Development and Application of a Structural Health Monitoring System Based on Wireless Smart Aggregates

Structural health monitoring (SHM) systems can improve the safety and reliability of structures, reduce maintenance costs, and extend service life. Research on concrete SHMs using piezoelectric-based smart aggregates have reached great achievements. However, the newly developed techniques have not been widely applied in practical engineering, largely due to the wiring problems associated with large-scale structural health monitoring. The cumbersome wiring requires much material and labor work, and more importantly, the associated maintenance work is also very heavy. Targeting a practical large scale concrete crack detection (CCD) application, a smart aggregates-based wireless sensor network system is proposed for the CCD application. The developed CCD system uses Zigbee 802.15.4 protocols, and is able to perform dynamic stress monitoring, structural impact capturing, and internal crack detection. The system has been experimentally validated, and the experimental results demonstrated the effectiveness of the proposed system. This work provides important support for practical CCD applications using wireless smart aggregates.


Introduction
Concrete structures, such as civil infrastructures, residential buildings, and industrial structures, are possibly the most common structures. However, the health of concrete structures is constantly affected and degraded throughout their service life by various kinds of factors, such as aging, fatigue, corrosion, and even natural disasters. The accumulated degradation will reduce the structures' disaster resistance ability, and can sometimes lead to partial failure, failure, or even the complete collapse of the structures. The potential for such accidents is a direct threat to the safety of both lives and properties. With these rising concerns, health monitoring technology of concrete structures has become one of the top research focus areas in recent years. Among them, the use of piezoceramic transducers for the health monitoring of concrete structures has achieved good initial results.
At present, piezoelectric-based structural health monitoring technology can be classified into two main categories: active monitoring and passive monitoring. Based on the damage diagnosis methods, active monitoring technology can also be divided into the stress wave-based method and the mechanical impedance-based method. The basic principle of the wave-based method is to use the active sensing property of piezoelectric ceramic to establish the actuator-sensor arrays inside structures. Through scanning and analyzing the difference between the original signal and the current

Selection of Wireless Network
Currently, many types of wireless technologies are available. Table 1 provides a comparison among the most commonly used wireless technologies, namely Zigbee, Bluetooth, Wi-Fi, and mobile communications. Although Wi-Fi has a high data transfer rate, it is not suitable for long-term SHM due to its high power consumption. Bluetooth also has a relatively high data transfer rate. However, the effective transmission distance is short (about 10 m). Mobile communication technology covers long distance communications. However, the service fees increase the cost for practical large scale SHM. Zigbee, based on the IEEE802.15.4 protocol, has many advantages over other wireless technologies for SHM-type applications, such as self-organization, low power, low cost, and high network capacity. These advantages make Zigbee a preferred wireless network for SHM. Moreover, smart aggregate-based SHM requires a relatively high data rate to cover the high bandwidth

Selection of Wireless Network
Currently, many types of wireless technologies are available. Table 1 provides a comparison among the most commonly used wireless technologies, namely Zigbee, Bluetooth, Wi-Fi, and mobile communications. Although Wi-Fi has a high data transfer rate, it is not suitable for long-term SHM due to its high power consumption. Bluetooth also has a relatively high data transfer rate. However, the effective transmission distance is short (about 10 m). Mobile communication technology covers long distance communications. However, the service fees increase the cost for practical large scale SHM. Zigbee, based on the IEEE802.15.4 protocol, has many advantages over other wireless technologies for SHM-type applications, such as self-organization, low power, low cost, and high network capacity. These advantages make Zigbee a preferred wireless network for SHM. Moreover, smart aggregate-based SHM requires a relatively high data rate to cover the high bandwidth features of PZTs. This is a challenge given the limited computational powers of the microcontrollers and the limited power budget from the batteries. This paper develops a Zigbee-based wireless smart aggregate CCD system that solves such conflict between the power and data rates, and hence enables to be a high sensitive device for large scale concrete SHM.

