Version of Record: https://www.sciencedirect.

com/science/article/pii/S0141029617335526
Manuscript_824f38aeb6473dfd4189789143a454f1

Emerging artificial intelligence methods in structural engineering

Hadi Salehi1, Rigoberto Burgueño1,2

1
Department of Civil and Environmental Engineering, Michigan State University, East Lansing, Michigan, USA
2
Department of Mechanical Engineering, Michigan State University, East Lansing, Michigan, USA
Email: [email protected]

Abstract

Artificial intelligence (AI) is proving to be an efficient alternative approach to classical modeling

techniques. AI refers to the branch of computer science that develops machines and software with human-
like intelligence. Compared to traditional methods, AI offers advantages to deal with problems associated
with uncertainties and is an effective aid to solve such complex problems. In addition, AI-based solutions
are good alternatives to determine engineering design parameters when testing is not possible, thus
resulting in significant savings in terms of human time and effort spent in experiments. AI is also able to
make the process of decision making faster, decrease error rates, and increase computational efficiency.
Among the different AI techniques, machine learning (ML), pattern recognition (PR), and deep learning
(DL) have recently acquired considerable attention and are establishing themselves as a new class of
intelligent methods for use in structural engineering. The objective of this review paper is to summarize
techniques concerning applications of the noted AI methods in structural engineering developed over the
last decade. First, a general introduction to AI is presented and the importance of AI in structural
engineering is described. Thereafter, a review of recent applications of ML, PR, and DL in the field is
provided, and the capability of such methods to address the restrictions of conventional models are
discussed. Further, the advantages of employing such algorithmic methods are discussed in detail. Finally,
potential research avenues and emerging trends for employing ML, PR, and DL are presented, and their
limitations are discussed.

Keywords: structural engineering, artificial intelligence, machine learning, pattern recognition, deep
learning, soft computing

1. INTRODUCTION
Civil engineering is fraught with problems that defy solution via traditional computational techniques.
However, they can often be solved by an expert with proper training. Classical artificial intelligence (AI)
has targeted this class of problems by capturing the essence of human cognition at the highest level. The
term “AI” was introduced at a workshop held in Dartmouth college in 1956 [1]. AI is a computational
method attempting to simulate human cognition capability through symbol manipulation and
symbolically structured knowledge bases to solve engineering problems that defy solution using
conventional methods. AI has been developed based on the interaction of various disciplines; namely,
computer science, information theory, cybernetics, linguistic, and neurophysiology.
Several terms referring to artificial intelligence can be found in the literature, and they need to be
identified to further elaborate on the state of the art. One of those terms is machine intelligence (MI). AI

1
© 2018 published by Elsevier. This manuscript is made available under the Elsevier user license
https://www.elsevier.com/open-access/userlicense/1.0/
and MI are almost identical terms [2,3] and are often used interchangeably. MI is often considered a
synonym of AI; yet it deals with different types of intelligent problems, e.g., clustering, classifications,
computer vision, etc. In general, MI refers to machines with human-like intelligent behavior and
reasoning, while AI refers to a machine’s ability to mimic the cognitive functions of humans to perform
tasks in a smart manner. Another important term is cognitive computing (CC), which is inspired by
human mind’s capabilities [4]. Cognitive systems are able to solve problems in a form mimicking humans
thinking and reasoning. Such systems are based on the ability of machines to measure, reason, and adapt
using learned experience[4,5]. The main characteristics of CC systems are their ability to interpret big
data, dynamic training and adaptive learning, probabilistic discovery of relevant patterns. Technically, AI
refers to computers and machines that can behave intelligently, while CC concentrates on solving the
problems using humanlike thinking. The most significant difference between AI and CC can be defined in
terms of interacting normally with humans. For any AI system, there is an agent that decides what actions
need to be taken. However, CC systems learn, reason, and interact like humans. Therefore, it can be
concluded that CC is essentially an AI agent, and as such CC is considered a sub-set of AI. Expert
systems, on the other hand, is a branch of AI. As noted, AI is defined as the ability of a machine to mimic
intelligent human behavior, seeking to use human-inspired algorithms to solve problems. Similarly, an
expert system is defined as a computer program attempting to mimic human experts to solve problems
demanding human/expert knowledge. It follows from the noted definitions that AI includes different
branches such as expert systems, machine learning, pattern recognition, fuzzy logic, etc.
In recent years, there has been a growing interest in the use of AI in all engineering domains, and it
has fueled many visions and hopes. While the civil engineering community has witnessed an extensive
growth in the use of different AI branches/methods in its diverse areas, the present study concentrates on
the AI methods that have gained significant attention over the last decade, namely machine learning
(ML), pattern recognition (PR), and deep learning (DL) with a focus on their application to the structural
engineering discipline. The scope of the review is to summarize the theoretical background of the
methods, provide a historical context on their use, summarize the latest research developments, and
discuss promising paths for future research.
The use of AI in civil engineering has been the topic of previous review articles. Adeli et al. [6]
presented a multiparadigm learning technique, where the authors demonstrated that the performance can
be notably enhanced by skillful integration of different AI branches, including neural networks, genetic
algorithms, fuzzy sets, and parallel processing. An extensive study of evolutionary computation, a branch
of AI, in the context of structural design was conducted by Kicinger et al. [7]. Lio et al. [8] carried out a
review of studies concerning the application of metaheuristics as optimization techniques to address
issues faced in the lifetime of a construction or engineering project. A survey on different AI methods
(e.g., fuzzy logic, evolutionary computation, neural networks, swarm intelligence, expert systems, etc.)
for civil engineering was conducted by Lu et al. [9]. Shahin et al. [10] studied applications of AI in
geotechnical engineering; and Saka et al. [11] conducted a survey on mathematical and metaheuristic
algorithms in design optimization of steel frame structures. Adeli et al. [12] carried out a review on
progress in the optimization of high-rise buildings; and a survey on the applications and methodologies of
the fuzzy multiple criteria decision-making techniques was conducted by Mardani et al [13].
Recently, a survey on the application of multi-criteria decision making methods for the analysis of
suspension bridges was conducted by Garcia-Segura et al. [14]; Sanchez et al. [15] presented a review on
the applications of artificial neural networks, a branch of AI, for civil infrastructure that includes

2
structural health monitoring, structural system identification, structural design and optimization, etc.; and
a comprehensive state-of-the-art overview of sustainable structural design in green buildings rating
systems and building codes was conducted by Pongiglione et al. [16]. Further, a survey on different AI
methods (e.g., artificial neural networks, Bayesian, genetic algorithms, case-based reasoning, and fuzzy
logic) for the field of fracture mechanics was performed by Khosravani et al. [17], while a literature
review of application of multi-criteria decision analysis for aging-dam management was carried out by
Mieza et al. [18]. Additionally, Sierra et al. [19] conducted a review on multi-criteria assessment of the
social sustainability of infrastructures and Zavadskas et al. [20] surveyed the state-of-the-art methods
applied to sustainable decision-making in civil engineering, construction, and building technology.
Although the noted review articles highlighted applications of AI in civil engineering
structures/infrastructure, they mainly focused on traditional techniques and do not cover recent methods,
such as PR, ML, and DL. Yet, these intelligent methods have experienced notable developments and
increased use in structural engineering during the last few years. Therefore, this review paper presents a
broad perspective of research efforts on the use of such emerging AI methods (i.e., PR, ML, and DL) in
structural engineering during the last decade. Due to space limitations, the review emphasis for each
paper was on the problem/issue being addressed, the domain and case structure being considered, and the
AI method being used. The contributions of this review paper are: 1) study and summarize techniques
concerning the applications of PR, ML, and DL in structural engineering over the last decade, 2) identify
future directions and emerging trends for employing PR, ML, and DL in structural engineering
applications, and 3) highlight current limitations of the reviewed AI methods in structural engineering.
The review paper is structured as follows. Section 2 presents the approach followed for selecting the
reviewed literature and conducting the content analysis. A general introduction to AI is presented in
Section 3, and the significance of AI in structural engineering is also described. New AI techniques
(namely ML, PR, and DL) are introduced and highlighted in Section 4, where the differences of these
techniques are elaborated. Section 5 reviews the application of such techniques in structural engineering.
Further, Section 6 identifies potential research avenues and emerging trends for using the noted AI
methods in future innovations, while highlighting the current limitations of such methods. Finally,
conclusions are provided in Section 7.

2. RESEARCH METHOD

The present study used content analysis [21] to select the reviewed literature. Content analysis is
commonly used to objectively make valid inferences according to collected data with the aim of
disclosing central aspects of previous studies. It further allows for qualitative and quantitative operations.
As a result, content analysis is able to provide an inclusive disclosure of AI applications in structural
engineering, leading to reliable results from the study.
Sample collection was performed in this study through the search and selection of peer-reviewed
articles. Articles were collected from prominent and well-accepted academic databases. The procedure of
literature search and selection for this study can be summarized as follows:
 The academic databases Web of Science, Scopus, Science Direct, ASCE Library, Engineering
Village, Wiley Online Library, Sage, and Emerald were used for article search and selection.
 Keywords such as “artificial intelligence”, “artificial intelligence in civil and structural
engineering”, “pattern recognition structural engineering”, “machine learning structural

3
 engineering”, “deep learning structural engineering”, “convolutional neural networks
structural engineering”, and “computational intelligence” were used to search the databases.
This resulted in the identification of academic articles concerning the application of AI methods
in structural engineering. The time period under review was from 2009 to 2017, which led to the
identification of approximately 430 candidate articles.
 The criteria for selecting the identified articles was the application of pattern recognition,
machine learning, and deep learning in structural engineering. In accordance with such criteria, a
two-round article selection technique was employed. That is, titles, abstract, and keywords of
the noted articles were checked in the first round to ascertain if they meet the criteria. The
second round consisted of reading and analyzing the entire article, thus ensuring that all of the
selected papers were closely related to the review objective. Finally, 284 articles were selected
and used for the present review.
For the review, qualitative and quantitative analyses were performed to identify the applications of
emerging AI methods in structural engineering, the AI algorithms used for such applications, and analyze
the applicability of these algorithms for the noted applications. This approach led to the identification of
the most promising applications of emerging AI techniques and future research directions.

3. OVERVIEW OF ARTIFICIAL INTELLIGENCE

In general, there are two types of machine intelligence: hard computing and soft computing methods.
Hard computing, which is based on binary logic, crisp systems, and numerical analysis, requires a
precisely stated analytical model and is capable of producing precise answers. Soft computing differs
from conventional computing in that, unlike hard computing, it can deal with ambiguous and noisy data,
incorporates stochastic information, and allows parallel computations. Soft computing is based on fuzzy
logic, neural nets, and probabilistic reasoning; where the methods are able to evolve their own programs
and yield approximate answers [22].
Soft computing is commonly considered a synonym of computational intelligence (CI). In fact, CI or
soft computing can be expressed by the capability of a computer to learn a specific task from sample data
or experimental observation. Mathematical or conventional modelling are useless in many complex real-
life problems due to factors such as: complexity of the processes for mathematical reasoning,
uncertainties during the process, and the stochastic nature of the process. The set of nature-inspired
computational techniques defining CI provides solutions for such problems [23]. CI uses a combination of
supplementary techniques such as artificial neural networks, fuzzy logic, learning theory, evolutionary
computing, and probabilistic methods, and is capable of solving and approximating nonlinear problems
while introducing human knowledge into the areas of computing.
Artificial intelligence (AI) is essentially defined as the ability of a machine to mimic intelligent human
behavior, thus seeking to use human-inspired algorithms for approximating conventionally defiant
problems. The main goals of AI research involve knowledge representation, reasoning, automated
planning, learning, natural language processing, perception, robotics, and general intelligence [24–28].
Although AI and CI/soft computing pursue a similar goal, there is a slight difference between them.
According to Bezdek [24], CI is a subset of AI. It is also important to distinguish AI from data science
and big data. There is indeed a substantial overlap among these methods. Data mining/science is a cross-
disciplinary field used to discover valuable insights and trends in a data set. Data mining techniques focus

4
on the discovery of unknown properties in an area where there is limited knowledge. The data set, on the
other hand, is called big data if it is big in terms of volume (i.e., number of data points or features per data
point), velocity (i.e., large portions of data arriving in a small amount of time for analysis and mining), or
variety (i.e., different types of data such as text, speech, images, etc.). Big data thus refers to large or
complex data sets that are difficult to represent using conventional data processing techniques. Machine
learning, a subfield of AI, is used to design a model to learn the trends, thus focusing on prediction based
on known properties learned from the training data. Deep learning, a subset of machine learning, is a tool
that concentrates on learning the representations and features of the data. Figure 1 schematically presents
the noted different intelligent techniques and their correlation.

Figure 1. Illustration of the interrelation of different intelligent computational techniques

In the field of structural engineering, there are numerous problems that are influenced by uncertainties,
e.g., those related to design, analysis, condition monitoring, construction management, decision making,
etc. Such problems need mathematics, physics, and mechanics calculations to be solved, and their
solution strongly depends on the practitioners’ experience. It can be further said that computers are yet to
be fully utilized for many tasks. This is essentially because of the need for logical reasoning, problems
tend to be unique, feasibility constraints, and the need to use prior experiences in the analysis and design
process. However, AI techniques can be effectively used to enhance these efforts and can also be
considered to check the general validity of laboratory or field test results. AI methods can also help
minimize (and potentially avoid) time-consuming laboratory or field tests to determine design parameters.
Uncertainties are an unavoidable part of structural engineering problems. For example, in seismic
design earthquake demands are not known with precision. In structural health monitoring there are
uncertainties in the amplitude of the input excitation, measurement noise, and spatial density of
measurements. Many uncertainties also exist in the models used to predict structural response, as well as
those defining constitutive behavior. Geotechnical information for foundation design purposes is
determined with limited information and/or based on laboratory tests with high levels of uncertainty. All
of the aforementioned problems can be modeled and treated as uncertainties [29]. AI is able to deal with
such uncertainty problems. For instance, AI methods have been used to solve uncertainty problems
defined within the context of damage detection and system identification using finite element model
updating [30]. Model updating can be used to identify physical parameters (e.g., stiffness of a structural
component) for which a reduction in value is taken to indicate damage. However, such reduction may be
simply due to statistical uncertainty. Thus, it is of importance to compute the uncertainty of the estimation
to distinguish whether the reduction of a parameter is due to actual damage. The use of AI methods can

5
also result in significant time and cost savings, as well as increasing computational efficiency in many
structural engineering tasks.
Many of the AI branches, such as machine learning (ML), pattern recognition (PR), neural networks,
fuzzy logic, evolutionary computation, deep learning (DL), expert systems, probability theory,
discriminant analysis, swarm optimization, metaheuristic optimization, and decision trees, have been used
in structural engineering. The number of research publications showing the use of these AI methods in
structural engineering over the last decade is presented in Figure 2. As can be seen, the use of most
methods has increased during the last decade. Nevertheless, the number of studies featuring techniques
such as evolutionary computation, fuzzy logic, and expert systems has not had a notable change. Even
though the use of neural networks has drawn a great attention from researchers, new studies on the use of
such method has also remained rather constant over the last decade. In contrast, the significant increase in
studies featuring the use of ML and PR is evident. Further, deep learning architectures, e.g., convolutional
neural networks (CNNs), are gaining remarkable attention among the research community over the last
few years. These observations motivated the authors to concentrate this review on ML, PR, and DL, as
they are emerging as the new computational intelligence paradigms in structural engineering.

Figure 2. Research publications on the use of different AI branches in structural engineering

6
4. EMERGING AI METHODS

As previously discussed, pattern recognition, machine learning, and deep learning are among the new
artificial intelligence methods that are increasingly emerging as reliable and efficient tools in the field of
structural engineering. This section provides technical background on the noted methods and insight
regarding the use of such algorithms for structural engineering problems.

