Artificial intelligence techniques for smart city applications

Daniel Luckey, Henrieke Fritz, Dmitrii Legatiuk, Kosmas Dragos, and Kay Smarsly

Bauhaus University Weimar, Chair of Computing in Civil Engineering, Weimar, Germany daniel.luckey@uni-weimar.de

Abstract. Recent developments in artificial intelligence (AI), in particular ma-chine learning (ML), have been significantly advancing smart city applications. Smart infrastructure, which is an essential component of smart cities, is equipped with wireless sensor networks that autonomously collect, analyze, and communicate structural data, referred to as “smart monitoring”. AI algorithms provide abilities to process large amounts of data and to detect patterns and fea-tures that would remain undetected using traditional approaches. Despite these capabilities, the application of AI algorithms to smart monitoring is still limited due to mistrust expressed by engineers towards the generally opaque AI inner processes. To enhance confidence in AI, the “black-box” nature of AI algo-rithms for smart monitoring needs to be explained to the engineers, resulting in so-called “explainable artificial intelligence” (XAI). However, when aiming at improving the explainability of AI algorithms through XAI for smart monitor-ing, the variety of AI algorithms requires proper categorization. Therefore, this review paper first identifies objectives of smart monitoring, serving as a basis to categorize AI algorithms or, more precisely, ML algorithms for smart monitor-ing. ML algorithms for smart monitoring are then reviewed and categorized. As a result, an overview of ML algorithms used for smart monitoring is presented, providing an overview of categories of ML algorithms for smart monitoring that may be modified to achieve explainable artificial intelligence in civil engineer-ing.

Keywords: Artificial intelligence (AI), machine learning (ML), smart cities, smart infrastructure, smart monitoring, explainable artificial intelligence (XAI).

1 Introduction

In the last decade, developments within the ongoing socioeconomic digitalization have created the vision of smart cities, which aspires to connect all aspects of urban life. The basis for connecting aspects of urban life in smart cities is being built around contemporary and emerging technologies, such as cloud computing, the Internet of Things and cyber-physical systems, representing the latest chain in industrial revolu-tion, referred to as Industry 4.0 (Acatech, 2015). A key aspect of the aforementioned technologies is adopting and advancing artificial intelligence (AI) techniques, which have proven their ability to process large amounts of data towards developing learn-

2

ing rules, e.g. via machine learning (ML), making complex associations, and predict-ing outcomes of complex physical processes. Although applications related to smart cities are expected to become a trillion-dollar market in the next five years (PWC, 2019), the term “smart city” has not been officially defined (OECD, 2019; Johnson, et al., 2019). However, several key components of smart cities have already been well-established, such as smart living, smart governance, smart citizen (people), smart mobility, smart economy, and smart infrastructure (Mohanty, et al., 2016).

Smart infrastructure is of particular importance for civil engineering and provides the foundation for key components of smart cities (UN Economic and Social Council, 2016). Therefore, smart infrastructure is considered the backbone of smart cities. Smart infrastructure is realized via smart (wireless) structural health monitoring (SHM) systems, referred to as “smart monitoring”, which enables timely detection of structural degradation, thus resulting in low maintenance, repair, and disruption costs (Ogie, et al., 2017). Because of aging infrastructure, smart monitoring has been gain-ing increasing popularity for leveraging the aforementioned benefits of smart infra-structure.

Smart monitoring fosters automation in SHM; therefore, aspects of SHM are essen-tial for defining objectives in smart monitoring. SHM is typically associated with structural condition assessment using structural response data and encompasses data acquisition, data communication, data analysis, data storage, and data retrieval. Spe-cifically, data analysis leads to conclusions drawn from structural response data, with respect to damage detection, damage classification, damage localization, condition assessment, and life-time prediction (Kabalci & Kabalci, 2019). Data analysis is usu-ally performed using data-driven models that extract information from structural re-sponse data. While several data-driven models draw from statistical processing and experimental mechanics, the increasing amounts of data in long-term monitoring sys-tems have fueled research in adopting AI algorithms for data analysis and processing. The intelligence inherent to AI algorithms is compatible with the automation neces-sary for smart monitoring, as part of smart infrastructure. Moreover, several AI algo-rithms used in smart monitoring are commonly referred to as “big data” algorithms and therefore serve a twofold purpose, (i) to detect patterns representing complex physical processes that otherwise would remain undetected, and (ii) to exploit, to the best possible extent, large amounts of data available in long-term SHM systems that are otherwise only partially utilized.

