Impact localization by a multi-radio sink–based wireless sensor network for large-scale structures

Abstract

With the rapid development of wireless sensor network technology, more and more researchers have been interested in taking advantages of wireless sensor network to reduce weight and cost and to solve the installing problem of the wired structural health monitoring systems. A number of wireless-based structural health monitoring methods have been developed over the years to ensure the safety of large-scale structures. However, little research has been reported on wireless impact monitoring of large-scale composite structures due to the limitations of ordinary monitoring methods and wireless sensor networks. In this article, a wireless multi-radio sink which can access multiple communication channels is developed and adopted to build an impact monitoring wireless sensor network. Besides, a corresponding network architecture with an energy-weighted factor–based localization method adopted is presented to enable impact localization within the whole monitoring scope of the network. To verify the performance of the impact monitoring wireless sensor network, the experiments are performed on complex aircraft composite structures and the experimental results prove the effectiveness of the proposed wireless sensor network.

Keywords

impact localization large-scale composite structures multi-radio sink structural health monitoring wireless sensor network

Introduction

As one of the most important technologies in the 21st century, wireless sensor network (WSN) has been widely researched and applied throughout the world (Wang et al., 2012; Xu et al., 2004). A WSN usually consists of a number of self-organized wireless sensor nodes working together to monitor a region to obtain data about the environment and accomplish specific tasks. Benefiting from the advantages of enabling dense in situ sensing and simplifying deployment of instrumentation of WSN, structural health monitoring (SHM) becomes a typical area among the many possible applications of WSN (Chen et al., 2015; Dürager et al., 2013; Sundaram et al., 2013; Wang et al., 2007; Zhou and Yi, 2013). Compared to traditional wire-based SHM systems, wireless SHM systems offer a number of benefits, including lighter additional weight due to the wires, more flexible sensor placement strategy on the structure, smaller influence of the measurement data due to cables’ capacitance, and fewer installation costs for the wires (Dos Santos et al., 2014). In the last decade, a numerous number of researches have been reported to take advantages of WSN for SHM applications (Liu and Yuan, 2008; Mohammad and Huang, 2011; Nguyen et al., 2014; Wu et al., 2009; Zhou et al., 2015).

WSN-based SHM systems have emerged as a promising alternative solution for rapid, accurate, and low-cost monitoring for large-scale structures. However, little wireless-related SHM research has been presented on impact monitoring of large-scale aircraft composite structures, which is becoming increasingly urgent with the increasing prevalence of composite materials in aircraft structures (Bond et al., 2014), since impact events may cause internal damages and lead to serious property degradation of the composite structures (Diamanti and Soutis, 2015; Katnam et al., 2013). This kind of situation may result from the following characteristics of impact monitoring: (1) impact is an instant event and needs to be monitored online (Gibson, 2010; Soutis, 2005); (2) in order to realize impact localization, a large amount of analog sensing signal should be acquired and transmitted to a monitoring center for locating process, and for large-scale structures, a sensor network with numbers of sensors is needed; and (3) ordinary impact monitoring systems usually pursue high-precision impact localization resulting in high hardware and software requests for SHM systems (Chen et al., 2012; Markmiller and Chang, 2009; Staszewski et al., 2009). Besides, with the requirements of low cost and low power, the design of WSN has mainly been for low bandwidth and non/non-urgent real-time applications, which further limits the development of wireless impact monitoring.

Aiming at conducting wireless impact monitoring for large-scale structures, a piezoelectric transducer (PZT)–based digital wireless impact monitor is developed by Liu et al. (2012). The monitor turns the outputs of PZT sensors into digital sequences instead of directly processing the original analog signals, which greatly simplifies the system, including its size and power consumption. Then the impact occurring sub-region can be located by analyzing specific feature parameters extracted from the digital sequences. In this way, further impact damage inspection can be instructed to the detected impact sub-regions instead of the whole structure by ordinary non-destructive testing (NDT) methods, which can effectively reduce the maintenance cost of aircraft structures. Based on the digital wireless impact monitor, a preliminary WSN is built for impact monitoring of aircraft composite structures (Yuan et al., 2014). By adopting the monitor, a large portion of signal processing and computation can be done locally, and simultaneously, the amount of information that needs to be transmitted over the network is largely reduced. However, although the proposed WSN has shown its potential, there still exist some issues that need to be addressed, such as the reliability and transmission efficiency of the network when large-scale impact monitoring is needed and the ability of enabling impact localization within the whole monitoring scope of the network.