4.1. Pattern recognition

Pattern recognition (PR) is a technique in which the main goal is to classify objects into a number of
classes or categories. The objects, depending on the applications, could be images, signals, hand writing,
speech, or measurements to be classified [31,32]. In PR, a pattern is represented by a set of features.
Concepts from statistical decision theory are used to establish decision boundaries between pattern
classes. The recognition system in PR consists of two modes, namely learning (training) and classification
(testing), as shown in Figure 3. In the learning/training mode the proper features for representing the input
patterns are discovered by means of the feature extraction/selection module, and the classifier is
trained/calibrated to partition the feature space. In the classification mode the input patterns are assigned
to one of the classes using the trained classifier; while the performance of the designed classifier, i.e.,
classification error rate, is evaluated by the system evaluation module.

Figure 3. Schematic of a pattern recognition system

In general, PR methods can be categorized into two main categories: supervised PR and unsupervised
PR. The supervised term refers to the condition when a set of labeled training samples are available.
When there is no prior information regarding the class labels and the training data are not labeled, this is
known as unsupervised PR, or clustering. These terms are further discussed in the following section.
Another difference in PR methods is that of generative models versus discriminative models. If the aim is
to discover the distribution of patterns in the model, this denotes the generative models in PR. The task
for this case is to find out how the patterns can be modeled in the class. In this regard, the density function
needs to be determined based on training data. On the other hand, the goal in discriminative PR models is
to determine the model that discovers the decision boundary, thus learning the function and parameters of
the decision boundary. Generative and discriminative PR models along with the algorithms used are
shown in Figure 4.

7
 Figure 4. Tree structure of generative and discriminative pattern recognition models and algorithms

4.2. Machine learning

Machine learning (ML) is a major subfield of artificial intelligence (AI) (see Figure 1) dealing with the
study, design, and development of algorithms that can learn from the data itself and make predictions
using learned data [33–36]. In fact, ML refers to the capability of computers to learn without being
explicitly programmed. ML based models can be predictive or descriptive to achieve knowledge from the
data [37,38]. The scope and potential of ML is much more general than other AI methods, although it is a
subset of AI and used in various disciplines; including computer science, information theory, control
computational complexity, probability and statistics, financial market, and theory and philosophy [35]. It
is of importance to differentiate ML from other similar AI subsets including pattern recognition (PR) and
deep learning (DL). In general, PR and ML are closely related areas, as they fundamentally overlap in
their scope. However, PR deals with methods for classification tasks, while ML focuses on algorithms
utilized for learning. In fact, the major task of PR is recognition of patterns in data and to classify them,
and it does not necessarily imply learning. ML systems, on the other hand, are designed to learn by
themselves. Further, DL is considered a subset of ML (see Figure 1), in which the system has the ability
to learn features from the data. Deep learning, in fact, is a tool to learn the representation of data. Once
the representation is determined, the ML problem can be solved. Indeed, deep learning transforms a
problem/representation with high dimensionality to a lower dimensional representation. Depending on the
resources of the training dataset, ML can be categorized as supervised, unsupervised, or reinforcement
learning [33,36].

4.2.1. Supervised learning

The goal of supervised learning is to build a model/function to accurately predict the unknown target
output of future examples. Training samples in supervised learning are labeled and the key characteristic
of the learning is the existence of a teacher that provides a cost or category label for each pattern in a
training dataset, thus seeking to decrease the added cost for these patterns. If the objective of the ML
model is to forecast continuous target variables, the task is said to be regression. However, if the aim is to
predict discrete target variables the task is known as classification.

8
4.2.2. Unsupervised learning
The objective of unsupervised learning is to separate the training dataset into clusters such that the data in
all clusters exhibits a high level of proximity. Unlike supervised learning, the labels for data are
unavailable and there is no explicit teacher. Thus, the system itself forms the clusters from the input
patterns.

4.2.3. Reinforcement learning

In reinforcement learning, or learning with a critic, no information is given regarding the desired category
signal or explicit goals. Reinforcement algorithms are forced to learn optimal goals through trial and
error. In fact, in order to maximize the model’s performance, reinforcement learning allows an agent to
determine the ideal behavior within a specific context. Agents receive a numerical reward as a
reinforcement signal encoding the success of an action’s outcome. The goal for the agent is then to learn
to select actions maximizing the accumulated reward over time.

Figure 5. Machine learning categories with commonly adopted algorithms

Recent research reveals the successful practical applications of ML in different fields, such as:
computer vision and image processing [39–44], speech recognition [45–50], computational finance [51–
53], energy production [54–56], and computational biology [57–59]. In a machine learning domain an
algorithm has to be developed to solve problems. Different methods from various fields have been

9
adopted for such a purpose [60,61]. Therefore, ML enables exploiting the interaction form all these fields,
which in turn leads to robust solutions using various domains of knowledge. Figure 5 illustrates some of
most prominent algorithms used in the ML domain.

4.3. Deep learning

Deep learning (DL), a branch of machine learning, is composed of networks that can learn unsupervised
from unstructured/unlabeled data. DL architecture aims to learn the feature representation of the input
data. In fact, DL is based on deep neural networks, i.e., neural networks with more than one hidden layer.
In such an architecture, increasing the number of layers results in a deeper network. Examples of DL
architectures include convolutional neural networks (CNNs), recurrent neural networks (RNNs),
autoencoders, deep belief nets, etc. Among these, CNNs are the DL architectures that have gained the
most attention among the structural engineering community during last few years. CNNs are inspired by
the visual cortex of animals [62]. They have been mainly used in computer science and engineering for
image recognition [63–68]. Unlike standard neural networks, CNNs are capable of capturing the 2D
topology of pixels, while demanding fewer computations because of a pooling process and sparsely
connected neurons. Further, CNNs are able to simultaneously extract and learn optimal features from the
raw data. Recent studies [69,70] have demonstrated that CNNs can outperform conventional artificial
intelligence methods in both accuracy and speed. Generally, CNNs leverage the following ideas: local
connectivity, parameter sharing, and pooling/subsampling of hidden units. The network consists of three
layer types, namely convolution, pooling, and fully connected layers. CNNs alternate between the
convolutional and pooling layers and the output is a fully-connected layer with a nonlinear classifier, e.g.,
softmax classifier, thus estimating the conditional probability of each class. To introduce nonlinearity in
the CNNs, a rectified linear unit (ReLU) is typically used as a nonlinear activation function. In addition,
among the different optimization algorithms, gradient descent algorithms are mainly used to train CNNs.
The basic components of CNNs are described in the following sub-sections. A schematic of a CNN
architecture for image recognition is presented in Figure 6, where the network consists of three
convolutional layers, three pooling layers, and three fully connected layers. For all layers in the network,
ReLU is used as the activation function. Further, a softmax loss layer is appended to the fully connected
layers for each classification task.

Figure 6. Schematic of a typical CNN architecture

10
5. APPLICATIONS

5.1. Pattern recognition

During the last decade, there has been a growing interest in the application of pattern recognition (PR) to
structural engineering for purposes such as structural health monitoring (SHM)/damage detection,
earthquake engineering and seismic design, structural reliability, structural identification, and
performance evaluation. This activity is illustrated in Figure 7 and a listing of works is summarized in
chronological order (i.e., date of publication) in Table 1. The applications are classified with respect to the
domain/problem type, the case structure, and the AI method/algorithm used for PR. The classification
reveals that the most common use of PR in structural engineering has been for SHM and damage
identification.

Two main approaches are commonly considered for damage detection: the inverse approach, known as
system identification, and the forward approach, which relies on extracting information from the
monitored structure. The computational complexity of the inverse approach, along with the physical
importance of model updating, have motivated researchers to investigate methods from the second type of
approach (forward) [71]. Therefore, PR is being most frequently utilized in the context of a forward
approach for damage detection and SHM.

According to Sohn et al. [72], sensors measuring strain and vibration of a structure produce signals
responding to the variation of environmental and operational conditions. Each group of signals can be
considered as a pattern having a relationship with structural and ambient environments. The change in
physical properties, mainly stiffness, is then reflected on the processed signals or patterns. Thus, the
interpretation of signals/patterns can be performed by PR. The idea of using statistical PR for SHM was
introduced by Farrar et al. [73,74]. Statistical PR can be described as collecting and processing data from
sensors mounted on the structure to remove/filter environmental effects. In this context, statistical pattern
comparison and statistical model development methods have been used to evaluate structural condition
[75,76]. Most of the studies that focus on the application of statistical PR on SHM are based on the
combination of time series modeling with a statistical detection method, such as outlier detection. As a
result of using such methodologies only data from the undamaged structure is required in the
training/calibration phase [77]. Sohn et al. [75,78] casted SHM within the context of statistical PR. For
this purpose, they adopted autoregressive models and outlier analysis with the Mahalanobis distance
measure to extract features and construct a reliable statistical model to assess structural conditions of the
boat. Farrar and Sohn [79] studied the applicability of statistical PR for vibration-based SHM and
described the relevant steps for such process. Worden et al. [80] adopted the methods of outlier analysis
to the problem of damage detection, for which they used the Mahalanobis distance in order to detect
damage. Further, Manson and Worden [81–83] studied the effectiveness of statistical PR using auto-
associative neural networks, outlier analysis, and density estimation through numerical and experimental
tests for SHM of an aircraft wing panel. They proved the applicability of the proposed statistical PR
damage detection approach through these tests. In addition, Nair and Kiremidjian [84] introduced a time
series algorithm based on an autoregressive moving average model for damage assessment of a
benchmark structure, where they showed that the algorithm was able to identify and localize small to
severe levels of damage. The authors also used a Gaussian Mixture Model to model the feature vectors
and incorporated it with a time series-based damage detection algorithm for SHM [85]. They concluded
that the proposed framework is useful, especially when several measurements can be used for robust

11
damage identification. Cheung et al. [86] also studied the applicability of statistical PR for SHM of real-
life structures, namely a bridge, for which they used autoregressive models. These studies have shown
that by using statistical methods a single vibration signal can be analyzed separately from all other signals
accumulated in the structure, thus allowing damage detection algorithms to be embedded at the sensor
level. This results in significant savings in power and computational time, which are essential for the
implementation of a wireless sensor network.

Figure 7. Research publications on the use of machine learning and pattern recognition

An SHM procedure cast within a statistical PR context is implemented in four phases [75]: (i)
operational evaluation, (ii) data acquisition and networking, (iv) feature selection and extraction, and (v)
statistical model development and discrimination. According to the recent literature, several studies
concerning applications of PR in SHM have addressed all of these four phases. Regarding the operation
evaluation phase, numerous techniques have been studied that are either based on linear or non-linear
regression models among actions and effects [87], or based on latent variable techniques. Posenato et al.
[88] proposed methodologies for model-free data analysis using moving principal component analysis
(PCA) and robust regression analysis to identify and localize anomalous behavior in civil structures. Zhou
et al. [89] introduced an approach to reconstruct input to back-propagation neural networks used for
modeling the temperature-caused modal variability with long-term monitoring data. In addition, a
technique based on symbolic data analysis for classifying the structural behavior of railway bridges was
developed by Cury et al. [90]. The method was shown to be efficient to discriminate structural
modifications based on vibration data. Further, a data-driven strategy integrating PCA, symbolic data, and
cluster analysis was proposed by Santos et al. [91], where the method was demonstrated to be effective
for early-damage detection. To take into account the effect of environmental conditions, e.g., temperature,
humidity, dead load redistribution effects, etc., for damage detection of bridge structures, Hsu et al. [92]
performed nonlinear PCA using an auto-associative neural network. They showed that the approach is
capable of dealing with both non-increasing features (stiffness) and non-decreasing features (damage
index).

12
 Methods having the ability to extract and fuse information from data in a sensor network are mainly
based on autoregressive models, time frequency analysis, modal analysis, or PCA. Lautour et al. [93]
presented a damage classification approach using time series analysis and PR, in which artificial neural
networks (ANN) were used to determine the coefficients of the autoregressive models. Results suggest
that ANN and autoregressive models are efficient tools for damage estimation. Additionally, an SHM
strategy employing cepstral features as damage sensitive parameters was proposed by Balsamo et al. [94]
for which the squared Mahalanobis distance was used. A decentralized damage detection approach using
signal analysis (wavelet transform) based on wireless sensor data was developed by Yun et al. [95], while
Kesavan et al. [96] proposed a wavelet-based damage diagnosis algorithm based on the combination of
PCA and wavelet transform. Results revealed that both approaches were able to consistently detect and
quantify damage. Further, most works related to statistical model development report the use of a
statistical process control [97,98]. However, some of these methods are based on learning algorithms, i.e.,
support vector machines, neural networks, decision trees, and clustering algorithms [91,99].

Numerous algorithms, including autoregressive models, artificial neural networks (ANNs), support
vector machine (SVM), etc., have been adopted and used for PR in structural engineering discipline.
Autoregressive models have been extensively considered for feature extraction in numerous studies
concerning the use of PR in structural engineering. Kiremidjian et al. [86,100] used autoregressive
models for SHM of bridge structures. They demonstrated that damage detection algorithms based on PR
methods can effectively detect structural damage. Gul et al. [77] and Yao et al. [101] utilized
autoregressive models with a Mahalanobis distance-based outlier detection algorithm for damage
detection in civil structures. These studies showed the superior performance of the proposed algorithms in
terms of identifying damage with high-dimensional data sets. In addition, ANNs have been widely
adopted for structural engineering applications, e.g., SHM and damage detection. Ng et al. [102,103]
incorporated an ANN with a Bayesian method for health assessment of a four-story steel frame structure.
ANNs have also been integrated with SVM and PCA, a method for dimensionality reduction, to develop
PR-based damage detection techniques in civil structures. As a further example on the application of such
PR methods in structural engineering, Radhika et al. [104] proposed a wavelet-based change detection
method in which damage buildings are recognized using wavelet-extracted statistical feature and
classification using ANN and SVM. They proved that the proposed damage classification method was
accurate compared to methods employing conventional feature extraction. Further, they showed that SVM
outperformed ANN in terms of damage classification accuracy. SVM has also been integrated with ANN
for supervised learning classification [105] in structural modification assessment using vibration data
from a bridge structure, and the method was found to be effective for continuous monitoring. As
previously noted, PCA has been utilized and adapted for damage detection based on PR. An approach for
damage detection in plate structures using a multi-layer perceptron network, in which PCA was utilized to
retain the principal features, has been proposed [106]. Also, Bandara et al. [107] introduced a frequency
response based damage detection method using a combination of PR and PCA. Ramos et al. [108]
presented a methodology employing a Bayesian data fusion technique for non-destructive and destructive
structural damage detection. They showed that the proposed method, within the context of PR, is able to
decrease uncertainties for structural parameter estimation.

Recently, Perez et al. [109] introduced a hierarchical nonlinear PCA method for damage diagnosis in
wind turbine blades. The authors demonstrated the effectiveness of the methodology based on strain
measurements and PR for SHM. Further, Alavi et al. [110] proposed a damage assessment approach

13
based on probabilistic neural networks and Bayesian decision theory, where they proved that the

Reference Domain Case structures AI method used for PR

supervised classification method can be utilized for SHM purposes. In addition, Loh et al. [111] proposed
an SHM methodology for damage identification and localization based on PCA, and investigated the
applicability of the proposed PR method on a steel tower structure. Salehi et al. [112–114] presented an
image-based PR approach based on integrating anomaly detection and a Bayesian method. They also
utilized a nearest neighbor classifier, along with a two-dimensional principal component analysis and a
two-dimensional linear discriminant analysis (well-established feature extraction techniques), for SHM in
plate-like structures. Additionally, Datteo et al. [115] proposed a statistical PR approach integrating
autoregressive models and principal component analysis, and explored the applicability of such approach
for long-time health monitoring of large-scale structures. Zhou et al. [116] also introduced a damage
detection technique using a cosine similarity measure. The authors demonstrated that the presented PR-
based method can be effectively used in the context of SHM.

