Smartphone Application for Structural Health Monitoring of Bridges · Figueiredo E, Moldovan I, Alves P et al.

1. Introduction

Investment in road and rail infrastructure has been extensive in most developed countries, and bridges have proved to be among the most vulnerable parts of these assets, as observed by the number of collapses and their dire consequences.

Bridge owners use bridge management systems (BMS) to ensure a coordinated approach for optimizing costs, risks, serviceability, and sustainability of bridges. Initially, BMSs were simply inventories of basic information about bridges, such as construction date, location, and some structural details [1]. Then, they evolved to incorporate information derived from scheduled inspections and maintenance activities. Currently, BMSs intend to cover all activities performed during the service life of bridges, from design to demolition, taking into account public safety, budgetary constraints, and transport network functionality. They possess mechanisms to ensure that bridges are regularly inspected, evaluated, and maintained in a systematic way [2]. Broadly speaking, a BMS aims to ensure safety while minimizing costs. However, BMSs still rely heavily on bridge inspections, especially on the qualitative, and not necessarily consistent, visual inspections. This shortcoming may compromise the structural evaluation, the maintenance decisions, and ultimately the prevention of bridge failures [1].

Following the increased awareness of the limitations of visual inspection, in the last three decades, the non-destructive evaluation (NDE) and the structural health monitoring (SHM) fields have emerged to add more quantitative information into BMSs. NDE concerns the health assessment of a structure, or its components, through offline non-damaging procedures [3]. On the other hand, SHM assumes an online approach, continuous in time and global in nature, with the aim of autonomous monitoring. It should be noted that SHM is more than just monitoring, as simply collecting data does not constitute SHM. Rather, SHM assumes a continuous strategy of damage identification based on monitoring and interpreting the collected data. Therefore, it is fair to observe that NDE may be incorporated into SHM systems, but not vice-versa [1].

In order to facilitate the automatic interpretation of monitoring data and damage identification, SHM has been posed in the context of a statistical pattern recognition (SPR), which is traditionally split into four stages: (i) operational evaluation, (ii) data acquisition, (iii) feature extraction and generation, and (iv) statistical modeling for feature classification. This paradigm has been rooted in the artificial intelligence field, where machine learning algorithms play an essential role for pattern recognition [4]. Damage identification has been defined as a five-level hierarchy [5]: detection, localization, type, severity, and prognosis.

Despite the potential benefits of SHM for the authorities, it is widely recognized that bridge SHM is still not fully capable of producing reliable information on the presence of damage in a full-field monitoring strategy [6]. According to Cawley [7], in order to speed up the knowledge transfer from research to practical application of SHM in bridges, better-focused research and development is needed. Therefore, the scientific community has tried to find out new ways to improve the damage identification process. Recent technologies have unveiled alternative perspectives to manage and observe the response of monitored structures. In particular, we have observed recent breakthroughs in digital advancements, namely in (1) mobile device-assisted monitoring, (2) unmanned aerial vehicles, (3) virtual and augmented realities, and (4) digital twins [8].

In line with the general digitalization trend, the broad availability and low cost of smartphones have justified the recent investment in their use for SHM of bridges. Smartphones contain built-in sensors, storage capabilities, processors, and broadband communication networks, such as the cellular 4/5G network, Wi-Fi, and Bluetooth. Built-in sensors include accelerometers, cameras, microphones, gyroscopes, magnetometers, Global Positioning System (GPS) systems, LiDAR sensors, etc. A summary of the smartphone-related sensor technology and its application in SHM is made in reference [8]. This type of technology is relevant, as it can be used either as SHM systems or as NDE during bridge inspections.

Two main sensing strategies have been pursued, namely the use of a smartphone’s camera for displacement, deformation, and crack measurements through digital image correlation [9,10], and the use of a smartphone’s accelerometer to measure accelerations and estimate natural frequencies of vibration [11]. Even though the use of cameras has gained significant traction in SHM [12], especially in the last decade, this paper is especially concerned with the use of built-in accelerometers.

Since 2015, several publications have reported the applicability of smartphones for SHM of bridges using vibrations measured from the internal accelerometers. Zhao et al. [13] showed the applicability of smartphones to measure internal forces of cables, both in laboratory conditions and on the Xinghai Bay Bridge in China. In 2016, Zhao et al. [14] carried out a comprehensive experimental study on the deck and cables of a laboratory-scale suspended bridge to verify the feasibility of smartphones for a rapid damage assessment of bridges after an extreme event, such as an earthquake. The vibration responses obtained from smartphone accelerometers and conventional sensors were compared in both time and frequency domains. The measurements from smartphones agreed well with the measurements from reference accelerometers in terms of amplitude characteristics. The fundamental natural frequency of the bridge model was estimated with a maximum error of 0.44%.