Smart monitoring, thus smart infrastructure, has taken advantage of distributed arti-ficial intelligence, a subfield of artificial intelligence. In particular, multi-agent tech-nology, representing a major branch of distributed artificial intelligence, has been deployed to advance different fields of smart monitoring, such as dam monitoring (Mittrup, et al., 2003), wind turbine monitoring (Hartmann, et al., 2011), and bridge monitoring (Smarsly, et al., 2007). Multi-agent systems have also been reported as an enabling technology of self-managing smart monitoring systems (Smarsly, et al., 2012) and process scheduling in smart infrastructure applications (Bilek, et al., 2003). Facilitating wireless smart infrastructure, multi-agent technology has been extended towards mobile multi-agent systems, as reported in (Smarsly & Law, 2013), proposed to enable agent-based software modules to autonomously migrate from one wireless

3

sensor node to another in an attempt to analyze smart infrastructure on demand. As could be demonstrated in a study presented in (Smarsly, et al., 2011), the mobile mul-ti-agent approach leads to significantly reduced resource consumption in wireless smart monitoring systems, as compared to traditional approaches. Further artificial intelligence techniques, such as neural networks (Dragos & Smarsly, 2016), support vector regression (Steiner, et al., 2019) and evolutionary algorithms (Nguyen, et al., 2007), have been implemented into smart monitoring systems in a decentralized man-ner.

Most recent approaches have as common ground machine learning algorithms, a subcategory of AI, that have been adopted for smart monitoring purposes (Smarsly, et al., 2016). Generally, ML algorithms in civil engineering may be distinguished by their application into (i) ML algorithms used for so-called surrogate modeling, where ML algorithms substitute conventional algorithms to achieve higher computational efficiency and (ii) ML algorithms used to solve abstract problems pertaining to data analysis, such as pattern recognition or classification problems, in which ML algo-rithms are deployed to analyze large amounts of data to classify given signals (or pictures) with respect to predefined classes.

In general, artificial intelligence algorithms, and, by extension, machine learning algorithms, may be categorized into symbolic AI, which includes inference and search algorithms using explicit symbolic programming, and into subsymbolic AI, which is generally considered “black-box” in terms of internal mechanisms. Subsymbolic AI, such as deep learning neural networks, shows good performance in analyzing com-plex engineering problems that involve large data sets and is therefore widely used in smart monitoring. However, the widespread adoption of subsymbolic AI/ML algo-rithms in smart monitoring is still limited, due to mistrust expressed by engineers towards the opaque inner mechanisms of subsymbolic AI/ML algorithms, and, by extension, to the reasoning and reproducibility of the outputs. While an explanation of the algorithms is inherent in symbolic AI, there is a strong need to explain subsymbol-ic AI/ML algorithms. The need for explaining the reasoning behind decisions made by subsymbolic black-box AI/ML algorithms has led to the development of “explain-able artificial intelligence (XAI)” (Gunning & Aha, 2019; Barredo Arrieta, et al., 2019). XAI is a technical discipline aiming to comprehensibly present AI systems and to clarify why and how AI systems generate certain outputs (Adadi & Berrada, 2018).

Addressing the explainability of AI/ML algorithms for smart monitoring requires a concise overview of existing approaches using AI/ML algorithms in smart monitor-ing. From the broader perspective of smart cities, AI/ML algorithms for smart city applications have been reported in reviews and summary papers, for example by Guo et al. (2019) and Mohapatra (2019). Soomro et al. (2019) have reviewed big data ana-lytics for smart cities and Martins (2018) has discussed the impact of ML algorithms on innovations in smart cities. Furthermore, Nosratabadi et al. (2019) have surveyed deep learning and ML models for smart cities. Regarding smart monitoring, Bao et al. (2019) have presented a review on data science approaches in SHM, and Joshuva et al. (2019) have reviewed machine learning algorithms for monitoring wind turbines. However, to the knowledge of the authors, no review has focused on categorizing AI/ML algorithms for highlighting the need for XAI in smart monitoring.

4

This paper essentially constitutes a preliminary step towards adapting XAI ap-proaches for smart monitoring. By reviewing and categorizing AI/ML algorithms for smart monitoring and discussing general XAI concepts, an overview of which AI/ML algorithms used in SHM may be modified towards adopting XAI in smart monitoring is shown. Subsequently, the review presented herein is summarized, and a concise outlook on potential future work is provided.

2 Machine learning algorithms in civil engineering

Because of the ability to recognize and to classify patterns in large data sets, ML algo-rithms are of increasing interest in civil engineering. In the following subsections, a categorization of ML algorithms is provided, and ML algorithms of particular rele-vance to smart monitoring applications are reviewed and categorized.