WSA Monitoring Nodes
The WSA health monitoring node is built with two main parts: the piezoceramic smart aggregates and the signal transceiver module. The piezoceramic smart aggregates are used to perform active sensing. They can be used to generate excitations and send out stress waves across the concrete structure, and can also be used to sense the propagated stress waves for concrete health analysis. The piezoelectric smart aggregate has the distinct properties of sensing and actuating dual-functions, a wide frequency band, quick responses, a good combination effect with concrete, a low cost, and a long service time [12][13][14][15][16][17]. These properties of the smart aggregate are obviously different from strain sensing cement-based materials doped with carbon nanotubes in fabrication. The smart aggregates need a waterproof and insulating glue layer in fabrication to prevent the embedded PZT chip from being damaged for it to play a role in producing piezoelectric effects for damage identification or SHM. The strain sensing cement-based materials' key technique in fabrication is how to disperse carbon nanotubes in the cement-based materials to ensure the electrical conductivity of the composite for SHM [31,32].
The signal transceiver module is used to acquire the stress wave signals from the smart aggregate and transmit the result wirelessly to the station. The structure of the WSA node is illustrated in Figure 2. The monitoring nodes are encapsulated in white steel protection shells, with dimensions of 150 mm × 100 mm × 50 mm, as shown in Figure 3.
features of PZTs. This is a challenge given the limited computational powers of the microcontrollers and the limited power budget from the batteries. This paper develops a Zigbee-based wireless smart aggregate CCD system that solves such conflict between the power and data rates, and hence enables to be a high sensitive device for large scale concrete SHM.

WSA Monitoring Nodes
The WSA health monitoring node is built with two main parts: the piezoceramic smart aggregates and the signal transceiver module. The piezoceramic smart aggregates are used to perform active sensing. They can be used to generate excitations and send out stress waves across the concrete structure, and can also be used to sense the propagated stress waves for concrete health analysis. The piezoelectric smart aggregate has the distinct properties of sensing and actuating dual-functions, a wide frequency band, quick responses, a good combination effect with concrete, a low cost, and a long service time [12][13][14][15][16][17]. These properties of the smart aggregate are obviously different from strain sensing cement-based materials doped with carbon nanotubes in fabrication. The smart aggregates need a waterproof and insulating glue layer in fabrication to prevent the embedded PZT chip from being damaged for it to play a role in producing piezoelectric effects for damage identification or SHM. The strain sensing cement-based materials' key technique in fabrication is how to disperse carbon nanotubes in the cement-based materials to ensure the electrical conductivity of the composite for SHM [31,32].
The signal transceiver module is used to acquire the stress wave signals from the smart aggregate and transmit the result wirelessly to the station. The structure of the WSA node is illustrated in Figure 2. The monitoring nodes are encapsulated in white steel protection shells, with dimensions of 150 mm × 100 mm × 50 mm, as shown in Figure 3.   features of PZTs. This is a challenge given the limited computational powers of the microcontrollers and the limited power budget from the batteries. This paper develops a Zigbee-based wireless smart aggregate CCD system that solves such conflict between the power and data rates, and hence enables to be a high sensitive device for large scale concrete SHM.

WSA Monitoring Nodes
The WSA health monitoring node is built with two main parts: the piezoceramic smart aggregates and the signal transceiver module. The piezoceramic smart aggregates are used to perform active sensing. They can be used to generate excitations and send out stress waves across the concrete structure, and can also be used to sense the propagated stress waves for concrete health analysis. The piezoelectric smart aggregate has the distinct properties of sensing and actuating dual-functions, a wide frequency band, quick responses, a good combination effect with concrete, a low cost, and a long service time [12][13][14][15][16][17]. These properties of the smart aggregate are obviously different from strain sensing cement-based materials doped with carbon nanotubes in fabrication. The smart aggregates need a waterproof and insulating glue layer in fabrication to prevent the embedded PZT chip from being damaged for it to play a role in producing piezoelectric effects for damage identification or SHM. The strain sensing cement-based materials' key technique in fabrication is how to disperse carbon nanotubes in the cement-based materials to ensure the electrical conductivity of the composite for SHM [31,32].
