Damage Detection of Common Timber Connections Using Piezoceramic Transducers and Active Sensing

Timber structures have been widely used due to their low-cost and environmental-friendly properties. It is essential to monitor connection damage to ensure the stability and safety of entire timber structures since timber connection damage may induce catastrophic incidents if not detected in a timely manner. However, the current investigations on timber connections focus on mechanical properties and failure modes, and the damage detection of timber connection receives rare attention. Therefore, in this paper, we investigate the damage detection of four common timber connections (i.e., the screw connection, the bolt connection, the decussation connection, and the tooth plate connection) by using the active sensing method. The active sensing method was implemented by using a pair of lead zirconate titanate (PZT) transducers: one PZT patch is used as an actuator to generate stress waves, and the other works as a sensor to detect stress waves after propagating across the timber connection. Based on the wavelet packet energy analysis, the signal energy levels of received stress waves under different damage extent are quantified. Finally, by comparing the signal energy between the intact status and the damage status of the timber connection, we find that the energy attenuates with increasing severity of the connection damage. The experimental results demonstrate that the active sensing method can realize real-time monitoring of timber connection damage, which can guide further investigations.


Introduction
Timber structures have a long history of use globally [1] since timber is a ubiquitous and environmental-friendly material that satisfies the requirements for sustainable development. Generally, to construct large-scale structures, individual timber elements are held together through different types of connections. However, the deformation and shrinkage of timber members caused by various issues (i.e., excessive load, wind vibration, or earthquake) can lead to damage to timber connections over time. Prior investigation has demonstrated that the stability and safety of entire timber structures greatly depend on the integrity of connections [2], and the insufficient integrity of connections is the major cause of failures in timber structures [3]. In addition, some researchers noticed that the peak stress, rotational stiffness, and bearing capacity of joints deteriorated due to damages of timber connections [4]. Thus, the monitoring of timber connection damage holds importance in timber and construction industries.
propagation path, the signal energy decreases significantly with damage expansion. Then, to quantify the damage extent, a wavelet packet energy index was applied to compute signal energy of received stress waves. The results indicate that the active sensing method can estimate the damages of different timber connections efficiently.

Active Sensing
In this paper, the active sensing method based on PZT transducers is used to detect the damage of different types of timber connections, and the schematic is depicted in Figure 1. PZT transducers in the form of patches are adopted in this paper since they can be easily bonded on the out surface of interested structure. As shown in Figure 1, the PZT1 patch generates stress waves that propagate across the interface of the timber connection, and the received signal is captured by the PZT2 patch. Both patches are bonded on the top surfaces of the timber specimens. Due to the damage of timber connections, the actual contact area of structures will change and lead to a corresponding energy change of the stress wave (i.e., attenuation). The wavelet packet energy approach is used to quantify the signal energy attenuation, and thus the damage extent of timber connections can be detected. to quantify the damage extent, a wavelet packet energy index was applied to compute signal energy of received stress waves. The results indicate that the active sensing method can estimate the damages of different timber connections efficiently.

Active Sensing
In this paper, the active sensing method based on PZT transducers is used to detect the damage of different types of timber connections, and the schematic is depicted in Figure 1. PZT transducers in the form of patches are adopted in this paper since they can be easily bonded on the out surface of interested structure. As shown in Figure 1, the PZT1 patch generates stress waves that propagate across the interface of the timber connection, and the received signal is captured by the PZT2 patch. Both patches are bonded on the top surfaces of the timber specimens. Due to the damage of timber connections, the actual contact area of structures will change and lead to a corresponding energy change of the stress wave (i.e., attenuation). The wavelet packet energy approach is used to quantify the signal energy attenuation, and thus the damage extent of timber connections can be detected.

Wavelet Packet Energy
The wavelet-packet approach has been widely used in analyzing signals, due to its satisfactory characteristics, such as high time-frequency resolution and effective decomposed capacity. In order to monitor the damage of different timber connections in real time, the wavelet packet analysis is applied in this paper to quantify the attenuation of signal energy at different frequency bands. Even though formulas of the wavelet packet energy analysis are very well-known, we still introduce their principle briefly in this section to clarify the procedure of establishing index, which is given as follow.
Firstly, by a n -level wavelet packet decomposition, the original signal S received by the PZT sensor is decomposed into 2 n signal subsets with different frequency bands. The signal subset X j can be expressed as [43], where m is the data sampling of the decomposed signal subset; j is the frequency band ( ) 1, 2,..., 2 n j = .
Subsequently, the energy of the signal subset , i j E is defined as, where i was the th i measurement.

