IoT Framework Application to SHM Systems

Abstract

This chapter presents and details a complete IoT solution for secure, synchronous, low-power, wireless and unattended structural health monitoring (SHM), whose architecture can be parameterised to any type of structure to be monitored, thanks to its modular design based on an embedded real-time operating system at the node level and a microservices architecture at the application level. This solution is validated by means of its deployment, together with two commercial reference systems, in two experimental use cases in two different structures: on the one hand, a bridge located at the Engineering School of the University of Seville (Spain) and, on the other hand, the Homage Tower of the Utrera’s Castle (Spain). The results obtained place this work as a novel solution in the current state of the art.

1. Introduction

Over the last few decades, the relationship between humans and the environment surrounding us, as well as with the different frameworks, has changed considerably due to the emergence of new technologies that have made it possible increasing the quantity and heterogeneity of the information acquired from the environment, as well as from the manufacturing processes. However, this has resulted in a variety of non-standardised, and therefore non-compatible technologies and architectures that do not allow information exchange between them and limit the impact of these advances on the global state of the art. It is precisely this heterogeneous framework of needs and available technologies that gives rise to the emergence of the IoT (Internet of Things) paradigm.

According to IBM [1], IoT is the concept of interconnecting any device, thing, or person (things) to the Internet and to other devices and people to share data on their behaviour and that of the surrounding environment. Oracle [2] specifies how these ‛ thingsʼ can range from everyday objects to sophisticated industrial tools. This global integration is produced thanks to the definition of a standardised layered architecture [3, 4] shown in Figure 1 and has a direct impact on the technology and its use, offering new capabilities and advantages over the previous framework, such as very low-cost and low-power devices with Internet connectivity and their integration in cloud processing platforms where they can host from simple applications to complex analysis processes based on machine learning techniques. The ability to integrate connected networked devices into a single system is one of the main features that make IoT a very attractive technology in the era of digital transformation.

Figure 1.

Standardised IoT four-layer architecture.

The evolution of wireless communication technologies, the standardisation of IoT architecture and the globalisation of manufacturing processes invite the implementation of the new paradigm in these production lines, which has given rise to the concept of Industry 4. 0, the fourth industrial revolution characterised by the integration of advanced technologies to create smart factories, cyber-physical systems and IIoT (Industrial Internet of Things) [5]. The implementation of IoT in industry is becoming more and more common due to its numerous advantages, generating several industry-beneficial applications: (1) Predictive analytics and preventive maintenance improving by enabling data collection and subsequent analysis and helping predict machine failures and perform preventive maintenance before they occur, which in turn reduces costs and downtime [6], (2) Improving product quality by monitoring and controlling critical factors during the production process allowed for greater control over processes and reduction of defects [7], (3) Edge computing [8], technology responsible for processing data in real time at the point, where it is generated rather than sending it to a central location for processing, reducing processing time and load on central networks and servers and (4) Digital twin generation [9], enabling simulation and testing of different scenarios virtually, which reduces the costs and risks associated with testing and validation processes.

All these technological advances are reflected in the 5.5 billion devices connected in 2022 to operated networks (2G/3G, NB-IoT/CatM, 4/5G) according to the manufacturer Ericsson [10] on the 43 billion connected devices estimated for 2023 by Forbes magazine [11]. On the other hand, this impact translates directly to the market, as predicted by the market analysis organisation Strait Research in its report [12], where it estimates a market evolution from USD 813 billion in 2021 to more than USD 2483 billion in 2030.

Thanks to the IoT benefits, it is applicable in different sectors such as smart agriculture [13, 14, 15], where more economically efficient ecosystems are generated by automating tasks such as cultivation and its monitoring allowing early detection of pests and diseases, in Healthcare [16, 17, 18], where remote and real-time monitoring, especially after the global pandemic of COVID-19, enables early and accurate detection of pathologies, while the application layer allows direct integration of patients with all links in the healthcare chain, resulting in new care models that improve the health and experience of patients, as well as the performance of healthcare professionals and smart homes [19, 20], where the integration of domestic elements in a secure and real-time network allows the optimization of different aspects such as energy management (even generating energy communities), the improvement of the feeling of comfort and the adaptation of the environment to the emotional state of the inhabitants. This framework has given rise to the smart city concept [21], which integrates structural monitoring of buildings, traffic control, vehicle-to-infrastructure (V2I) communications, waste management, noise monitoring or air quality, among others; and the automotive sector [22, 23], where the emergence of 5G communications allows real-time monitoring of vehicles and cloud computing the generation of digital twins [24] application of machine learning algorithms for vehicle and fleet management tasks from behavioural analysis, accident detection [25] and predictive maintenance.

