Low-Cost Wireless Structural Health Monitoring of Bridges · Komarizadehasl S, Lozano F, Lozano-Galant JA et al.

1. Introduction

Civil infrastructures and structures might be considered the leading basis of modern society; consequently, their health state is of the highest importance [1]. However, data provided by the American Society of Civil Engineers (ASCE) infrastructure indicates about 9.1% of bridges in the United States of America are structurally inefficient. Every day, over 188 million trips are taken across structurally deficient bridges. In addition, in average, a bridge in the United States of America is 43 years old, which is close to its expected life span [2].

Evaluating these structures along with their health-state assessment is needed to reduce reparation costs, maintenance, and, eventually, for certifying infrastructure/structure safety [3,4,5]. Structural Health Monitoring (SHM) applications issue statistics on structures’ functioning states and their structural responses [6,7,8].

For measuring the structural responses (such as deflections, strains, rotation, temperature, humidity, and accelerations) over time, sensors are widely used in SHM systems [9,10]. The recorded information of the sensors is then used for structural performance estimations [11,12,13]. In fact, natural phenomena can be categorized as static or quasi-static and dynamic [14,15,16]. On the one hand, some environmental elements (such as temperature or humidity) change very slowly, so they can be considered to perform either as quasi-static or static [17,18]. On the contrary, some events (such as traffic-induced vibrations, ambient activities, and waves from seismic activities) are considered dynamic because their changing rate is significantly affected by the time [19,20,21].

Accelerometers are force-sensors that are broadly used for measuring vibrations [22]. According to [23], accelerometers are frequently set up on one of the following principles:(1)Piezoelectricity: This is one of the most popular types of accelerometers in the industry for measuring [24], based on specific materials’ piezoelectric effect. For measuring the dynamic changes in mechanical variables, the piezoelectric effect of particular materials is used [25]. The main advantage of these devices is that they can operate on a wide range of frequencies (up to 12 kHz) [26];(2)Piezoresistivity: The second most popular vibration acquisition sensors are piezoresistive (strain gauge) accelerometers. This kind of accelerometer measures the change in electrical resistance of a particular piezoresistive element when vibrations are induced [23];(3)Differential capacitive measurement: These accelerometers identify vibrations by measuring the capacitance changes of a seismic mass [23];(4)Micro-Electro-Mechanical Systems (MEMS) are silicon-based microsensors built from any principles mentioned above [23].

It is essential to mention that the use of accelerometers varies in different SHM applications [27,28]. According to the literature (see, e.g., [29]), most civil structures’ significant natural frequencies vary between 0.2 and 100.0 Hz. The eigenfrequency range of short-span bridges (up to 40 m of span length [30]) usually ranges between 3 and 30 Hz [31,32,33]. Medium- and long-span bridges have eigenfrequencies ranging between 0.1 and 8.0 Hz (such as [34,35,36]). It should also be noted that most of the ambient vibrations in civil structures are characterized by low amplitudes [37,38]. In fact, the acceleration amplitude of these structures can be as low as 0.04 g. This characteristic shows the necessity of using accelerometers with low noise density and high sensitivity for SHM of bridges [39]. The commercial MEMS accelerometers have a bandwidth of a few kHz. However, their design for higher frequencies has a lower noise density [23,40]. Finally, it should be highlighted that the sensitivity of an accelerometer has an indirect ratio with the squared value of the resonant frequency. For that reason, the higher the accelerometer’s sampling frequency, the lower its sensitivity [41]. For that, many researchers tried to develop low-cost MEMS accelerometers with limited sampling frequency and vibration amplitude range for civil SHM [42].

A literature review on various low-cost MEMS accelerometers is organized in the following paragraphs. Even though many applications use low-cost accelerometers, only a handful of them were used in civil engineering.

  1. Grimmelsman et al. [43] studied a low-cost accelerometer (ADXL335). In their work, this low-cost accelerometer’s performance and capability were compared with conventional instrument-grade accelerometers (PCB 393A03 and 3741E122G);
  2. For providing the structural modal analysis from several synchronized low-cost accelerometers (LIS344ALH) distributed along with a structure, a system was proposed by Girolami et al. [44];
  3. The research of Ozdagli et al. [45] used the accelerometer MPU6050 for a real-time, Low-cost, Efficient Wireless Intelligent Sensor (LEWIS). In the performed experiments, LEWIS was placed on a breadboard for signal acquisition. Moreover, its measurements were compared with a Linear Variable Differential Transformer (LDVT) sensor and a commercial accelerometer (PCB 3711B1110G);
  4. An updated version of LEWIS is proposed by Aguero et al. [46]. This work describes LEWIS 2, which solved three main drawbacks of the first LEWIS. It is mentioned in the introduction that LEWIS 2 has data storage, is battery-powered, and is more accurate than LEWIS. The measurements were compared to those of a LDVT sensor;
  5. Meng et al. [47] propose a low-cost acquisition system composed of a LSM9DS1 accelerometer and a Raspberry Pi 4. This system was evaluated with a commercial accelerometer (PCB 356B18);
  6. Recently a Cost Hyper Efficient Arduino Product (CHEAP) was introduced by Komarizadehasl et al. [23]. Different laboratory tests were performed to validate CHEAP’s performance, and the results of CHEAP were compared with those of two commercial seismic accelerometers (PCB 393A03 and PCB 356B18).