2.1 Categorization of machine learning algorithms

The term “intelligence” in “artificial intelligence” denotes the ability of an entity to capture, to process, and to respond to input of different kind (Legg & Hutter, 2007). Extending the definition of intelligence, the term “artificial intelligence” describes the ability of an artificial entity (e.g., a software or computer system) to achieve specific goals under a variety of environmental conditions. However, to qualify as “intelli-gent”, a system needs to possess the ability to respond to previously unknown (envi-ronmental) conditions through learning and adaption (Hutter, 2005). In a broader sense, AI is the ability of a computer system to approximate the intellectuality of hu-man beings. To mimic human behavior, Russel & Norvig (2016) have defined six categories of AI: machine learning, robotics, computer vision, natural language pro-cessing, knowledge representation, and automated reasoning.

In intelligent systems, ML helps adapt a system to new circumstances through pro-cessing and analyzing data, extrapolating patterns, and making predictions. By com-bining concepts of computer science with optimization and statistical concepts (Mohri, et al., 2018), ML essentially represents the learning processes of AI, often described as converting experience into expertise or knowledge (Shalev-Shwartz & Ben-David, 2014). In summary, ML algorithms show two distinct advantages, as compared to traditional algorithms (Russel & Norvig, 2016; Shalev-Shwartz & Ben-David, 2014):

1. ML algorithms operate with previously unknown (i.e., newly derived) data on

which the system has not been trained, and 2. ML algorithms are adaptable to changes in the data.

However, ML algorithms need to learn from experience or knowledge of domain

experts (Shalev-Shwartz & Ben-David, 2014). Depending on the type of learning, ML algorithms may be categorized into

5

i. supervised learning, ii. unsupervised learning, and

iii. reinforcement learning,

as shown in Figure 1. In supervised learning, ML algorithms use labeled input-output pairs as training data, and the system learns based on given examples. Typical learning problems in supervised learning are classification and regression. In classifi-cation, data is sorted into predefined categories, while in regression, the outputs to corresponding input data are calculated. In contrast to supervised learning, the train-ing data in unsupervised learning is not labeled. A typical problem of unsupervised learning is clustering, where data is grouped according to commonalities. In rein-forcement learning, no training data is provided. Instead, the system develops a strat-egy to maximize a predefined cumulative reward. Figure 1 shows the categorization of AI and ML algorithms as well as the subcategories mentioned above. In addition, examples of ML algorithms, corresponding to the subcategories, are illustratively provided in Figure 1.

Fig. 1. Categorization of machine learning algorithms.

Depending on the category of ML algorithms, mixed forms of (i), (ii), and (iii) are likely to be used (Burkov, 2019; Salehi & Burgueno, 2018). For example, artificial neural networks are trained with different specifications and, depending on the pur-pose and structure of the artificial neural network (ANN), may fit into any of the three categories. Therefore, the categorization presented in Figure 1 is regarded as a starting point to approach the basic concepts of machine learning but cannot be considered generally valid for any ML specification.

6

2.2 Machine learning algorithms for smart monitoring

The application areas of machine learning in smart monitoring are manifold. This paper focuses on ML algorithms applied to data analysis, which, according to Bisby & Briglio (2005), may pursue the following goals,

i. damage detection, ii. damage classification,

iii. damage localization, iv. condition assessment, and v. life-time prediction,

to be achieved primarily by supervised and unsupervised ML algorithms as well as

algorithms that combine both categories. In the remainder of this subsection, ML algorithms addressing the above goals are reviewed, distinguishing between super-vised and unsupervised/hybrid ML algorithms. Supervised machine learning algorithms for smart monitoring For damage detection and damage classification, support vector machines (SVM) are common. For example, Li et al. (2019) have identified damage based on SVMs and Lamb waves in smart monitoring. Gui et al. (2017) have compared different SVM-based optimization techniques for damage detection with a Gaussian radial basis func-tion (RBF) chosen as kernel function. Gardner et al. (2016) have proposed an RBF-kernel based SVM, fed by a finite element-based damage model to generate output data, while Pan et al. (2018) have proposed a framework for data-driven structural diagnosis and damage detection using SVM with wavelet transform, Hilbert-Huang transform, and Teager-Huang transform as feature extraction methods. Ghiasi et al. (2016) have reported on a new kernel function for least square support vector ma-chines using multidimensional orthogonal-modified Littlewood-Paley wavelets and a thin plate spline radial basis function. Abdeljaber et al. (2018) have presented an ap-proach based on a 1-D convolutional neural network (CNN) to detect damage with two labeled sets of data, regardless of the size of the structure. Gunawan et al. (2018) have examined k-nearest neighbors (k-NN) algorithms, stating that the accuracy of the algorithms strongly depends on the amount of training data, which is often not sufficiently available for solving engineering problems in smart monitoring.