The signal transceiver module is used to acquire the stress wave signals from the smart aggregate and transmit the result wirelessly to the station. The structure of the WSA node is illustrated in Figure 2. The monitoring nodes are encapsulated in white steel protection shells, with dimensions of 150 mm × 100 mm × 50 mm, as shown in Figure 3.   The smart aggregate is an aggregate with an embedded piezoceramic disk [33,34]. The function of a piezoceramic is illustrated in Figure 4. When the piezoceramic disk inside the smart aggregate is compressed or stretched, the mechanical energy will be transformed into electrical charges, and this piezoelectric effect can be used as a sensor to sense stresses, shown in Figure 4a. On the other hand, under a changing electric field, the piezoelectric ceramic can convert the electric energy into mechanical vibration, where the smart aggregate can be used as actuator to produce excitations, shown in Figure 4b. The mechanical function of the smart aggregate, shown in Figure 4c, can be described as a PZT chip protected by two thin layers and embedded in concrete. As a stress wave is applied from the concrete to the layer, the PZT chip can be compressed or stretched, and an electrical signal will be generated; on the contrary, as an electrical field is applied on the PZT chip, it will generate a stress wave which will propagate in the host concrete, shown in Figure 4d. The orientation of the smart aggregate is defined in the same way as the PZT disk. Therefore, the smart aggregate is designed and installed at the given position, which makes the orientation of the smart aggregate perpendicular to the assumed concrete crack direction. The smart aggregate is an aggregate with an embedded piezoceramic disk [33,34]. The function of a piezoceramic is illustrated in Figure 4. When the piezoceramic disk inside the smart aggregate is compressed or stretched, the mechanical energy will be transformed into electrical charges, and this piezoelectric effect can be used as a sensor to sense stresses, shown in Figure 4a. On the other hand, under a changing electric field, the piezoelectric ceramic can convert the electric energy into mechanical vibration, where the smart aggregate can be used as actuator to produce excitations, shown in Figure 4b. The mechanical function of the smart aggregate, shown in Figure 4c, can be described as a PZT chip protected by two thin layers and embedded in concrete. As a stress wave is applied from the concrete to the layer, the PZT chip can be compressed or stretched, and an electrical signal will be generated; on the contrary, as an electrical field is applied on the PZT chip, it will generate a stress wave which will propagate in the host concrete, shown in Figure 4d. The orientation of the smart aggregate is defined in the same way as the PZT disk. Therefore, the smart aggregate is designed and installed at the given position, which makes the orientation of the smart aggregate perpendicular to the assumed concrete crack direction. The PZT type of piezoceramic is used in this research to build the smart aggregates due to its strong piezoelectric effect. In this paper, PZT-4, with high sensitivity, and applying the transverse piezoelectric effect and orientation of d33, is used. The fabrication process of the smart aggregate is illustrated step by step in Figure 5. The main steps include the soldering of electric wires onto the PZT The PZT type of piezoceramic is used in this research to build the smart aggregates due to its strong piezoelectric effect. In this paper, PZT-4, with high sensitivity, and applying the transverse piezoelectric effect and orientation of d 33 , is used. The fabrication process of the smart aggregate is illustrated step by step in Figure 5. The main steps include the soldering of electric wires onto the PZT disk, waterproofing the PZT disk, installing the PZT disk into a mold, pouring the concrete, and removing the mold. disk, waterproofing the PZT disk, installing the PZT disk into a mold, pouring the concrete, and removing the mold. A calibration test is performed after the concrete is cured to obtain the sensitivity coefficient of each smart aggregate. The electro-hydraulic servo testing machine MTS (CSS-280I-250) is used to calibrate PZT-based smart aggregates. The charge amplifier from Hengke (HK9038) is selected to perform signal conditioning. The signal is recorded using the national instrument data acquisition system (NI9234) [35]. Figure 6 illustrates the calibration process. By applying sinusoidal loads at different frequencies on the smart aggregates, the output from the smart aggregate is correlated with the load applied. Figure 7 shows a case where a sinusoidal load of 10 Hz is applied on the wireless smart aggregates, and the correlation of the wireless smart aggregate's voltage output and the applied MTS load is shown in Figure 8. It can be seen that the smart aggregate bears an almost linear correlation with the load applied. The slope of the curve in Figure 8 represents the sensitivity coefficient of the wireless smart aggregate.  