These applications provide an overview of how SHM calculation could be automated with current technology. Today, using the range of components that comprise the IoT architecture, data can be obtained in an automated way, and, in real time, relevant indicators for determining structural health.

2. Structural health monitoring

Once a given structure has been built, whether it is a building or civil infrastructure, it is necessary to identify any damage that may compromise structural safety at an early stage. In this context, the structural health monitoring (SHM) plays a key role. According to the International Society for Structural Health Monitoring of Intelligent Infrastructures, it can be defined as: ‘a type of system that provides on-demand information about any significant change or damage to the structure’ [26].

This discipline has experienced great growth in recent decades. Equipping, a structure with a structural monitoring system, has many advantages. A direct consequence of this is better control of structural safety, detecting in advance situations that could lead to collapse. Moreover, a proper structural diagnosis is vital to design an adequate repair or reinforcement intervention and, consequently, extend the useful life of the construction.

Basically, a structural health monitoring system consists of several elements, which can be grouped into six blocks [27]: (i) a sensory system (SS), (ii) a data acquisition and transmission system (DATS), (iii) a data processing and control system (DPCS), (iv) a data management system (DMS), (v) a structural health evaluation system (SHES) and (vi) an inspection and maintenance system (IMS). The first two blocks are located in the structure itself, while the next three blocks are normally located outside the structure. The last block supports all the others. The relationship between the listed blocks is graphically depicted in Figure 2.

Figure 2.

Composition of a SHM system. Source: Author, based on Ref. [27].

Depending on the number of variables under study, the SHM system can be made up of a wide variety of sensors. In some cases, the focus is on taking data relating to environmental conditions, such as humidity or temperature. In other cases, the interest extends to data related to the response of the structure. This response can be expressed in terms of displacements, velocities or accelerations. These three variables have a very small magnitude when the structure is subjected to environmental excitations. Among them, acceleration is the most commonly used parameter in the dynamic characterisation of the construction by means of operational modal analysis.

The sensor used to measure the rate of change of velocity with respect to time is known as an accelerometer. Currently, these sensors can be classified according to the technology used, with the following types being distinguished: piezoelectric, piezoresistive, capacitive, balanced force, optical and thermal, among others. In each category, there are multiple variants depending on the number of axes, the range, the bandwidth or the price.

Traditionally, the collection of acceleration readings in the structure is performed with commercial accelerographs [28, 29]. This equipment is characterised by high accuracy, which usually involves high costs. Accelerograph prototypes developed around micro-electro-mechanical systems (MEMS) accelerometers are becoming increasingly common [30, 31]. In this way, it is possible to obtain devices that are very economical and at the same time very low in energy consumption [32]. The equipment constructed in this way is also lightweight and very small in size, which makes it easy to transport and place in structures.

Acceleration data collected with accelerometers are used for the estimation of the dynamic properties of structures. For this purpose, a large number of well-known operational modal analysis methods are available. These methods are usually classified according to the domain in which the identification of modal parameters takes place, with two groups being distinguished: (i) time domain methods, which operate directly on the signals recorded in tests or on correlation functions, such as natural excitation technique [33], covariance-driven stochastic subspace identification [34] and data-driven stochastic subspace identification [35] and (ii) frequency domain methods, which require the calculation of spectral density functions. This second group includes, among others, the following methods: maximum likelihood [36], frequency domain decomposition [37] and enhanced frequency domain decomposition [38].

The complete dynamic characterisation of a structure requires the determination of the three basic dynamic properties of the structure, namely natural frequencies, modal displacements and modal damping factors. When the structure deteriorates, either due to the passage of time or a sudden event such as an earthquake or a vehicle collision, a variation of these dynamic properties occurs. Therefore, in order to assess the deterioration of the building, it is necessary to compare two states: (i) initial or reference, which generally corresponds to a situation in which the structure has not been damaged and (ii) final, in which it is expected to have suffered some type of damage.

Perhaps the simplest way to detect damage in a construction is by the reduction of the natural frequencies of the structure. This is evidenced by the large number of scientific papers published in the last quarter of a century related to structural diagnostics. Under the assumption that the mass remains constant, the progression of deterioration will lead to a decrease in the stiffness of the structure. Consequently, there will be a reduction of the natural frequencies. However, the observation of significant changes in the natural frequencies requires a considerable deterioration of the structure. In the other direction, an increase in this dynamic property is useful to check the effectiveness of certain repair or reinforcement actions on the structure under study.

