2% for detecting
damage location and severity in an FE model, respectively. This grid-based methodology is very similar to the
proposed model in this paper. One of the significant differences is that our model takes wave responses from
sparsely distributed sensors as input and produces very dense predictions.

Problem definition
The crack detection problem of a solid structure can be formalized as follows: Given a solid structure repre-
sented by a physical domain  ⊂ Rn (n = 2, 3), an internal crack can be written as a function φ(x) = 1, where
x ∈ R n is a point in the domain. The intact area is represented as φ(x) = 0. Given N sensors at fixed locations
x 0 , · · · , x i , · · · , x N , the recorded wave data at time t of sensor location x i is denoted as u(i,t). For example, if the
observation period is given from t0 = 0 to t = T , we omit the temporal index without any ambiguity. The crack
detection problem can be addressed by finding a functional G : f (u0 , · · · , uN ) � → φ(x).
In practice, we can use a neural network of trainable parameters θ to approximate the functional

Ĝθ (u0 , · · · , uN )(x), (1)

Training such a neural network means finding optimal parameters θ that ∗


∗
 
θ = arg min Ĝθ (u0 , · · · , uN )(x) − φ(x)dx. (2)
θ �

To simplify the problem, a 2D rectangular domain is assumed. The physical domain is decomposed into a regular
grid tessellation. The cell size of the grid determines the spatial resolution of the decomposition. Determining
cell size requires considering the size of the crack region, material properties, excitation source intensity, etc. Too
small cell size increases the difficulty in prediction, while too large cell size makes it difficult to determine the
exact location of the crack. Grid cell can be indexed by a row-column pair i, j. Then, we can convert the above
optimization problem to a discretized version,

θ ∗ = arg min
 
Ĝθ (u0 , · · · , uN )(x i,j ) − φ(x i,j ).
θ (3)
i,j

The simplified problem is comparable to the image semantic segmentation task. Semantic segmentation is a com-
puter vision task that aims to classify each pixel of an input image into a fixed set of semantic categories without
distinguishing object instances. Deciding what class a pixel belongs to requires information from the pixel itself
and the context information from its neighbouring pixel. This requirement leads to a widely accepted encoder-
decoder architecture in deep learning. The encoder learns highly abstracted but coarse semantic information in
deep layers and fine appearance information in shallow ­layers47. The decoder network semantically projects the
information learnt by the encoder (lower resolution) onto the pixel space (higher resolution) to get a dense clas-
sification. Some models also employ “skip” connections to combine shallow and deep ­features47–50. The encoder
usually employs well-established CNN architectures, such as V ­ GGNet51 or Residual Network (ResNet)52; in the
decoder, the combinations of upsampling and trainable layers achieve the pixel-level prediction.
This encoder-decoder architecture has been widely used in computer vision tasks like image segmenta-
tion. Similar architectures are also applied to visually inspect surface cracks, where hand-holding devices or
autonomous platforms can collect p ­ hotos34,53,54. Image segmentation requires an understanding of the image
from both semantics and location, which induces the tension between capturing global information (semantics)

Scientific Reports | (2024) 14:4439 | https://doi.org/10.1038/s41598-024-54494-y 3

Vol.:(0123456789)
 www.nature.com/scientificreports/

and capturing local information (location)47. Many models use the spatial correspondence between the input
and output domains and add a skip connection between the deep features in the encoder network and decoder
­network47,48.
Crack detection from observed wave data is comparable to the binary semantic segmentation task, where
each image is classified as damaged or intact, where each pixel of the material domain requires an assignment
of the class. The model requires learning semantic and locational information from condensed wave patterns.
Thus, similar ideas were borrowed from image segmentation models. However, there is an essential difference
between wave-based crack detection and image semantic segmentation, i.e., the spatial mismatch between input
and output. The ultimate aim of a crack detection model is to use the data from the sparsely distributed sensor
array to make dense predictions for the solid domain. Though regularly distributed sensor arrays imply locational
information in the data, they can not be easily related to the spatial locations of the output. The spatial dimension
of sensors (sparse) is mismatched with the desired spatial dimension of crack predictions (dense). Besides, the
measured wavefield also involves temporal dimension and x-/y-displacements instead of image channels. In our
model, this information is condensed and combined by the encoder in a temporal-to-spatial way.

The crack detection dataset

The wave-crack interaction dataset consists of homogeneous 2D samples of a rectangular domain. Each sample
contains a randomly assigned crack (or no crack). The previous studies built a small dataset and illustrated the
influence of crack size on the model’s p ­ erformance13,14. Specifically, much smaller cracks are more difficult to
detect than larger cracks. To facilitate a fair evaluation, it’s necessary to have a similar distribution of samples of
different crack sizes in training, validation, and testing datasets. The previous dataset and the random selection
of the validation set from the training set fail to address the concerns. In this work, we expanded the dataset
size by adding more samples and re-sampling them for the three subsets. The dynamic lattice element method
simulates their dynamic behaviour under an external force.

Dynamic lattice element method

Compared to crack generation in real-world samples, numeric simulation is a fast and efficient alternative for
producing massive data to meet the data requirements of training deep neural networks. On the other hand,
producing a large number of real-world samples is prohibitively ­expansive55. To be compatible with the previous
dataset, we adopted the same technique for simulating the dynamic behaviour of the samples, i.e., dynamic Lattice
Element Method (dLEM)56,57. The lattice element method (LEM) consists of mesoscale models that decompose a
material sample into lattice cells. The lattice nodes can be considered as the centres of the lattice particles, which
are connected by beams that can carry normal force, shear force, and bending moment. Cracks can be repre-
sented by simply removing the lattice particles. We can simulate the wave propagation in the samples by solving
the equation of motion under an external force (Eq. 5). We adjust the external force so that the wavelength is
greater than the length of the elements to ensure an accurate r­ esult58. Details of the equation of motion of the
dLEM are referred to in a previous s­ tudy13. The most relevant formulas for deriving dLEM are given below. In the
dLEM, the resistance to axial, transverse and rotational displacements are modelled using three spring elements:
A A l2
ks = G , kn = E , kφ = kn , (4)
l l 12
where A is the cross-section area, l is the length of elements or Euclidean distance between nuclei, ks is the trans-
verse stiffness, kn is the axial stiffness, kφ is the rotational stiffness, G is the shear modulus and E is the Young’s
modulus. These stiffness components are composed and transformed into the global coordinate system to obtain
the global stiffness matrix. The general equation of motion under an external excitation is given in the following:
M ü(t) + C u̇(t) + K u(t) = f (t), (5)
where ü , u̇ , u represent accelerations, velocities, and displacements respectively, M is the mass matrix, C is the
damping matrix, and K the global stiffness matrix, f is an external excitation. The contact damping is neglected
in the simulation. The mass matrix is assembled using the consistent mass matrix (CMM) by lumping the mass
at the nodes. The global stiffness matrix is assembled by assuming a linear elasticity (according to Eq. 4).
The Newmark-β method with the incremental formulation is adopted for solving the equation of motion.
1 1 1
δ ü = δu − u̇ − ü,
β�t 2 β�t 2β
γ γ γ (6)
δ u̇ = δu − u̇ + �t(1 − ü),
β�t β 2β

δ ü = ü(ti+1 ) − ü(ti ),
δ u̇ = u̇(ti+1 ) − u̇(ti ), (7)
δu = u(ti+1 ) − u(ti ),
where β and γ are the Newmark-β parameters; t = ti+1 − ti ; and δ ü , δ ü , δ u̇ , and δu are the increments of
accelerations, velocities, and displacements. The incremental form of the Equation 5 becomes
Mδ ü + Cδ u̇ + K δu = δf , (8)

Scientific Reports | (2024) 14:4439 | https://doi.org/10.1038/s41598-024-54494-y 4

Vol:.(1234567890)
www.nature.com/scientificreports/

By solving the motion equation, the displacements of the excited wave field are calculated for each time step. The
wave displacements at specific locations are extracted as the observed wave data.