The research studies noted above indicate the significance of PR in structural engineering.
Nevertheless, an in-depth analysis of some of the highest cited publications was performed to further
investigate their contributions and limitations. Lautour et al. [117] proposed a damage assessment method
using autoregressive (AR) models in which the computational burden of the approach was lessened by a
dimensionality reduction technique (i.e., PCA). The authors showed that the AR coefficients form
separable clusters by increasing the number of principal components, leading to good classification
accuracy. Zhang et al. [118,119] introduced a structural identification method employing PR and support
vector regression (SVR). SVR was integrated with autoregressive time series analysis for linear and
nonlinear structural parameter identification (for damage detection) with vibration data of a five-floor
structure shaking table test. Lautour et al. [120] presented an approach using ANN to predict seismic-
induced damage on 2D reinforced concrete frames. Relations between parameters that describe the
structure, ground motion, and damage were modeled using ANN. Laory et al. [121] developed a
methodology to predict natural frequency responses of a suspension bridge with measurements of
temperature, wind, and traffic loading, within the context of vibration-based SHM. Multiple linear
regression, ANN, SVR, regression tree, and random forest were used to distinguish changes in natural
frequency due to structural damage and environmental variations, and the method’s prediction accuracy
was compared. Bandara et al. [107,122] proposed a damage detection method using frequency response
functions in which ANN, PCA, and frequency response functions were combined to detect various levels
of nonlinearity using identified patterns. The authors applied the algorithm to a three-story structure and
demonstrated the method’s applicability for large amounts of data. Tibaduiza et al. [123] developed an
SHM method in which PR, feature extraction, and sensor data fusion were examined with different
damage indices. Performance of the proposed approach using PCA was tested for an aircraft skin panel
and turbine blade. The effectiveness of the approach was validated; yet, the effect of environmental and
operational conditions on the damage identification method was not considered. As a final example,
Elwood et al. [124] proposed an approach based on fuzzy PR for seismic damage detection in concrete
structures. The input to the fuzzy classifier was post-earthquake building damage data to determine the
existence of building damage patterns. It follows that the noted studies highlight the emerging
applications of PR in structural engineering.

14
 Statistical PR based on time series
[125] SHM Flexible risers
Reference Domain Case structures AI methodanalysis
used for PR
[86,100] SHM Bridge structure Autoregressive models
Autoregressive models with principal
[117] Damage detection Bookshelf structure
component analysis (PCA)
Fiber-reinforced polymer
[126] Failure mechanism Self-organizing map
(FRP) structures
Seismic damage
[127] Frame structure Artificial neural networks (ANN)
detection
Five-story structure resting on Support vector regression and
[118,119] Structural identification
shaking table autoregressive time series model
Bayesian method incorporated with
[102,103] SHM Four-story steel frame
ANN
[128,129] Damage detection Four-story steel frame Artificial immune PR method
Autoregressive model with
Steel grid structure and simply
[77] SHM Mahalanobis distance-based outlier
supported steel beam
detection
[120] SHM Reinforced concrete frames ANN
Outlier analysis with Mahalanobis
[130] SHM Railroad steel structure
squared distance
Three-story steel frame and a Nearest neighbor classifier and learning
[131,132] Damage detection
bookshelf structure vector quantization
Structural modification Clustering techniques (unsupervised
[90] Bridge structure
assessment PR)
Modeling temperature-
[89] Bridge structure Back-propagation neural networks
caused modal variability
Statistical PR based on the Mahalanobis
Prestressed reinforced concrete
[133,134] Damage detection and Euclidean distance decision
beams
functions
Reinforced concrete water
[135] Seismic performance Statistical PR
tanks
Principal component analysis and
[88] SHM Beam structure
robust regression analysis
Three-story steel frame and a
[93] Damage detection Artificial neural networks
bookshelf structure
Statistical PR based on an
[136] SHM Simply supported steel beam
autoregressive model
Statistical PR based on an
[101] Damage detection Bridge slab and space truss
autoregressive models
Feed-forward multi-layer neural
[137] Earthquake engineering Earthquake risk evaluation
network
[96] Damage detection ASCE benchmark structure PCA
Table 1. Applications of pattern recognition (PR) in structural engineering

15
 [138] Damage detection Three-story steel structure Statistical PR
Feed-forward multi-layer neural
[139] Damage detection Cantilever plate
network
Statistical PR based on a pattern
[76] SHM Bridge structure
comparison approach
Statistical PR based on a multivariate
[140] Failure detection Posttensioned concrete beam
outlier analysis
[98] SHM PSC box girder bridge Symbolic clustering method
Bridge structure and simply Support vector machines (SVM) and
[141] SHM
supported beam neural networks
Sparse representation and fourier
[142] SHM Bridge structure
discriminant method
Principal component analysis (PCA)
[91] Damage detection Bridge structure
and symbolic data clustering
[143–145] Damage detection Bridge structure Supervised statistical PR
Pretensioned prestressed
[146] Performance evaluation Feed-forward neural regression
concrete members
[147] SHM Cable-stayed bridge structure PCA and Mahalanobis squared distance

[148] SHM Aluminum beams Bayesian approach

Artificial neural networks, support
[121] SHM Suspension bridge structure vector regression, random forest,
regression tree
[106] Damage detection Plate structures Multi-layer neural network and PCA
Three-story prototype building Unsupervised PR based on an outlier
[149] Damage detection
structure analysis
Steel reinforced concrete Statistical PR with autoregressive
[150] SHM
structure models
Artificial neural network (ANN) with
[107] Damage detection Two-story frame structure
PCA
[151] Performance evaluation Steel beam structure Statistical PR
Non-destructive
[108] Concrete structure Bayesian fusion model
evaluation
[104] Damage detection Building structures ANN and SVM
Structural modification Bayesian decision trees, neural
[105] Simply supported steel beam
assessment network, and SVM
Statistical PR based on a Mahalanobis
[94,152] SHM Four-story steel frame
squared distance
[122] Damage detection Three-story building structure ANN

[153] Damage detection Steel beams ANN

Reference Domain Case structures AI method used for PR

16
 Probabilistic neural networks and
[110] SHM Bridge gusset plate
Bayesian approach
Statistical PR based on a Mahalanobis
[71,154] Damage detection Cable-stayed bridge structure
squared distance
Seismic damage
[124] Concrete structures Fuzzy PR
detection
Bayesian method, nearest neighbor,
two-dimensional principal component
[112–114] SHM Plate-like structures analysis (2DPCA), and two-
dimensional linear discriminant analysis
(2DLDA)
[155] Risk-based management Bridge structures Statistical PR

[156] Damage detection Cable-stayed bridge structure Multi-layer perceptron neural network

[157] SHM Plate-like structures K-nearest neighbor method

Artificial neural network (ANN) and
[158] Damage detection Steel grid structure
self-organizing maps (SOM)
[111] Damage detection Steel tower structures Principal component analysis (PCA)

[123] Damage detection Aircraft skin panel PCA

Neural network with back propagation
[159] SHM Composite cantilever beam
based learning mechanism
Truss bridge and two-story
[160] SHM PCA and frequency response function
frame structure
Hierarchical nonlinear principal
[109] Damage detection Wind turbine blades
component analysis
Autoregressive models with principal
[115] SHM Stadium structure
component analysis
[161] Damage detection Bridge structure Statistical PR

[116] Damage detection Three-story frame structure Cosine similarity measure

[162] SHM Steel beam Principal component analysis

5.2. Machine learning

Machine learning (ML) methods have been increasingly adopted over the last decade for modelling real-
world problems concerning structural engineering (see Figure 2). This is because of their enormous
capacity to capture relations among input and output data that are nonlinear or complicated to formulate
mathematically. The first uses of ML techniques in structural engineering have dealt with problems such
as the development of management tools for structural safety [163], and information acquisition for the
design of steel members [164]. In general, ML methods have been used for SHM and damage
identification, optimization, performance evaluation, structural reliability and reliability assessment, and
structural parameter identification (e.g., modeling material properties of concrete). Among these, SHM

17
and concrete property modeling are the uses to attain most attention during the last decade. This can be
seen in Figure 7 and a listing works organized by year of publication is provided in Table 2.

5.2.1. Structural health monitoring and optimization

SHM involves monitoring of a structure through data collected from sensors, extracting damage sensitive
features, and interpreting the extracted features for condition assessment of the structure. Significant
progress has been made over the past two decades in the development of SHM models for different kinds
of structures. The numerous studies carried out in this field can be categorized as model-driven and data-
driven approaches [165–167]. A model-driven approach uses a numerical model of the structure, e.g.,
based on the finite element (FE) method, that correlates inconsistencies between the measured and model-
generated data for damage detection. Although numerous studies have been conducted to develop model-
driven approaches, these methods suffer from several shortcomings. First, the approach is
computationally inefficient because it requires an iterative analysis of a computer simulation model.
Second, results obtained from the simulation might not be accurate enough for precise evaluation of the
structure. By contrast, in a data-driven approach the model is created through the learning gained from
measured/sensed data. Damage can thus be detected by conducting a comparison among the measured
data and a model. In fact, a data-driven model uses information from previously collected sensor data
(i.e., training data). It is worth pointing out that data-driven approaches are beneficial if: (i) large volumes
of data exist, (ii) the physical characteristics of the structure are unknown or complicated to model, and
(iii) the aim is to decrease the computational effort.

A data-driven approach commonly adopts techniques from pattern recognition (PR) and machine
learning (ML). ML in the context of SHM is expressed as creating knowledge from previous experiences,
learning the model parameters, and then focusing on predicting new input data. Different learning
schemes, such as supervised and unsupervised learning, have been used in SHM applications. Algorithms
including artificial neural networks (ANNs) [168], support vector machine (SVM) [169], k-nearest
neighbor method (k-NN) [170], principal component analysis (PCA) [123], and low-rank matrix
decomposition [171] are attractive for structural damage identification within the context of ML due to
their effectiveness and robustness while dealing with insufficient information, noise, and uncertainty. As
mentioned before, there has been a growing interest in the use of ML for SHM models during the last
decade. As an example, Figueiredo et al. [172] investigated ANNs, Mahalanobis distance, singular value
decomposition techniques, and factor analysis to study environmental variability and its effect on damage
detection in civil structures. They used a three-story frame structure as a case study to obtain time-series
data from accelerometers, and the data was fed into different ML algorithms. They showed that the
Mahalanobis distance provided the best classification accuracy. Dervilis et al. [173] investigated the SHM
of wind turbine blades using neural networks. Yan et al. [174] reported on the use of a back-propagation
neural network and SVM for damage assessment in beams mounted on ocean platforms. Butcher et al.
[175] examined the use of ANNs and extreme learning machine methods for SHM in mesh-reinforced
concrete structures. The study revealed that these algorithms can outperform traditional ANN methods.
Liu et al. [176] studied SVM for damage detection of a long span cable-stayed bridge and demonstrated
that SVM is more accurate compared to a back-propagation neural network. Gui et al. [169] presented a
data-driven SVM approach using optimization algorithms, i.e., grid-search and particle swarm
optimization, for damage diagnosis of a three-story frame structure. They proved that a genetic algorithm-
based SVM yields a better prediction than other methods. Gong et al. [177] also evaluated the
applicability of SVM, random forest, and k-NN methods for earthquake-induced damage identification in

18
buildings employing images. Results showed that the proposed approach was capable of differentiating
collapsed and standing buildings. Lederman et al. [178] used PCA along with a kernel regression method,
within the context of signal processing and ML, for damage quantification and localization in bridges.
Results suggested that PCA can be effectively used to decrease the dimensionality of the signal, while a
kernel regression can be employed to map the signals to the bridge condition. A wavelet SVM-based
neural network metamodel for reliability analysis was proposed by Dai et al. [179] to expand the
application of wavelet neural network to higher dimensions. The authors used a set of wavelet SVM with
various resolution as the activation function of wavelet neural network, where they tested the applicability
of the proposed method on five-story structures, truss string structures, and cylindrical shell roof. Further,
Diez et al. [180] presented a clustering-based approach incorporated k-NN, k-means, and Fourier
transform for vibration signal processing to detect damage and abnormal behavior in bridge joints. The
clustering approach helped to group joints with similar behavior, increasing the SHM performance. Zheng
et al. [181] introduced a probabilistic classification framework using vibration measurements to assess the
probability of barge collision damage on bridge piers, where Bayesian inference combined with Markov
Chain Monte Carlo simulations and PCA were used to extract the feature vectors from variations in modal
properties due to damage. Results revealed that the approach can be effectively used to determine the
probability of structural damage locations. Santos et al. [182] proposed a hybrid approach based on
Gaussian mixture models (GMM) to discover the normal state of a bridge, in which the GMM parameters
were estimated through a hybrid method based on an expectation-maximization algorithm. Results
confirmed that the proposed algorithm was more stable than other genetic algorithms in terms of damage
detection performance. A model-free damage assessment method based on ANN was presented by Neves
et al. [183] for SHM of bridges. The ANN was trained with an unsupervised learning algorithm using
accelerations from a bridge, the prediction errors were characterized using a Gaussian process, and
damage indices were compared with a threshold to identify damage. The noted studies highlight the
importance of ML for data-driven SHM and damage assessment techniques.

Recently, a new class of ML methods, namely low-rank matrix decomposition and singular value
decomposition, having the ability of dealing with sparse and incomplete data have been adopted by the
SHM community. Structural response measurements from mounted sensors can be represented as a data
matrix. These measurements possess a low-rank structure and sparsity nature, which can be processed by
emerging mathematical tools such as sparse representation and low-rank matrix decomposition. Salehi et
al. [170] presented a machine learning framework for health monitoring of an aircraft stabilizer based on
the integration of low-rank matrix decomposition and k-NN techniques. They validated the proposed
approach through the interpretation of self-powered wireless sensor data generated from a network
communication protocol using energy-efficient pulse switching technology [184]. The authors further
employed low-rank matrix decomposition and statistical methods for health monitoring and localized
damage identification in plate-like structures, and used a data fusion concept to combine the information
obtained from a network of self-powered sensors [185]. Nagarajaiah et al. [171] presented a new
paradigm for damage detection based on modelling and harnessing sparse and low-rank data structures.
They demonstrated that the proposed method is able to effectively address structural dynamics,
identification, monitoring, data sensing and management problems. Yang et al. [186,187] also utilized
low-rank matrix decomposition along with nuclear-norm-minimization methods for recovering structural
vibration responses from a steel tower and a cable-stayed bridge. The authors developed a global
computational approach to analyze sparse sets of 2D strain measurements for damage localization. The
proposed data-driven approach increased the effectiveness of SHM when limited numbers of strain

19
sensors were deployed. The studies discussed in this paragraph show the use of numerous types of ML
algorithms for assessing structural health; and that under specific circumstances these methods can
outperform model-driven approaches while being satisfactorily accurate.