Typically, structural damage is classified into four levels according to the degree of information that is available about it. In this context, it is common practice in the scientific community to use the damage levels proposed by Rytter [39], which are summarised in Table 1.

Level	Name	Description
Level 1	Detection	Indication of the presence of damage to the structure
Level 2	Localisation	Level I + Geometric location of damage
Level 3	Assessment	Level II + Estimation of the extent of damage
Level 4	Prediction	Level III + Structural safety prediction

Table 1.

Levels of damage on structures.

Source: [39].

In terms of the level of damage, methods based on changes in natural frequencies are only useful to indicate the presence of damage to the structure (Level I). When a higher degree of information on the deterioration of the building is required, techniques using variations in modal displacements are often used. In this way, it is possible to locate the most likely area in which the damage is located (Level II).

On the other hand, civil engineering has not been transparent to the technological evolution prior to the emergence of the IoT and, as a consequence, numerous solutions have appeared for different fields of application [40, 41, 42], which have not been able to mature enough to become widely adopted solutions by the community, mainly due to the changing and constant emergence of new technologies and, therefore, have not generated systems for the extraction of advanced information from structures (Level III or Level IV).

However, the emergence of the IoT paradigm and its standardised architecture as an integrative ecosystem eliminates dependencies between layers, and thus allows for expert contributions from each knowledge area to its corresponding layer. In other words, this allows the development and evolution of each of the functional blocks of a SHM system (data processing, hardware design, firmware design, high-level applications, etc.) independently in a framework that ensures the integration of all of them and the bridging of desired levels III and IV.

3. Proposed solution

Once the application framework has been analysed, the design of the proposed SHM-IoT solution is carried out. To do this, initially, an analysis of the main requirements to be met by the resulting design will be carried out and, subsequently, the most important implementation aspects that give rise to the solution will be detailed.

3.1 Requirements

Based on the usual IoT device architecture model, the proposed solution is composed of four functional layers (Figure 1) with the following scope:

• Perception layer: These are the devices or nodes in charge of carrying out the monitoring of the structure under analysis by acquiring the magnitudes of interest.

• Transport layer: This is the infrastructure responsible for guaranteeing a communication channel between the monitoring nodes and the processing and application layers.

• Processing layer: It is in charge of housing the data center with the capacity to execute data processing algorithms (traditional or machine learning) and offer its services to the application layer.

• Application layer: Together with the processing layer, it makes up the application server (physical or virtual) and hosts the applications necessary to carry out both the exploitation of the information generated by the processing layer and the management of the monitored infrastructures.

In this way, the integration of the solution in the IoT paradigm is guaranteed. The specific requirements of each of the proposed layers are detailed below.

3.1.1 Perception layer

The measurement devices are one of the most important elements of the proposed solution as they are responsible for monitoring the behaviour of the structure under test and the surrounding environment. Ultimately, therefore, they are the interface between the rest of the system and the infrastructure. In terms of technological requirements and challenges, the following are the most important.

3.1.1.1 Versatility and scalability

From a functional point of view, for frequency analysis, it is necessary for the nodes to have the capacity to monitor structural accelerations. Nevertheless, each structure has different monitoring needs depending on its architecture, typology or impact on society. For this reason, the proposed solution must be versatile in terms of the type of magnitudes to be monitored, which would allow it to be adapted to any civil infrastructure.

In the same way, each structure will have different needs in terms of the number of monitoring points, so the scalability of the solution becomes one of the enabling factors to ensure its success.

3.1.1.2 Wireless connection

Even though the IoT framework allows for the integration of both wireless and wired devices, the proposed application involves the monitoring of civil, strategic and historical heritage infrastructure, where the minimisation of visual pollution is an important factor. This implies devices as small as possible. On the other hand, from an economic point of view, the elimination of the high cost of cabling and its installation and maintenance, added to the low cost of wireless communications, makes wireless communications a must.

3.1.1.3 Time synchronisation

Frequency analysis of structures implies the need for time-synchronised devices so that the structural behaviour of the monitored building can be reconstructed for subsequent analysis using operational modal analysis (OMA) techniques.

