1. Introduction
Traditional methods for monitoring bridge health primarily depended on physical inspections, which come with inherent limitations. They are time-consuming, labor-intensive, and probably unreliable [
1]. Furthermore, any delays in taking action or neglecting maintenance may lead to substantial future costs, especially for infrastructures with critical importance [
2].
To overcome these drawbacks, techniques for bridge structural health monitoring (BSHM) have been developed, incorporating both static and dynamic approaches. Similar to static, dynamic analysis is divided into short- and long-term evaluations. The first one focuses on analyzing structural behavior during or after specific events, such as earthquakes, load testing of bridges, and responses to live loads. In contrast, the objective of the long-term assessment is to identify any deviations through the structural dynamic parameters over time, which would probably result in recognizing the presence, determining the severity and type, and pinpointing the location of damage or deterioration that may require maintenance or repair. This is achieved through a combination of data collection, analysis, and reporting [
3].
Structural, environmental, and operational sensors serve different purposes in this process [
4,
5]. The structural sensors commonly used in BSHM include strain gauges [
6], displacement transducers [
7], velocity transducers [
8], and accelerometers [
9]. Among these, accelerometers play a key role in capturing the structural dynamic response, which are primarily categorized into three types: piezoresistive [
10], piezoelectric [
11], and microelectromechanical system (MEMS) [
12,
13]. With the advancements in MEMS accelerometers, their integration into BSHM has markedly increased. MEMS emerges as a practical alternative to other types, given its attributes such as low power consumption, high sensitivity, cost-effectiveness, and sufficient sampling frequency [
4].
From another standpoint, sensing systems can be classified into wired and wireless. While the first employ physical cables for data transmission [
14], wireless sensors utilize radio frequency or Bluetooth [
15]. Despite offering significant advantages, such as ease of installation, relocation flexibility, remote monitoring, and cost-effectiveness, wireless sensing systems present distinct challenges. One of the primary concerns is their reliance on batteries as the primary power source. This constraint restricts the continuous operation of wireless sensors, limiting the duration of data collection and imposing limitations on their deployment in remote or inaccessible locations. Furthermore, wireless data transmission is inherently less reliable compared to its wired counterpart, often resulting in lower data transfer rates. Another challenge lies in synchronizing the data captured by multiple sensors. Unlike wired systems, where a central system clock housed at the data server provides a unified time reference, wireless setups face the complexity of coordinating data timestamps [
16]. This issue can introduce errors in data interpretation and hinder the aggregation of sensor readings into meaningful insights. To address these probable transmission delays [
17], various techniques have been developed, including time-stamp estimation and correlation functions. These methods aim to establish a consistent time reference across the network, enabling accurate synchronization and facilitating reliable data collection [
4].
Assessing long-term behavior in bridges involves dynamic approaches to determine natural frequencies, mode shapes, and damping ratios [
3]. Experimental modal analysis (EMA) and operational modal analysis (OMA) are commonly employed techniques. EMA uses controlled excitation sources for induced vibrations, while OMA relies on structural vibration during operational states, proving useful for continuous monitoring in huge and hard-to-access structures such as bridges and towers [
4]. OMA offers economic advantages by eliminating the need for controlled excitation equipment. Common OMA methods include peak picking (PP), frequency domain decomposition (FDD), and enhanced frequency domain decomposition (EFDD) in frequency and stochastic subspace identification (SSI) in time domain analysis, offering two methods: SSI-data and SSI-covariance. To assess structural damage, damage-sensitive features (DSFs) are derived from modal identification techniques, relying on changes in dynamic responses. DSFs include shifts in natural frequencies, alterations in mode shapes, variations in strain levels, adjustments in damping properties, and modifications in energy dissipation [
4]. Among these, mode shapes, less affected by environmental factors, offer valuable spatial information for damage localization.