Dataset generation
The dataset consists of homogeneous 2D plates with various cracks in terms of size, orientation, and location.
Some samples containing no crack are generated as reference samples. The homogeneity means the lattice par-
ticles share the same density and Young’s Modulus. Generating the lattice particles involves a random drifting
of regular grid nodes. Then, the nodes are used to generate Voronoi cells to represent the geometric shape of the
lattice particles. To simulate damaged cases, randomly generated linear cracks are placed in the domain, and
corresponding lattice particles are removed accordingly. Each sample contains 100 × 100 lattice particles. The
wave field is excited by a 1000N load in a short time span (in 3 time steps), and the displacements in the x- and
y-direction are recorded for 2000 time steps. Figure 1 shows the setup for the virtual sensors and their recorded
wave field history (in the x-direction). The red dot on the left middle edge indicates the location where the exci-
tation is induced. The 9 × 9 orange dots in the material domain are the locations of the sensors. The resulting
wave field forms a 9 × 9 × 2000-shaped array.
For each sample, a binarized image shows the ground-truth crack existence. The original image resolution
is 100 × 100, in alignment with the mesh size used by dLEM. The original image is then downsampled to the
resolution required 16 × 16 for prediction. The image pixels are set to 1 at the locations where the lattice nodes
are removed to represent the presence of a crack. In contrast, 0 valued pixels indicate the intact material domain.
The following figure (Fig. 2) shows some ground-truth images for the corresponding 2D plate samples.

Dataset facts
The dataset consists of randomly generated samples in the same 2D domain, with a single, straight crack of
random length, location, and orientation. A small portion of undamaged samples is also generated for the train-
ing dataset as references. The dataset has 4200 samples for training, 960 for validation, and 1600 for testing. An
overview of the crack size distribution is given in Fig. 3. The crack size is specified as the number of damaged
pixels among the ground-truth images’ total pixel numbers (100×100). Stratified drafting ensures that the dis-
tributions of crack sizes among training, validation, and test datasets are similar.

Preprocessing
Normalization is a crucial step to guarantee convergence and numerical stability since the value of the measured
wave signal is very small. Previously, a scaling function for normalization was used. The min-max normalization
function maps the input data X to the interval [−1, 1] ∈ R . The normalized wave signals are then given to the
neural network as input.
Besides a direct link between the measured wave signals and the crack’s existence, we developed an alterna-
tive preprocessing method using a reference wavefield. The idea is to learn from the difference between the wave
signals measured from undamaged and damaged samples. We generated the reference wave signals as the average
of the measured wave signals of all the undamaged samples in the training dataset. Then, the input becomes a
normalized difference of the signals, subtracting the reference wave signals.
The insignificant change in the wavefield causes difficulty in detecting small cracks. By normalizing the differ-
ence, we scale the insignificant signals measured from samples with small cracks. We believe this preprocessing
method helps the crack detection models to distinguish undamaged samples and samples with small cracks.
Through empirical study, we can illustrate the efficiency of the preprocessing method.

Figure 1.  The placement of sensor array in sample domain and the recorded wave removal in the x-direction.

Figure 2.  Examples of generated samples.

Scientific Reports | (2024) 14:4439 | https://doi.org/10.1038/s41598-024-54494-y 5

Vol.:(0123456789)
 www.nature.com/scientificreports/

Figure 3.  Distribution of crack size in the three sub-dataset (a) absolute number, (b) density.

Deep neural networks for crack detection

In this work, the proposed models follow the encoder-decoder a­ rchitecture13,14. Different backbone networks
were tested for the encoder. Upsampling layers for the decoder were tested as well. We also developed a model
with only an encoder followed by dense layers. It is also attractive since it uses fewer trainable parameters than
all other models and achieves quite similar performance compared to other models. The proposed models are
tested on the crack detection dataset to establish a benchmark.
Figure 4 illustrates a schematic drawing of the architecture. The encoder network extracts the spatiotemporal
features from measured wave signals recorded at fixed sensor locations. We adopt the strategy first to handle
the temporal and spatial dimensions. Thus, we implement the encoder network with a 1D version of backbone
networks for extracting temporal information. Two 2D-CNN layers are followed to extract spatial information.
Besides the previously tested networks, including the 1D version of stacked CNN (like VGG) and ResNet, we
also tested a 1D version of DenseNet. The application of the 2D-convolutional layer for spatial information relies
on the regular distribution of sensors. The decoder network reconstructs the probability of crack existence for
the whole material domain. The primary operator in a decoder network is upsampling. The ideas of learnable
upsampling layers come from computer vision tasks, such as image generation and segmentation. These layers

Figure 4.  Schema of deep crack detection model.

Scientific Reports | (2024) 14:4439 | https://doi.org/10.1038/s41598-024-54494-y 6

Vol:.(1234567890)
www.nature.com/scientificreports/

include transpose convolution, resize-convolution, and pixel-shuffle. Additionally, we also discussed the pos-
sibility of making predictions without a decoder.