3.1.1.4 Low energy consumption

One of the key aspects for the success of the solution is to meet very strict low-power requirements, so as to guarantee the operability of the deployed solution without the need for an external power supply (many infrastructures do not have an electrical installation) or to carry out maintenance work on it. These tasks would include the replacement of batteries, which has a direct impact on the cost of the solution and avoids the management of access to the infrastructures, which presents an added difficulty in the case of structures belonging to historical heritage.

3.1.1.5 Remote updating

Since this is a novel solution and a line of research in constant evolution, added to the fact that the structural particularity of each construction implies a parameterisation of the monitoring systems (full scale, sampling frequency, preprocessing, etc.), the need to carry out firmware updates of the devices deployed is guaranteed. Therefore, it is necessary for the measurement nodes to have FOTA (Firmware over the air) functionality that allows them to be updated without the need for physical action (usually located in areas that are difficult to access), avoiding interaction with strategic structures or historical heritage, as well as the management of complex bureaucratic permits and helping to reduce maintenance costs.

3.1.2 Transport layer

Another of the key aspects for the implementation of the proposed solution is the communications infrastructure used to send information between its different functional elements.

On this point, it is important to note that the solution presented must be suitable for monitoring a structure or a set of structures that may be located in one geographical point or distributed throughout the entire regional, autonomous or national geography (depending on the organisation that operates them), which, due to costs, rules out the option of using local networks maintained by the owners of the structures and highlights the need for a solution that offers communications to all of them globally, which makes it necessary to think of operated networks.

On the other hand, given that it may be a strategic infrastructure, it is necessary that the communications network offers sufficient security mechanisms to guarantee the integration and privacy of the information transmitted, as well as long-term operability.

3.1.3 Processing layer

The third main element is the application server, which will be in charge of providing secure services to both the processing layer and the application layer.

While integration in an IoT framework involves hosting services on a server (physical or virtual) in the cloud, it is necessary to think of a design based on microservices that allow scaling the functionalities of the deployed solution without the need to redesign the previous deployment and, therefore, will facilitate the integration of new value-added applications. This feature will also allow the migration of deployed services with minimum effort from one service provider to another if necessary.

On the other hand, the large amount and heterogeneous nature of data collected from infrastructures require an architecture and engine capable of integrating, moving and processing such data. For this reason, a non-relational database architecture and big data should be chosen.

The processing layer will oversee housing the processing algorithms in charge of extracting the structural characteristics of the monitored buildings from the data collected by the perception layer. Finally, these results will be offered to the application layer for subsequent exploitation.

3.1.4 Application layer

Finally, like the processing layer and the application layer should be composed of those microservices necessary to ensure the management of the monitored assets and provide the end user with visualisation of results and secure access based on roles and permission schemes.

3.2 IoT solution implementation details

3.2.1 Perception layer

From a hardware design point of view, the monitoring nodes have an architecture of two printed circuit boards (PCB) with different functionality. On the one hand, the core board is responsible for the transversal functionalities of any infrastructure to be monitored such as power supply, local processing and control of the node (STM32L152 microprocessor), local storage of information (1024Kb EEPROM memory and MicroSD card) and management of wireless communications (NB-IoT-SIM7080G). On the other hand, the sensor board is responsible for housing the necessary sensorization in each structure under analysis. Although the fundamental elements will be the accelerometer (ADXL355) and the GPS module (A2235H) that will guarantee the synchronisation of all the elements of the perception layer, this board has analogue and digital interfaces (SPI, IIC and UART) for the connection of other useful sensors depending on the type of structure (strain gauges, temperature and humidity, pressure, etc.). Finally, both boards are housed in an IP68 enclosure together with a PCB antenna for NB-IoT communication.

From a firmware design point of view, the need of time synchronisation and the parallelisation of the tasks of measurement acquisition, local storage, data processing and information sending have implied the use of a real-time operating system, opting for FreeRTOS. The parallelization of the tasks of the monitoring nodes and the use of the low consumption modes of the microprocessor and the sensors, as well as the efficient management of the microSD card, have allowed the optimisation of the energy consumption of the devices, maximising the autonomy of the device up to 2 years and 4 months with a 17 Ah battery capacity [32].

Additionally, the devices have firmware over the air (FOTA) functionality as another task of the node, which allows, after a connection is initiated by the node, checking for the existence of new firmware version and, if necessary, updating its functionality remotely and wirelessly, without the need to access them physically, guaranteeing the maintenance of the deployed solution.