For damage localization, Zhao et al. (2019) have presented an algorithm based on ANN regression using acoustic emission sensors for carbon fiber reinforced polymer composite materials. The training data required for the artificial neural network has been obtained from a finite element model.

To advance condition assessment of smart structures, Nazarian et al. (2018) have combined SVMs, ANNs, and Gaussian naïve Bayes techniques to assess the condition of a masonry building with timber frames. The ML model has been trained by finite element model simulation data to relate the change of stiffness of different building components to intensity and location of the damage sources.

7

Aiming at life-time prediction, Sysyn et al. (2019) have addressed a railway cross-ing based on features extracted by principal component analysis and partial least square regression. Hoang et al. (2018) have predicted the scour depth at bridges by using support vector regression, for which several feature selection algorithms have been combined, with the variable neighborhood search feature selection method providing the best outcome.

A number of studies have been reported that aim at combinations of the data analy-sis goals within smart monitoring, for example pursuing damage detection and dam-age classification together. Vitola et al. (2016) have presented a combination of prin-cipal component analysis (PCA) with k-NN and PCA with bagged trees. Vitola et al (2017a) have compared different k-NN algorithms to detect and to classify damage based on identical data sets, linked with the research conducted by Tibaduzia et al. (2018) and Vitola et al. (2017b), who combine PCA and k-NN components to detect und to classify damage of sandwich structures and composite plates. Joshuva & Sugumaran (2018) have compared classification and regression algorithms with re-spect to damage detection and damage classification, including a sequential minimal optimization classifier, a simple logistic algorithm classifier, a multilayer perceptron in terms of a feedforward artificial neural network, logistic regression, and an RBF network. The authors have been able to define five different damage classes. Vashisht et al. (2018) have compared Bayesian ANNs, CNNs, and long short-term memory ANNs to identify and to localize damage in a cantilever beam with training data for the ANNs provided by finite element simulations.

Unsupervised and hybrid machine learning algorithms for smart monitoring

Studies applying unsupervised /hybrid ML algorithms to achieve the goals of data analysis in smart monitoring are less common than supervised learning approaches, because labeled training data is usually available in smart monitoring. For example, Sierra-Perez et al. (2017) have presented a multi-layer ANN-based damage detection methodology for strain field pattern recognition, using a hierarchical non-linear PCA dimensionality reduction technique. Santos et al. (2016) have improved Gaussian mixture models to detect and to classify damage of bridges. Senniappan et al. (2017) have applied fuzzy cognitive maps to categorize cracks in reinforced concrete col-umns.

Furthermore, Diez et al. (2016) have used a k-NN outlier detector for performing k-means clustering on data in an attempt to isolate and to localize damaged joints of a bridge. Das et al. (2019) have used Gaussian mixture models for clustering unlabeled data and for feature separation by an SVM-calculated hyperplane for crack mode classification.

8

3 Results and discussion: Towards explainable artificial intelligence

The result of the review presented in the previous section is shown in Figure 2 in terms of an overview of ML algorithms for smart monitoring. As can be seen from Figure 2, the ML algorithms are assigned to the goals of data analysis in smart moni-toring, with the thickness of the lines connecting an ML algorithm and a data analysis goal denoting the quantity of papers found in literature. Regardless of the ML algo-rithm and the data analysis goal, it has been concluded that intransparency and mis-trust in ML algorithms that are black-box in nature are hindering the widespread adoption of the algorithms in civil engineering practice. Particularly following the enforcement of the European data protection regulation, which requires comprehensi-ble decision making in AI, the incomprehensibility of ML-based decision making further limits the distribution and implementation of ML algorithms.

Fig. 2. Review of machine learning algorithms for smart monitoring.

9

XAI has the potential to overcome implementation obstacles and provide explana-tions as well as additional information regarding decision-making processes, hence offering more comprehensible ML algorithms. However, when designing XAI-based ML algorithms, different levels of explanations must be considered, ranging from “comprehensive explanation” in case of complex subsymbolic ML algorithms to “no explanation” in case of symbolic ML algorithms, as implemented in expert systems that inherently explain themselves. Further distinctions must be made with respect to the experience and expertise of human individuals that are addressed by the explana-tions, such as technicians using ML algorithms in engineering practice or computer scientists implementing data analysis into smart monitoring systems.