Within the context of optimization, ML has been used for infrastructure maintenance and durability
assessment. Yepes et al. [188] proposed a cognitive method for selecting an optimal solution for the
multi-objective optimization of high-strength reinforced concrete beams, where different Minkowsky
metrics were used for the optimization task. Garcia-Segura et al. [189] presented a reliability-based
method employing a modified harmonic search algorithm to optimize the design of post-tensioned
concrete box-girder bridges under corrosion attack. The authors demonstrated that lower life-cycle cost is
correlated to designs with longer corrosion initiation time. The same research group further used multi-
objective harmony search integrated with ANN to decrease computational demand for the finite element
analysis of post-tensioned box-girder bridges [190]. Mondoro et al. [155] proposed an approach for
optimal risk-based management strategies for bridges in which they considered the uncertainties
associated with hazards, economic, social, and environmental outcomes of failure under traffic loads and
hurricanes. Chatterjee et al. [191] employed a multi-objective genetic algorithm for calibration of a neural
network model to minimize the root mean squared error and maximum error of the network. Results of
structural failure classification for reinforced concrete buildings indicated that the proposed optimization
algorithm outperformed a multi-layer perceptron feed-forward network. The mentioned studies indicate
the wide applicability of ML for structural optimization.

5.2.2. Mechanical properties of concrete

The design of concrete structures requires considering several key mechanical properties of the material,
such as compressive strength, splitting tensile strength, shear strength, and elastic modulus. Linear or
nonlinear regression models to these material parameters have been proposed to save time and costs
associated with material testing [192,193]. However, the mechanical properties of concrete are known to
have strong nonlinear relations between the constituents and the macroscale material characteristics
[167,194]. Therefore, the development of reliable models is of interest to explore material mechanical
properties in a way that optimizes cost and time. The potential of ML algorithms has been harnessed to
model such properties and address the noted issues.
Several ML algorithms, such as neural networks, genetic programming, fuzzy logic, and support
vector machines (SVM) have been used to develop accurate models to forecast the mechanical properties
of concrete. Most significantly, ML algorithms have been used for modeling the properties of self-
compacting concrete [195,196] (e.g., strength, elastic modulus), as well as modeling the tensile and
compressive strength of normal concrete [197,198]. As an example, Yeh et al. [193] proposed a genetic
operation tree composed of an operation tree and a genetic algorithm to generate formulas that predict the
compressive strength of high-performance concrete. Cheng et al. [199] used a genetic weighted pyramid
operation tree to construct a model for determining the compressive strength of high-performance
concrete. The obtained model gave better results, using benchmark tests, in comparison to ANN, SVM,
and evolutionary support vector inference models. Xu et al. [200] established an SVM-based model to
assess the relation between the strength and mechanical properties of concrete obtained from non-
destructive testing. They showed that the proposed method is less computationally demanding, while also
providing high accuracy in its predictions compared to other numerical methods. Yan and Shi [201]
investigated the applicability of SVM for predicting the elastic modulus of normal and high strength

20
concrete. They discovered that SVM has superior performance compared to ANN models. Yan et al.
[202] developed an SVM model with experimental data from the literature and compared the results with
empirical design equations. They showed that an SVM model is capable of accurately estimating the
splitting tensile strength from compressive strength. Parsad et al. [203] used a neural network to predict
the compressive strength of self-compacting and high performance concrete. An artificial intelligence
system based on combination of fuzzy logic, weighted SVM, and fast messy genetic algorithms was
developed by Cheng et al. [204] to predict high-performance concrete compressive strength, where results
showed that the method achieved higher performance compared to SVM. Saridemir [205] used gene
expression programming to determine the splitting tensile strength concrete from its compressive
strength. Results showed that the proposed formulations led to the best accuracy and were able to predict
splitting tensile strength similar to experimental results. Nedushan [206] introduced an adaptive network-
based fuzzy inference system (ANFIS) model and an SVM for predicting the elastic modulus of normal
and high strength concrete, and found that the ANFIS model outperformed nonlinear regression models
and the predictive models in the literature. Lee et al. [207] presented a theoretical model using ANNs to
predict the shear strength of slender fiber reinforced polymer reinforced concrete beams, which was
shown to perform better than other existing equations. All these studies concluded that ML methods are
influential tools for evaluating the mechanical properties of concrete without being affected by data
complexity and incoherence.

Reference Domain Case structures AI method used for ML

Table 2. Applications of machine learning (ML) in structural engineering

21
 [208] Damage detection Concrete slabs Support vector machine (SVM)
Reference Domain Case structures AI method used for ML
[176] SHM Cable-stayed bridge structure SVM
Long-span arch bridge
[218] Damage detection
Modelling of concrete SVM
[200] structure
Concrete beam SVM
strength
[219] Damage detection
Earthquake Transmission tower SVM and RBF neural network
[209] Seismic evaluation Bayesian method
engineering
[220] Damage detection Bridge structure Neural network, SVM, and SOM
Modelling of concrete
[197] Concrete beam SVM
Performance
strength
[221] Steel beams Linear genetic programming
evaluation
Structural reliability SVM-based radial basis function (RBF)
[210] Truss structures
Performance
analysis network
[222] Concrete dam Artificial neural network (ANN)
evaluation
Seismic damage Buildings with steel moment-
[211] Neural network
Structural
detection frame structure
[223] Steel-box girder bridge Principal component analysis (PCA)
identification
Structural
[212] Concrete bridge Neural network
Modeling concrete
identification
[224] Concrete cube SVM
strength
Structural
[213] Concrete dam Neural network
[225] identification
SHM Cantilever beam Dynamic Bayesian networks
Prediction of concrete
[198] Concrete blockstructural
Concrete with fly ash Support vector regression
[226] properties
SHM SVM
components
[214] SHM
Prediction of concrete Metallic
Concrete withstructures
construction Adaboost machine learning
[227,228] ANN
strength
Performance and demolishing waste
[215,216] Self-compacting concrete Artificial neural network (ANN)
[168] evaluation
Damage detection Beams on ocean platform Neural network
Performance
[217] Concrete dam ANN and linear regression
[229] evaluation
SHM Steel pipes SVM and adaptive boosting
Singular value decomposition,
[172] SHM Three-story frame structure Mahalanobis distance, auto-associative
neural network, and factor analysis

22
 Prediction of concrete Self-compacting concrete
[195] SVM
properties block
Reference Domain Case structures AI method used for ML
[230] SHM Cantilever concrete beam SVM
Seismic damage Reinforced concrete slab Multiclass support vector machine and
[236] Prediction of concrete
[202] identification column concrete
Cylinder frames multi-layer perceptron
SVM neural network
strength Kernel regression method and principal
[178] Damage detection Bridge structure
[231] Damage detection Steel structures component analysis
Multi-objective genetic (PCA)
algorithm
Modelling concrete Reinforced and unreinforced Multivariate adaptive regression splines
[237] Concrete strength ANN,SVM, Classification
[192] shear strength concrete joints
High performance concrete and symbolic regressionand
simulations
Structural regression tree, linear regression
[238] Bridge structure Gaussian process model
[232] identification
Damage detection Steel frame structure SVM with Gaussian kernel
[239,240] Structural reliability
Performance Truss structure Gaussian process machine learning
[233] Concrete dam Support vector regression
evaluation
Predicting concrete Support vector regression and adaptive
[240] Prediction of strength
concrete Concrete specimen
[196] compressive Self-compacting concrete neuro-fuzzy
Least square supportinference
vector machine
strength
[241] Damage of
Prediction detection
concrete Bridge structure SVM, regression, random forest
[234] Corroded reinforced concrete SVM
properties Least square support vector machine
[242] SHM Truss structure
Mesh-reinforced concrete
[175] SHM with a mixed
ANN kernel
structure
Three-story steel frame
[169] SHM
Earthquake SVM
[235] structure
Two-story building Gaussian process regression
engineering
Seismic damage SVM, K-nearest neighbor method (K-
[177] Prediction of shear Seismic
Fiber performance
reinforced polymer
[207] detection NN), andANN
random forest
strength concrete
[180] SHM Bridge structure K-NN and k-means clustering

23
 Gaussian mixture models and genetic
[182] Damage detection Bridge structure
algorithm
Bridge design Post-tensioned concrete road Artificial neural network (ANN) and
[190]
optimization bridge structure harmony search algorithm
Reliability-based Post-tensioned box-girder
[189] Modified harmony search algorithm
optimization bridge structure
[191] Damage detection Reinforced concrete buildings Neural network and genetic algorithm
Risk-based
[155] Coastal bridge structure Optimization algorithm
management
Tensile strength
[243] Steel plates ANN
prediction
Six-story reinforced concrete
[244] Seismic performance Support vector regression
frame structure
Bayesian inference and Markov Chain
[181] Damage detection Bridge structure
Monte Carlo simulation
[183] SHM Bridge structure ANN

[157] SHM Plat-like structures K-NN

Earthquake
[245] Seismic performance SVM and neural networks
engineering
Shear capacity Fiber-reinforced polymer
[246] Least square support vector machine
estimation concrete slabs
[247] Reliability assessment Steel-box girder bridge Support vector regression
Moving principal component analysis
[248,249] SHM Movable bridge structures
and robust regression analysis

24
Reference Domain Case structures AI method used for ML

[250] Crack categorization Reinforced concrete columns Fuzzy logic

Performance Steel-concrete composite Extreme learning machine models,
[251]
evaluation beams ANN, and genetic programming
Predicting concrete SVM, Gaussian processes regression,
[252] Concrete structures
compressive strength and ANN
[253] SHM Wind turbine systems Affinity propagation clustering
Low-rank matrix decomposition and
[170] SHM Aircraft wing structure
K-NN
Structural parameter Three story structure and
[254] Particle swarm optimization
identification three-span continuous beam
Structural tower and cable-
[187] SHM l1 minimization sparse recovery
stayed bridge
Structural parameter
[255] Three-story steel frames Independent component analysis
identification
Five-story structures, truss
Structural reliability
[179] string structures, and SVM-based neural network
assessment
cylindrical shell roof

25
[256,257] SHM Steel plate SVM

[171] SHM Reinforced concrete beams Low-rank matrix decomposition

Compressed sensing based random
[258] SHM Cable stayed bridge
encoding
[186] Damage detection Plate structure Low-rank matrix decomposition
Low-rank matrix decomposition, K-NN,
[259] SHM Aircraft wing structure
SVM, and ANN

5.3. Deep learning

During the last few years, there has been a growing interest in the use of deep learning, e.g., convolutional
neural networks (CNNs) for structural engineering applications, mainly in structural health monitoring
(SHM). The application of CNNs is very new in the field of SHM and damage detection. CNNs within
the context of SHM are defined as learning and extracting optimal features and classification using
learned features. As previously discussed, CNNs are primarily designed for two-dimensional signals (e.g.,
images, video frames, etc.), thus leading to an efficient image recognition method. Therefore, CNNs are
categorized and used as vision-based SHM techniques in which dataset are images captured at various
states of the structure being monitored.

The first use of CNNs in structural engineering was conducted by Sarkar et al. [260] for characterizing
crack damage on composite materials. Further, Abdeljaber et al. [261,262] introduced a vibration-based
structural damage detection approach using one-dimensional CNNs. They proved that the method was
capable of learning directly from the measured acceleration data, yielding an accurate approach for health
monitoring of civil structures. However, the proposed system, especially for large civil structures,
suffered from the fact that a large number of measurement sessions was required to generate the training
data. To overcome this drawback, they proposed a nonparametric damage identification method using
CNNs that required two measurement sessions to generate the training data [263]. They showed that the
SHM system was effective in estimating the actual amount of damage. Cha et al. [264] presented a deep
learning network to detect concrete cracks in the tunnels without the need for computing defect features.
They also conducted a comparative study to show how the proposed deep learning-based damage
assessment approach was able to detect concrete cracks in a robust manner compared to traditional image
detection methods. Gulgec et al. [265] proposed a structural damage identification method using CNNs to
discover the unknown relation between the measurements and patterns representing damage. Lee at al.
[266] also investigated the applicability of deep learning and CNNs for structural analysis in a ten bar
planar truss and proved that such techniques are more efficient compared to conventional neural
networks. All these studies suggest that deep learning/CNNs architectures are effective tools for
monitoring structural health, and that these frameworks are establishing themselves as viable methods for
a new generation of vision-based SHM systems.

6. DISCUSSION AND FUTURE DIRECTIONS

This study reviewed papers published during the last decade concerning the applications of emerging AI
methods, namely, pattern recognition (PR), machine learning (ML), and deep learning (DL), in structural
engineering. The papers were thoroughly reviewed to identify the nature of the problem, the AI
algorithms adopted and used, and to assess the methods’ applicability for the given problem. The survey

26
showed that PR and ML are being widely used by the structural engineering community for applications,
such as SHM, structural identification, earthquake engineering, etc. Yet, the most common use for PR and
ML has been for SHM. The review further indicated that DL architectures have also been utilized for
SHM and damage identification. It is to be expected that the use of AI in structural engineering will
increase as their potential is better understood and as new methods are developed.

Figure 8. Applications of ML, PR, and DL in structural engineering

Current and emerging applications of ML, PR and DL in structural engineering are shown in Figure 8.
The following sub-sections discuss future directions for AI-based methods, including emerging
applications and issues for improving their efficiency and robustness.

6.1. Data-driven SHM systems with self-powered sensing technology

Performance of the noted AI methods for SHM applications strongly depends on the amount of data
collected through the monitoring system. Wireless sensor networks (WSNs) have emerged to overcome
the drawbacks of wires in dense sensor arrays, and have increasingly become an alternative to traditional
SHM systems. Durability monitoring using WSNs transforms the way of inspecting structure s to an
automated, rapid, and objective manner. Additionally, continuous remote monitoring using WSNs for
long periods of time is more economical than conducting periodic field experiments or inspections.
Recently, self-powered sensors have evolved to be able to harvest the needed power (for computational,
storage and transmission requirements) from the signal being sensed as well as form ambient vibrations,
thus providing a promising alternative to traditional sensor systems. PR, ML, and DL methods can then
be integrated with self-powered wireless sensor networks to present the new type of data-driven SHM
systems that are energy-lean.

27
 Data-driven approaches are nowadays combined with empirical models to monitor the state of a
structure. Although these approaches enhance performance prediction, they still depend on empirical
formulas, which have the previously discussed limitations. However, data-driven approaches for SHM
solutions are expected to rely on data collected from embedded/mounted sensors along with artificial
intelligence techniques. ML and PR are powerful tools to extract information and develop predictive
models from large data. Furthermore, the increased use of these intelligent methods in structural
engineering clearly indicates that these methods are becoming predominant approaches for SHM. The
incorporation of WSNs and the noted AI methods for structural engineering purposes could result in the
efficient inspection and assessment of civil structures, as the evaluation can be performed remotely
through sensors with wireless data transmission capabilities and by interpreting data using PR and ML
techniques. Furthermore, ML algorithms are able to learn the complex interrelation among influencing
factors, thus performing predictions without the need for empirical models, while also being able to
improve on their predictive capability. Advancements in self-powered sensors have also promoted the
development of energy-efficient network technologies, such as the pulse switching protocol [184,267],
which can be coupled with ML algorithms for SHM and damage identification [170,185,259,114]. As a
result of using such an intelligent system, the constraint of a communication power budget for an SHM
sensor network can be addressed, thus leading to a reliable and efficient SHM system.