3.2.2 Transport layer

Considering the requirements defined for the transport layer, the NB-IoT-operated network offers a level of coverage [43] and sufficient operators [44] to guarantee a service that satisfies them. Moreover, the market prices [45] guarantee the cost-effectiveness of the service compared to the deployment of proprietary local networks, such as LORA, which require the maintenance of the owner and a gateway to the operated network.

Another important aspect of SHM is to guarantee the integrity and security of the transmitted information, which may be sensitive information pertaining to strategic structures or the historical heritage of a city, region or country. For this reason, a secure communications stack is proposed, consisting of a transmission control protocol (TCP) connection-oriented communication service that guarantees that the transmitted information always reaches its destination and a message queuing telemetry transport (MQTT) service that minimises the number of established connections (light service), offers different QoS (Quality of Service) levels and guarantees the integration of the information in the IoT services offered by all the reference operators (Amazon WebServices, Azure, etc.). In addition, the selection of MQTT allows the use of technologies, such as SSL/TLS and user authentication via password or certificate.

Therefore, by opting for NB-IoT as the communication technology and the TCP/MQTT/SSL stack, we have a service that satisfies all the proposed security requirements.

3.2.3 Processing and application layers

As discussed in the requirements for these layers in Section 4.1, its implementation as a microservices deployment is a fundamental aspect. In this sense, Docker is selected as the container technology because it is the most mature of the solutions offered in the current catalogue.

The microservices that make up the solution at server level are the following:

• Reverse proxy: This is a service for port forwarding that enables communication between the different microservices. Nginx technology is chosen as it is accessible from the outside with HTTPS protocol, protected by the use of certificates.

• Decoder: Programmed in Python, this is the service in charge of receiving the information collected by the perception layer and interpreting its information before storing it in the database. It is also responsible for encoding the information downlinked to the nodes.

• Backend: This is the server core. It uses node.js image and it is in charge of the management and maintenance tasks of the processing and application layers, as well as being the interconnection with the frontend.

• Data processing (Processing layer): This is the data processing service. Developed in Python, it hosts the information processing algorithms, in this case, the detection of fundamental frequencies, and has the capacity to host the processing and machine learning services that may be considered in the future.

• Redis DB (Processing layer): Service whose function is purely the intercommunication of the backend with the data processing module, communication necessary to receive orders for training and updating behavioural models (deterministics or machine learning models) from the frontend and backend, and to transmit these orders to the Data processing module.

• Mongo DB: This is a non-relational database data storage service. This type of database engine allows the hosting of heterogeneous data that does not respond to a fixed structural pattern (different kinds of structures and different sensors), as occurs in relational databases. In this case, a Mongo 5.0.9 image is used and allows the creation of a big data architecture, as desired.

• Frontend: It is the service for the client side, and it uses a node.js 18.6.0 image. It is in this service where the web interface will be mounted to which the user will have access and where they will be able to consult the data received, manage the set of monitored infrastructures, as well as the nodes deployed in the field and their firmware updates.

Therefore, the proposed design results in a versatile and complete IoT solution for SHM, which will be validated in the use cases proposed in section V.

4. Case studies

Currently, there is a great deal of concern among researchers about historic buildings as they have a high patrimonial value. In this sense, it is worth highlighting the research carried out by Zini et al. [46], in which the ‘Torre Grossa’ (San Gimignano, Italy) and the Mastio di Matilde (Livorno, Italy) were studied. In order to obtain a more complete structural diagnosis, it is common not only to take sensor data but also to carry out visual inspections of the structure. This procedure was successfully applied in the Zuccaro’s Tower, a medieval masonry building located in the Italian city of Mantua [47]. The execution of environmental vibration tests can also be applied to ancient bridges. As an example of this, it is worth noting the study of the Turkish bridge spanning the Firtina Creek [48] and the Roman bridge near the Italian town of Todi [49].

In other cases, the identification of the dynamic properties of these historical structures experimentally is intended for the calibration of finite element models. That is, the adjustment of the model parameters to achieve a higher degree of approximation to the real behaviour of the structure. In this way, the resulting model allows for a more rigorous assessment of future scenarios, such as earthquakes or ground settlements. In the field of historic buildings, it is worth mentioning the work carried out by Venanzi et al. [50] on the Sciri Tower (Perugia, Italy) and the study carried out by Standoli et al. [51] on the Civic Tower (Ostra, Italy). With regard to historic civil infrastructures, it is interesting to mention the research carried out on the St. Lázaro and the Lagoncinha bridges [52] and the Durrães bridge [53], all of them located in Portugal.