In general, an explanation is considered a collection of human-interpretable fea-tures, relevant to decisions provided by ML algorithms. The explainability of ML algorithms is often referred to as interpretability, with “interpretation” denoting a mapping of abstract concepts that are comprehensible for human individuals (Monta-von, et al., 2017). Different efforts towards implementing XAI approaches have been reported. For example, LIME, a local interpretable explanation, presents a model-independent approach towards approximating black-box models around any classifier of interest and explaining the predictions of the classifier in an interpretable manner (Ribeiro, et al., 2016). The layer-wise relevance propagation (LRP) algorithm for image classification serves as another XAI implementation example. LRP decompos-es the classifier and iterates the relevance of each layer of a network backwards, start-ing with the output prediction (Bach, et al., 2015). Aiming to explain autonomous decisions made by smart monitoring systems with respect to sensor fault diagnosis, Fritz (2019) has implemented an XAI approach that extends deep learning NNs cou-pled with blockchain technology. In summary, representing an open research problem in smart city applications, it can be concluded that explanations must be adapted to the goal of data analysis, to the level of explainability, and to the target audience.

4 Summary and conclusions

Smart infrastructure is a key component of smart cities and requires smart monitoring to achieve more reliable, durable, and cost-efficient infrastructure as compared to the past. Smart monitoring is a combination of SHM and AI algorithms. ML algorithms, a subcategory of AI algorithms, are used to automatically analyze sensor data. Howev-er, the black-box nature of ML algorithms typically used in smart monitoring, alt-hough efficient in analyzing sensor data, causes intransparency and mistrust expressed by engineers, thus hindering the exploitation of the ML full potential in engineering practice.

XAI is supposed to enhance the transparency, thus the confidence, in ML algo-rithms. Drawing from trends in current ML applications for smart monitoring, this paper has presented a preliminary step towards adapting XAI approaches in smart monitoring. ML algorithms commonly deployed to smart monitoring have been re-viewed and XAI approaches have been presented, proposed to overcome the obstacles of incomprehensibility of ML algorithms. For smart monitoring, ML algorithms may

10

require different levels of explanations based on their purpose and the human individ-uals addressed. In conclusion, the overview of ML algorithms in smart monitoring provided in this paper has demonstrated that an in-depth analysis of explainability and levels of explanation for ML algorithms is required to advance smart monitoring and smart city developments.

Acknowledgments

The authors gratefully acknowledge the support offered by the German Research Foundation (DFG) under grants SM 281/9-1, SM 281/14-1, and SM 281/15-1. This research is also partially supported by the German Federal Ministry of Transport and Digital Infrastructure (BMVI) under grant VB18F1022A. Any opinions, findings, conclusions or recommendations expressed in this paper are those of the authors and do not necessarily reflect the views of DFG or BMVI.

References

1. Acatech – National Academy of Science and Engineering (2015). Industry 4.0, Urban de-velopment and German international development cooperation (Acatech position paper), Herbert Utz Verlag, Munich, Germany, 2015

2. Adadi, A., & Berrada, M. (2018). Peeking inside the black-box: A survey on explainable artificial intelligence (XAI). IEEE Access, 6(2018), 52138-52160.

3. Abdeljaber, O., Avci, O., Kiranyaz, S., Boashash, B., Sodano, H. & Inman, D. (2018). 1-D CNNs for structural damage detection: verification on a structural health monitoring benchmark data. Neurocomputing, 275(2018), 1308-1317.

4. Bach, S., Binder, A., Montavon, G., Klauschen, F., Müller, K.R. & Samek, W. (2015). On pixel-wise explanations for non-linear classifier decisions by layer-wise relevance propa-gation. PLoS One, 10(7), e0130140.

5. Bao, Y., Chen, Z., Wei, S., Xu, Y., Tang, Z. & Li, H. (2019). The state of the art of data science and engineering in structural health monitoring. Engineering 5(2), 234-242.

6. Barredo Arrieta, A., Diaz Rodriguez, N., Del Ser, J., Bennetot, A., Tabik, S., et al. (2019). Explainable artificial intelligence (XAI): Concepts, taxonomies, opportunities and chal-lenges toward responsible AI. Information Fusion, 58(2020), 82-115.

7. Bilek, J., Mittrup, I., Smarsly, K. & Hartmann, D. (2003). Agent-based concepts for the holistic modeling of concurrent processes in structural engineering. In: Proceedings of the 10th ISPE International Conference on Concurrent Engineering: Research and Applica-tions, Madeira, Portugal, July 26, 2003.