In recent years, multi-channel communication is proposed as an efficient method to alleviate the effects of interference and maximize concurrency and parallel transmission capacity of WSN (Incel, 2011). Wu et al. (2008) presented a tree-based multi-channel scheme to improve communication performance of WSN, which was verified to be able to significantly enhance the reliability of the network. Incel et al. (2011) designed a multi-channel media access control (MAC) protocol with the objective of maximizing the throughput of WSN by coordinating transmission over multiple frequency channels. The simulation results show that the throughput of the WSN with multi-channel MAC protocol is far greater than the single-channel one. Multi-channel communication has been proved to be an efficient method to improve the data transmission ability and the reliability of the WSN by enabling parallel transmission among different frequency channels.

In this article, a multi-radio sink which supports up to eight frequency channels provided by IEEE 802.15.4 protocol is developed to construct an impact monitoring WSN for large-scale aircraft composite structures. With the multi-radio sink, the network can accommodate multiple clusters with different channels, enlarge the monitoring scope, and improve the transmission efficiency through parallel transmission at the same time. Besides, an energy-weighted factor (EWF)–based localization method which enables impact localization within the whole monitoring scope of the network is proposed in this article. According to the methods mentioned above, an online impact monitoring WSN is built and verified on complex aircraft composite structures.

Development of the multi-radio sink

In this section, aiming at constructing an impact monitoring WSN for large-scale structures, a multi-radio sink with eight radio frequency (RF) modules is developed.

Hardware design of the multi-radio sink

The hardware structure of the multi-radio sink is shown in Figure 1. Different from ordinary sink nodes, this one consists of a field programmable gate array (FPGA)–based processor module, a communication module, a power module, a universal serial bus (USB) module, and an interface module.

Figure 1.

Hardware structure of the multi-radio sink.

It is the high-speed parallel working ability of FPGA chip that is adopted to control several RF modules to receive data from different communication channels simultaneously. The FPGA chip should be chosen according to the following aspects: (1) having sufficient number of available input/output (I/O) pins, (2) supporting high enough crystal frequency for working, and (3) small size. The FPGA chip EP3C16Q240C8N from ALTERA Company is chosen here. This chip contains 240 pins which are enough to connect eight RF modules.

The CC2420 chip-based RF module from TI is chosen to realize wireless communication in the sink. The communication module contains eight CC2420 RF modules. To control one CC2420 RF module, FPGA needs to configure four available I/O pins to receive working condition of it and another four available I/O pins as analog serial peripheral interface (SPI) to exchange data, send commands, and access internal registers and storage of CC2420.

The USB chip FX2 CY7C68013 from CYPRESS and an electrically erasable programmable read-only memory (E2PROM) are chosen to build the USB module. The FPGA module controls the first-in-first-out (FIFO) buffer in CY7C68013 chip to read and upload data.

Three kinds of power management chip: AMS1117-1.2, AMS1117-2.5, and AMS1117-3.3 are adopted in the power module to meet the demands of different modules of the sink. And the I/O module is responsible for restarting and programming.

Figure 2 shows the developed eight-radio sink with a size of 180 × 115 mm². Since the sink is used near the monitoring center in the WSN application, no strict weight and energy consumption limitation exists for the service of this kind of node; the size of the developed sink is acceptable.

Figure 2.

Developed multi-radio sink.

Software design of the multi-radio sink

Aiming at managing the encoding, decoding, and error detection/correction when receiving data from multiple RF radios, a double-layer parallel-to-serial data caching mechanism and a redundancy processing unit–based high-speed data transfer mechanism are designed. Based on these mechanisms, the software framework is built with wireless receiving layer (WRL), data processing buffer layer (DPBL), logic control layer (LCL), and data communication layer (DCL) from the bottom to the top layer, as shown in Figure 3.

Figure 3.

Software structure of the multi-radio sink.

The double-layer parallel-to-serial data caching mechanism is presented to coordinate the parallel data receiving from multiple radios and the serial data uploading through the USB port. This mechanism first builds multiple FIFOs as data caching units in DPBL for parallel data receiving. Then, in the LCL, a main FIFO unit is built to gather these data for uploading. The high-speed data transfer mechanism is designed by building two kinds of redundancy processing units, named as RF module management state machine and data packet processing state machine in DPBL for each communication channel. When the sink receives data from different channels, the software distributes the data to their corresponding redundancy processing units. These processing units process the data in parallel. By this mechanism, less time is consumed comparing to ordinary queue processing mechanism. The data processing speed can be improved greatly.

In the working process of the multi-radio sink, WRL is able to accommodate multiple RF modules, which means different communication channels. The multi-radio sink downloads data from several different channels through WRL at the same time.