Structural health monitoring is also of interest for more recent constructions, especially those that are considered to be unique or vital for the society. The common denominator of these buildings is the use of contemporary materials, such as concrete or steel. In this context, without being exhaustive, the following examples can be highlighted: the Infante D. Henrique Bridge, a concrete arch bridge in Portugal [54]; a steel cable-stayed footbridge over the Hron River in Slovakia [55]; the roof of the Municipal Sports Stadium of Braga, also in Portugal [56]; the Morrow Point Dam in the United States [57]; a tall building in Qatar [58].

As it can be seen, there is a wide variety of structures that are equipped with a structural health monitoring system. The results obtained are extremely useful, both to protect the lives of users and to properly manage the resources allocated to the conservation of the structures.

The solution presented in this chapter is an IoT system, which has all the components of a structural health monitoring system. The following subsections focus on node validation. For this purpose, two case studies located in Spain have been selected, namely: the Homage Tower of the Utrera’s Castle and a bridge close to the Engineering School of the University of Seville. These two structures are completely different in their characteristics, which is intended to demonstrate the versatility of the presented node. Specifically, the first case study represents a historical masonry building with a predominant vertical dimension and no structural code was used in its design. However, the other case study consists of a more recent concrete structure, which has the largest dimension in the horizontal plane and follows strict design criteria.

4.1 Methodology

In both case studies, the same procedure is used for the experimental determination of the dynamic properties of the structures. In particular, the focus is on the determination of the natural frequencies.

The process for the dynamic characterisation of these structures consists of two phases. In the first of these, a set of environmental vibration tests is carried out. For this purpose, natural sources of excitation, such as wind or traffic, are used. These sources of excitation cause the structure to vibrate. The structural response in terms of accelerations is recorded using accelerographs. This data collection campaign is carried out simultaneously with the node presented in this chapter and commercial equipment.

In each environmental vibration test, acceleration readings are taken at two points on the structure. One of these points is common to all the tests and serves as a reference. Proceeding in this way, in each test there are a total of six series of accelerations since the accelerometers are triaxial. Ideally, it would be desirable to have as many accelerographs as measurement points on the structure. This significantly reduces the duration of the experimental campaign but is unfeasible due to the high cost of the commercial measuring equipment. However, it would be feasible with the node presented due to its low cost.

In the second phase, operational modal analysis methods are applied in both the time and frequency domain. This allows the identification of the dynamic properties of the structures. The commercial software ARTeMIS Modal Pro is used to carry out this task, which is widely used in the scientific community.

4.2 Expo-92 bridge

This structure is located a few metres from the Engineering School of the University of Seville and, specifically, in the Water Street. It consists of a reinforced concrete bridge with a curved direction, with a length of 38 metres. The deck is 16.20 metres wide, which allows for the passage of both people and vehicles. The deck is supported by the abutments and two groups of isolated piers, which are also made of reinforced concrete. The piers have a hexagonal cross section and are separated by a distance of 4.80 m in the transverse direction of the bridge. A panoramic view of the bridge under study is shown in Figure 3.

Figure 3.

Expo-92 bridge. Seville, Spain.

4.2.1 Dynamic characteirsation

According to Figure 4, data collection on the bridge is carried out at 28 points, distributed in four rows in the longitudinal direction of the bridge. The accelerometers are placed on the bridge deck, with the z-axis in the direction of gravity and upwards. Point P1 was common to all ambient vibration tests. In order to measure all the points of the structure, eight tests were carried out with the commercial equipment and eight tests with the presented node.

Figure 4.

Measurement points on the bridge.

As a reference, eight 352C33 accelerometers from the manufacturer PCB piezotronics were used. The output signals from the sensors are collected by an eight-channel data acquisition system (DAQ) and, in particular, the SIRIUS-8xACC model from the manufacturer Dewesoft. The transformation of the analogue to digital signal takes place by means of a 24-bit ADC embedded in the DAQ. For this equipment, a measuring range of ±2 g was set, as with the node presented. A sampling frequency of 100 Hz was adopted for the commercial equipment. In contrast, a value of 31.25 Hz was adopted for the low-cost equipment since the interest frequencies are below 10 Hz and this sampling frequency satisfies Nyquist theorem. Prior to the experimental campaign, a finite element model was prepared. This preliminary study made it possible to define the duration of the time series with a better criterion, establishing a length of 8 minutes. Figure 5a illustrates the collection of data with the commercial equipment, while Figure 5b shows the low-cost equipment.