A calibration test is performed after the concrete is cured to obtain the sensitivity coefficient of each smart aggregate. The electro-hydraulic servo testing machine MTS (CSS-280I-250) is used to calibrate PZT-based smart aggregates. The charge amplifier from Hengke (HK9038) is selected to perform signal conditioning. The signal is recorded using the national instrument data acquisition system (NI9234) [35]. Figure 6 illustrates the calibration process. By applying sinusoidal loads at different frequencies on the smart aggregates, the output from the smart aggregate is correlated with the load applied. Figure 7 shows a case where a sinusoidal load of 10 Hz is applied on the wireless smart aggregates, and the correlation of the wireless smart aggregate's voltage output and the applied MTS load is shown in Figure 8. It can be seen that the smart aggregate bears an almost linear correlation with the load applied. The slope of the curve in Figure 8 represents the sensitivity coefficient of the wireless smart aggregate. disk, waterproofing the PZT disk, installing the PZT disk into a mold, pouring the concrete, and removing the mold. A calibration test is performed after the concrete is cured to obtain the sensitivity coefficient of each smart aggregate. The electro-hydraulic servo testing machine MTS (CSS-280I-250) is used to calibrate PZT-based smart aggregates. The charge amplifier from Hengke (HK9038) is selected to perform signal conditioning. The signal is recorded using the national instrument data acquisition system (NI9234) [35]. Figure 6 illustrates the calibration process. By applying sinusoidal loads at different frequencies on the smart aggregates, the output from the smart aggregate is correlated with the load applied. Figure 7 shows a case where a sinusoidal load of 10 Hz is applied on the wireless smart aggregates, and the correlation of the wireless smart aggregate's voltage output and the applied MTS load is shown in Figure 8. It can be seen that the smart aggregate bears an almost linear correlation with the load applied. The slope of the curve in Figure 8 represents the sensitivity coefficient of the wireless smart aggregate.   The signal transceiver module mainly includes four sub-modules: a wireless radio module, a power supply module, a data acquisition module, and a charge amplifier module. Each sub-module is a separate printed circuit board (PCB).
For the wireless radio module, the RF chip MC13193 from Freescale is chosen as the solution for the single chip Zigbee implementation, as shown in Figure 9. Code warrior is used as the software development platform for this microprocessor. For the power module, the MSP430F2 chip was used as a microcontroller to perform battery monitoring and control. A lithium battery with a capacity of 1000 mAh was chosen. The PCB is shown  The signal transceiver module mainly includes four sub-modules: a wireless radio module, a power supply module, a data acquisition module, and a charge amplifier module. Each sub-module is a separate printed circuit board (PCB).
For the wireless radio module, the RF chip MC13193 from Freescale is chosen as the solution for the single chip Zigbee implementation, as shown in Figure 9. Code warrior is used as the software development platform for this microprocessor. The signal transceiver module mainly includes four sub-modules: a wireless radio module, a power supply module, a data acquisition module, and a charge amplifier module. Each sub-module is a separate printed circuit board (PCB).
For the wireless radio module, the RF chip MC13193 from Freescale is chosen as the solution for the single chip Zigbee implementation, as shown in Figure 9. Code warrior is used as the software development platform for this microprocessor.  The signal transceiver module mainly includes four sub-modules: a wireless radio module, a power supply module, a data acquisition module, and a charge amplifier module. Each sub-module is a separate printed circuit board (PCB).