The encoder network

The encoder network uses the 1D version of backbone networks for extracting temporal information. In Fig. 5,
we show three basic blocks of the three backbone nets used for the encoder. The VGG-style network consists of
stacked convolution layers followed by pooling layers to reduce the size of the temporal dimension. The ResNet
and DenseNet employ skip connections to preserve the information from input data. Each backbone network
consists of several such blocks accordingly. We want to point out that the 1D version of these networks can be
implemented with 2D CNN, where the filter size is set to 1 along the sensory dimension to handle multiple
sensors.
The VGG-style network consists of multiple convolutional layers. VGG-net proves that deeply stacked neural
network layers can perform various computer vision t­ asks51. The previous study showed that such a stacked
1D-CNN could effectively extract spatial-temporal patterns from simulated wave s­ ignals13.
The following work showed that an advanced network, i.e. ResNet, facilitated better performance and reduced
the required computational resource on the same dataset used for training the model with 1D-VGG-net14.
ResNets are characterised by their skip connections to enable identity mapping and residual l­earning52. Using
skip connections enables CNNs to be substantially deeper, more accurate, and more efficient to train. The residual
block (ResBlock), consisting of two stacked convolution layers with non-linearity each, performs a mapping
from input x to learned features F(x). Adding a skip connection changes the output feature to the following form,
Hl (x) = Fl (x) + x, (9)
where Fl is the transformation of the residual block, and Hl (x) is the corresponding output containing the
l th
transformed features and the identity of the input features.
DenseNet extends the idea of using skip connections by connecting each layer to every other layer in a
feed-forward fashion. DenseNets has shown impressive results in many computer vision tasks, including image
classification, semantic segmentation, and object detection. They have also been used in other domains, such as
natural language processing and reinforcement l­earning59–63. We have modified the DenseNet with 1D convo-
lutional layers, which are responsible for extracting the temporal features from wave signals, thus reducing the
time dimension completely.
The batch normalization and activation functions are essential for these 1D version CNNs. Batch Normali-
zation normalizes the prior layer outputs to a standard distribution, such that the distribution of output values
remains while the weights of a layer are u­ pdated64. By reducing internal covariate shifts inside each batch, batch
normalization stabilizes the training process and helps to accelerate the convergence of deep models.
The non-linearity Layer is also known as the activation layer. It performs a non-linear transformation on its
input features. Many such functions exist, and the most relevant ones are discussed in this paper. Previous work
showed that the Rectified Linear Unit (ReLU) could not work well in the backbone networks. It has a 0 gradient
for negative input, i.e., the dying ReLu problem. LeakyReLU applies a parameter 0 < α < 1 for its negative input
to guarantee a small gradient. It has been used for the 1D-VGG in previous work and performs well. In this work,
we use a Scaled Exponential Linear Unit (SELU) that scales the Exponential Linear Unit (ELU) output with a
scaling parameter 65. ELU modifies ReLU by changing the constant 0 output for negative values to a weighted
exponential ­function66. SELU is a self-normalizing activation function, meaning that it can maintain the mean
and variance of activations close to a fixed point, which allows deep neural networks to be trained effectively.
Figure 6 shows a complete model implementation focusing on encoder that is mainly relied on 1D-DensNet.
Four 1D Dense blocks are stacked to process the wave signals in the time domain. Then, the temporal features
are reshaped according to their sensor locations and passed to two 2D convolution layers to learn spatial infor-
mation, yielding a maximally condensed feature representation of 8 × 8 spatial extent. Finally, the spatially
rearranged features are passed to the decoder (marked in the green box) to predict crack existence at the desired
spatial resolution.

Figure 5.  Tested backbone nets, (a) simply-stacked 1DCNN, (b) 1D-ResNet, (c) 1D-DenseNet.

Scientific Reports | (2024) 14:4439 | https://doi.org/10.1038/s41598-024-54494-y 7

Vol.:(0123456789)
 www.nature.com/scientificreports/

Figure 6.  The crack detection network implemented with 1D-DenseCDNet as encoder.

The decoder network

The decoder network consists of stacked upsampling layers to scale the input features to the desired spatial
resolution. In this work, three types of trainable upsampling methods are evaluated. They are transposed convo-
lution, resize and convolution, and pixel shuffle. The three upsampling methods are described in detail and also
depicted in Fig. 4. Transposed convolution is also known as fractionally-stridden convolution or mistakenly as
deconvolution. It rises from the need to use a transformation going in the opposite direction of a normal convo-
lution, i.e., from the output space of convolution to its input space, while maintaining the connectivity pattern
of the c­ onvolution67. The transposed convolution works by adding padding to expand the spatial dimension of
input to compete with the spatial dimension of output and then performing convolution on the expended input.
In transposed convolution, padding and stride are applied to output instead of input. Choosing stride equals
2; transposed convolution resizes the input with an up-scale factor of 2. Transposed convolutions suffer from
the well-known Checkerboard Effect, i.e., the artefacts caused by uneven overlap at some parts of the features.
Pixel shuffle is an upsampling method used in super-resolution models to implement efficient sub-pixel
convolutions with a stride of 1/r68. It rearranges elements in a tensor of shape (H, W, C × r 2 ) to a tensor of shape
(H × r, W × r, C). In detail, this operation treats the elements in the channel as sub-pixels and places these sub-
pixels alternately in corresponding spatial locations. Usually, we need convolutional layer(s) before the pixel
shuffle operation to transform the features into the desired number of channels.
Resize & convolution is proposed as an alternative to transpose convolution. The resizing layer upsamples
the input feature map. The convolutional layer analyses the extracted features and prepares the spatial output. It
is advantageous to avoid checkerboard a­ rtefacts69. Resize &convolution block consists of an interpolation layer
(e.g., nearest-neighbour interpolation or bilinear interpolation) followed by a convolutional layer.
The encoder extracts the features of temporal and spatial dimensions and produces a feature map in the shape
of (4,4,64). It is then forwarded to the decoder consisting of two upsampling blocks of the same type. Each of
these blocks expands the spatial dimension of the input feature map. The first block is responsible for bringing
the data into the shape of (8,8,64). The second block performs similar upsampling but produces an output in the
shape of (16,16,16). The final convolutional layer produces the prediction in the shape of (16,16,1) with sigmoid
activation. This is important because the output is expected to be between 0 and 1 as a possible crack existence.

Benchmarking experiments
This section presents the benchmarking results for the neural network models on the crack detection task i.e. to
predict the probability of crack existence in a sample. Firstly, we describe the loss function, metrics, and hyper-
parameters we employed for our experiments.

The loss functions

The loss function measures the difference between the expected output and the actual output a neural network
gives. It defines an optimization target for training the neural network. In image segmentation, loss functions
are classified into distribution-based, region-based, boundary-based, and compound losses, i.e., a combination
of several losses. Previous works employ focal loss, which belongs to distribution-based ­losses70. It is a variant of
the binary cross-entropy loss that addresses the issue of class imbalance, where the target class consists of only
a tiny portion of the image domain. It focuses more on the problematic cases and less consideration of the easy
cases. Though proven powerful in previous studies, focal loss introduces two hyperparameters, which require
unavoidable intensive experiments to find the optimal combination of the two parameters.
In this work, we use Dice loss. Dice loss is a region-based loss function, which is based on the concept of the
Sørensen-Dice ­index71. The Sørensen-Dice index, known as the Dice similarity coefficient (DSC), is the most
commonly used metric for evaluating accuracy in classification problems. The DSC is defined as the per voxel
classification of true positives (TP), false positives (FP), and false negatives (FN):
2TP
DSC = . (10)
2TP + FP + FN
Then, the Dice loss is defined as
LDSC = 1 − DSC. (11)

Scientific Reports | (2024) 14:4439 | https://doi.org/10.1038/s41598-024-54494-y 8

Vol:.(1234567890)
www.nature.com/scientificreports/

The Dice loss can handle class imbalance (to some degree) while keeping a simple form. The DSC value is always
between 0 and 1, while its gradient is inherently unstable with highly class-imbalanced data where gradient
calculations involve small denominators. We adopt the following form of Dice loss to reduce the risk of unstable,
2I + ε
LDSC = 1 − , (12)
U+ε
N N
where I = 1 yi pi , U = 1 (yi + pi ), and ǫ is the smooth term.
We noticed that there are other possible loss functions, including Tversky loss, Focal Tversky loss, Combo loss,
and so on. The Tversky index is modified from DSC by re-weighting the false positives and false negatives to
address the imbalance issue. Focal Tversky loss modifies the Tversky loss by adding the focal parameters. The
Combo loss is an example of compound loss, which sums up the Dice and cross-entropy loss. Since a review of
loss functions is beyond the scope of this paper, we refer to the detailed review a­ rticle72.