There are two units in DPBL: data processing unit and data storage unit. The former is responsible for receiving and processing data packets from WRL, and the latter stores the data into its FIFO. Figure 4 gives out the data reception and storage process of every channel. When the sink begins to work, multiple RF modules receive wireless data packets from their own channels in parallel. Once one of the RF modules receives a packet, the data processing unit reads it from the receive FIFO (RXFIFO) in DPBL connecting to this module. The packet is stored in the FIFO of data storage unit after it is processed by the high-level data link control (HDLC) and added the cyclic redundancy check (CRC) code. Meanwhile, the other RF modules and their corresponding data processing units and data storage units work independently.

Figure 4.

Data reception and storage of each channel in DPBL.

LCL contains FIFO scanning state machine, FIFO management state machine, and dataflow generation module. The FIFO scanning state machine scans FIFOs of different channels constantly and circularly and drives the FIFO management state machine to read data when some FIFOs are not empty. Then these data will be packaged and wait to be uploaded to the dataflow generation module. The specific workflow of LCL is shown in Figure 5. When data storage is found in an FIFO, the action of data processing unit connected with this FIFO is suspended, and the data are read out immediately. Then the data processing unit is activated again to enable data reading. This workflow works in a serial manner. After finishing the data reading of a channel, the FIFO scanning state machine instantly starts to scan the next channel.

Figure 5.

Specific workflow of the LCL.

The USB communication management state machine in the DCL is developed to provide data transmission channels and manage the software interface of the USB chip. This state machine is able to control the dataflow generation module to transfer data into the cache FIFO in the USB chip with a high speed. Then it uses bulk transfer mode to upload data to the monitoring center.

Design of the multi-radio sink–based impact monitoring WSN

For real aircraft, usually large-scale and multiple parts of the structures need to be monitored, such as wing, vertical fin, and fuselage. Therefore, it is of critical importance to organize a number of digital wireless impact monitors to form a network to fulfill the monitoring. To realize this and keep the efficiency and reliability of the network at the same time, a multi-radio sink–based impact monitoring WSN is designed in this section.

Digital wireless impact monitor

Different from traditional SHM systems, the digital wireless impact monitor makes a compromise to satisfy the requirements of online and on-board impact monitoring (Liu et al., 2012; Yuan et al., 2012). The monitor focuses on achieving approximate localization results rather than high-precision ones, and the complexity of the monitor is greatly simplified in return. The circuit board and the well-packaged digital wireless impact monitor are illustrated in Figure 6. The monitor has the following characters: (1) small size (8 × 6 × 3 cm³), light weight (120 g), and low-power request (100 mW); (2) access up to 24 PZT sensors; (3) online response and rapid impact localization and storage; (4) wireless transmission and networking monitoring; (5) dual power supply (airborne power supply or rechargeable battery supply); and (6) strong electro magnetic compatibility (EMC).

Figure 6.

Digital wireless impact monitor.

In addition, a preliminary localization algorithm is designed for the monitor to locate impact occurring sub-region. As shown in Figure 7(a), six PZT sensors are placed on a composite structure, forming two impact monitoring sub-regions: sub-region 1 and sub-region 2. Figure 7(b) gives the typical response signals of the PZT sensors caused by the impact occurring on sub-region 1, denoted as V₁(t) to V₆(t). By comparators designed in the monitor, all these analog signals are taken in comparison with a predefined threshold. The comparator will output a high digital level “1” if the voltage value of the signal is larger than the threshold. Otherwise it outputs a low digital level “0.”Figure 7(c) shows the achieved six digital sequences. According to the arrival time of the first rising edge in each digital sequence, the algorithm recognizes the first three response sensors (namely, PZT1, PZT4, and PZT3 in this case) to locate impact. Hence, sub-region 1 surrounded by PZT1, PZT2, PZT3, and PZT4 is determined as the impact occurring sub-region.

Figure 7.

Explanation of the sub-region and digital sequences: (a) schematic layout of sub-regions, (b) impact response signals, and (c) digital sequences.

Architecture of the impact monitoring WSN

Figure 8 describes the architecture of the proposed impact monitoring WSN. As shown in this figure, the network contains impact monitoring layer, network management layer, and monitoring center layer. There are four kinds of devices adopted in the network, namely, monitor, management node, multi-radio sink, and monitoring center. In order to avoid communication confliction, the monitors are arranged into several clusters, and different clusters work in different channels provided by IEEE 802.15.4 protocol. Each cluster contains one management node and m monitors. A commercial TelosB node with low-power design is taken as the management node. A multi-radio sink is adopted in the designing to receive data from multiple clusters simultaneously and deliver them to the monitoring center. This can greatly improve the data transmission efficiency.

Figure 8.