6.2. Vision-based SHM systems and computational mechanics

Deep learning methods emerged to interpret big data in order to identify implicit features from it, and to
classify the learned features. Deep learning-based damage detection techniques have been found to be
computationally efficient. Unlike conventional ML techniques that use hand-crafted features that result in
high computational complexity, DL and CNNs use optimal features learned by the network, thus
increasing the classification accuracy significantly. Further, the structure of the DL architecture,
specifically one-dimensional CNNs, make their mobile and low-cost hardware implementation quite
feasible. Therefore, it is expected that DL will play important role in the future generation of vision-based
SHM systems, i.e., those based on computer vision techniques [268–270]. Another interesting potential
application of ML and DL is in the computational mechanics domain [271]. In computational mechanics,
problem solving rules strongly depends on an expert’s insight. Such rules are valid when certain
assumptions hold, thus indicating a limitation of the expert’s ability. To cope with this difficulty, ML and
DL can be used to automatically discover the rules required to solve computational mechanics problems
such as those using the finite element method (FEM). DL methods are able to generate implicit rules and
discover mapping relations among the input-output data. For instance, optimizing numerical quadrature is
an essential problem to the FEM that requires great amount of computation. However, DL can be used as
a tool to address such problem. A framework of computational mechanics methods enriched by DL/CNNs
can be developed and applied to optimize numerical quadrature in order to compute the FEM element
matrices. That is, the number of integration points in the numerical quadrature of the element stiffness
matrix can be minimized for the prescribed range of error predicted thru DL. The DL architecture can also
estimate the most accurate result by optimizing the numerical quadrature parameters.

6.3. SHM systems with IoT

The durability of civil infrastructures has nowadays become a big issue given the number of structures
that need to be repaired, and concerns on the efficiency of traditional techniques used to manage
maintenance and repair actions. This situation is creating a paradigm shift toward cutting-edge

28
technologies such as the Internet of Things (Iot) [272]. The IoT refers to a system in which WSNs
mounted with intelligent software and local computing power could be effectively used for the monitoring
of structures. IoT aims to increase machine-to-machine communication thru wireless integrated sensors
with the goal of monitoring devices remotely and efficiently. In this new paradigm, smart devices collect
data, transmit information, and process information collaboratively using cloud computing techniques.
Software is also needed to extract useful information from the large amount of data that is generated. On
this basis, ML could be integrated with IoT for SHM purposes [273–275]. ML can thus become an
essential tool that can be applied to expand the boundaries of IoT. On the other hand, the important issue
regarding the SHM of structures, such as bridges, is to constantly monitor the installed sensors and to
compare the new data with previous readings. It is, however, a challenging task to visit all monitored
bridges given the fact that they are typically geographically distant from each other. Thus, a technology
that links all sensors on the bridges to a common recording device is needed. Further, it is essential to link
collected information to a centralized monitoring station that could receive all the data from the sensors
through the internet. The IoT and noted artificial intelligence methods could be used to effectively address
the noted difficulties. Accordingly, the IoT will enable engineers to collect data from several bridges for
further analysis. ML can then be used for data analysis and interpretation. Structural health assessment
employing IoT could provide a promising solution for rapid, accurate, and low-cost SHM systems. The
integration of SHM, IoT, and cloud computing can lead to powerful processing of the sensed data
compared to traditional SHM systems. In fact, cloud platforms can enable an SHM system to store and
use data from smart monitoring devices. The structure’s health status can then be sent to an Internet
server, and data stored on the server can then be monitored remotely from a mobile device and interpreted
using ML.

6.4. Smart cities with IoT

The concept of smart cities has been recently gaining attention in diverse engineering communities, and
the application of the IoT paradigm to smart cities is generating research interest [276–279]. The main
aim of a smart city is to make better use of public services and to reduce operational costs. In other words,
the goal of a smart city is to make infrastructure smarter in order to use resources efficiently. The
achievement of this goal depends on a data provided by the wireless sensor networks deployed in cities.
The IoT for a smart city can provide distributed data of structural integrity measurements of monitored
structures using data collected by sensors, where DL architectures, e.g., CNNs, can be used as tools to
interpret data and classification [280,281]. The data collated from a city varies so much in format and
quality that it is difficult for one given system to effectively process all such data. The fact that every city
is unique and has a different set of problems yields the need for smart data interpretation techniques.
Thus, robust layers for data collection, communication protocols, data storage, etc. need to be built. DL
can then be used as a viable tool for interpreting such large amounts of data. DL can be utilized to train
systems to recognize patterns for large numbers of real-time networks and provide early recognition of
developing network performance issues. On the other hand, the big challenge for the smart cities concept
is how to deal with the large amount of time series data, a particular form of sequential data, received
from connected sensors. DL architectures (e.g., CNNs) are very efficient in the analysis of sequential
data. DL platforms can thus enable a system to solve optimization problems relating to smart cities and
structures.

The notion of a smart city is to use sensors within the city’s infrastructures to ensure sustainability,
safety, and efficiency. Recent progress in nanotechnology have led to the emergence of a new class of

29
sensors, e.g., self-sensing materials that can provide smart cities with methods to assess and monitor the
condition of the infrastructure. Smart concrete, having the ability of enabling any concrete structure with
self-sensing capabilities, is one of the most promising technologies [282–284]. Such functional property
is achieved by correlating the variation on internal strain with the variation of appropriate material
properties, e.g., electrical resistance. Sensors fabricated using a cementitious matrix with nanoinclusions
of carbon nanotubes can be used for condition assessment of concrete structures and traffic monitoring in
smart cities. Consequently, AI methods can be effective in the interpretation of sensor data. Other
examples include new developing approaches to detect the first stages of corrosion in concrete structures.
The aim is to monitor the state of concrete during the curing period, leading to concrete structures with
increased lifetime and safety. To accurately monitor the strength and temperature of concrete during
curing, sensors are embedded in the concrete at the time of placement and measurements are
communicated to smartphones through IoT. AI methods such as ML and DL can then be used to interpret
the collected data for structural assessment.

6.5. Improving the performance of AI methods in structural engineering

The findings that make AI methods such valuable tools have been particularly highlighted. However, it is
well known that all methods and models have limitations. Table 3 summarizes some general advantages
and disadvantages of PR, ML, and DL for structural engineering applications. Further, there are aspects of
the implementation of the noted AI methods that could help enhance their performance. First, it is clear
that use of AI methods for solving structural engineering problems is no longer at the initial phase.
Therefore, it becomes important to shift from exploratory uses to well targeted and rational
implementation of the diverse algorithmic options, since different AI methods can lead to various levels
of performance and accuracy depending on the application. It is thus important that future studies present
a clear rationale for the chosen AI method(s). Another important issue is computational efficiency.
Commonly, the performance of an AI method can be defined in terms of accuracy and computational
efficiency (i.e., less simulation/computational time). It should be noted that some of the publications
studied in this review indicated the good performance of AI method being used, even though the method
was found to be computationally expensive. Hence, it is of importance that future studies consider this
issue such that the AI methods being used result in good accuracy while also being computationally
efficient.

Measurement noise, modeling errors, environmental effects, etc., are unavoidable factors that could
significantly affect data availability. It is thus essential to use AI methods that can effectively interpret
incomplete and noisy data, and to assess their performance under these influences. Uncertainty analyses
could be used for this purpose. The selection of optimal parameters/hyper-parameters can also
significantly affect the performance of AI methods. Thus, future studies implementing AI techniques
should take into account the noted issue such that optimum algorithmic parameters are chosen. Finally,
clear presentation of the process by which the dataset is prepared and pre-processed (i.e., training,
validation, and testing) is essential to properly assess the performance of the implemented AI-based
methodology.
Table 3. Comparison of different AI methods for structural engineering applications

AI Methods

30
 Pattern Recognition Machine Learning Deep Learning
Applicable for traditional and Applicable for vision-based
Applicable for traditional and
data-driven SHM systems SHM systems
data-driven SHM systems
Can be integrated with IoT for Effective while dealing with
Do not necessarily need vast
smart applications large amount of dataset
amount of data
Advantages Applicable for optimization Applicable for interpretation
Can be effectively used for
problems of Big data in smart cities
classification and recognition
Do not necessarily need vast Can be integrated with IoT
problems
amount of data for smart applications
Computationally efficient Computationally efficient
Cannot be directly integrated Cannot be effectively used
Cannot be used for new vision-
with Iot for smart and for traditional SHM systems
Disadvantages based SHM systems based on
intelligent applications Need vast amount of data
images
It does not imply learning for efficient performance

7. CONCLUSIONS
This review paper presented the significance of emerging AI methods for structural engineering
applications during the last decade. The survey indicated that among the numerous AI methods, pattern
recognition (PR), machine learning (ML), and deep learning (DL) have been increasingly adapted and
used for SHM and damage identification, optimization, modeling concrete properties, structural
identification, earthquake engineering, etc. Yet, the common use of the noted methods has been for
interpreting sensor data in SHM. The survey revealed that ML, PR, and DL algorithmic techniques have
the ability to learn complicated interrelations among the contributing parameters, and thus allow solving a
diversity of problems that are difficult, or not possible, to solve with traditional methods.

Based on the literature survey, potential research avenues for employing PR, ML, and DL were also
presented. Considering the emerging use of wireless sensor networks (e.g., self-powered sensor
networks), ML- and PR-based models could become the next generation approaches to conduct non-
destructive structural and material evaluation in SHM. This review showed that ML methods are able to
discover hidden information about the structure’s performance by learning the influence of various
damage or degrading mechanisms and the data collected from sensors, leading to reliable and efficient
SHM frameworks. The literature further suggests that ML and DL techniques could also be applied to the
computational mechanics domain, such as to optimize processes in the finite element method to enhance
computational efficiency. These methods can also be used to solve complex problems through the novel
concept of the Internet of Things (IoT). On this basis, ML and DL architectures (e.g., convolutional
neural networks) within the context of IoT can be used to analyze and interpret complex and big data.
Further, the integration of ML and IoT can result in the creation of novel SHM systems employing
diverse and noisy sensor data. DL architectures can also be incorporated with IoT to develop unique
frameworks for use in smart cities. Data interpretation systems, which are part of the noted frameworks in
smart cities, can thus be optimized using such intelligent architectures.

Finally, the review was also used to identify general challenges and limitations on the use of AI
techniques. Among those limitations is the lack of rational selection of the AI method, disregarding the
effect of missing/incomplete and noisy data, discarding considerations for computational efficiency,
reporting classification accuracy without exploring alternative solutions to increase performance, and
insufficiency presentation of the process to select optimal parameters for the AI technique. However, it
was concluded that by addressing the noted issues/limitations in future studies, ML, PR, and DL could

31
represent pioneering methods to increase the efficiency of many current structural engineering
applications as well as for the creation of innovative uses.

ACKNOWLEDGMENTS
The research described in this paper was carried out with funding from the U.S. National Science
Foundation under grant number CNS-1405273.

REFERENCES
[1] Russell SJ, Norvig P, Canny JF, Malik JM, Edwards DD. Artificial intelligence: a modern approach. vol. 2.
Prentice hall Upper Saddle River; 2003.
[2] Back T. Evolutionary computation: Toward a new philosophy of machine intelligence. Complexity
1997;2:28–30. doi:10.1002/(SICI)1099-0526(199703/04)2:4<28::AID-CPLX7>3.0.CO;2-2.
[3] Fadlullah ZM, Tang F, Mao B, Kato N, Akashi O, Inoue T, et al. State-of-the-Art Deep Learning: Evolving
Machine Intelligence Toward Tomorrow’s Intelligent Network Traffic Control Systems. IEEE Commun
Surv Tutor 2017;19:2432–55. doi:10.1109/COMST.2017.2707140.
[4] Modha DS, Ananthanarayanan R, Esser SK, Ndirango A, Sherbondy AJ, Singh R. Cognitive computing.
Commun ACM 2011;54:62–71.
[5] Noor AK. Potential of cognitive computing and cognitive systems. Open Eng 2015;5:75–88.
[6] Adeli H, Hung S-L. Machine learning: neural networks, genetic algorithms, and fuzzy systems. John Wiley
& Sons, Inc.; 1994.
[7] Kicinger R, Arciszewski T, Jong KD. Evolutionary computation and structural design: A survey of the state-
of-the-art. Comput Struct 2005;83:1943–78. doi:10.1016/j.compstruc.2005.03.002.
[8] Liao TW, Egbelu P, Sarker B, Leu S. Metaheuristics for project and construction management–A state-of-
the-art review. Autom Constr 2011;20:491–505.
[9] Lu P, Chen S, Zheng Y. Artificial Intelligence in Civil Engineering. Math Probl Eng 2012;2012:1–22.
doi:10.1155/2012/145974.
[10] Shahin MA. Artificial intelligence in geotechnical engineering: applications, modeling aspects, and future
directions. Metaheuristics Water Geotech. Transp. Eng., Elsevier; 2013, p. 169–204.
[11] Saka MP, Geem ZW. Mathematical and metaheuristic applications in design optimization of steel frame
structures: an extensive review. Math Probl Eng 2013;2013.
[12] Aldwaik M, Adeli H. Advances in optimization of highrise building structures. Struct Multidiscip Optim
2014;50:899–919.
[13] Mardani A, Jusoh A, Zavadskas EK. Fuzzy multiple criteria decision-making techniques and applications–
Two decades review from 1994 to 2014. Expert Syst Appl 2015;42:4126–48.
[14] Penadés-Plà V, García-Segura T, Martí JV, Yepes V. A review of multi-criteria decision-making methods
applied to the sustainable bridge design. Sustainability 2016;8:1295.
[15] Amezquita-Sanchez J, Valtierra-Rodriguez M, Aldwaik M, Adeli H. Neurocomputing in civil infrastructure.
Sci Iran Trans C Chem Chem Eng 2016;23:2417.
[16] Pongiglione M, Calderini C. Sustainable structural design: Comprehensive literature review. J Struct Eng
2016;142:04016139.
[17] Nasiri S, Khosravani MR, Weinberg K. Fracture mechanics and mechanical fault detection by artificial
intelligence methods: A review. Eng Fail Anal 2017;81:270–93.
[18] Zamarrón-Mieza I, Yepes V, Moreno-Jiménez JM. A systematic review of application of multi-criteria
decision analysis for aging-dam management. J Clean Prod 2017;147:217–30.
[19] Sierra-Varela LA, Yepes V, Pellicer E. A Review of Multi-Criteria Assessment of the Social Sustainability
of Infrastructures. J Clean Prod 2018.
[20] Zavadskas EK, Antucheviciene J, Vilutiene T, Adeli H. Sustainable decision-making in civil engineering,
construction and building technology. Sustainability 2017;10:14.
[21] Krippendorff K. Content analysis: An introduction to its methodology. Sage; 2012.
[22] Pedrycz W. Fuzzy sets in pattern recognition: methodology and methods. Pattern Recognit 1990;23:121–46.
[23] Siddique N, Adeli H. Computational intelligence: synergies of fuzzy logic, neural networks and evolutionary
computing. John Wiley & Sons; 2013.