For the wireless radio module, the RF chip MC13193 from Freescale is chosen as the solution for the single chip Zigbee implementation, as shown in Figure 9. Code warrior is used as the software development platform for this microprocessor. For the power module, the MSP430F2 chip was used as a microcontroller to perform battery monitoring and control. A lithium battery with a capacity of 1000 mAh was chosen. The PCB is shown in Figure 10. The MSP430 series was chosen due to its advantages in power saving. The corresponding software platform is IAR EW for MSP430. in Figure 10. The MSP430 series was chosen due to its advantages in power saving. The corresponding software platform is IAR EW for MSP430. As for the sampling module, the STM32F103 series chips were chosen, as shown in Figure 11. The STM32F103 is cost-effective and rapid-responding. The corresponding software platform is EWARM. As for the charge amplifier, the single-channel charge amplifier produced by Beijing Bichuang Company of China was used, as shown in Figure 12. The conversion coefficient is 0.1 mv/pc. The simplified circuit diagram is shown in Figure 13, where Q is the charge quantity produced by the smart aggregate from external force; qin is the quantity of electric charge of input node; qinp is the quantity of electric charge at the charge amplifier input capacitance; qf is the quantity of electric charge at the feedback capacitor; Cinp, Cf, Rf, Uinp, and Uout are the input capacitance, the feedback capacitance, the feedback resistance, the input voltage, and the output voltage of the charge amplifier, respectively. According to the node charge conservation theorem, the input and output charge at every node must be the same, as shown in Equations (1) and (2). In normal operation conditions, the input potential difference of in-phase and phase inversion of operational amplifier is zero; therefore, the input voltage of charge amplifier Uinp is the grounding voltage, namely Uinp = 0. Equation (2) can be simplified to Equation (3); then, the output voltage of the operational amplifier can be obtained. As shown in Equation (4), for a given charge amplifier, the feedback capacitance is constant, and the output voltage of the charge amplifier is proportional to the quantity of electric As for the sampling module, the STM32F103 series chips were chosen, as shown in Figure 11. The STM32F103 is cost-effective and rapid-responding. The corresponding software platform is EWARM. in Figure 10. The MSP430 series was chosen due to its advantages in power saving. The corresponding software platform is IAR EW for MSP430. As for the sampling module, the STM32F103 series chips were chosen, as shown in Figure 11. The STM32F103 is cost-effective and rapid-responding. The corresponding software platform is EWARM. As for the charge amplifier, the single-channel charge amplifier produced by Beijing Bichuang Company of China was used, as shown in Figure 12. The conversion coefficient is 0.1 mv/pc. The simplified circuit diagram is shown in Figure 13, where Q is the charge quantity produced by the smart aggregate from external force; qin is the quantity of electric charge of input node; qinp is the quantity of electric charge at the charge amplifier input capacitance; qf is the quantity of electric charge at the feedback capacitor; Cinp, Cf, Rf, Uinp, and Uout are the input capacitance, the feedback capacitance, the feedback resistance, the input voltage, and the output voltage of the charge amplifier, respectively. According to the node charge conservation theorem, the input and output charge at every node must be the same, as shown in Equations (1) and (2). In normal operation conditions, the input potential difference of in-phase and phase inversion of operational amplifier is zero; therefore, the input voltage of charge amplifier Uinp is the grounding voltage, namely Uinp = 0. Equation (2) can be simplified to Equation (3); then, the output voltage of the operational amplifier can be obtained. As shown in Equation (4), for a given charge amplifier, the feedback capacitance is constant, and the output voltage of the charge amplifier is proportional to the quantity of electric As for the charge amplifier, the single-channel charge amplifier produced by Beijing Bichuang Company of China was used, as shown in Figure 12. The conversion coefficient is 0.1 mv/pc. The simplified circuit diagram is shown in Figure 13, where Q is the charge quantity produced by the smart aggregate from external force; q in is the quantity of electric charge of input node; q inp is the quantity of electric charge at the charge amplifier input capacitance; q f is the quantity of electric charge at the feedback capacitor; C inp , C f , R f , U inp , and U out are the input capacitance, the feedback capacitance, the feedback resistance, the input voltage, and the output voltage of the charge amplifier, respectively. According to the node charge conservation theorem, the input and output charge at every node must be the same, as shown in Equations (1) and (2). In normal operation conditions, the input potential difference of in-phase and phase inversion of operational amplifier is zero; therefore, the input voltage of charge amplifier U inp is the grounding voltage, namely U inp = 0. Equation (2) can be simplified to Equation (3); then, the output voltage of the operational amplifier can be obtained. As shown in Equation (4), for a given charge amplifier, the feedback capacitance is constant, and the output voltage of the charge amplifier is proportional to the quantity of electric charge produced by the piezoelectric sensing element without attenuation and lag phenomenon. Since the quantity of electric charge produced by the piezoelectric sensor is proportional to the external force, the output voltage of the charge amplifier will also be proportional to the external force, which defines the basic principle for smart aggregate-based health monitoring.