Figure 5.

Ambient vibration test on the bridge: (a) 352C33 accelerometer and (b) node.

As mentioned above, the identification of the natural frequencies of the bridge is done with ARTeMIS Modal Pro. Only the bridge deck is introduced in this software. The 28 points measured on the structure define a grid, with all points at the same elevation. The x and y coordinates of the points match Figure 4. By proceeding in this way, the model shown in Figure 6 is obtained. The acceleration series recorded during the tests are assigned to each of these nodes.

Figure 6.

Bridge geometry in ARTeMIS modal pro.

The dynamic characterisation of the bridge under study has been carried out using an operational modal analysis method in the frequency domain and another in the time domain. In particular, the EFDD and SSI-UPC-Merged methods are used, obtaining practically the same results. The first four modes of vibration of the structure are located in the range between 0 Hz and 10 Hz. A summary of the natural frequencies obtained with the EFDD method, both with commercial and low-cost equipment, is given in Table 2. The maximum relative error in the identification of this modal parameter with the equipment presented here is 9.68%.

Mode	Campaign N° 1—352C33 (Hz)	Campaign N° 2—This work (Hz)	Deviation (%)
Mode 1	2.808	2.960	5.41
Mode 2	5.493	5.439	0.98
Mode 3	7.448	7.075	5.01
Mode 4	8.060	8.840	9.68

Table 2.

Dynamic characterisation of the bridge.

The low-cost equipment is able to estimate the second natural frequency of the bridge with an excellent degree of approximation, with a relative deviation of less than 1%. In the identification of the fundamental frequency and the frequency associated with the third mode of vibration, the relative error increases to about 5%. This difference is even greater in the last detected vibration mode. It should be noted that the equipment presented has a significantly lower cost than commercial equipment but, even so, it allows the identification of the first three modes of vibration of the structure with a good degree of approximation. Although the deviation is higher in other modes of vibration, this is not very relevant as the higher modes are less significant from a structural point of view.

4.3 The homage tower of the Utrera’s castle

This building belongs to the Utrera’s Castle and is located in the town of the same name in Seville. The castle is considered to be built between the thirteenth and fourteenth centuries. This justifies the use of older building materials such as rammed earth, brick and stone. The tower under study has a rectangular cross section, with a smaller side of 11.50 m and a larger side of 12.12 m. The distance between the base and the parapet of the roof is 18.55 m. The walls are 1.90 m thick, although the side with the stairs is thicker. The tower is divided in height into two chambers, which are limited at the top by two vaults. In Figure 7, it can be seen a photograph of the current state of the Homage Tower.

Figure 7.

Homage tower of the Utrera’s castle.

4.3.1 Dynamic characterisation

The recording of accelerations in the tower is carried out at eight points, which are shown in Figure 8. In this last figure, the orientation of the accelerographs can be seen, with the z-axis in the direction of the gravitational acceleration, but in an upward direction. Point A was taken as the reference point for all tests so that seven ambient vibration tests were carried out at the Homage Tower.

Figure 8.

Measurement points on the tower.

The first experimental campaign was done with two GMS Plus6-73 devices, manufactured by GeoSIG. This device contains, among other components, a triaxial accelerometer, a 24-bit analogue-to-digital converter, a GPS and a battery. The second experimental campaign was carried out with the low-cost equipment. In both cases, a measurement range of ±2 g was taken. However, there are differences with regard to the sampling frequency. In the GMS Plus6-73 station, a value of 200 Hz was used, while in the developed node, a value of 125 Hz was adopted. Each test has a duration of 10 minutes. Figure 9 shows photographs taken during the execution of the tests with the two devices.

Figure 9.

Ambient vibration test in the homage tower: (a) GMS Plus6-73 and (b) node.

The acceleration series recorded during the tests are transferred to the ARTeMIS Modal Pro. Previously, it is necessary to generate the geometry of the structure in a simplified form. This is composed of the eight points, where readings have been taken. In addition, the base of the tower is represented by four points. These have the same x and y coordinates as the points of the first floor but are located at zero elevation. The model generated in this way can be seen in Figure 10.

Figure 10.

Tower geometry in ARTeMIS modal pro.