These papers have significantly contributed to the progress of wireless monitoring systems for structures. Guérineau et al. [
18] proposed a wireless SHM framework utilizing a seismic MEMS accelerometer for bridge monitoring under ambient excitation. In parallel, Araujo et al. [
19] worked on a wireless sensing system with high synchronization for OMA, which was developed under a Spanish ministry-funded project. He et al. [
20], Hu et al. [
21], and Whelan et al. [
22] independently explored vibration-based monitoring on a highway bridge using wireless measurements. Komarizadehasl et al. [
23] investigated a low-cost triaxial accelerometer based on Arduino technology and applied it in the eigenfrequency analysis of a footbridge. Reviews by Abdulkarem et al. [
24], Zhou and Yi [
25], Lynch [
26], and Sabato et al. [
27] underscore the growing importance of wireless sensor networks in structural health monitoring. Kim et al. [
28] applied a wireless sensing system on the Yeondae Bridge in Korea. Lynch et al. [
29] compared a low-cost wireless sensor network in the Geumdang Bridge, Korea, to a commercial system, showing comparable results. Dai et al. [
30] designed, implemented, and tested a wireless sensor network (WSN) for BSHM on the ZhengDian viaduct bridge in China. Reyer et al. [
31] proposed a wireless sensor network using readily available sensors for capturing and analyzing vibration features of bridges. Gutiérrez and Garita [
32] developed a cost-effective wireless bridge monitoring system prototype with a web interface for vibration data analysis. Asadollahi et al. [
33] deployed a wireless smart sensor network on a cable-stayed bridge in South Korea, collecting extensive data for statistical analysis of modal properties and exploring temperature–natural frequency relationships. Furthermore, Chae et al. [
34] proposed a wireless sensor system for BSHM using ZigBee for short-distance and code division multiple access (CDMA) for long-distance communication, integrating a versatile one-channel data logger. The collection of these papers demonstrates the extensive range of applications and innovative advancements in wireless monitoring systems, which serve to improve bridge structural health assessment.
On the other hand, evaluating the displacement response of the bridge, including each member, is necessary to detect fatigue damage and preempt brittle fracture. It is important to specify displacement responses in members that are susceptible to stress concentrations. Accurate assessment of displacement response under live loads, a dominant factor in fatigue damage, is vital for effective bridge maintenance [
35]. Converting acceleration to displacement allows for determining the bridge’s movement in response to external forces. While acceleration reveals dynamic forces, displacement response provides crucial information about the actual deformation and movement of bridge elements.
Amidst the challenges faced by wireless monitoring systems, which often engage with issues related to system lifetime, hardware affordability, system configurability, and data reliability, this study illustrates an innovative and cost-effective wireless data acquisition system designed for OMA in bridges to address these concerns. The system comprises a series of custom-designed, battery-powered wireless sensors equipped with tri-axial MEMS accelerometers (ADXL355 version, from Analog Devices Inc., Cambridge, MA, USA), a cost-effective Wi-Fi module, a commercial Wi-Fi 4G LTE router powered by a solar panel, and a web-based graphical interface to configure the test and receive, store, and download data from the field. To optimize energy consumption, the system employs low-power strategies by using energy-efficient hardware components and a standby mode to minimize continuous power usage. The gateway, along with the solar panel utilized in the bridge monitoring, boasts versatile functionality, suitable for various off-grid applications requiring 12 V batteries. It includes a long-lasting lithium battery capable of over 3000 cycles, significantly surpassing lead-acid batteries. The 25-watt waterproof panel is durable and equipped with a user-friendly connector for easy plug-and-play. The paper also addresses various challenges inherent in wireless communication, including data preprocessing, synchronization, system lifespan, and ease of remote configuration.
This paper initially details the hardware design of the sensor in
Section 2.1.1, emphasizing the participation of the components such as MEMS accelerometer, wireless module, memory, host microcontroller, USB communication, battery, antenna, and enclosure box on the printed circuit board (PCB). Then, the workflow and the operational process upon sensor awakening are elaborated in
Section 2.1.2. The involvement of the other two physical components of the wireless system—router and cloud platform—is discussed in the subsequent sections (
Section 2.2 and
Section 2.3).
Section 3 covers the pre-analytical process involved in signal processing of the monitoring system, focusing on sensor calibration, and data preprocessing steps for eliminating the undesired frequencies and implementing the synchronization technique to multiple wireless sensors. A laboratory test is then conducted (
Section 4.3) to validate the functionality and precision of the monitoring system, using a 4-story shear-type structure on a shaking table, before its real-world application. After preprocessing steps, such as filtering and synchronization, its modal parameters are extracted using frequency and time domain modal identification methods. Additionally, a computational model is employed to validate the experimental results with finite element modeling (FEM). Following the verification, in
Section 4.4, a system composed of 30 wireless sensors is installed on a concrete arch bridge for continuous OMA, and the dynamic results are reported. Highlighting the system’s adaptability and efficiency, displacement is determined from the acceleration data using conventional methods during laboratory testing and an alternative approach based on the Kalman filter for the real-world case study.
The procedure for assessing the dynamic behavior of a structure using a wireless monitoring system is illustrated in
Figure 1. The process initiates with data acquisition, progressing through the pre-processing stage to appropriately prepare the data for subsequent analysis. It then advances to the modal identification, with the objective of extracting the dynamic characteristics inherent to the structure. The procedure concludes with the post-processing and damage detection step. In these sequential stages, the data acquisition system (DAQ) component assumes a crucial role in detecting the targeted vibrations emanating from the structure. The precision of the DAQ holds utmost significance, as it directly impacts the accuracy of the ensuing data analysis.