The sampling frequency of the wireless sensor used in the experiment is 1 kHz, and the capacity of the wireless sensor is 2 GB. If it is necessary, the data can be uploaded to the gateway directly.
Sensors 2017, 17, 1641 9 of 16 charge produced by the piezoelectric sensing element without attenuation and lag phenomenon.
Since the quantity of electric charge produced by the piezoelectric sensor is proportional to the external force, the output voltage of the charge amplifier will also be proportional to the external force, which defines the basic principle for smart aggregate-based health monitoring.
The sampling frequency of the wireless sensor used in the experiment is 1 kHz, and the capacity of the wireless sensor is 2 GB. If it is necessary, the data can be uploaded to the gateway directly. charge produced by the piezoelectric sensing element without attenuation and lag phenomenon.
Since the quantity of electric charge produced by the piezoelectric sensor is proportional to the external force, the output voltage of the charge amplifier will also be proportional to the external force, which defines the basic principle for smart aggregate-based health monitoring.
The sampling frequency of the wireless sensor used in the experiment is 1 kHz, and the capacity of the wireless sensor is 2 GB. If it is necessary, the data can be uploaded to the gateway directly.

Auxiliary Equipment
To meet the needs of different monitoring functions, the developed system is also equipped with some auxiliary equipment. A function generator, a power amplifier, a solar-cell panel and a rechargeable battery, whose capacity is 50 Ah and nominal voltage is 12 V, are also used for excitation signal generation, amplification, extra power supply, and power storage, respectively.
The power amplifier and function generator are only used in the active monitoring mode. The power amplifier is used to amplify the excitation signals generated from function generator. Considering the fact that line power may not be available in the monitoring field, the system is also equipped with a solar-cell panel and an accumulator.

Active Monitoring Mode
Two kinds of passive and active monitoring modes are used to validate the effectiveness of the proposed WSAs-based SHM system. The proposed passive mode is validated by an internal load monitoring test of a concrete bridge model under impact loading [36]. The experimental results show that the passive mode in the developed wireless system is stable and reliable, and can be applied in concrete bridge's structure health monitoring under impact loading. This paper focuses on the validation of the active monitoring mode.

Active Monitoring Method
The active monitoring mode is based on wave-based active monitoring technology to detect cracks inside a concrete structure. The basic principle is that the smart aggregates, which are used as transducers, are embedded in different locations inside the concrete structure. The actuator generates an excitation signal, which propagates through the concrete structure reaching the sensor receiving side. Based on the analysis of the changes in the signal and the extraction of the damage factor, the on-line system for the diagnosis of the crack damage in concrete structures can be realized. The hardware of the module includes a function generator, a power supply, WSA nodes, and a computer, respectively.
According to the results of the research by [25], the amplitude of the analyzed signal is sensitive to the crack damage in the concrete structure. The attenuation in the amplitude will increase as the damage becomes more severe. Therefore, the signal amplitude can be used as a damage identification parameter. It can also be referred to as the energy value through quantification. The energy value represented by the absolute value of the square integral is the amplitude of the signal after discretization. A damage index can be defined by Equations (5)- (7). In the equations, E j is the energy value at a moment, as shown in Equation (5); and x i is the discrete signal at each collecting point by the sensors. If E h is the energy value in a healthly state, then the relative state of the structure can be expressed by Equation (6). x h (n) and x i (n) are discrete signals in the healthly and damaged states, respectively, which are collected by the sensor.