The estimation of the natural frequencies of the Homage Tower is also carried out with the EFDD and SSI-UPC-Merged methods. The results obtained with the EFDD method in the two experimental campaigns are summarised in Table 3. In the range from 0 Hz to 10 Hz, two modes of vibration were identified. As it can be seen in this table, the maximum relative error between the natural frequencies obtained with both experimental campaigns is 0.08%.

Mode	Campaign N° 1—GMS Plus6-73 (Hz)	Campaign N° 2—This work (Hz)	Deviation (%)
Mode 1	3.359	3.357	0.06
Mode 2	7.208	7.202	0.08

Table 3.

Dynamic characterisation of the tower.

The identification of natural frequencies with the reported node is less than 0.1% compared to the commercial equipment. This highlights the usefulness of low-cost equipment as a diagnostic tool in heritage buildings.

5. Conclusions

This chapter introduces the IoT paradigm and its standardisation as a strategic framework for the development of large-scale deployable structural health monitoring systems. Following the presentation of the emergence and evolution of the IoT, as well as the opportunity of its application to the field of structural monitoring, this work has made it possible to reflect the importance of structural monitoring in the field of civil engineering and architecture. This relevance is even more important in the field of heritage and historical buildings, where construction techniques and technical solutions are not usually standardised.

Based on these premises, the requirements that would enable such a mass deployment of SHM solutions have been presented, which have resulted in a secure, modular and scalable architecture that allows its adaptability to different types of potentially monitorable structures, such as bridges and towers.

Subsequently, the proposed architecture has been validated through its development and subsequent real field tests (TRL7 level) such as the Expo92 bridge located at the Engineering School of the University of Seville (Spain) and the Homage Tower of the Utrera’s Castle (Spain). Parallel monitoring with two traditional commercial reference systems has made it possible to validate the results obtained by the developed solution. With these data sets, the natural frequencies of the two structures were estimated. The error in the estimation of the fundamental frequency is around 5% in the first case study and below 0.1% in the second case study. With regard to the natural frequencies associated with the second and third modes of vibration, deviations within the range of 5% are observed. These deviations are more pronounced for the natural frequencies associated with the higher modes. With these results, it can be concluded that the node presented in this paper can be used to carry out a dynamic characterisation of structures and, at least, to detect damage on them.

From the point of view of signal processing, the architecture presented proposes a deployment of independent microservices that allow the application of different processing techniques depending on the monitored structure, which is presented as an opportunity, taking advantage of the trend of recent years, to integrate artificial intelligence and deep learning algorithms, such as [51] that allow Levels III and IV of SHM to be achieved. On the other hand, minimising energy consumption is always a technological challenge in any wireless monitoring solution such as the one presented, so minimising this energy consumption, mainly through Edge Computing techniques that minimise the amount of data to be transmitted wirelessly, is one of the most relevant current and future lines of research in this field of application.

The presented solution, which has been developed according to the stack of layers proposed by the IoT framework, allows centralised efforts to be applied in the different layers without altering the relationship between them, which guarantees the future integration of the different SHM solutions that may appear in the state of the art, both for signal processing in the application layer and in the perception layer, as well as for the migration to other types of wireless communication that may appear. Therefore, it can be concluded that this work presents and validates an IoT-SHM solution for low-cost, secure, scalable, low-power, wireless and large-scale deployable structural monitoring, which positions it as a novel reference in the current state of the art.

Acknowledgments

This work was fully supported by proyecto PID2019-107258RB-C31 financiado por MCIN/ AEI /10.13039/501100011033. The authors of this work also thank the Ayuntamiento de Utrera for their collaboration in the research.

Appendices and nomenclature

DAQ

data acquisition

DATS

data acquisition and transmission system

DMS

data management system

DPCS

data processing and control system

EFDD

enhanced frequency domain decomposition

FOTA

firmware over the air

IMS

inspection and maintenance system

IIC

inter-integrated circuit

IoT

Internet of Things

LoRaWAN

long range wide area network

LTE

long-term evolution

MEMS

micro-electro-mechanical systems

MQTT

message queuing telemetry transport

NB-IoT

narrow band internet of things

OMA

operational modal analysis

PCB

printed circuit board

RTOS

real-time operating system

QoS

quality of service

SHES

structural health evaluation system

SHM

structural health monitoring

SPI

serial peripheral interface

SS

sensory system

SSL

secure sockets layer

SSI-UPC

stochastic subspace identification-unweighted principal component

TCP

transmission control protocol

TLS

transport layer security

TRL

technology readiness level

UART

universal asynchronous receiver-transmitter