Metrics
We follow the metrics often used in semantic segmentation to evaluate the models’ performance. Precision is the
portion of correctly identified damage (TP) among all predicted damage (TP+FP), while recall is the portion of
correctly identified damage (TP) among all actual damages (TP+FN). The crack detection task is a significant
class imbalance problem, i.e., many more undamaged regions than damaged regions in each sample. Precision
and recall are inappropriate metrics if we consider only damaged and undamaged classes since the model can
trivially achieve high performance when most (or all) pixels are labelled as undamaged. In this case, we must
use them to quantify a model’s performance. The harmonic mean of precision and recall (Eq. 13) is called the
F-1 score, which is equivalent to the Dice coefficient (DSC).
2 · precision · recall
F= (13)
precision + recall
Besides DSC, Intersection over Union (IoU) metric is another commonly used metric for semantic segmenta-
tion tasks. The IoU metric is closely related to the Dice coefficient. It provides a geometric interpretation for the
performance by quantifying the percentage overlap between the true damaged area and the predicted damaged
area. The IoU metric is computed by the following equation (Eq. 14). Compared to the Dice coefficient, IoU is
more sensitive to the change of FP and FN, while both metrics agree on the evaluation. If one model performs
well under one metric, it is also evaluated as working well under another.
TP
IoU = . (14)
TP + FP + FN
The correctness is defined based on the IoU value as its predicted IoU is more significant than a given threshold.
The output of the models can be interpreted as confidence (probability) of crack existence. Given the threshold,
this soft prediction can be transformed into a binary one. The accuracy of the whole dataset is calculated as the
ratio of the number of correctly predicted cases to the total number of evaluated datasets.

Settings and hyperparameters

The Adam optimizer is a commonly used algorithm for training deep neural ­networks73. We employed it in our
experiments with its default learning rate of 0.001 in all experiments. Each training lasts for 200 epochs, and we
chose the best performing model w.r.t. DSC on the validation data.
Dropout is a widely used regularization technique to prevent overfitting in deep neural n ­ etworks74,75. Its main
idea is to drop “neurons” randomly and their connections during training to prevent complex co-adaptations
on training data. We randomly “drop” all features from the same sensor, resulting in an “average” of models
generated by the random drop strategy. In experiments, we set the dropout rate to 0.8, requiring ca. 65 out of 81
sensors not to be used during training.

Results
We adopt the same activation function (SELU) and dropout rate (0.8) across all models to compare different
encoders and decoders. This can be particularly useful when isolating the impact of the architectural differences
between models. We modified the previous models by switching LeakyReLU to SELU, where 1D-CNN and
1D-ResNet are used as the encoder and trained on the new dataset. The 1D-DenseNet encoder with different
decoders and without a decoder are also developed and tested.
First, it’s necessary to explain that we have binarized predictions at the pixel level. The chosen threshold
(binarising threshold) is 0.5, meaning that a prediction value higher than 0.5 for each pixel indicates the possible
presence of a crack. Whether the prediction of a crack is correct is defined through the IoU value. This means
accurately predicted crack pixels should exceed a certain threshold for the total number of crack pixels. We have
selected an IoU threshold 0.5 as the accuracy threshold (or IoU threshold).
To illustrate the rationale behind the threshold selection, let’s take the model with 1D-ResNet as the encoder
as an example. In Fig. 7, we illustrate the relationship between different thresholds and the final accuracy. If we
demand that every pixel be accurately classified to represent correct crack identification (IoU→1), the accuracy
will drop sharply. Though this could ensure absolute accuracy, it’s challenging to achieve in practice. Considering
a very low threshold (IoU→0), even if a very small number of damaged pixels are identified, it could be consid-
ered a correct crack identification. However, this ignores many falsely identified pixels, which is also impractical.

Scientific Reports | (2024) 14:4439 | https://doi.org/10.1038/s41598-024-54494-y 9

Vol.:(0123456789)
 www.nature.com/scientificreports/

Figure 7.  Accuracy influenced by the choices of IoU threshold and binarising threshold (taken from the
evaluation of model 1D-ResNet-Resize &Conv ).

We found that around the threshold of 0.5, we can ensure a certain number of correctly identified pixels while
maintaining a reasonable accuracy level. Figure 8 further illustrates the impact of different IoU thresholds and
different binarizing thresholds on accuracy when the binarizing threshold is set to 0.5 and when the IoU threshold
is set to 0.5. Regarding the binarizing threshold, accuracy is minimally affected except for extreme values (close
to 0 or 1). This indicates that the model’s pixel-level predictions have a high level of confidence.
Table 1 presents all metrics on the selected model, including IoU, Dice Similarity Coefficient (DSC), Preci-
sion, Recall, and accuracy. Accuracy is calculated based on whether each damaged region (pixel) is predicted
accurately above 50%. IoU, DSC, Precision, and Recall are calculated by assessing whether each region (pixel)
is accurately classified (has a crack or no crack). Networks based on encoder-decoder architecture perform bet-
ter in crack detection. For the decoder network, whether using a VGG-like CNN or a more complex network,
such as ResNet and DenseNet, relevant features associated with cracks can be extracted from wave field signals.

0.863125 0.8625
0.8575
0.83375 0.826875
0.823125
0.850625 0.831875

0.8375 0.82125
Accuracy

Accuracy

0.826875
0.7925 0.83 0.826875
0.713125 0.80625 0.828125

0.5775 0.83375
0.753125
0.33875
0.24375

IoU Threshold Binarising Threshold

Figure 8.  Accuracy influenced by (a) IoU threshold when binarising threshold is 0.5, and (b) binarising
threshold when IoU threshold is 0.5 (taken from the evaluation of model 1D-ResNet-Resize &Conv ).

Scientific Reports | (2024) 14:4439 | https://doi.org/10.1038/s41598-024-54494-y 10

Vol:.(1234567890)
www.nature.com/scientificreports/

Accu IoU DSC Prec Recall

1D-CNN-Resize &Conv 0.81875 0.756758 0.861539 0.869824 0.853411
1D-ResNet-Resize &Conv 0.826875 0.772105 0.871399 0.883311 0.859803
1D-DenseNet-no decoder 0.803125 0.717579 0.83557 0.83317 0.837984
1D-DenseNet-TConv 0.836875 0.76014 0.863727 0.875395 0.852366
1D-DenseNet-Subpixel 0.83625 0.757815 0.862224 0.880307 0.844868
1D-DenseNet-Resize &Conv 0.834375 0.768578 0.869148 0.881086 0.857529

Table 1.  Evaluated metrics for the benchmarking models. The best values are in bold.