As shown in Equation (7), it is easy to observe that, when the monitored member is in a healthy condition, H i = 1; when the monitored member is severely damaged, H i tends to zero. In the actual health monitoring, the more severe the damage is, the higher the damage index the structure will have. Therefore, selecting D values as damage indexes in this paper is meaningful.

Validation Test of Active Monitoring Function
To verify the developed active CCD function of the system, a reinforced concrete (RC) beam-column joint with a partial RC floor embedded with WSAs is selected as the object of the validation test. The dimensions and the test site are shown in Figures 14 and 15, respectively. WSAs are embedded in the specimen before the casting of the concrete. Two WSAs are separately arranged at 100 mm and 600 mm to the beam's end, as shown in Figure 16.
The main devices of the test include a reaction wall, an MTS 50 t loading system, and a 200 t hydraulic jack. During the test, the actuator and hydraulic jack are connected horizontally on the reaction wall, and then the specimen is installed horizontally. The column end of the specimen is connected with the hydraulic jack, and the other end is connected with the end plate. Through the pull rod, the reacting-force wall, the hydraulic jack, the specimen, and the end plate are connected. The MTS actuator is linked with the specimen through an adapting piece. The diagram of the loading device is shown in Figure 16. During the loading process, the vertical axial load of the column is applied through the 200 t hydraulic jack. Through the voltage stabilizer and strain gauge, the vertical load is maintained at about 80 t during the test. Meanwhile, the horizontal pseudo-static load is applied by the 50 t actuator on the end of the beam. The first step is applying 20 kN pre-load for the validity check of the component and other facilities. During the test, the combination loading plan of the force control and displacement control is adopted as shown in Figure 17. When the specimen is in the elastic stage, the force control loading plan is used, with the increment of force of 20 kN for one cycle in every loading level; when the specimen yields, the loading plan switches to displacement control. The displacement increment is 1∆, namely ∆/∆ y = ±1, ±2, ±3, ±4 (where ∆ is the applying lateral displacement, and ∆ y is the yield displacement), for one cycle in every level, until the applying force falls to 80% of the bearing capacity or the specimen is directly destroyed. To verify the developed active CCD function of the system, a reinforced concrete (RC) beam-column joint with a partial RC floor embedded with WSAs is selected as the object of the validation test. The dimensions and the test site are shown in Figures 14 and 15, respectively. WSAs are embedded in the specimen before the casting of the concrete. Two WSAs are separately arranged at 100 mm and 600 mm to the beam's end, as shown in Figure 16.
The main devices of the test include a reaction wall, an MTS 50 t loading system, and a 200 t hydraulic jack. During the test, the actuator and hydraulic jack are connected horizontally on the reaction wall, and then the specimen is installed horizontally. The column end of the specimen is connected with the hydraulic jack, and the other end is connected with the end plate. Through the pull rod, the reacting-force wall, the hydraulic jack, the specimen, and the end plate are connected. The MTS actuator is linked with the specimen through an adapting piece. The diagram of the loading device is shown in Figure 16. During the loading process, the vertical axial load of the column is applied through the 200 t hydraulic jack. Through the voltage stabilizer and strain gauge, the vertical load is maintained at about 80 t during the test. Meanwhile, the horizontal pseudo-static load is applied by the 50 t actuator on the end of the beam. The first step is applying 20 kN pre-load for the validity check of the component and other facilities. During the test, the combination loading plan of the force control and displacement control is adopted as shown in Figure 17. When the specimen is in the elastic stage, the force control loading plan is used, with the increment of force of 20 kN for one cycle in every loading level; when the specimen yields, the loading plan switches to displacement control. The displacement increment is 1Δ, namely Δ/Δy = ±1, ±2, ±3, ±4 (where Δ is the applying lateral displacement, and Δy is the yield displacement), for one cycle in every level, until the applying force falls to 80% of the bearing capacity or the specimen is directly destroyed.     