32
[24] Bezdek J. What is Computational Intelligence? Comput Intell Imitating Life JM Zurada RJ Marks II CJ
Robinson Eds IEEE Press 1994:1–12.
[25] Artificial Intelligence: A New Synthesis. Elsevier; 1998. doi:10.1016/C2009-0-27773-7.
[26] Computational intelligence: a logical approach. Choice Rev Online 1998;35:35-5701-35–5701.
doi:10.5860/CHOICE.35-5701.
[27] Kurzweil R. The Age of Spiritual Machines: When Computers Exceed Human Intelligence. New York, NY:
Penguin Books; 2000.
[28] Luger GF. Artificial Intelligence: Structures and Strategies for Complex Problem Solving. 6 edition. Boston:
Pearson; 2008.
[29] Yuen K-V. Bayesian methods for structural dynamics and civil engineering. John Wiley & Sons; 2010.
[30] Lombaert G, Moaveni B, He X, Conte JP. Damage identification of a seven-story reinforced concrete shear
wall building using Bayesian model updating. Proc IMAC-XXVII 2009:9–12.
[31] Duda RO, Hart PE, Stork DG. Pattern Classification. 2 edition. New York: Wiley-Interscience; 2000.
[32] theodoridis S, Koutroumbas K. Pattern Recognition. 4th Edition. Elsevier; 2008.
[33] Bishop C. Pattern Recognition and Machine Learning. Springer N Y 2007.
[34] Cherkassky V, Mulier FM. Learning from data: concepts, theory, and methods. John Wiley & Sons; 2007.
[35] Michalski RS, Carbonell JG, Mitchell TM. Machine learning: An artificial intelligence approach. Springer
Science & Business Media; 2013.
[36] Alpaydin E. Introduction to machine learning. MIT press; 2014.
[37] Robert C. Machine learning, a probabilistic perspective 2014.
[38] Marsland S. Machine learning: an algorithmic perspective. CRC press; 2015.
[39] Tong S, Chang E. Support vector machine active learning for image retrieval, ACM Press; 2001, p. 107.
doi:10.1145/500141.500159.
[40] Sebe N. Machine learning in computer vision. vol. 29. Springer Science & Business Media; 2005.
[41] Rosten E, Drummond T. Machine Learning for High-Speed Corner Detection. In: Leonardis A, Bischof H,
Pinz A, editors. Comput. Vis. – ECCV 2006, vol. 3951, Berlin, Heidelberg: Springer Berlin Heidelberg;
2006, p. 430–43. doi:10.1007/11744023_34.
[42] Lézoray O, Charrier C, Cardot H, Lefèvre S. Machine learning in image processing 2008.
[43] Bradski G, Kaehler A. Learning OpenCV: Computer vision with the OpenCV library. O’Reilly Media, Inc.;
2008.
[44] Di K, Li W, Yue Z, Sun Y, Liu Y. A machine learning approach to crater detection from topographic data.
Adv Space Res 2014;54:2419–29. doi:10.1016/j.asr.2014.08.018.
[45] Duygulu P, Barnard K, de Freitas JF, Forsyth DA. Object recognition as machine translation: Learning a
lexicon for a fixed image vocabulary, Springer; 2002, p. 97–112.
[46] Dede G, Sazlı MH. Speech recognition with artificial neural networks. Digit Signal Process 2010;20:763–8.
[47] Hinton G, Deng L, Yu D, Dahl G, Mohamed A, Jaitly N, et al. Deep Neural Networks for Acoustic
Modeling in Speech Recognition: The Shared Views of Four Research Groups. IEEE Signal Process Mag
2012;29:82–97. doi:10.1109/MSP.2012.2205597.
[48] Deng L, Hinton G, Kingsbury B. New types of deep neural network learning for speech recognition and
related applications: an overview, IEEE; 2013, p. 8599–603. doi:10.1109/ICASSP.2013.6639344.
[49] Graves A, Mohamed A, Hinton G. Speech recognition with deep recurrent neural networks, IEEE; 2013, p.
6645–9. doi:10.1109/ICASSP.2013.6638947.
[50] Deng L, Li X. Machine learning paradigms for speech recognition: An overview. IEEE Trans Audio Speech
Lang Process 2013;21:1060–89.
[51] LeBaron B. Agent-based computational finance. Handb Comput Econ 2006;2:1187–233.
[52] Brabazon A, O’Neill M. Natural computing in computational finance. vol. 100. Springer Science & Business
Media; 2008.
[53] Harris T. Credit scoring using the clustered support vector machine. Expert Syst Appl 2015;42:741–50.
[54] Sharma N, Sharma P, Irwin D, Shenoy P. Predicting solar generation from weather forecasts using machine
learning, IEEE; 2011, p. 528–33.
[55] Marvuglia A, Messineo A. Monitoring of wind farms’ power curves using machine learning techniques.
Appl Energy 2012;98:574–83.
[56] Wan C, Xu Z, Pinson P, Dong ZY, Wong KP. Probabilistic forecasting of wind power generation using
extreme learning machine. IEEE Trans Power Syst 2014;29:1033–44.
[57] Needham CJ, Bradford JR, Bulpitt AJ, Westhead DR. A primer on learning in Bayesian networks for
computational biology. PLoS Comput Biol 2007;3:e129.

33
[58] Ben-Hur A, Ong CS, Sonnenburg S, Schölkopf B, Rätsch G. Support vector machines and kernels for
computational biology. PLoS Comput Biol 2008;4:e1000173.
[59] Che D, Liu Q, Rasheed K, Tao X. Decision tree and ensemble learning algorithms with their applications in
bioinformatics. Softw. Tools Algorithms Biol. Syst., Springer; 2011, p. 191–9.
[60] Reich Y. Machine learning techniques for civil engineering problems. Comput Civ Infrastruct Eng
1997;12:295–310.
[61] Kanevski M, Pozdnoukhov A, Timonin V. Machine learning for spatial environmental data: theory,
applications, and software. EPFL press; 2009.
[62] Ciresan DC, Meier U, Masci J, Maria Gambardella L, Schmidhuber J. Flexible, high performance
convolutional neural networks for image classification. vol. 22, Barcelona, Spain; 2011, p. 1237.
[63] Zeiler MD, Fergus R. Visualizing and understanding convolutional networks, Springer; 2014, p. 818–33.
[64] Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition. ArXiv
Prepr ArXiv14091556 2014.
[65] Russakovsky O, Deng J, Su H, Krause J, Satheesh S, Ma S, et al. Imagenet large scale visual recognition
challenge. Int J Comput Vis 2015;115:211–52.
[66] Szegedy C, Liu W, Jia Y, Sermanet P, Reed S, Anguelov D, et al. Going deeper with convolutions, Cvpr;
2015.
[67] He K, Zhang X, Ren S, Sun J. Deep residual learning for image recognition, 2016, p. 770–8.
[68] Gu J, Wang Z, Kuen J, Ma L, Shahroudy A, Shuai B, et al. Recent advances in convolutional neural
networks. Pattern Recognit 2017.
[69] Krizhevsky A, Sutskever I, Hinton GE. Imagenet classification with deep convolutional neural networks,
2012, p. 1097–105.
[70] Karpathy A, Toderici G, Shetty S, Leung T, Sukthankar R, Fei-Fei L. Large-scale video classification with
convolutional neural networks, 2014, p. 1725–32.
[71] Santos JP, Cremona C, Orcesi AD, Silveira P. Early damage detection based on pattern recognition and data
fusion. J Struct Eng 2016;143:04016162.
[72] Sohn H, Czarnecki JA, Farrar CR. Structural health monitoring using statistical process control. J Struct Eng
2000;126:1356–63.
[73] Farrar CR, Sohn H, Park GH. A statistical pattern recognition paradigm for structural health monitoring. Los
Alamos National Laboratory; 2004.
[74] Farrar CR, Worden K. Fundamental axioms of structural health monitoring. Struct Health Monit Mach
Learn Perspect 2005:439–60.
[75] Sohn H, Farrar CR, Hunter NF, Worden K. Structural health monitoring using statistical pattern recognition
techniques. J Dyn Syst Meas Control 2001;123:706–11.
[76] Noman AS, Deeba F, Bagchi A. Health monitoring of structures using statistical pattern recognition
techniques. J Perform Constr Facil 2012;27:575–84.
[77] Gul M, Catbas FN. Statistical pattern recognition for structural health monitoring using time series
modeling: Theory and experimental verifications. Mech Syst Signal Process 2009;23:2192–204.
[78] Sohn H, Farrar C, Hunter N, Worden K. Applying the LANL statistical pattern recognition paradigm for
structural health monitoring to data from a surface-effect fast patrol boat. Los Alamos National Lab., NM
(US); 2001.
[79] Farrar CR, Sohn H. Pattern recognition for structural health monitoring, 2000.
[80] Worden K, Manson G, Fieller NR. Damage detection using outlier analysis. J Sound Vib 2000;229:647–67.
[81] Worden K, Manson G, Allman D. Experimental validation of a structural health monitoring methodology:
Part I. Novelty detection on a laboratory structure. J Sound Vib 2003;259:323–43.
[82] Manson G, Worden K, Allman D. Experimental validation of a structural health monitoring methodology:
Part II. Novelty detection on a Gnat aircraft. J Sound Vib 2003;259:345–63.
[83] Manson G, Worden K, Allman D. Experimental validation of a structural health monitoring methodology:
Part III. Damage location on an aircraft wing. J Sound Vib 2003;259:365–85.
[84] Nair KK, Kiremidjian AS, Law KH. Time series-based damage detection and localization algorithm with
application to the ASCE benchmark structure. J Sound Vib 2006;291:349–68.
[85] Nair KK, Kiremidjian AS. Time series based structural damage detection algorithm using Gaussian mixtures
modeling. J Dyn Syst Meas Control 2007;129:285–93.
[86] Cheung A, Cabrera C, Sarabandi P, Nair K, Kiremidjian A, Wenzel H. The application of statistical pattern
recognition methods for damage detection to field data. Smart Mater Struct 2008;17:065023.

34
[87] Lanata F, Schoefs F. Multi-algorithm approach for identification of structural behavior of complex structures
under cyclic environmental loading. Struct Health Monit 2012;11:51–67.
[88] Posenato D, Kripakaran P, Inaudi D, Smith IF. Methodologies for model-free data interpretation of civil
engineering structures. Comput Struct 2010;88:467–82.
[89] Zhou H, Ni Y, Ko J. Constructing input to neural networks for modeling temperature-caused modal
variability: mean temperatures, effective temperatures, and principal components of temperatures. Eng
Struct 2010;32:1747–59.
[90] Cury A, Crémona C, Diday E. Application of symbolic data analysis for structural modification assessment.
Eng Struct 2010;32:762–75.
[91] Santos JP, Crémona C, Orcesi AD, Silveira P. Multivariate statistical analysis for early damage detection.
Eng Struct 2013;56:273–85.
[92] Hsu T, Loh C. Damage detection accommodating nonlinear environmental effects by nonlinear principal
component analysis. Struct Control Health Monit 2010;17:338–54.
[93] de Lautour OR, Omenzetter P. Damage classification and estimation in experimental structures using time
series analysis and pattern recognition. Mech Syst Signal Process 2010;24:1556–69.
[94] Balsamo L, Betti R, Beigi H. A structural health monitoring strategy using cepstral features. J Sound Vib
2014;333:4526–42.
[95] Yun GJ, Lee S-G, Carletta J, Nagayama T. Decentralized damage identification using wavelet signal
analysis embedded on wireless smart sensors. Eng Struct 2011;33:2162–72.
[96] Kesavan KN, Kiremidjian AS. A wavelet‐ based damage diagnosis algorithm using principal component
analysis. Struct Control Health Monit 2012;19:672–85.
[97] Mujica L, Rodellar J, Fernandez A, Güemes A. Q-statistic and T2-statistic PCA-based measures for damage
assessment in structures. Struct Health Monit 2011;10:539–53.
[98] Cury A, Cremona C, Dumoulin J. Long-term monitoring of a PSC box girder bridge: operational modal
analysis, data normalization and structural modification assessment. Mech Syst Signal Process 2012;33:13–
37.
[99] Worden K, Staszewski WJ, Hensman JJ. Natural computing for mechanical systems research: A tutorial
overview. Mech Syst Signal Process 2011;25:4–111.
[100] Andre J, Kiremidjian A, Liao Y, Georgakis CT. Structural health monitoring approach for detecting ice
accretion on bridge cables using the autoregressive model, Taylor & Francis; 2016, p. 431–431.
[101] Yao R, Pakzad SN. Autoregressive statistical pattern recognition algorithms for damage detection in civil
structures. Mech Syst Signal Process 2012;31:355–68.
[102] Lam HF, Ng CT. The selection of pattern features for structural damage detection using an extended
Bayesian ANN algorithm. Eng Struct 2008;30:2762–70.
[103] Ng C-T. Application of Bayesian-designed artificial neural networks in Phase II structural health monitoring
benchmark studies. Aust J Struct Eng 2014;15:27–36.
[104] Radhika S, Tamura Y, Matsui M. Cyclone damage detection on building structures from pre-and post-
satellite images using wavelet based pattern recognition. J Wind Eng Ind Aerodyn 2015;136:23–33.
[105] Alves V, Cury A, Roitman N, Magluta C, Cremona C. Structural modification assessment using supervised
learning methods applied to vibration data. Eng Struct 2015;99:439–48.
[106] Goswami S, Bhattacharya P. A Scalable Neural-Network Modular-Array Architecture for Real-Time Multi-
Parameter Damage Detection in Plate Structures Using Single Sensor Output. Int J Comput Intell Appl
2012;11:1250024.
[107] Bandara RP, Chan TH, Thambiratnam DP. Frequency response function based damage identification using
principal component analysis and pattern recognition technique. Eng Struct 2014;66:116–28.
[108] Ramos LF, Miranda T, Mishra M, Fernandes FM, Manning E. A Bayesian approach for NDT data fusion:
The Saint Torcato church case study. Eng Struct 2015;84:120–9.
[109] Sierra-Pérez J, Torres-Arredondo MA, Güemes A. Damage and nonlinearities detection in wind turbine
blades based on strain field pattern recognition. FBGs, OBR and strain gauges comparison. Compos Struct
2016;135:156–66.
[110] Alavi AH, Hasni H, Lajnef N, Chatti K, Faridazar F. An intelligent structural damage detection approach
based on self-powered wireless sensor data. Autom Constr 2016;62:24–44.
[111] Loh C-H, Huang Y-T, Hsueh W, Chen J-D, Lin P-Y. Visualization and Dimension Reduction of High
Dimension Data for Structural Damage Detection. Procedia Eng 2017;188:17–24.