More advanced networks like DenseNet exhibit superior performance regarding crack prediction accuracy,
while ResNet performs better in pixel-wise predictions. When employing different decoder networks, conven-
tional Transpose Convolution yields satisfactory results. However, Resize & Convolution demonstrate advantages
in pixel-level predictions. We speculate that this is due to the limited impact of the chessboard effect on the whole
crack detection but a larger influence on pixel-level predictions.
It’s worth noting that even though we included a model with an upsampling decoder, it can be understood
as employing fully connected layers for decoding. While this approach omits an upsampling decoder network
and still achieves around 80% accuracy, its performance remains weaker than the upsampling version. Another
drawback of using fully connected layers for decoding is the challenge of expanding spatial resolution. For predic-
tions at higher resolutions, such as 32×32 or 64×64, the parameters that must be trained for the fully connected
layers will grow exponentially.
In general, using a more powerful encoder network can better extract features from wave field signals relevant
to cracks. However, the selection of the encoder network regarding the scale and complexity still requires careful
consideration. Larger networks have more trainable parameters, which demand more data and robust training
methods. Additionally, larger networks have the risk of overfitting when training data is limited. On the other
hand, by comparing the performance of models on cracks of different sizes, we found that predicting smaller
cracks is more challenging than larger ones. In Fig. 9, the accuracies of the models trained with the unnormal-
ized dataset are shown. To show the influence of crack size to the performance, the results for the whole test
set as well as for the adjusted test set of excluding smaller cracks are given. For larger cracks, the differences in
the performance (e.g., accuracy) among different models (including various encoder networks and no decoder
versions) are minimal. In contrast, for smaller cracks, more advanced networks tend to produce better results
for predicting cracks. We attribute this to the fact that the wavefield changes caused by smaller cracks are very
insignificant, making them difficult to capture by sensors and more susceptible to irregularities in particle shapes
and distributions within the material and measurement noise. This consideration also prompts us to contemplate
amplifying such wavelength variations in the data.
The evaluation of training the models with the normalization method is shown in Table 2. It presents all
metrics on the selected model trained using a reference wave field. The reference wave field is an average wave
field from different intact samples. We trained each encoder network with five models and took the median value
of their metrics. We have discovered that training with reference wavefields significantly improves the models
employing different encoder networks. Moreover, more advanced encoder networks, such as 1D-DenseNet,
appear to benefit more from this approach. Figure 10 illustrates the accuracies of the models trained with nor-
malized dataset over different crack sizes. The model of 1D-DenseNet with resize and convolutional upsampling
method is chosen as the reference model (marked as “*” in the Figure). It is observable that the 1D-CNN model
with the Resize & convolution method trained with normalized data out performs the model of 1D-DenseNet
and the same upsampling method trained with unnormalized data. It even outperforms other models trained
without using the reference wavefield, showing the advantage of using the normalization method. We can aslo
observe improved accuracy when we exclude the test samples with smaller cracks. This agrees with the previ-
ous finding that smaller cracks are more difficult to detect. The most important is using a reference wavefield in
normalization improves the performance largely on the samples with smaller cracks.

Accu IoU DSC Prec Recall

1D-CNN-Resize &Conv 0.8675 0.760874 0.8642 0.836427 0.893882
1D-ResNet-Resize &Conv 0.884375 0.767203 0.868268 0.841303 0.897019
1D-DenseNet-Resize &Conv 0.88625 0.7856 0.879929 0.877088 0.882788

Table 2.  Evaluated metrics for the normalization method with reference wave field. The best values are in
bold.

Scientific Reports | (2024) 14:4439 | https://doi.org/10.1038/s41598-024-54494-y 11

Vol.:(0123456789)
 www.nature.com/scientificreports/

0.95

Accuracy
0.9

0.85

0.8
≥0 >0.001 >0.002 >0.003 >0.004
crack size

1D-CNN-Resize&Conv 1D-ResNet-Resize&Conv 1D-DenseNet-no decoder

1D-DenseNet-Subpixel 1D-DenseNet-TConv 1D-DenseNet-Resize&Conv

Figure 9.  The prediction accuracy of tested models evaluated across various crack size groups.

0.95
(using reference wave field)
Accuracy

0.9

0.85

0.8
≥0 >0.001 >0.002 >0.003 >0.004
crack size

1D-DenseNet-Resize&Conv* 1D-CNN-Resize&Conv
1D-ResNet-Resize&Conv 1D-DenseNet-Resize&Conv
* trained with unnormalized data (not using the reference wave field)

Figure 10.  The prediction accuracies assessed across different crack size groups for the models trained with
with reference wavefield.

Conclusion and future work

Numerical simulation is a cost-efficient way of creating vast data for training deep neural networks. It is a flex-
ible way of producing data in different damage scenarios, rarely seen in a real-world structure. This work shows
an upgraded dataset for the crack detection problem generated by running many numerical simulations. We
noticed difficulties detecting small cracks, so we kept a similar distribution of crack sizes in the training, valida-
tion, and testing datasets.

Scientific Reports | (2024) 14:4439 | https://doi.org/10.1038/s41598-024-54494-y 12

Vol:.(1234567890)
www.nature.com/scientificreports/

We defined crack detection for plates as a semantic segmentation problem. We proposed a general neural
network architecture employing an encoder and a decoder. The encoder and decoder’s backbone networks can be
implemented as well-developed neural networks, such as ResNet and DenseNet for the encoder and transposed
convolution for the decoder. The backbone of encoder networks is modified to the 1D version for measured
wave signal data. The decoder consists of stacked upsampling layers to scale the spatial-temporal feature map to
a desired spatial resolution. Though it is possible to predict cracks at a coarse spatial resolution, upsampling lay-
ers facilitate prediction at a finer resolution. We generated a dataset to benchmark the variations of architecture
and identified the main influence factors of the model’s performance. Variations of the model’s components have
been tested, including different encoder backbone networks (VGG, ResNet, DenseNet) and upsampling blocks
(transposed convolution block, sub-pixel block, and resize-convolution block). We identified that normalizing
the input data using reference wave signals significantly increases the model’s performance for small cracks. We
point out that this method inherently scales the modified wave signals of small cracks. The paper reaches its
limits in detecting extremely small microcracks caused by two main challenges. First, the model is designed for
cracks of different sizes, which makes precise control harder and is not specific to a single crack group or size.
Secondly, the weak signals associated with small cracks could challenge the model in this crack category, as the
waves can be easily disturbed in the case of more minor cracks. A possible solution to this problem could be to
adjust the data processing or modify the complexity of the model, perhaps by combining multiple models to
consider different crack groups.
Future work could include a two-fold improvement of the model’s performance by applying more powerful
components and focusing on small cracks. More neural networks would be tested for spatial-temporal feature
extraction. For example, the network that handles sequential data, such as RNNs, LSTM, and attention blocks,
can extract temporal patterns. It is also possible to fine-tune a well-trained model on a dataset that consists of
small cracks. We also noted that our proposed method relies on predefined sample dimensions and shapes.
Future extension of our method from these two aspects is needed, i.e., extending from 2D plate to 3D samples
and considering different sample shapes. The ultimate goal of our research is to provide knowledge that helps to
build a real-time, cost-efficient diagnosing system for civil structures.