8. Bisby, L.A. & Briglio, M.B. (2005). ISIS Educational Module 5: An introduction to struc-tural health monitoring. SAMCO Final Report 2006. Winnipeg, Manitoba, Canada: ISIS Canada.

9. Burkov, A. (2019). The hundred-page machine learning book. ISBN-13: 978-1999579500 10. Cha, Y.-J., Choi, W. & Büyüköztürk, O. (2017). Deep learning-based crack damage detec-

tion using convolutional neural networks. Computer-Aided Civil and Infrastructure Engi-neering, 32(5), 361-378.

11

11. Das, A., Suthar, D. & Leung, C. (2019). Machine learning based crack mode classification from unlabeled acoustic emission waveform features. Cement and Concrete Research, 121(2019), 42-57.

12. Diez, A., Khoa, N.L.D., Makki Alamdari, M., Wang, Y., Chen, F. & Runcie, P. (2016). A clustering approach for structural health monitoring on bridges. Journal of Civil Structural Health Monitoring, 6(2016), 1-17.

13. Dragos, K. & Smarsly, K. (2016). Distributed adaptive diagnosis of sensor faults using structural response data. Smart Materials and Structures, 25(10), 105019.

14. Fritz, H. (2019). An explainable artificial intelligence model coupling deep learning and blockchain technology. Bachelor thesis. Chair of Computing in Civil Engineering, Bau-haus University Weimar, Germany.

15. Gardner, P., Barthorpe, R.J. & Lord, C. (2016). The development of a damage model for the use in machine learning driven SHM and comparison with conventional SHM meth-ods. In: Proceedings of the International Conference on Noise and Vibration Engineering 2016 (ISMA 2016) and International Conference on Uncertainty in Structural Dynamics (USD 2016), Leuven, Belgium, September 13, 2016.

16. Ghiasi, R., Torkzadeh, P. & Noori, M. (2016). A machine-learning approach for structural damage detection using least square support vector machine based on a new combinational kernel function. Structural Health Monitoring, 15(3), 302-316.

17. Gunawan, F., Soewito, B., Surantha, N. & Tuga, M. (2018). One more reason to reject manuscript about machine learning for structural health monitoring, In: Proceedings of the 2018 Indonesian Association for Pattern Recognition (INAPR) International Conference, Jakarta, Indonesia, September 7, 2018.

18. Gui, G., Pan, H., Lin, Z., Li, Y. & Yuan, Z. (2017). Data-driven support vector machine with optimization techniques for structural health monitoring and damage detection. KSCE Journal of Civil Engineering. 21(2), 523-534.

19. Gunning, D. & Aha, D.W. (2019). DARPA’s explainable artificial intelligence program. AI Magazine, 40(2), 44-58.

20. Guo, X., Shen, Z., Zhang, Y. & Wu, T. (2019). Review on the application of artificial in-telligence in smart homes. Smart Cities, 2(3), 402-420.

21. Hartmann, D., Smarsly, K. & Law, K. H., 2011. Coupling sensor-based structural health monitoring with finite element model updating for probabilistic lifetime estimation of wind energy converter structures. In: Proceedings of the 8th International Workshop on Struc-tural Health Monitoring, Stanford, CA, USA, September 13, 2011.

22. Haugeland, J. (1987). Artificial intelligence. The very idea. Cambridge, MA, USA: MIT Press.

23. Hoang, N.-D., Liao, K.-W. & Tran, X.-L. (2018). Estimation of scour depth at bridges with complex pier foundations using support vector regression integrated with feature selection. Journal of Civil Structural Health Monitoring, 8(3), 431–442.

24. Hutter, M. (2005). Universal artificial intelligence – sequential decisions based on algo-rithmic probability. Heidelberg, Germany: Springer-Verlag GmbH.

25. Johnson, P., Robinson, P. & Philpot, S. (2019). Type, tweet, tap, and pass: How smart city technology is creating a transactional citizen. Government Information Quarterly, 37(1), 101414.

26. Joshuva, A. & Sugumaran, V. (2018). A study of various blade fault conditions on a wind turbine using vibration signals through histogram features. Journal of Engineering Science and Technology, 13(1), 102-121.

27. Joshuva, A., Aslesh, A. & Sugumaran, V. (2019). State of the art of structural health moni-toring of wind turbines. International Journal of Mechanical Sciences, 9(5), 95-112.

12

28. Kabalci, E. & Kabalci, Y. (2019). From smart grid to Internet of Energy (1st edition). Lon-don, UK: Academic Press.

29. Kelley, T. (2003). Symbolic and sub-symbolic representations in computational models of human cognition: What can be learned from biology? Theory & Psychology, 13(6), 847-860.