Architecture of the multi-radio sink–based impact monitoring WSN.

The impact monitoring layer adopts numbers of digital wireless impact monitors to detect and record impact events. One monitor takes charge of monitoring one part of the structure independently. The main functions of the monitor include the following: (1) detecting impact and turning the responding signals of PZT sensors into digital sequences, (2) extracting specific feature parameters for impact localization, and (3) transmitting impact-related information to the corresponding management node.

The network management layer contains multiple management nodes, which are responsible for the following aspects in the network: (1) managing the operation of monitors, (2) keeping time synchronization of the network, (3) receiving impact records produced by the monitors, (4) locating impact occurring sub-regions based on the records and storing them temporarily, and (5) uploading impact localization results to the monitoring center layer during the fixed interval requested by engineering inspection.

The monitoring center with a multi-radio sink connected is designed as the monitoring center layer. As the control core of the presented WSN, the monitoring center layer is mainly responsible for (1) receiving network data from multiple clusters in parallel, (2) managing history data of the network, and (3) instructing further NDT-based damage inspection.

A LabVIEW platform-based integrated software is developed as the user interface in the monitoring center, as shown in Figure 9. The software stores information about all the structures monitored in the network, including the corresponding relationship between the structures and clusters and the distribution of sub-regions of each structure. After receiving data packages from the multi-radio sink, the software will parse these packages and obtain impact records. By confirming the structure where impact occurs according to the cluster number, the software updates the impact history record of the WSN. When the impact times of certain sub-region exceed a preset value, NDT-based damage inspection will be instructed to this sub-region to search impact damage; the maintenance time and cost of aircraft composite structures can be greatly reduced. Since this process is carried out on ground, there is no need to install the multi-radio sink and the monitoring center on aircraft. This is very important since the aerospace application has a very strict limitation on the weight caused by the monitoring system.

Figure 9.

User interface of the monitoring center.

Time synchronization mechanism of the WSN

The digital wireless impact monitor adopted in the WSN uses a crystal-driven timer to record impact occurring time when triggered by an impact. If multiple monitors are triggered, the management node needs to make sure that these monitors are triggered by the same impact through comparing their impact occurring time. However, it is hard to maintain a consistent timing of all monitors after working for a period of time because of the crystal drift. So the management node needs to conduct time synchronization to ensure that the timers of the monitors in a cluster have enough consistency.

Figure 10 gives out the working process of the time synchronization mechanism. The management node sends out a synchronization packet with its own clock time stamp T₀₁. Each monitor working in this cluster receives this packet and records its own clock time stamp T₁₁. When next synchronization action is performed, similarly, the time stamp of the management node T₀₂ and the corresponding time stamp of the monitor T₁₂ are recorded. The clock time difference between the management node T₀ and the monitor T₁ can then be calculated by equation (1)

T_{0} - T_{1} = (T_{02} - T_{01}) - (T_{12} - T_{11})

(1)

Figure 10.

Time synchronization process in a cluster.

Each monitor corrects its timer according to equation (1). This synchronization mechanism can ensure that the impact occurring time recorded by different monitors has a little difference, and enable the management node to judge whether the impact records it receives belong to the same impact. The experiments conducted on a 2-mm composite structure show that the responding signals of the PZT sensors will attenuate below 3 V, which is unable to trigger the monitor to detect impact, after propagating a distance about 1000 mm. Given an approximate group velocity of 800 m/s of the signals, this propagation takes 1.25 ms, which is the maximal time difference of the adjacent monitors triggered by the same impact. Considering that the frequency stability of the crystal adopted in the monitor is 20 ppm, the time synchronization interval of the network is chosen to be 20 s to achieve a synchronization accuracy of 400 µs. General WSNs usually pursue synchronous data acquisition, resulting in the demand of high synchronization accuracy. The proposed impact monitoring WSN just needs to distinguish the recorded impact occurring time of different monitors, which is not a high requirement. The synchronization accuracy of 400 µs is available in this work.

EWF-based localization method

In order to enable impact localization of the proposed impact monitoring WSN, a EWF-based localization method and its implementation in the network are presented in this section.

Basic principle of the method

A number of digital wireless impact monitors should be organized to form a monitoring network to fulfill the monitoring of large-scale structures. When the monitoring regions of different monitors are separated far away from each other, these monitors can work individually. However, if their monitoring regions are close, the impact occurring in one region may cause the nearby monitors to respond. In this case, according to the preliminary localization algorithm designed for the monitor, all the triggered monitors will give out their own localization results, in which only one is correct. Hence, a new method which is not limited within one single monitor and has the ability to unite multiple monitors to locate impact within the whole monitoring scope of the network is needed. This work puts forward a EWF-based localization method, which is capable of locating all sub-regions under monitoring accurately. Compared with the preliminary localization algorithm, this method is more suitable for impact networking monitoring.