Data availability
The datasets generated during and/or analysed during the current study are available from the corresponding
author on reasonable request.

Received: 25 August 2023; Accepted: 13 February 2024

References
1. Rytter, A. Vibrational based inspection of civil engineering structures. Ph.D. thesis, University of Aalborg, Denmark. Ph.D.-Thesis
defended publicly at the University of Aalborg, April 20, 1993 PDF for print: 206 pp (1993).
2. Malekloo, A., Ozer, E., AlHamaydeh, M. & Girolami, M. Machine learning and structural health monitoring overview with emerg-
ing technology and high-dimensional data source highlights. Struct. Health Monit.https://​doi.​org/​10.​1177/​14759​21721​10368​80
(2022).
3. Bao, Y. & Li, H. Machine learning paradigm for structural health monitoring. Struct. Health Monit. 20, 1353–1372. https://​doi.​
org/​10.​1177/​14759​21720​972416 (2021).
4. LeCun, Y., Bengio, Y. & Hinton, G. Deep learning. Nature 521, 436–444. https://​doi.​org/​10.​1038/​natur​e14539 (2015).
5. Zhang, L., Yang, F., Daniel Zhang, Y. & Zhu, Y. J. Road crack detection using deep convolutional neural network. In: 2016 IEEE
International Conference on Image Processing (ICIP), 3708–3712, https://​doi.​org/​10.​1109/​ICIP.​2016.​75330​52 (2016).
6. Dung, C. V. & Anh, L. D. Autonomous concrete crack detection using deep fully convolutional neural network. Autom. Constr.
99, 52–58. https://​doi.​org/​10.​1016/j.​autcon.​2018.​11.​028 (2019).
7. Deng, J., Lu, Y. & Lee, V.C.-S. Imaging-based crack detection on concrete surfaces using you only look once network. Struct. Health
Monit. 20, 484–499. https://​doi.​org/​10.​1177/​14759​21720​938486 (2021).
8. Abdeljaber, O. et al. 1-d CNNs for structural damage detection: Verification on a structural health monitoring benchmark data.
Neurocomputing 275, 1308–1317. https://​doi.​org/​10.​1016/j.​neucom.​2017.​09.​069 (2018).
9. Rautela, M. & Gopalakrishnan, S. Ultrasonic guided wave based structural damage detection and localization using model assisted
convolutional and recurrent neural networks. Expert Syst. Appl. 167, 114189. https://​doi.​org/​10.​1016/j.​eswa.​2020.​114189 (2021).
10. Hung, D. V., Hung, H. M., Anh, P. H. & Thang, N. T. Structural damage detection using hybrid deep learning algorithm.. J. Sci.
Technol. Civ. Eng. 14, 53–64. https://​doi.​org/​10.​31814/​stce.​nuce2​020-​14(2)-​05 (2020).
11. Tang, Z., Chen, Z., Bao, Y. & Li, H. Convolutional neural network-based data anomaly detection method using multiple informa-
tion for structural health monitoring. Struct. Control. Health Monit. 26, e2296. https://​doi.​org/​10.​1002/​stc.​2296 (2019).
12. Khodabandehlou, H., Pekcan, G. & Fadali, M. S. Vibration-based structural condition assessment using convolution neural net-
works. Struct. Control. Health Monit. 26, e2308. https://​doi.​org/​10.​1002/​stc.​2308 (2019).
13. Wuttke, F., Lyu, H., Sattari, A. S. & Rizvi, Z. H. Wave based damage detection in solid structures using spatially asymmetric
encoder-decoder network. Sci. Rep. 11, 20968. https://​doi.​org/​10.​1038/​s41598-​021-​00326-2 (2021).
14. Moreh, F. et al. Crack detection and localization based on spatio-temporal data using residual networks. in SSDBM ’22: Proceedings
of the 34th International Conference on Scientific and Statistical Database Management, https://​doi.​org/​10.​1145/​35387​12.​35387​43
(2022).
15. Simonyan, K. & Zisserman, A. Very deep convolutional networks for large-scale image recognition. Preprint at arXiv:​1409.​1556
(2014).
16. He, K., Zhang, X., Ren, S. & Sun, J. Deep residual learning for image recognition. in Proceedings of the IEEE Conference on Computer
Vision and Pattern Recognition (CVPR) (2016).
17. Yuan, S., Wang, L. & Peng, G. Neural network method based on a new damage signature for structural health monitoring. Thin-
Walled Struct. 43, 553–563. https://​doi.​org/​10.​1016/j.​tws.​2004.​10.​003 (2005).
18. Golato, A., Santhanam, S., Ahmad, F. & Amin, M. G. Extended defect localization in sparsity-based guided wave structural health
monitoring. in 2017 IEEE 7th International Workshop on Computational Advances in Multi-Sensor Adaptive Processing (CAMSAP),
1–4, https://​doi.​org/​10.​1109/​CAMSAP.​2017.​83131​84 (2017).

Scientific Reports | (2024) 14:4439 | https://doi.org/10.1038/s41598-024-54494-y 13