30. Langley, P. (2011). The changing science of machine learning. Machine Learning, 82(3), 275-279.

31. Legg, S. & Hutter, M. (2007). Universal intelligence: A definition of machine intelligence. Minds & Machines, 17(4), 391-444.

32. Li, R., Gu, H., Hu, B. & She, Z. (2019). Multi-feature fusion and damage identification of large generator stator insulation based on Lamb wave detection and SVM method. Sen-sors, 19(7), 3733.

33. Martins, J. (2018). Towards smart city innovation under the perspective of software-defined networking, artificial intelligence and big data. RTIC – Revista de tecnologia da informação e comunicação, 8(2), 1-7.

34. Mittrup, I., Smarsly, K., Hartmann, D. & Bettzieche, V. (2003). An agent-based approach to dam monitoring. In: Proceedings of the 20th CIB W78 Conference on Information Technology in Construction, Auckland, New Zealand, April 23, 2003.

35. Mohanty, S. (2016). Everything you wanted to know about smart cities. IEEE Consumer Electronics Magazine, 5(3), 60-70.

36. Mohapatra, B. (2019). Machine learning applications to smart city. ACCENTS Transac-tions on Image Processing and Computer Vision, 5(14), 1-6.

37. Mohri, M., Rostamizadeh, A. & Talwalkar, A. (2012). Foundations of machine learning. Cambridge, MA, USA: MIT Press.

38. Montavon, G., & Samek, W. & Müller, K.-R. (2018). Methods for interpreting and under-standing deep neural networks. Digital Signal Processing, 73(2018), 1-15.

39. Nazarian, E., Taylor, T., Weifeng, T. & Ansari, F. (2018). Machine-learning-based ap-proach for post event assessment of damage in a turn-of-the-century building structure. Journal of Civil Structural Health Monitoring, 8(2), 237-251.

40. Nguyen, V.V., Smarsly, K. & Hartmann, D. (2007). A computational steering approach towards sensor placement optimization for structural health monitoring using multi-agent technology and evolutionary algorithms. In: Proceedings of the 6th International Work-shop on Structural Health Monitoring, Stanford, CA, USA, September 11, 2007.

41. Nomura, Y. & Shigemura, K. (2019). Development of real-time screening system for structural surface damage using object detection and generative model based on deep learning. Journal of the Society of Materials Science, 68(3), 250-257.

42. Nosratabadi, S., Mosavi, A., Keivani, R., Ardabili, S. & Aram, F. (2019). State of the art survey of deep learning and machine learning models for smart cities and urban sustaina-bility. In: Proceedings of the 18th International Conference on Global Research and Edu-cation Inter-Academia, Budapest, Hungary, September 4, 2019.

43. Ogie, R.I., Perez, P. & Dignum, V. (2017). Smart infrastructure: An emerging frontier for multidisciplinary research. ICE Smart Infrastructure and Construction, 170(1), 8-16.

44. Organisation for Economic Co-operation and Development (OECD) (2019). Enhancing the contribution of digitalisation to the smart cities of the future. Online: https://one.oecd.org/document/CFE/RDPC/URB(2019)1/REV1/en/pdf, last accessed: Jan-uary 20, 2020.

45. Pan, H., Azimi, M., Lin, Z. & Yan, F. (2018). Time-frequency based data-driven structural diagnosis and damage detection for cable-stayed bridges. Journal of Bridge Engineering, 23(6), 04018033.

13

46. PricewaterhouseCoopers (2019). Creating the smart cities of the future. Online: https://www.pwc.com/gx/en/sustainability/assets/creating-the-smart-cities-of-the-future.pdf, last accessed: January 21, 2020.

47. Ribeiro, M.T., Singh, S. & Guestrin, C. (2016). Why should I trust you?: Explaining the predictions of any classifier. In: Proceedings of the 22nd ACM International Conference on Knowledge Discovery and Data Mining, San Francisco, CA, USA, August 13, 2016.

48. Russel, S.J. & Norvig, P. (2014). Artificial intelligence: A modern approach (3rd edition). Harlow, Essex, UK: Pearson Education Ltd.

49. Salehi, H., & Burgueno, R. (2018). Emerging artificial intelligence methods in structural engineering. Engineering Structures, 171(2018), 170-189.

50. Santos, A., Figueiredo, E., Silva, M., Santos, R., Sales, C. & Costa, J. (2016). Genetic-based EM algorithm to improve the robustness of Gaussian mixture models for damage de-tection in bridges. Structural Control and Health Monitoring, 24(3), e1886.