35
[112] Salehi H, Das S, Chakrabartty S, Biswas S, Burgueño R. Structural assessment and damage identification
algorithms using binary data, American Society of Mechanical Engineers; 2015, p. V002T05A011-
V002T05A011.
[113] Salehi H, Burgueño R, Das S, Biswas S, Chakrabartty S. Structural health monitoring from discrete binary
data through pattern recognition. Insights Innov Struct Eng Mech Comput 2016:1840–5.
[114] Salehi H, Das S, Chakrabartty S, Biswas S, Burgueño R. Structural damage identification using image-based
pattern recognition on event-based binary data generated from self-powered sensor networks. Struct Control
Health Monit 2018:e2135. doi:10.1002/stc.2135.
[115] Datteo A, Lucà F. Statistical pattern recognition approach for long-time monitoring of the G. Meazza
stadium by means of AR models and PCA. Eng Struct 2017;153:317–33.
[116] Zhou Y-L, Wahab MA. Cosine based and extended transmissibility damage indicators for structural damage
detection. Eng Struct 2017;141:175–83.
[117] de Lautour OR, Omenzetter P. Classification of damage using time series analysis and statistical pattern
recognition, 2008, p. 1055–63.
[118] Zhang J, Sato T, Iai S, Hutchinson T. A pattern recognition technique for structural identification using
observed vibration signals: Linear case studies. Eng Struct 2008;30:1439–46.
[119] Zhang J, Sato T, Iai S, Hutchinson T. A pattern recognition technique for structural identification using
observed vibration signals: Nonlinear case studies. Eng Struct 2008;30:1417–23.
[120] De Lautour OR, Omenzetter P. Prediction of seismic-induced structural damage using artificial neural
networks. Eng Struct 2009;31:600–6.
[121] Laory I, Trinh TN, Smith IF, Brownjohn JM. Methodologies for predicting natural frequency variation of a
suspension bridge. Eng Struct 2014;80:211–21.
[122] Bandara RP, Chan TH, Thambiratnam DP. Structural damage detection method using frequency response
functions. Struct Health Monit 2014;13:418–29.
[123] Tibaduiza DA, Mujica LE, Rodellar J, Güemes A. Structural damage detection using principal component
analysis and damage indices. J Intell Mater Syst Struct 2016;27:233–48.
[124] Elwood E, Corotis RB. Application of Fuzzy Pattern Recognition of Seismic Damage to Concrete
Structures. ASCE-ASME J Risk Uncertain Eng Syst Part Civ Eng 2015;1:04015011.
[125] Riveros CA, Utsunomiya T, Maeda K, Itoh K. Damage detection in flexible risers using statistical pattern
recognition techniques. Int J Offshore Polar Eng 2008;18.
[126] Li X, Ramirez C, Hines EL, Leeson MS, Purnell P, Pharaoh M. Pattern recognition of fiber-reinforced
plastic failure mechanism using computational intelligence techniques, IEEE; 2008, p. 2340–5.
[127] Alvanitopoulos P, Andreadis I, Elenas A. A new algorithm for the classification of earthquake damages in
structures, 2008, p. 151–6.
[128] Chen B, Zang C. Artificial immune pattern recognition for structure damage classification. Comput Struct
2009;87:1394–407.
[129] Chen B, Zang C. Artificial immune pattern recognition for damage detection in structural health monitoring
sensor networks. vol. 7293, International Society for Optics and Photonics; 2009, p. 72930K.
[130] Park S, Inman DJ, Yun C-B. An outlier analysis of MFC-based impedance sensing data for wireless
structural health monitoring of railroad tracks. Eng Struct 2008;30:2792–9.
[131] Omenzetter P, de Lautour OR. Classification of damage in structural systems using time series analysis and
supervised and unsupervised pattern recognition techniques. vol. 7647, International Society for Optics and
Photonics; 2010, p. 76474S.
[132] de Lautour OR, Omenzetter P. Nearest neighbor and learning vector quantization classification for damage
detection using time series analysis. Struct Control Health Monit 2010;17:614–31.
[133] Ren W, Lin Y, Fang S. Structural damage detection based on stochastic subspace identification and
statistical pattern recognition: I. Theory. Smart Mater Struct 2011;20:115009.
[134] Lin Y, Ren W, Fang S. Structural damage detection based on stochastic subspace identification and
statistical pattern recognition: II. Experimental validation under varying temperature. Smart Mater Struct
2011;20:115010.
[135] Soroushnia S, Tafreshi ST, Omidinasab F, Beheshtian N, Soroushnia S. Seismic performance of RC elevated
water tanks with frame staging and exhibition damage pattern. Procedia Eng 2011;14:3076–87.
[136] Mosavi AA, Dickey D, Seracino R, Rizkalla SH. Time-series models for identifying damage location in
structural members subjected to ambient vibrations. vol. 7650, International Society for Optics and
Photonics; 2010, p. 76502N.

36
[137] Reddy TA, Devi KR, Gangashetty SV. Multilayer feedforward neural network models for pattern
recognition tasks in earthquake engineering, Springer; 2011, p. 154–62.
[138] Qiao L, Esmaeily A, Melhem HG. Signal pattern recognition for damage diagnosis in structures. Comput
Civ Infrastruct Eng 2012;27:699–710.
[139] Goswami S, Bhattacharya P. Pattern Recognition for damage detection in aerospace vehicle structures,
IEEE; 2012, p. 178–82.
[140] Salamone S, Veletzos MJ, Lanza di Scalea F, Restrepo JI. Detection of initial yield and onset of failure in
bonded posttensioned concrete beams. J Bridge Eng 2011;17:966–74.
[141] Cury A, Crémona C. Pattern recognition of structural behaviors based on learning algorithms and symbolic
data concepts. Struct Control Health Monit 2012;19:161–86.
[142] Wang Z, Chen S, Lederman G, Cerda F, Bielak J, Garrett J, et al. Comparison of sparse representation and
fourier discriminant methods: damage location classification in indirect lab-scale bridge structural health
monitoring, 2013, p. 436–46.
[143] Cremona C, Cury A, Orcesi A. Supervised learning algorithms for damage detection and long term bridge
monitoring. Measurement 2012;4:2.
[144] Cunha A, Caetano E, Ribeiro P, Müller G. An indirect bridge inspection method incorporating a wavelet-
based damage indicator and pattern recognition n.d.
[145] Xiao H, Lu C, Ogai H, Roy K. A novel bridge structure damage diagnosis algorithm based on statistical
pattern recognition, IEEE; 2014, p. 775–80.
[146] Marti-Vargas JR, Ferri FJ, Yepes V. Prediction of the transfer length of prestressing strands with neural
networks. Comput Concr 2013;12:187–209.
[147] Ho H-N, Kim K-D, Park Y-S, Lee J-J. An efficient image-based damage detection for cable surface in cable-
stayed bridges. Ndt E Int 2013;58:18–23.
[148] Ng C-T. On the selection of advanced signal processing techniques for guided wave damage identification
using a statistical approach. Eng Struct 2014;67:50–60.
[149] Hwang J-H, Joo B-C, Yoo Y-J, Park K-T, Lee C-H. Damage detection of a prototype building structure
under shaking table testing using outlier analysis. vol. 8695, International Society for Optics and Photonics;
2013, p. 86953H.
[150] Mustapha S, Hu Y, Nguyen K, Alamdari MM, Runcie P, Dackermann U, et al. Pattern recognition based on
time series analysis using vibration data for structural health monitoring in civil structures. Electron J Struct
Eng 2015.
[151] McGetrick P, Kim C. Experimental investigation of a wavelet based drive-by bridge inspection system
incorporating pattern recognition. Life-Cycle Struct Syst Des Assess Maint Manag 2014:333.
[152] Balsamo L, Betti R. Data‐ based structural health monitoring using small training data sets. Struct Control
Health Monit 2015;22:1240–64.
[153] Hakim S, Razak HA, Ravanfar S, Mohammadhassani M. Structural Damage Detection Using Soft
Computing Method. Struct. Health Monit. Vol. 5, Springer; 2014, p. 143–51.
[154] Heo G, Kim C, Lee C, Hur J, Seo S. A damage assessment technique based on a revised Statistical Pattern-
recognition Technique (SPRT). KSCE J Civ Eng 2017;21:882–8.
[155] Mondoro A, Frangopol DM, Soliman M. Optimal risk-based management of coastal bridges vulnerable to
hurricanes. J Infrastruct Syst 2016;23:04016046.
[156] Crémona C, da Silveira APC, de Oliveira Martins LC. Real‐ time damage detection based on pattern
recognition. Struct Concr 2016;17:338–54.
[157] Vitola J, Pozo F, Tibaduiza DA, Anaya M. A sensor data fusion system based on k-nearest neighbor pattern
classification for structural health monitoring applications. Sensors 2017;17:417.
[158] Abdeljaber O, Avci O. Nonparametric structural damage detection algorithm for ambient vibration response:
utilizing artificial neural networks and self-organizing maps. J Archit Eng 2016;22:04016004.
[159] Mallik N, Wali A, Kuri N. Damage location identification through neural network learning from optical
fiber signal for structural health monitoring, ACM; 2016, p. 157–61.
[160] Khoshnoudian F, Talaei S. A new damage index using FRF data, 2D-PCA method and pattern recognition
techniques. Int J Struct Stab Dyn 2017;17:1750090.
[161] Bonessio N, Benzoni G, Lomiento G. A multi-mode approach for multi-directional damage detection in
frame structures. Eng Struct 2017;147:505–16.
[162] Shahsavari V, Chouinard L, Bastien J. Wavelet-based analysis of mode shapes for statistical detection and
localization of damage in beams using likelihood ratio test. Eng Struct 2017;132:494–507.

37
[163] Stone J, Blockley D, Pilsworth B. Towards machine learning from case histories. Civ Eng Syst 1989;6:129–
35.
[164] Arciszewski T, Mustafa M, Ziarko W. A methodology of design knowledge acquisition for use in learning
expert systems. Int J Man-Mach Stud 1987;27:23–32.
[165] Khoa NL, Zhang B, Wang Y, Chen F, Mustapha S. Robust dimensionality reduction and damage detection
approaches in structural health monitoring. Struct Health Monit 2014;13:406–17.
[166] Smarsly K, Dragos K, Wiggenbrock J. Machine learning techniques for structural health monitoring, 2016,
p. 5–8.
[167] Taffese WZ, Sistonen E. Machine learning for durability and service-life assessment of reinforced concrete
structures: Recent advances and future directions. Autom Constr 2017;77:1–14.
[168] Yan B, Cui Y, Zhang L, Zhang C, Yang Y, Bao Z, et al. Beam structure damage identification based on BP
neural network and support vector machine. Math Probl Eng 2014;2014.
[169] Gui G, Pan H, Lin Z, Li Y, Yuan Z. Data-driven support vector machine with optimization techniques for
structural health monitoring and damage detection. KSCE J Civ Eng 2017;21:523–34.
[170] Salehi H, Das S, Chakrabartty S, Biswas S, Burgueño R. A machine-learning approach for damage detection
in aircraft structures using self-powered sensor data. In: Lynch JP, editor., 2017, p. 101680X.
doi:10.1117/12.2260118.
[171] Nagarajaiah S, Yang Y. Modeling and harnessing sparse and low‐ rank data structure: a new paradigm for
structural dynamics, identification, damage detection, and health monitoring. Struct Control Health Monit
2017;24.
[172] Figueiredo E, Park G, Farrar CR, Worden K, Figueiras J. Machine learning algorithms for damage detection
under operational and environmental variability. Struct Health Monit 2011;10:559–72.
[173] Dervilis N. A machine learning approach to structural health monitoring with a view towards wind turbines
2013.
[174] Yan B, Cui Y, Zhang L, Zhang C, Yang Y, Bao Z, et al. Beam structure damage identification based on BP
neural network and support vector machine. Math Probl Eng 2014;2014.
[175] Butcher JB, Day CR, Austin JC, Haycock PW, Verstraeten D, Schrauwen B. Defect Detection in Reinforced
Concrete Using Random Neural Architectures: Defect detection in reinforced concrete using random neural
architectures. Comput-Aided Civ Infrastruct Eng 2014;29:191–207. doi:10.1111/mice.12039.
[176] Liu C, Liu J, Liu L. Study on the Damage Identification of Long-Span Cable-Stayed Bridge Based on
Support Vector Machine, IEEE; 2009, p. 1–4.
[177] Gong L, Wang C, Wu F, Zhang J, Zhang H, Li Q. Earthquake-induced building damage detection with post-
event sub-meter vhr terrasar-X staring spotlight imagery. Remote Sens 2016;8:887.
[178] Lederman G, Wang Z, Bielak J, Noh H, Garrett J, Chen S, et al. Damage quantification and localization
algorithms for indirect SHM of bridges, 2014.
[179] Dai H, Cao Z. A Wavelet Support Vector Machine‐ Based Neural Network Metamodel for Structural
Reliability Assessment. Comput Civ Infrastruct Eng 2017;32:344–57.
[180] Diez A, Khoa NLD, Alamdari MM, Wang Y, Chen F, Runcie P. A clustering approach for structural health
monitoring on bridges. J Civ Struct Health Monit 2016;6:429–45.
[181] Zheng W, Qian F. Promptly assessing probability of barge–bridge collision damage of piers through
probabilistic-based classification of machine learning. J Civ Struct Health Monit 2017;7:57–78.
[182] Santos A, Figueiredo E, Silva M, Santos R, Sales C, Costa JC. Genetic‐ based EM algorithm to improve the
robustness of Gaussian mixture models for damage detection in bridges. Struct Control Health Monit
2017;24.
[183] Neves A, González I, Leander J, Karoumi R. Structural health monitoring of bridges: a model-free ANN-
based approach to damage detection. J Civ Struct Health Monit 2017;7:689–702.
[184] Das S, Salehi H, Shi Y, Chakrabartty S, Burgueno R, Biswas S. Towards packet-less ultrasonic sensor
networks for energy-harvesting structures. Comput Commun 2017;101:94–105.
doi:10.1016/j.comcom.2016.11.001.
[185] Salehi H, Burgueño R, Das S, Biswas S, Chakrabartty S. Localized Damage Identification of Plate-Like
Structures With Time-Delayed Binary Data From a Self-Powered Sensor Network, American Society of
Mechanical Engineers; 2017, p. V002T05A008-V002T05A008.
[186] Yang Y, Sun P, Nagarajaiah S, Bachilo SM, Weisman RB. Full-field, high-spatial-resolution detection of
local structural damage from low-resolution random strain field measurements. J Sound Vib 2017;399:75–
85.

38
[187] Yang Y, Nagarajaiah S. Harnessing data structure for recovery of randomly missing structural vibration
responses time history: Sparse representation versus low-rank structure. Mech Syst Signal Process
2016;74:165–82.
[188] Yepes V, García-Segura T, Moreno-Jiménez J. A cognitive approach for the multi-objective optimization of
RC structural problems. Arch Civ Mech Eng 2015;15:1024–36.
[189] García-Segura T, Yepes V, Frangopol DM, Yang DY. Lifetime reliability-based optimization of post-
tensioned box-girder bridges. Eng Struct 2017;145:381–91.
[190] García-Segura T, Yepes V, Frangopol DM. Multi-objective design of post-tensioned concrete road bridges
using artificial neural networks. Struct Multidiscip Optim 2017;56:139–50.
[191] Chatterjee S, Sarkar S, Hore S, Dey N, Ashour AS, Shi F, et al. Structural failure classification for
reinforced concrete buildings using trained neural network based multi-objective genetic algorithm. Struct
Eng Mech 2017;63:429–38.
[192] Chou J-S, Tsai C-F, Pham A-D, Lu Y-H. Machine learning in concrete strength simulations: Multi-nation
data analytics. Constr Build Mater 2014;73:771–80.
[193] Yeh I-C, Lien L-C. Knowledge discovery of concrete material using genetic operation trees. Expert Syst
Appl 2009;36:5807–12.
[194] Demir F, Korkmaz KA. Prediction of lower and upper bounds of elastic modulus of high strength concrete.
Constr Build Mater 2008;22:1385–93.
[195] Cao YF, Wu W, Zhang HL, Pan JM. Prediction of the elastic modulus of self-compacting concrete based on
SVM. vol. 357, Trans Tech Publ; 2013, p. 1023–6.
[196] Aiyer BG, Kim D, Karingattikkal N, Samui P, Rao PR. Prediction of compressive strength of self-
compacting concrete using least square support vector machine and relevance vector machine. KSCE J Civ
Eng 2014;18:1753–8.
[197] Chen B-T, Chang T-P, Shih J-Y, Wang J-J. Estimation of exposed temperature for fire-damaged concrete
using support vector machine. Comput Mater Sci 2009;44:913–20.
[198] Bin C, Xuemang G, Guohua L. Prediction of concrete properties based on rough sets and support vector
machine method. J Hydroelectr Eng 2011;6:045.
[199] Cheng M-Y, Firdausi PM, Prayogo D. High-performance concrete compressive strength prediction using
Genetic Weighted Pyramid Operation Tree (GWPOT). Eng Appl Artif Intell 2014;29:104–13.
[200] XU C, YUN S, SHU Y. Concrete Strength Inspection Conversion Model Based on SVM [J]. J Luoyang Inst
Sci Technol Nat Sci Ed 2008;2:025.
[201] Yan K, Shi C. Prediction of elastic modulus of normal and high strength concrete by support vector
machine. Constr Build Mater 2010;24:1479–85.
[202] Yan K, Xu H, Shen G, Liu P. Prediction of splitting tensile strength from cylinder compressive strength of
concrete by support vector machine. Adv Mater Sci Eng 2013;2013.
[203] Prasad BR, Eskandari H, Reddy BV. Prediction of compressive strength of SCC and HPC with high volume
fly ash using ANN. Constr Build Mater 2009;23:117–28.
[204] Cheng M-Y, Chou J-S, Roy AF, Wu Y-W. High-performance concrete compressive strength prediction
using time-weighted evolutionary fuzzy support vector machines inference model. Autom Constr
2012;28:106–15.
[205] Sarıdemir M. Empirical modeling of splitting tensile strength from cylinder compressive strength of concrete
by genetic programming. Expert Syst Appl 2011;38:14257–68.
[206] Ahmadi-Nedushan B. Prediction of elastic modulus of normal and high strength concrete using ANFIS and
optimal nonlinear regression models. Constr Build Mater 2012;36:665–73.
[207] Lee S, Lee C. Prediction of shear strength of FRP-reinforced concrete flexural members without stirrups
using artificial neural networks. Eng Struct 2014;61:99–112.
[208] HIROKANE M, NOMURA Y, KUSUNOSE Y. Damage detection using support vector machine for
integrity assessment of concrete structures. Doboku Gakkai Ronbunshuu A 2008;64:739–49.
doi:10.2208/jsceja.64.739.
[209] Oh CK. Bayesian learning for earthquake engineering applications and structural health monitoring.
California Institute of Technology; 2008.
[210] Dai H, Zhao W, Wang W, Cao Z. An improved radial basis function network for structural reliability
analysis. J Mech Sci Technol 2011;25:2151.
[211] González MP, Zapico JL. Seismic damage identification in buildings using neural networks and modal data.
Comput Struct 2008;86:416–26.