Wavelet Packet Energy
The wavelet-packet approach has been widely used in analyzing signals, due to its satisfactory characteristics, such as high time-frequency resolution and effective decomposed capacity. In order to monitor the damage of different timber connections in real time, the wavelet packet analysis is applied in this paper to quantify the attenuation of signal energy at different frequency bands. Even though formulas of the wavelet packet energy analysis are very well-known, we still introduce their principle briefly in this section to clarify the procedure of establishing index, which is given as follow.
Firstly, by a n-level wavelet packet decomposition, the original signal S received by the PZT sensor is decomposed into 2 n signal subsets with different frequency bands. The signal subset X j can be expressed as [43], where m is the data sampling of the decomposed signal subset; j is the frequency band ( j = 1, 2, . . . , 2 n ). Subsequently, the energy of the signal subset E i,j is defined as, where i was the ith measurement. The energy vector for the signal at the ith measurement can be given as, Finally, based on the definition of the energy vector E i , the total energy E of the received original signal at the ith measurement can be computed as,

Experimental Setup
The most used connection parts in timber constructions are mechanical fasteners, and four common types of timber connections are used in this paper, as shown in Figure 2. The energy vector for the signal at the th i measurement can be given as, , , , Finally, based on the definition of the energy vector i E , the total energy E of the received original signal at the th i measurement can be computed as,

Experimental Setup
The most used connection parts in timber constructions are mechanical fasteners, and four common types of timber connections are used in this paper, as shown in Figure 2.

Timber Specimen
Four groups of timber specimens (pine wood from North America) with the same dimensions (200 mm × 40 mm × 90 mm) are used in this experiment. Each group has two pieces of the specimens fastened by one connection type, and thus there are a total of eight specimens, as shown in Figure 3: Group A is a connection with four screws (#6*3/4in); Group B is a bolt connection (HEX Bolt 3/8in); Group C is a metal tooth plate connection (MP24), and Group D is a decussation connection (RTB22). For each group, two PZT patches ( φ 10 mm × 2mm), which are sandwiched structures with two electrode layers and one layer of PZT, were mounted at the predetermined location using epoxy (Loctite Heavy Duty 5 min epoxy). Moreover, in this paper, the monitoring of timber connection damage based on active sensing method was performed with artificial damages. The practical damage levels are described as follows: (1) for Group A, Case1-A is the initial health status with four screws all tightened, then we damage the integrity by loosening one, two, and three screws, respectively. In accordance with the different damage levels, the damage cases are called Case2-A, Case3-A, Case4-A, respectively. (2) for Group B, a torque wrench is used to apply the pre-load from 5 to 20 N m. Case1-B is considered as the intact case with the maximal applied torque 20 N m, then the damage cases, namely Case2-B, Casee3-B, and

Timber Specimen
Four groups of timber specimens (pine wood from North America) with the same dimensions (200 mm × 40 mm × 90 mm) are used in this experiment. Each group has two pieces of the specimens fastened by one connection type, and thus there are a total of eight specimens, as shown in Figure 3: Group A is a connection with four screws (#6*3/4in); Group B is a bolt connection (HEX Bolt 3/8in); Group C is a metal tooth plate connection (MP24), and Group D is a decussation connection (RTB22). For each group, two PZT patches (φ 10 mm × 2mm), which are sandwiched structures with two electrode layers and one layer of PZT, were mounted at the predetermined location using epoxy (Loctite Heavy Duty 5 min epoxy). The energy vector for the signal at the th i measurement can be given as, Finally, based on the definition of the energy vector i E , the total energy E of the received original signal at the th i measurement can be computed as,

Experimental Setup
The most used connection parts in timber constructions are mechanical fasteners, and four common types of timber connections are used in this paper, as shown in Figure 2.