Feng et al. [15] performed a comprehensive experimental study, involving seismic shaking-table tests and bridge field tests. Three standard smartphones with different accelerometers were compared with a high-quality reference/conventional sensor. The shaking-table tests confirmed that the smartphone sensors were capable of accurately measuring sinusoidal vibration between 0.5 Hz and 20 Hz. The authors also verified that the accelerometers embedded in newer generation smartphones are significantly more accurate than those in the older generations, which suggests that accelerometers are quickly evolving over time. The laboratory and field tests also showed the advantages of smartphone sensors over conventional sensors, such as the ease of installation and data acquisition, as well as wireless transmission.

Using two smartphones in the laboratory, Ozer et al. [11] showed how response measurements from cameras and accelerometers can be blended for experimental modal analysis, benefitting from the accessibility of smartphone technology to provide innovative SHM solutions with mobile, programmable, and low-cost interfaces.

Mei and Gul [16] proposed a crowdsourcing-based methodology for monitoring and evaluation of a population of bridges using the accelerometers mounted in smartphones in a large number of moving vehicles as mobile sensors. The authors showed that the methodology could provide damage features correlated with the severity of damage. Other authors have shown in laboratory experiments the feasibility of bridge frequency estimation using multi-vehicle (i.e., crowdsourced) smartphone accelerations [17].

Shrestha et al. [18] implemented a smartphone-based monitoring system for more than a year covering one bridge in Japan. Six smartphones were streaming accelerations to a cloud server to verify their viability and accuracy. Dynamic properties estimated from this system were compared with those of reference seismometers. Inconsistent sampling rates and faulty sensor readings were identified to be two major problems with the use of the proposed system.

Recently, Ozer et al. [19] used multiple smartphones throughout the Golden Gate Bridge in the United States. The accelerometer data were collected to retrieve modal frequencies and mode shapes of the bridge. Despite the low vibration frequency and amplitude of a long-span suspension bridge combined with limited sensing and acquisition quality of the smartphones, the authors found significant correlation with the modal properties estimated with conventional sensing technology.

In conclusion, early studies assessed the applicability, precision, and resolution of smartphone vibration measurements and concluded that they are reliable in the frequency range of interest for bridges (<20 Hz).

Previous attempts to harvest the capability of the smartphones to recognize and measure frequencies that lay in the effective frequency range have been reported. An effort which is relatively similar to App4SHM was reported by Ozer et al. in 2015 [20]. The authors created CS4SHM, an application that crowdsourced smartphone measurements on flexible bridges and extracted the frequency with the largest spectral coefficient from the acceleration time series. Besides the fact that CS4SHM was developed under IoS, the main difference in respect to App4SHM is that it included no damage detection component, although the authors plan to add that in future developments. Another excellent effort is Orion-CC, an IoS smartphone application for cable force measurement based on monitoring the vibration of the cable [13].

This paper proposes a novel smartphone application for damage detection in bridges under ambient vibration conditions, making use of the built-in accelerometer and following the SPR paradigm. The application is not limited to cables, nor to frequency extraction, as was the case of the applications cited above. Instead, it includes a machine learning algorithm capable of recognizing damage after being trained with data from the undamaged structure. The application is first tested on data sets from a laboratory steel beam to verify the sensing and damage detection capabilities and then it is tested on two twin post-tensioned concrete bridges for damage detection. In both cases, the natural frequencies obtained from the application are compared with those estimated using data sets from a conventional data acquisition system.

App4SHM is particularly well-suited to periodically monitor bridges under operational and environmental conditions. It was developed in a joint effort between the Civil Research Group and the Computer Science Department from the Lusófona University in Portugal. It can be applied to perform SHM or NDE during scheduled inspections, or to assess the structural condition after a catastrophic event, or when required by authorities and stakeholders. It may also be regarded as NDE technology integrated into the SHM of bridges. In the damage identification hierarchy, this application only addresses the first level, related to damage detection.

App4SHM is an Android application already available in the Google Play Store to a limited group of testers, selected upon request after the application was presented at the 10th European Workshop on Structural Health Monitoring [21]. We are planning to make it generally available as soon as the testing phase ends. App4SHM fits into a larger effort to transpose SHM technology from research to practice by developing innovative, low cost data acquisition systems along with supporting software (e.g., [22,23]).

The organization of this article is as follows: besides this section, Section 2 describes the basic architecture of the proposed smartphone application. Section 3 presents and discusses results from laboratory tests conducted on a simply supported steel beam. Section 4 presents the results obtained from ambient vibration tests carried out on two twin concrete bridges. Finally, Section 5 provides the main conclusions of this paper and gives some indications of future works.