Vol.:(0123456789)
 www.nature.com/scientificreports/

19. Gravenkamp, H., Saputra, A. A., Song, C. & Prager, J. Detection of defects in thin-walled structures by means of lamb waves. AIP
Conf. Proc. 1433, 443–446. https://​doi.​org/​10.​1063/1.​37032​23 (2012).
20. Kudela, P., Radzieński, M. & Ostachowicz, W. Identification of cracks in thin-walled structures by means of wavenumber filtering.
Mech. Syst. Signal Process. 50–51, 456–466. https://​doi.​org/​10.​1016/j.​ymssp.​2014.​05.​041 (2015).
21. Ono, K. Review on structural health evaluation with acoustic emission. Appl. Sci.https://​doi.​org/​10.​3390/​app80​60958 (2018).
22. Farhidzadeh, A., Salamone, S., Luna, B. & Whittaker, A. Acoustic emission monitoring of a reinforced concrete shear wall by
b-value-based outlier analysis. Struct. Health Monit. 12, 3–13. https://​doi.​org/​10.​1177/​14759​21712​461162 (2013).
23. Van Driessche, A., Aggelis, D. G. & Tsangouri, E. Complex fracture on thin-wall textile reinforced cement (TRC) shells monitored
by acoustic emission. Thin-Walled Struct. 167, 108216. https://​doi.​org/​10.​1016/j.​tws.​2021.​108216 (2021).
24. Argus, P., Gurka, M. & Kelkel, B. Development of a small-scale and low-cost SHM system for thin-walled CFRP structures based
on acoustic emission analysis and neural networks. In Nondestructive Characterization and Monitoring of Advanced Materials,
Aerospace, Civil Infrastructure, and Transportation XIII (ed. Gyekenyesi, A. L.) (SPIE, 2019).
25. Kim, B., Yuvaraj, N., Sri Preethaa, K. R. & Arun Pandian, R. Surface crack detection using deep learning with shallow CNN archi-
tecture for enhanced computation. Neural Comput. Appl. 33, 9289–9305. https://​doi.​org/​10.​1007/​s00521-​021-​05690-8 (2021).
26. Li, S. & Zhao, X. Convolutional neural networks-based crack detection for real concrete surface. In Sensors and Smart Structures
Technologies for Civil, Mechanical, and Aerospace Systems 2018 (ed. Sohn, H.) (SPIE, 2018).
27. Zhang, Y., Huang, J. & Cai, F. On bridge surface crack detection based on an improved yolo v3 algorithm. IFAC-PapersOnLine 53,
8205–8210. https://​doi.​org/​10.​1016/j.​ifacol.​2020.​12.​1994 (2020).
28. Zhang, C., Chang, C.-C. & Jamshidi, M. Concrete bridge surface damage detection using a single-stage detector. Comput. Aided
Civ. Infrastruct. Eng. 35, 389–409. https://​doi.​org/​10.​1111/​mice.​12500 (2020).
29. Li, Y., Li, H. & Wang, H. Pixel-wise crack detection using deep local pattern predictor for robot application. Sensorshttps://​doi.​
org/​10.​3390/​s1809​3042 (2018).
30. Deng, W. et al. Vision based pixel-level bridge structural damage detection using a link ASPP network. Autom. Constr. 110, 102973.
https://​doi.​org/​10.​1016/j.​autcon.​2019.​102973 (2020).
31. Qiao, W., Ma, B., Liu, Q., Wu, X. & Li, G. Computer vision-based bridge damage detection using deep convolutional networks
with expectation maximum attention module. Sensorshttps://​doi.​org/​10.​3390/​s2103​0824 (2021).
32. Pan, Y., Zhang, G. & Zhang, L. A spatial-channel hierarchical deep learning network for pixel-level automated crack detection.
Autom. Constr. 119, 103357. https://​doi.​org/​10.​1016/j.​autcon.​2020.​103357 (2020).
33. Lee, D., Kim, J. & Lee, D. Robust concrete crack detection using deep learning-based semantic segmentation. Int. J. Aeronaut. Space
Sci. 20, 287–299 (2019).
34. Huyan, J., Li, W., Tighe, S., Xu, Z. & Zhai, J. Cracku-net: A novel deep convolutional neural network for pixelwise pavement crack
detection. Struct. Control. Health Monit. 27, e2551. https://​doi.​org/​10.​1002/​stc.​2551 (2020).
35. Bang, S., Park, S., Kim, H. & Kim, H. Encoder-decoder network for pixel-level road crack detection in black-box images. Comput.
Aided Civ. Infrastruct. Eng. 34, 713–727 (2019).
36. Protopapadakis, E., Voulodimos, A., Doulamis, A., Doulamis, N. & Stathaki, T. Automatic crack detection for tunnel inspection
using deep learning and heuristic image post-processing. Appl. Intell. 49, 2793–2806 (2019).
37. Xu, H. et al. Automatic bridge crack detection using a convolutional neural network. Appl. Sci. 9, 2867 (2019).
38. Doebling, S. W., Farrar, C. R. & Prime, M. B. A summary review of vibration-based damage identification methods. Shock Vibr.
Dig. 30(2), 91–105 (1998).
39. Fan, G., Li, J. & Hao, H. Vibration signal denoising for structural health monitoring by residual convolutional neural networks.
Measurement 157, 107651. https://​doi.​org/​10.​1016/j.​measu​rement.​2020.​107651 (2020).
40. Hakim, S., Abdul Razak, H. & Ravanfar, S. Fault diagnosis on beam-like structures from modal parameters using artificial neural
networks.. Measurement 76, 45–61. https://​doi.​org/​10.​1016/j.​measu​rement.​2015.​08.​021 (2015).
41. Abdeljaber, O., Avci, O., Kiranyaz, S., Gabbouj, M. & Inman, D. J. Real-time vibration-based structural damage detection using
one-dimensional convolutional neural networks. J. Sound Vib. 388, 154–170. https://​doi.​org/​10.​1016/j.​jsv.​2016.​10.​043 (2017).
42. Lin, Y.-Z., Nie, Z.-H. & Ma, H.-W. Structural damage detection with automatic feature-extraction through deep learning. Comput.
Aided Civ. Infrastruct. Eng. 32, 1025–1046. https://​doi.​org/​10.​1111/​mice.​12313 (2017).
43. Wang, Z. & Jin Cha, Y. Automated damage-sensitive feature extraction using unsupervised convolutional neural networks. In
Sensors and Smart Structures Technologies for Civil, Mechanical, and Aerospace Systems 2018 (ed. Sohn, H.) (SPIE, 2018).
44. Goh, L. D., Bakhary, N., Rahman, A. A. & Ahmad, B. H. Prediction of unmeasured mode shape using artificial neural network for
damage detection. J. Teknologihttps://​doi.​org/​10.​11113/​jt.​v61.​1624 (2013).
45. Ding, Z., Li, J. & Hao, H. Structural damage identification by sparse deep belief network using uncertain and limited data. Struct.
Control. Health Monit. 27, e2522. https://​doi.​org/​10.​1002/​stc.​2522 (2020).
46. Sajedi, S. O. & Liang, X. Vibration-based semantic damage segmentation for large-scale structural health monitoring. Comput.
Aided Civ. Infrastruct. Eng. 35, 579–596. https://​doi.​org/​10.​1111/​mice.​12523 (2020).
47. Long, J., Shelhamer, E. & Darrell, T. Fully convolutional networks for semantic segmentation. in Proceedings of the IEEE Conference
on Computer Vision and Pattern Recognition (CVPR) (2015).
48. Ronneberger, O., Fischer, P. & Brox, T. U-net: Convolutional networks for biomedical image segmentation. In Medical Image
Computing and Computer-Assisted Intervention—MICCAI 2015 (eds Navab, N. et al.) (Springer, 2015).
49. Noh, H., Hong, S. & Han, B. Learning deconvolution network for semantic segmentation. in Proceedings of the IEEE International
Conference on Computer Vision (ICCV) (2015).
50. Badrinarayanan, V., Kendall, A. & Cipolla, R. SegNet: A deep convolutional encoder-decoder architecture for image segmentation.
IEEE Trans. Pattern Anal. Mach. Intell. 39, 2481–2495. https://​doi.​org/​10.​1109/​TPAMI.​2016.​26446​15 (2017).
51. Simonyan, K. & Zisserman, A. Very deep convolutional networks for large-scale image recognition. in 3rd International Confer-
ence on Learning Representations(eds Bengio, Y. & LeCun, Y) ICLR 2015, San Diego, CA, USA, May 7-9, 2015, Conference Track
Proceedings (2015).
52. He, K., Zhang, X., Ren, S. & Sun, J. Deep residual learning for image recognition. in 2016 IEEE Conference on Computer Vision
and Pattern Recognition (CVPR), 770–778, https://​doi.​org/​10.​1109/​CVPR.​2016.​90 (2016).
53. Dorafshan, S., Thomas, R. J. & Maguire, M. Fatigue crack detection using unmanned aerial systems in fracture critical inspection
of steel bridges. J. Bridg. Eng. 23, 04018078. https://​doi.​org/​10.​1061/​(ASCE)​BE.​1943-​5592.​00012​91 (2018).
54. Li, S. & Zhao, X. Automatic crack detection and measurement of concrete structure using convolutional encoder-decoder network.
IEEE Access 8, 134602–134618. https://​doi.​org/​10.​1109/​ACCESS.​2020.​30111​06 (2020).
55. Wiggenhauser, H., Köpp, C., Timofeev, J. & Azari, H. Controlled creating of cracks in concrete for non-destructive testing. J.
Nondestr. Eval. 37, 67. https://​doi.​org/​10.​1007/​s10921-​018-​0517-x (2018).
56. Rizvi, Z. H., Wuttke, F. & Sattari, A. S. Dynamic analysis by lattice element method simulation. In Wu, W. & Yu, H.-S. (eds.) Pro-
ceedings of China-Europe Conference on Geotechnical Engineering, 405–409 (Springer International Publishing, Cham, 2018).
57. Rizvi, Z. H. et al. Dynamic lattice element modelling of cemented geomaterials. In Advances in Computer Methods and Geomechan-
ics (eds Prashant, A. et al.) (Springer Singapore, 2020).
58. Sattari, A. S., Rizvi, Z. H., Aji, H. D. B. & Wuttke, F. Study of wave propagation in discontinuous and heterogeneous media with
the dynamic lattice method. Sci. Rep. 12, 6343. https://​doi.​org/​10.​1038/​s41598-​022-​10381-y (2022).