Timber Specimen
Four groups of timber specimens (pine wood from North America) with the same dimensions (200 mm × 40 mm × 90 mm) are used in this experiment. Each group has two pieces of the specimens fastened by one connection type, and thus there are a total of eight specimens, as shown in Figure 3: Group A is a connection with four screws (#6*3/4in); Group B is a bolt connection (HEX Bolt 3/8in); Group C is a metal tooth plate connection (MP24), and Group D is a decussation connection (RTB22). For each group, two PZT patches ( φ 10 mm × 2mm), which are sandwiched structures with two electrode layers and one layer of PZT, were mounted at the predetermined location using epoxy (Loctite Heavy Duty 5 min epoxy). Moreover, in this paper, the monitoring of timber connection damage based on active sensing method was performed with artificial damages. The practical damage levels are described as follows: (1) for Group A, Case1-A is the initial health status with four screws all tightened, then we damage the integrity by loosening one, two, and three screws, respectively. In accordance with the different damage levels, the damage cases are called Case2-A, Case3-A, Case4-A, respectively. (2) for Group B, a torque wrench is used to apply the pre-load from 5 to 20 N m. Case1-B is considered as the intact case with the maximal applied torque 20 N m, then the damage cases, namely Case2-B, Casee3-B, and Moreover, in this paper, the monitoring of timber connection damage based on active sensing method was performed with artificial damages. The practical damage levels are described as follows: (1) for Group A, Case1-A is the initial health status with four screws all tightened, then we damage the integrity by loosening one, two, and three screws, respectively. In accordance with the different damage levels, the damage cases are called Case2-A, Case3-A, Case4-A, respectively. (2) for Group B, a torque wrench is used to apply the pre-load from 5 to 20 N m. Case1-B is considered as the intact case with the maximal applied torque 20 N m, then the damage cases, namely Case2-B, Casee3-B, and Case4-B, are assigned as 15, 10, and 5 N m, respectively; (3) for Group C, two timber specimens are fastened by two metal tooth plates that have short and sharp nails, and the damage is simulated by prying up metal tooth plates under different levels. Specifically, Case1-C is considered as the intact case with two tightened tooth plates, Case2-C, Case3-C, and Case4-C are the damage cases with the tooth plates loose slightly, moderately and severely, respectively. (4) for Group D, two timber specimens are connected and fixed by a decussation part with screws, and we mimic different damage cases by loosening screws. Case1-D is considered as the intact case with the tightened decussation, then the damage cases, labeled as Case2-D, Case3-D, and Case4-D are damaged by loosening one, two, and three screws of the connection decussation, respectively. In order to describe the intact and damage cases more clearly, damages of timber connections in four groups are summarized in Table 1.
loose one screw loose two screws loose three screws

Experimental Setup and Experimental Procedure
The experimental apparatus includes a data acquisition system (NI USB-6363), a signal power amplifier (Trek model 2100 HF), a laptop and timber connections of Group A, B, C, and D (as depicted in Figure 4). Case4-B, are assigned as 15, 10, and 5 N m, respectively; (3) for Group C, two timber specimens are fastened by two metal tooth plates that have short and sharp nails, and the damage is simulated by prying up metal tooth plates under different levels. Specifically, Case1-C is considered as the intact case with two tightened tooth plates, Case2-C, Case3-C, and Case4-C are the damage cases with the tooth plates loose slightly, moderately and severely, respectively. (4) for Group D, two timber specimens are connected and fixed by a decussation part with screws, and we mimic different damage cases by loosening screws. Case1-D is considered as the intact case with the tightened decussation, then the damage cases, labeled as Case2-D, Case3-D, and Case4-D are damaged by loosening one, two, and three screws of the connection decussation, respectively. In order to describe the intact and damage cases more clearly, damages of timber connections in four groups are summarized in Table 1. Table 1. Test cases of connection damage in Group A, B, C, and D.

Group Number Intact Case Damage Case
Group A Case1-A Case2-A Case3-A Case4-A four tightened screws loose one screw loose two screws loose three screws