Table 1 summarizes the essential characteristics of the reviewed accelerometers. It includes the following information collected in columns (1) sensor number, (2) the name of the accelerometer, (3) acceleration range: the maximum acceleration amplitude capacity of the sensors, (4) sampling frequency: the sampling speed rate of the accelerometer—also, half of this amount corresponds to the accelerometer’s bandwidth, (5) spectral Noise Density (ND), (6) NTP: the capability of the sensor’s clock for getting constantly synchronized with the exact time of the Internet, and (7) Internet access: being connected to the Internet for data uploading or wireless access.

Table. Low-cost solutions in the literature to measure accelerations. (See the original article for the full table.)

The analysis of Table 1 shows that almost all the systems have more or less the same acceleration range and sampling frequency. LSM9DS1 (No. 5) is the only accelerometer with a very high sampling frequency that needs further investigation. Furthermore, the shown ND in Table 1 is reported from the datasheets provided by the sensor producer company. This has been calculated in a standard temperature (25 degrees Celsius for MPU6050) and perfect laboratory conditions. The presented low-cost solutions of Table 1 have not tested the ND of their representatives after they have wired them up. The device of Meng et al. [47] is the only intelligent system with the ability to synchronize its internal clock with the exact time of the Internet through the NTP procedure. The analysis of Table 1 further shows that only sensors numbered 2 and 5 can access the Internet for uploading their results in a cloud server.

Further study of the information as mentioned earlier presented in the literature review shows a lack of consistent work, including all the following points in single inception for a low-cost accelerometer: (1) access to the time through Internet, for accurate sensor synchronization, (2) wireless control, (3) scheduling for a synchronized data acquisition experiment, and, finally, (4) measuring the ND and data resolution after the system was wired and (5) performing experimental eigenfrequency analyses on actual infrastructures.

This paper aims to adapt CHEAP to solve all the detected problems and to develop a new Low-cost Adaptable Reliable Accelerometer (LARA). LARA has the following main improvements: (1) higher bandwidth: achieving a faster sampling frequency (333 Hz) by efficiently rewriting all the library codes, (2) independency: this accelerometer is not dependent on any connected computer for data acquisition—LARA has an attached low-cost microprocessor that allows the acquired data to be saved on an SD card or any portable flash or hard drive, (3) triaxial data acquisition: having all the accelerometers of LARA in x, y, and z direction aligned makes the final product suitable for triaxial data acquisition—besides, if needed, it can work as a uniaxial, biaxial accelerometer, (4) low Noise Density, (5) Internet timestamp and synchronization: the Network Time Protocol (NTP) of LARA is activated. Firstly, NTP synchronizes the clock of each LARA with the accurate time of the Internet constantly. Secondly, each acquired vibration receives an accurate timestamp. This is the feature that makes post-processing of sensor synchronization possible. The literature review reveals no other proper sensor synchronization solution that uses free software. Instead of having Data Acquisition (DAQ) system, LARA uses python software for scheduling a measurement. This way, different LARAs can start acquiring information simultaneously in other parts of a structure, (6) experimental analysis: in this paper, a short-span was instrumented and the eigenvalue analyses of LALA are compared with those of certified vibration sensing devices, and (7) wireless accessibility: the user can have access to the saved data wirelessly through the Internet. Therefore, LARA can be controlled, get upgraded through online commands and data can be recorded. In addition, LARA’s weight is around 207 g which makes it 150 g or 58% lighter than CHEAP.

A laboratory test campaign was designed to validate LARA’s performance against its older version. In this set of experiments, frequencies ranging from 0.1 Hz to 32 Hz and acceleration amplitudes in the range of [0.001–9.806 m/s2] were tested.

Furthermore, in this paper, another field test is performed on a short-span footbridge in Barcelona. In this experiment, the eigenfrequencies of LARA are compared with a commercial dynamic sensor (HI-INC) [53]. HI-INC is a dynamic inclinometer with a sampling frequency of 250 Hz. This sampling frequency is enough for acquiring the main mode shapes of the footbridge under study. It should be noted that the reported noise density of the used commercial sensor (HI-INC) in its datasheet is 0.0004 degrees/√Hz. As in other works in the literature (see, e.g., [54]), this MEMS inclinometer was only used for comparing the eigenfrequency analysis of LARA.

This paper is organized as follows: in Section 2, LARA (the updated version proposed by this paper) is presented with its detailed characteristics. In Section 3, the validating laboratory tests of CHEAP, LARA, and MPU9250 are presented together with their results. In Section 4, the eigenfrequency validation of LARA on a short-span bridge is presented. Finally, the main conclusions are drawn in Section 5.