39
[212] Soyoz S, Feng MQ. Long‐ Term Monitoring and Identification of Bridge Structural Parameters. Comput
Civ Infrastruct Eng 2009;24:82–92.
[213] Karimi I, Khaji N, Ahmadi M, Mirzayee M. System identification of concrete gravity dams using artificial
neural networks based on a hybrid finite element–boundary element approach. Eng Struct 2010;32:3583–91.
[214] Kim D, Philen M. Damage classification using Adaboost machine learning for structural health monitoring.
vol. 7981, International Society for Optics and Photonics; 2011, p. 79812A.
[215] Siddique R, Aggarwal P, Aggarwal Y. Prediction of compressive strength of self-compacting concrete
containing bottom ash using artificial neural networks. Adv Eng Softw 2011;42:780–6.
[216] Uysal M, Tanyildizi H. Estimation of compressive strength of self compacting concrete containing
polypropylene fiber and mineral additives exposed to high temperature using artificial neural network.
Constr Build Mater 2012;27:404–14.
[217] Mata J. Interpretation of concrete dam behaviour with artificial neural network and multiple linear regression
models. Eng Struct 2011;33:903–10.
[218] Liu C-C, Liu J. Damage identification of a long-span arch bridge based on support vector machine.
Zhendong Yu ChongjiJournal Vib Shock 2010;29:174–8.
[219] Chun-cheng L, Jiao L, Biao T. Damage Identification for Transmission Tower Based on Support Vector
Machine and RBF, 2010.
[220] Li Z, Burgueño R. Using soft computing to analyze inspection results for bridge evaluation and
management. J Bridge Eng 2010;15:430–8.
[221] Aminian P, Niroomand H, Gandomi AH, Alavi AH, Esmaeili MA. New design equations for assessment of
load carrying capacity of castellated steel beams: a machine learning approach. Neural Comput Appl
2013;23:119–31.
[222] Kao C, Loh C. Monitoring of long‐ term static deformation data of Fei‐ Tsui arch dam using artificial
neural network‐ based approaches. Struct Control Health Monit 2013;20:282–303.
[223] Kim H-J, Park W, Koh H-M, Choo JF. Identification of structural performance of a steel-box girder bridge
using machine learning technique. vol. 99, International Association for Bridge and Structural Engineering;
2013, p. 1313–20.
[224] Zhang W, Song Z. Prediction of Concrete Corrosion in Sulfuric Acid by SVM-Based Method, 2012.
[225] Bartram G, Mahadevan S. System Modeling for SHM Using Dynamic Bayesian Networks. Infotech Aerosp.
2012, 2012, p. 2423.
[226] Son H, Kim C, Kim C. Automated color model–based concrete detection in construction-site images by
using machine learning algorithms. J Comput Civ Eng 2011;26:421–33.
[227] Dantas ATA, Leite MB, de Jesus Nagahama K. Prediction of compressive strength of concrete containing
construction and demolition waste using artificial neural networks. Constr Build Mater 2013;38:717–22.
[228] Duan Z-H, Kou S-C, Poon C-S. Prediction of compressive strength of recycled aggregate concrete using
artificial neural networks. Constr Build Mater 2013;40:1200–6.
[229] Ying Y, Garrett Jr JH, Oppenheim IJ, Soibelman L, Harley JB, Shi J, et al. Toward data-driven structural
health monitoring: application of machine learning and signal processing to damage detection. J Comput Civ
Eng 2012;27:667–80.
[230] Satpal SB, Khandare Y, Guha A, Banerjee S. Structural health monitoring of a cantilever beam using
support vector machine. Int J Adv Struct Eng 2013;5:2.
[231] Cha Y-J, Buyukozturk O. Modal strain energy based damage detection using multi-objective optimization.
Struct. Health Monit. Vol. 5, Springer; 2014, p. 125–33.
[232] Long J, Buyukozturk O. Automated structural damage detection using one-class machine learning. Dyn. Civ.
Struct. Vol. 4, Springer; 2014, p. 117–28.
[233] Ranković V, Grujović N, Divac D, Milivojević N. Development of support vector regression identification
model for prediction of dam structural behaviour. Struct Saf 2014;48:33–9.
[234] Yang S, Fang CQ, Yuan ZJ. Study on Mechanical Properties of Corroded Reinforced Concrete Using
Support Vector Machines. vol. 578, Trans Tech Publ; 2014, p. 1556–61.
[235] Alimoradi A, Beck JL. Machine-learning methods for earthquake ground motion analysis and simulation. J
Eng Mech 2014;141:04014147.
[236] Kia A, Sensoy S. Classification of earthquake-induced damage for R/C slab column frames using multiclass
SVM and its combination with MLP neural network. Math Probl Eng 2014;2014.
[237] Jeon J, Shafieezadeh A, DesRoches R. Statistical models for shear strength of RC beam‐ column joints
using machine‐ learning techniques. Earthq Eng Struct Dyn 2014;43:2075–95.

40
[238] Wan H-P, Mao Z, Todd MD, Ren W-X. Analytical uncertainty quantification for modal frequencies with
structural parameter uncertainty using a Gaussian process metamodel. Eng Struct 2014;75:577–89.
[239] Su G, Yu B, Xiao Y, Yan L. Gaussian process machine-learning method for structural reliability analysis.
Adv Struct Eng 2014;17:1257–70.
[240] Motamedi S, Shamshirband S, Hashim R, Petković D, Roy C. Estimating unconfined compressive strength
of cockle shell–cement–sand mixtures using soft computing methodologies. Eng Struct 2015;98:49–58.
[241] Ataei N, Padgett JE. Fragility surrogate models for coastal bridges in hurricane prone zones. Eng Struct
2015;103:203–13.
[242] Ghiasi R, Torkzadeh P, Noori M. A machine-learning approach for structural damage detection using least
square support vector machine based on a new combinational kernel function. Struct Health Monit
2016;15:302–16.
[243] Karina CN, Chun P, Okubo K. Tensile strength prediction of corroded steel plates by using machine learning
approach. STEEL Compos Struct 2017;24:635–41.
[244] Mirhosseini RT. Seismic response of soil-structure interaction using the support vector regression. Struct
Eng Mech 2017;63:115–24.
[245] Rafiei MH, Adeli H. NEEWS: a novel earthquake early warning model using neural dynamic classification
and neural dynamic optimization. Soil Dyn Earthq Eng 2017;100:417–27.
[246] Vu D-T, Hoang N-D. Punching shear capacity estimation of FRP-reinforced concrete slabs using a hybrid
machine learning approach. Struct Infrastruct Eng 2016;12:1153–61.
[247] Lu N, Noori M, Liu Y. Fatigue reliability assessment of welded steel bridge decks under stochastic truck
loads via machine learning. J Bridge Eng 2016;22:04016105.
[248] Malekzadeh M, Catbas FN. A Machine Learning Framework for Automated Functionality Monitoring of
Movable Bridges. Dyn. Civ. Struct. Vol. 2, Springer; 2016, p. 57–63.
[249] Catbas FN, Malekzadeh M. A machine learning-based algorithm for processing massive data collected from
the mechanical components of movable bridges. Autom Constr 2016;72:269–78.
[250] Senniappan V, Subramanian J, Papageorgiou EI, Mohan S. Application of fuzzy cognitive maps for crack
categorization in columns of reinforced concrete structures. Neural Comput Appl 2017;28:107–17.
[251] Toghroli A, Suhatril M, Ibrahim Z, Safa M, Shariati M, Shamshirband S. Potential of soft computing
approach for evaluating the factors affecting the capacity of steel–concrete composite beam. J Intell Manuf
2016:1–9.
[252] Omran BA, Chen Q, Jin R. Comparison of data mining techniques for predicting compressive strength of
environmentally friendly concrete. J Comput Civ Eng 2016;30:04016029.
[253] Häckell MW, Rolfes R, Kane MB, Lynch JP. Three-tier modular structural health monitoring framework
using environmental and operational condition clustering for data normalization: Validation on an
operational wind turbine system. Proc IEEE 2016;104:1632–46.
[254] Chang J, Nagarajaiah S. Quantum-Behaved Particle Swarm Optimization-Based Structural Modal Parameter
Identification Under Ambient Excitation. Int J Struct Stab Dyn 2016;16:1550008.
[255] Chang J, Liu W, Hu H, Nagarajaiah S. Improved independent component analysis based modal
identification of higher damping structures. Measurement 2016;88:402–16.
[256] Hasni H, Alavi AH, Lajnef N, Abdelbarr M, Masri SF, Chakrabartty S. Self-powered piezo-floating-gate
sensors for health monitoring of steel plates. Eng Struct 2017;148:584–601.
[257] Hasni H, Alavi AH, Jiao P, Lajnef N. Detection of fatigue cracking in steel bridge girders: a support vector
machine approach. Arch Civ Mech Eng 2017;17:609–22.
[258] Yang Y, Nagarajaiah S. Robust data transmission and recovery of images by compressed sensing for
structural health diagnosis. Struct Control Health Monit 2017;24.
[259] Salehi H, Das S, Chakrabartty S, Biswas S, Burgueño R. A methodology for structural health diagnosis and
assessment using machine learning with noisy and incomplete data from self-powered wireless sensors.
Sens. Smart Struct. Technol. Civ. Mech. Aerosp. Syst., SPIE; 2018.
[260] Sarkar S, Reddy KK, Giering M, Gurvich MR. Deep Learning for Structural Health Monitoring: A Damage
Characterization Application n.d.
[261] Abdeljaber O, Avci O, Kiranyaz S, Gabbouj M, Inman DJ. Real-time vibration-based structural damage
detection using one-dimensional convolutional neural networks. J Sound Vib 2017;388:154–70.
[262] Avci O, Abdeljaber O, Kiranyaz S, Inman D. Structural Damage Detection in Real Time: Implementation of
1D Convolutional Neural Networks for SHM Applications. Struct. Health Monit. Damage Detect. Vol. 7,
Springer; 2017, p. 49–54.

41
[263] Abdeljaber O, Avci O, Kiranyaz MS, Boashash B, Sodano H, Inman DJ. 1-D CNNs for structural damage
detection: Verification on a structural health monitoring benchmark data. Neurocomputing 2018;275:1308–
17.
[264] Cha Y, Choi W, Büyüköztürk O. Deep Learning‐ Based Crack Damage Detection Using Convolutional
Neural Networks. Comput Civ Infrastruct Eng 2017;32:361–78.
[265] Gulgec NS, Takáč M, Pakzad SN. Structural Damage Detection Using Convolutional Neural Networks.
Model Valid. Uncertain. Quantif. Vol. 3, Springer; 2017, p. 331–7.
[266] Lee S, Ha J, Zokhirova M, Moon H, Lee J. Background Information of Deep Learning for Structural
Engineering. Arch Comput Methods Eng 2018;25:121–9.
[267] Huo Q, Dong B, Biswas S. A pulse switching paradigm for ultra low power cellular sensor networks.
Pervasive Mob Comput 2014;13:221–45.
[268] Cha Y, Choi W, Büyüköztürk O. Deep Learning‐ Based Crack Damage Detection Using Convolutional
Neural Networks. Comput Civ Infrastruct Eng 2017;32:361–78.
[269] Cha Y-J, Choi W. Vision-Based Concrete Crack Detection Using a Convolutional Neural Network. Dyn.
Civ. Struct. Vol. 2, Springer; 2017, p. 71–3.
[270] Kong X, Li J. Vision‐ Based Fatigue Crack Detection of Steel Structures Using Video Feature Tracking.
Comput Civ Infrastruct Eng 2018.
[271] Oishi A, Yagawa G. Computational mechanics enhanced by deep learning. Comput Methods Appl Mech
Eng 2017;327:327–51.
[272] Atzori L, Iera A, Morabito G. The internet of things: A survey. Comput Netw 2010;54:2787–805.
[273] Abdelgawad A, Yelamarthi K. Structural health monitoring: Internet of things application, IEEE; 2016, p. 1–
4.
[274] Abdelgawad A, Yelamarthi K. Internet of Things (IoT) Platform for Structure Health Monitoring. Wirel
Commun Mob Comput 2017;2017.
[275] Tokognon CA, Gao B, Tian GY, Yan Y. Structural health monitoring framework based on Internet of
Things: A survey. IEEE Internet Things J 2017;4:619–35.
[276] Schaffers H, Komninos N, Pallot M, Trousse B, Nilsson M, Oliveira A. Smart cities and the future internet:
Towards cooperation frameworks for open innovation, Springer; 2011, p. 431–46.
[277] Zanella A, Bui N, Castellani A, Vangelista L, Zorzi M. Internet of things for smart cities. IEEE Internet
Things J 2014;1:22–32.
[278] Perera C, Zaslavsky A, Christen P, Georgakopoulos D. Sensing as a service model for smart cities supported
by internet of things. Trans Emerg Telecommun Technol 2014;25:81–93.
[279] Kim T, Ramos C, Mohammed S. Smart city and IoT 2017.
[280] Wang L, Sng D. Deep learning algorithms with applications to video analytics for a smart city: A survey.
ArXiv Prepr ArXiv151203131 2015.
[281] Chin J, Callaghan V, Lam I. Understanding and personalising smart city services using machine learning,
The Internet-of-Things and Big Data, IEEE; 2017, p. 2050–5.
[282] Han B, Yu X, Ou J. Self-sensing concrete in smart structures. Butterworth-Heinemann; 2014.
[283] Konsta-Gdoutos MS, Aza CA. Self sensing carbon nanotube (CNT) and nanofiber (CNF) cementitious
composites for real time damage assessment in smart structures. Cem Concr Compos 2014;53:162–9.
[284] Gupta S, Gonzalez JG, Loh KJ. Self-sensing concrete enabled by nano-engineered cement-aggregate
interfaces. Struct Health Monit 2017;16:309–23.

42
Emerging artificial intelligence methods in structural engineering

Hadi Salehi1, Rigoberto Burgueño1,2

1
Department of Civil and Environmental Engineering, Michigan State University, East Lansing, Michigan, USA
2
Department of Mechanical Engineering, Michigan State University, East Lansing, Michigan, USA

Highlights:
 A review of recent applications of emerging artificial intelligence (AI) methods is
presented.
 The methods of pattern recognition, machine learning, and deep learning are studied.
 The advantages of employing novel AI methods in structural engineering are discussed.
 Potential research avenues for using AI methods in structural engineering are identified.

43