Experimental Setup and Experimental Procedure
The experimental apparatus includes a data acquisition system (NI USB-6363), a signal power amplifier (Trek model 2100 HF), a laptop and timber connections of Group A, B, C, and D (as depicted in Figure 4).
The sampling frequency of the data acquisition system is 1 MHz. For each group, a swept sine wave signal (from 100 kHz to 300 kHz) with amplitude of 5 V was applied to excite PZT actuator, and the generated stress wave propagated from one timber to the other. The PZT sensor captured the response signal from the other timber. On the other hand, all tests in this manuscript were finished within two hours in the laboratory to minimize the influence of the environmental issues on the results, and thus the humidity and temperature change during the tests could be ignored. Moreover, we will conduct further investigations on the temperature effects on PZTs in the future work.  The sampling frequency of the data acquisition system is 1 MHz. For each group, a swept sine wave signal (from 100 kHz to 300 kHz) with amplitude of 5 V was applied to excite PZT actuator, and the generated stress wave propagated from one timber to the other. The PZT sensor captured the response signal from the other timber. On the other hand, all tests in this manuscript were finished within two hours in the laboratory to minimize the influence of the environmental issues on the results, and thus the humidity and temperature change during the tests could be ignored. Moreover, we will conduct further investigations on the temperature effects on PZTs in the future work.

Results and Discussions
For the four Groups A, B, C, and D, the time-domain signals received by PZT sensors are shown in Figures 5-8, respectively. The results demonstrate that the amplitude of signals decrease with the increase of the damage. In other words, the intact status (Case1 in every group) has the largest amplitude, and amplitudes under the damage status (i.e., Case2, Case3, and Case4 in every group) decrease sequentially. Moreover, the Group B, namely the bolt connection, shows the most obvious tendency among four different groups, and the next was the Group A, i.e., the screw connection. The change of Group C (the metal tooth plate connection) and Group D (the decussation connection) are not as obvious as the first two. It is worth noting that the wave shapes under each case are different from each other. This phenomenon can be explained through the structural stiffness changes under different cases. We applied linear swept sine waves to excite the PZT patch, in other words, the time-domain waves also present frequency characteristics. It is well known that structural damages can induce corresponding stiffness changes, thus the wave shape under each case is different. The experimental results reveal that the active sensing method has potential to achieve real-time monitoring of the damage of timber connections.