The test takes advantage of a Structural Health Monitoring System based on WSAs to perform real-time monitoring. To increase the sensitivity to concrete cracks [9], a sweep sine wave is selected as the excitation signal. The amplitude is 10 V, the duration is 10 s, and the frequency progressed from 100 Hz to 10 kHz. When the specimen is loaded to a certain level, the loading state is maintained for several minutes. During this time, the signal generator (NI9074) produces the excitation signal on one of the smart aggregates, while at the same time the other smart aggregates function as sensors. Then, the specimen returns to the original position and the actuator stops loading. For each loading level, the measurements are continuously carried out 10 times before switching to the next loading level.   The test takes advantage of a Structural Health Monitoring System based on WSAs to perform real-time monitoring. To increase the sensitivity to concrete cracks [9], a sweep sine wave is selected as the excitation signal. The amplitude is 10 V, the duration is 10 s, and the frequency progressed from 100 Hz to 10 kHz. When the specimen is loaded to a certain level, the loading state is maintained for several minutes. During this time, the signal generator (NI9074) produces the excitation signal on one of the smart aggregates, while at the same time the other smart aggregates function as sensors. Then, the specimen returns to the original position and the actuator stops loading. For each loading level, the measurements are continuously carried out 10 times before switching to the next loading level.   The test takes advantage of a Structural Health Monitoring System based on WSAs to perform real-time monitoring. To increase the sensitivity to concrete cracks [9], a sweep sine wave is selected as the excitation signal. The amplitude is 10 V, the duration is 10 s, and the frequency progressed from 100 Hz to 10 kHz. When the specimen is loaded to a certain level, the loading state is maintained for several minutes. During this time, the signal generator (NI9074) produces the excitation signal on one of the smart aggregates, while at the same time the other smart aggregates function as sensors. Then, the specimen returns to the original position and the actuator stops loading. For each loading level, the measurements are continuously carried out 10 times before switching to the next loading level. The test takes advantage of a Structural Health Monitoring System based on WSAs to perform real-time monitoring. To increase the sensitivity to concrete cracks [9], a sweep sine wave is selected as the excitation signal. The amplitude is 10 V, the duration is 10 s, and the frequency progressed from 100 Hz to 10 kHz. When the specimen is loaded to a certain level, the loading state is maintained for several minutes. During this time, the signal generator (NI9074) produces the excitation signal on one of the smart aggregates, while at the same time the other smart aggregates function as sensors. Then, the specimen returns to the original position and the actuator stops loading. For each loading level, the measurements are continuously carried out 10 times before switching to the next loading level.
In the experiment, both the WSA and the original wired smart aggregate are used. As shown in Figures 18-20, respectively, with the development of the crack inside the specimen, the amplitudes of the monitoring signals decrease gradually. The damage indices at each loading level throughout the test process are shown in Figure 21. It can be seen that, as the structure progresses into different damage states, the corresponding damage index changes drastically to different levels. The increase of the damage index reflects that the structure has reached a new damage level. The experiment validates that the developed system can effectively monitor and identify the crack damage status inside the concrete structure. The result of the WSA is consistent with the wired smart aggregate [37]. In the experiment, both the WSA and the original wired smart aggregate are used. As shown in Figures 18-20, respectively, with the development of the crack inside the specimen, the amplitudes of the monitoring signals decrease gradually. The damage indices at each loading level throughout the test process are shown in Figure 21. It can be seen that, as the structure progresses into different damage states, the corresponding damage index changes drastically to different levels. The increase of the damage index

Conclusions
In this paper, a wireless smart aggregate-based concrete crack detection (CCD) system was designed and developed. Experiments were performed to validate the efficiency of the proposed system. The experimental results demonstrate that the developed system can not only effectively identify the development of crack damage inside concrete structures, but also realize the online CCD of concrete internal stress change. Moreover, the developed system is "wire-free" and power efficient, which makes it suitable for the complex environment of a construction site. This work provides important support for practical large scale concrete structural health monitoring, and helps to promote smart aggregate technology for SHM from the laboratory research level to a practical engineering application level.