51. Senniappan, V., Subramanian, J., Papageorgiou, E. & Mohan, S. (2016). Application of fuzzy cognitive maps for crack categorization in columns of reinforced concrete structures. Neural Computing and Applications, 28(1), 107-117.

52. Shalev-Shwartz, S., & Ben-David, S. (2014). Understanding machine learning. From theo-ry to algorithms. New York, NY, USA: Cambridge University Press.

53. Sierra-Perez, J.,Torres Arredondo, M.A. & Alvarez-Montoya, J. (2017). Damage detection methodology under variable load conditions based on strain field pattern recognition using FBGs, nonlinear principal component analysis, and clustering techniques. Smart Materials and Structures, 27(1), 015002.

54. Smarsly, K., Lehner, K. & Hartmann, D. (2007). Structural health monitoring based on ar-tificial intelligence techniques. In: Proceedings of the International Workshop on Compu-ting in Civil Engineering, Pittsburgh, PA, USA, July 24, 2007.

55. Smarsly, K., Law, K.H. & König, M. (2011). Resource-efficient wireless monitoring based on mobile agent migration. In: Proceedings of the SPIE (Vol. 7984): Health Monitoring of Structural and Biological Systems 2011, San Diego, CA, USA. March 06, 2011.

56. Smarsly, K., Law, K.H. & Hartmann, D. (2012). A multiagent-based collaborative frame-work for a self-managing structural health monitoring system. ASCE Journal of Compu-ting in Civil Engineering, 26(1), 76-89.

57. Smarsly, K. & Law, K.H. (2013). A migration-based approach towards resource-efficient wireless structural health monitoring. Advanced Engineering Informatics, 27(4), 625-635.

58. Smarsly, K., Dragos, K. & Wiggenbrock, J. (2016). Machine learning techniques for struc-tural health monitoring. In: Proceedings of the 8th European Workshop on Structural Health Monitoring (EWSHM), Bilbao, Spain, July 5, 2016.

59. Soomro, K., Bhutta, M., Khan, Z. & Tahir, M. (2019). Smart city big data analytics: An advanced review. Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discov-ery, 9(5), e1319.

60. Steiner, M., Legatiuk, D. & Smarsly, K. (2019). A support vector regression-based ap-proach towards decentralized fault diagnosis in wireless structural health monitoring sys-tems. In: Proceedings of the 12th International Workshop on Structural Health Monitoring. Stanford, CA, USA, September 10, 2019.

61. Suleiman, A.R. & Nehdi, M.L. (2017). Modeling self-healing of concrete using hybrid ge-netic algorithm-artificial neural network. Materials, 10(2), 135.

62. Sysyn, M., Gerber, U., Nabochenko, O., Li, Y. & Kovalchuk, V. (2019). Indicators for common crossing structural health monitoring with track side inertial measurements. Acta Polytechnica, 59(2), 170-181.

14

63. Tibaduiza, D., Torres Arredondo, M.A., Oyaga, J., Anaya, M. & Pozo, F. (2018). A dam-age classification approach for structural health monitoring using machine learning. Com-plexity, 2018, 5081283.

64. United Nations Economic and Social Council (2016). Smart cities and infrastructure. Online: https://unctad.org/meetings/en/SessionalDocuments/ecn162016d2_en.pdf, last ac-cessed: January 25, 2020.

65. Vashisht, R., Viji, H., Sundararajan, T., Mohankumar, D. & Sarada, S. (2018). Structural health monitoring of cantilever beam, a case study – using Bayesian neural network and deep learning. In: Proceedings of the 13th International Conference on Systems, Athens, Greece, April 22, 2018.

66. Vitola, J., Tibaduiza, D., Anaya, M. & Pozo, F. (2016). Structural damage detection and classification based on machine learning algorithms. In: Proceedings of the 8th European Workshop On Structural Health Monitoring (EWSHM), Bilbao, Spain, July 5, 2016.

67. Vitola, J., Pozo, F., Tibaduiza, D. & Anaya, M. (2017a). A sensor data fusion system based on k-nearest neighbor pattern classification for structural health monitoring applica-tions. Sensors, 17(2), 417.

68. Vitola, J., Pozo, F., Tibaduiza, D. & Anaya, M. (2017b). Distributed piezoelectric sensor system for damage identification in structures subjected to temperature changes. Sensors, 17(6), 1252.

69. Zhao, Z., Yua, M. & Dong, S. (2019). Damage location detection of the CFRP composite plate based on neural network regression. In: Proceedings of the 7th Asia-Pacific Work-shop on Structural Health Monitoring, Hong Kong, China, November 12, 2018.