Scientific Reports | (2024) 14:4439 | https://doi.org/10.1038/s41598-024-54494-y 14

Vol:.(1234567890)
www.nature.com/scientificreports/

59. Huang, G., Liu, Z., van der Maaten, L. & Weinberger, K. Q. Densely connected convolutional networks. in:Proceedings of the IEEE
Conference on Computer Vision and Pattern Recognition (CVPR) (2017).
60. Zhu, X., Dai, J., Ye, Y., Jia, J. & Su, H. Pad-net: Multi-branch network for object detection. in Proceedings of the IEEE International
Conference on Computer Vision, 7223–7232 (2019).
61. Chen, L. -C., Papandreou, G., Schroff, F. & Adam, H. Rethinking atrous convolution for semantic image segmentation. Preprint
at arXiv:​1706.​05587 (2018).
62. Li, H., Qi, X. & Ji, X. Improved densenet with efficient channel attention for object recognition. IEEE Access 7, 64342–64349 (2019).
63. Liu, Y., Zhang, X., Hou, Y. & Cheng, J. Multi-view convolutional neural networks for 3d shape recognition. IEEE Trans. Multimed.
21, 1518–1530 (2019).
64. Ioffe, S. & Szegedy, C. Batch normalization: Accelerating deep network training by reducing internal covariate shift. In: Bach, F.
& Blei, D. (eds.) Proceedings of the 32nd International Conference on Machine Learning, vol. 37 of Proceedings of Machine Learning
Research, 448–456 (PMLR, Lille, France, 2015).
65. Klambauer, G., Unterthiner, T., Mayr, A. & Hochreiter, S. Self-normalizing neural networks. In Advances in Neural Information
Processing Systems (eds Guyon, I. et al.) (Curran Associates Inc., 2017).
66. Clevert, D. -A., Unterthiner, T. & Hochreiter, S. Fast and accurate deep network learning by exponential linear units (elus). in:
International Conference on Learning Representations (ICLR) (2016).
67. Dumoulin, V. & Visin, F. A guide to convolution arithmetic for deep learning. Preprint at arXiv:​ 1603.​07285 (2016).
68. Shi, W. et al. Real-time single image and video super-resolution using an efficient sub-pixel convolutional neural network. in
Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) (2016).
69. Odena, A., Dumoulin, V. & Olah, C. Deconvolution and checkerboard artifacts. Distill (2016).
70. Lin, T.-Y., Goyal, P., Girshick, R., He, K. & Dollár, P. Focal loss for dense object detection. in Proceedings of the IEEE international
conference on computer vision, 2980–2988 (2017).
71. Sudre, C. H., Li, W., Vercauteren, T., Ourselin, S. & Cardoso, M. J. Generalised dice overlap as a deep learning loss function for
highly unbalanced segmentations. in Deep learning in medical image analysis and multimodal learning for clinical decision support,
240–248 (Springer, 2017).
72. Jadon, S. A survey of loss functions for semantic segmentation. in 2020 IEEE Conference on Computational Intelligence in Bioin-
formatics and Computational Biology (CIBCB), 1–7 (IEEE, 2020).
73. Kingma, D. P. & Ba, J. Adam: A method for stochastic optimization. in 3rd International Conference on Learning Representations
(ICLR) (2015).
74. Hinton, G. E., Srivastava, N., Krizhevsky, A., Sutskever, I. & Salakhutdinov, R. R. Improving neural networks by preventing co-
adaptation of feature detectors. Preprint at arXiv:​1207.​0580 (2012).
75. Srivastava, N., Hinton, G., Krizhevsky, A., Sutskever, I. & Salakhutdinov, R. Dropout: A simple way to prevent neural networks
from overfitting. J. Mach. Learn. Res. 15, 1929–1958 (2014).

Acknowledgements
The presented work is supported by the Competence Centre for Geo-Energy (KGE) (website: https://​www.​kge.​
uni-​kiel.​de/, last access: 14 February 2021). We gratefully acknowledge the funding of KGE by the European
Union - European Regional Development Fund (EFRE). We also thank the funding of the project ’Seismic
identification of concrete structures’ funded by the Federal Ministry of Economic Affairs and Industry - BMWI
and the German Federation of Industrial Research Associations - ZIM/ AIF with funding code ZF4016813FF8.

Author contributions
F.M. developed deep learning models and conducted benchmarking experiments. H.L. and F.M. wrote the cor-
responding parts. H.L. generated the dataset using the dynamic LEM. H.L. and Z.R. wrote the Crack Detection
Dataset section. H.L. studied and wrote the normalization method using a reference wave field. F.M., Z.R.,
and H.L. analyzed the experiment results and wrote corresponding parts. F.W. guided the writing. All authors
reviewed the manuscript.

Funding
Open Access funding enabled and organized by Projekt DEAL.

Competing interests
The authors declare no competing interests.

Additional information
Correspondence and requests for materials should be addressed to H.L.
Reprints and permissions information is available at www.nature.com/reprints.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International
License, which permits use, sharing, adaptation, distribution and reproduction in any medium or
format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the
Creative Commons licence, and indicate if changes were made. The images or other third party material in this
article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the
material. If material is not included in the article’s Creative Commons licence and your intended use is not
permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from
the copyright holder. To view a copy of this licence, visit http://​creat​iveco​mmons.​org/​licen​ses/​by/4.​0/.

© The Author(s) 2024

Scientific Reports | (2024) 14:4439 | https://doi.org/10.1038/s41598-024-54494-y 15

Vol.:(0123456789)