Results and Discussions
For the four Groups A, B, C, and D, the time-domain signals received by PZT sensors are shown in Figures 5-8, respectively. The results demonstrate that the amplitude of signals decrease with the increase of the damage. In other words, the intact status (Case1 in every group) has the largest amplitude, and amplitudes under the damage status (i.e., Case2, Case3, and Case4 in every group) decrease sequentially. Moreover, the Group B, namely the bolt connection, shows the most obvious tendency among four different groups, and the next was the Group A, i.e., the screw connection. The change of Group C (the metal tooth plate connection) and Group D (the decussation connection) are not as obvious as the first two. It is worth noting that the wave shapes under each case are different from each other. This phenomenon can be explained through the structural stiffness changes under different cases. We applied linear swept sine waves to excite the PZT patch, in other words, the timedomain waves also present frequency characteristics. It is well known that structural damages can induce corresponding stiffness changes, thus the wave shape under each case is different. The experimental results reveal that the active sensing method has potential to achieve real-time monitoring of the damage of timber connections.  In order to quantify the received stress waves, the energy of the received signal was computed through the wavelet packet energy method, and the results are shown in Figure 9. It can be found that the energy of received signals in four groups decreased with the increase of damages of timber connections. Additionally, there are some differences among the four groups. For instance, the energy in Group A and B decreased dramatically the more severe the damages to the timber connections were. However, the decreasing trend in Group C was not obvious, and energy attenuation in Group D tended to saturate. These differences may be attributed to the different types of timber connections. It is well known that the PZT-enabled active sensing method depends on stress wave propagation, particularly, the energy attenuation when stress wave propagates through the interface. Generally, a In order to quantify the received stress waves, the energy of the received signal was computed through the wavelet packet energy method, and the results are shown in Figure 9. It can be found that the energy of received signals in four groups decreased with the increase of damages of timber connections. Additionally, there are some differences among the four groups. For instance, the energy in Group A and B decreased dramatically the more severe the damages to the timber connections were. However, the decreasing trend in Group C was not obvious, and energy attenuation in Group D tended to saturate. These differences may be attributed to the different types of timber connections. It is well known that the PZT-enabled active sensing method depends on stress wave propagation, particularly, the energy attenuation when stress wave propagates through the interface. Generally, a In order to quantify the received stress waves, the energy of the received signal was computed through the wavelet packet energy method, and the results are shown in Figure 9. It can be found that the energy of received signals in four groups decreased with the increase of damages of timber connections. Additionally, there are some differences among the four groups. For instance, the energy in Group A and B decreased dramatically the more severe the damages to the timber connections were. However, the decreasing trend in Group C was not obvious, and energy attenuation in Group D tended to saturate. These differences may be attributed to the different types of timber connections. It is well known that the PZT-enabled active sensing method depends on stress wave propagation, particularly, the energy attenuation when stress wave propagates through the interface. Generally, Sensors 2019, 19, 2486 9 of 12 a larger interface area means that more stress wave energy can be transferred and received by the PZT patch. Thus, timber connections with bolts and screws, whose principle are similar, are more sensitive to active sensing, since preloads are proportional to interface areas in these two types of connections. On the other hand, the wave propagation paths of tooth plate connections and decussation joints are different. For example, the propagation paths of tooth plate connections during active sensing method are: PZT1→one timber specimen→teeth→plate→teeth→the other specimen→PZT2. Similarly, in the decussation joints, the stress wave will propagate across the paths: PZT1→one timber specimen→screws→decussation→screws→the other specimen→PZT2. We found that the loosening damages in these two connections (i.e., the debonding of teeth and screw looseness) have a smaller impact on wave propagation, since they affect the interface slightly. Therefore, the PZT-enabled active sensing has unsatisfactory performance in detecting the tooth plate connections and the decussation joints. larger interface area means that more stress wave energy can be transferred and received by the PZT patch. Thus, timber connections with bolts and screws, whose principle are similar, are more sensitive to active sensing, since preloads are proportional to interface areas in these two types of connections. On the other hand, the wave propagation paths of tooth plate connections and decussation joints are different. For example, the propagation paths of tooth plate connections during active sensing method are: PZT1→one timber specimen→teeth→plate→teeth→the other specimen→PZT2. Similarly, in the decussation joints, the stress wave will propagate across the paths: PZT1 → one timber specimen→screws→decussation→screws→the other specimen→PZT2. We found that the loosening damages in these two connections (i.e., the debonding of teeth and screw looseness) have a smaller impact on wave propagation, since they affect the interface slightly. Therefore, the PZT-enabled active sensing has unsatisfactory performance in detecting the tooth plate connections and the decussation joints. Overall, the experimental results demonstrated that the active sensing method has a great potential to monitor damages of timber connections. However, there are still many challenges in applying this technology for practical use. Firstly, the investigation in this paper considered the only four common connection types; other joints such as epoxy-bonded connection and mortise-tenon connection are required to be studied in future study. Additionally, some factors such as boundary Overall, the experimental results demonstrated that the active sensing method has a great potential to monitor damages of timber connections. However, there are still many challenges in applying this technology for practical use. Firstly, the investigation in this paper considered the only four common connection types; other joints such as epoxy-bonded connection and mortise-tenon connection are required to be studied in future study. Additionally, some factors such as boundary conditions, the effect of wood type, and geometry size of the specimen were not considered. Future investigations will be conducted to address these issues.

Conclusions
Timber structures are widely used in industries, due to their merits such as being low-cost. The damage of timber connections directly affects the safety and reliability of timber structures. Thus, this paper aims to apply a stress wave-based active sensing approach to monitor damages of common timber connections. Four common types of timber connections (screwed connection, bolted connection, metal tooth plate, and decussation) were considered in this paper. Surface-bonded PZT patches were used in the active approach. The experimental results showed that the amplitude of received signals decreased with the increase of damage extent in timber connections, and the energy of the received signals, which was computed by the wavelet packet energy method, could be used to quantify the damage of the timber connections. Moreover, it is worth noting that the energy decreased dramatically with more severe damages when we detected screw connections and bolted connections in timber structures. However, the decreasing trend was not obvious during the detection of tooth plate connection, and energy attenuation for monitoring the decussation connection tended to saturate. The difference may be attributed to the different types of timber connections, and we will propose a more advanced method to solve this issue. Overall, we demonstrated that the active sensing method based on PZT transducers is effective and sensitive to monitor the damages of timber connections in real time. Therefore, regarding the widely-used wooden houses in America and Japan, we can apply this method to detect connection damages to ensure structural integrity and protect property.
Recently, vision-based structural health monitoring (SHM) methods, particularly for bolt-loosening detection [44,45], have been reported. However, there are several demerits of the current visual-based detection of bolt looseness, for instance, a loose bolt that has exactly one circle rotation cannot be detected. Therefore, our future work will focus on the improvement of the current vision-based methods for bolt loosening detection, to overcome existing problems.