This method defines a feature parameter called EWF to perform impact localization. In order to obtain EWF, two parameters extracted from the digital sequences are adopted. They are the duration of the rise (DR) and index of the first rising edge (IFRE), as shown in Figure 11. DR is the total length of the high digital level 1 of the digital sequence and its unit is milliseconds. It represents the energy a PZT sensor obtains from the impact. Theoretically, the closer the distance between a sensor and the impact, the bigger its DR. IFRE describes the order of the arrival time of the first rising edge among all the sensors which connect to the same monitor. It characterizes the order of the triggering time of a sensor compared with other sensors. The sensor nearest to the impact is supposed to be first triggered and has the minimal IFRE.

Figure 11.

Illustration of the DR and IFRE.

In order to decrease the influence of the complexity of aircraft structures and the environmental noises and achieve a higher estimation reliability, EWF is defined by combining DR and IFRE, as shown in equation (2). EWF is capable of describing the impact influence on every PZT sensor within the monitoring scope of the whole network instead of a signal monitor. Based on the truth that PZT sensors in the impact occurring sub-region are the closest to the impact location, the EWF values of these sensors should be the maximal ones among all the responding sensors

EWF = \frac{DR}{IFRE}

(2)

In real applications, the placement strategy of the PZT sensors is known. Therefore, the number of impact monitoring sub-regions and the four sensors of each sub-region are also known. By calculating the sum of the four sensors’ EWF values of each sub-region, the impact influence on every sub-region which is expressed as EWF^s can be evaluated. Assuming there are M impact monitoring sub-regions, the EWF^s of each sub-region can be calculated by equation (3), in which EWF_i represents the EWF value of the i^th PZT of the sub-region.

{EWF}^{s} = \sum_{i = 1}^{4} {EWF}_{i}, s = 1, \dots, M

(3)

Considering that the four PZT sensors of the impact occurring sub-region should have much larger EWF values than sensors in other sub-regions, the one with the maximum EWF^s is considered as the impact occurring sub-region. In this way, the impact influence on all the sub-regions under monitoring can be uniformly compared after impact happening, and only one sub-region will be given as the localization result.

Implementation in the impact monitoring WSN

Taking a cluster of the proposed impact monitoring WSN as an example, Figure 12 shows the basic working process of the WSN when an impact occurs. The monitor which is triggered records the impact occurring time first. Then it turns the output of the PZT sensors it connects into digital sequences. After extracting the IFRE and the corresponding DR of every digital sequence, the triggered monitor calculates the EWF values of all the digital sequences and picks out the maximal four. Along with their PZT sensor numbers, the monitor number, and the impact occurring time, the maximal four EWF values will be packaged as an impact record and transmitted to the management node. Figure 13 gives out the data format of the package, which only has 17 bytes.

Figure 12.

Implementation of the energy-weighted factor–based localization method.

Figure 13.

Data format of the impact record.

Assuming there are n monitors that are triggered, the management node will receive n packages. Then it picks out the maximal four EWF values again from the received $4 \times n$ and set the remaining to 0. After calculating the EWF^s value of every sub-region according to equation (3), the management node determines that the sub-region with the maximal EWF^s is the impact occurring sub-region. Figure 14 shows the localization result generated by the management node, which only contains the cluster number and the sub-region number. The localization result will be stored in the management node temporarily and then be uploaded to the multi-radio sink at a fixed interval which is specified by the application request.

Figure 14.

Data format of the localization result.

Performance evaluation

In order to verify the performance of the methods proposed in this work, a multi-radio sink–based impact monitoring WSN is built and performed on a composite wing box structure.

Evaluation setup

The composite wing box structure and the arrangement of the WSN are shown in Figure 15. The area of the wing box structure is 1000 × 1800 mm². There are totally six T-shape stiffeners with a distance of 130 mm between each other and five lines of bolt holes vertical to the stiffeners with an interval of 280 mm. The network contains two clusters, two TelosB nodes which are used as management node 1 and management node 2, and four digital wireless impact monitors, namely, monitor 1 to monitor 4. Cluster 1 includes management node 1, monitor 1, and monitor 2, working in communication channel 11 provided by IEEE 802.15.4 protocol. Cluster 2 contains management node 2, monitor 3, and monitor 4, working in channel 13. Besides, considering the laying cable of the PZT sensors, a special piezoelectric transducer layer (PSL) (Qiu et al., 2012) is adopted to reduce the wire burden of PZTs. The connection cables of PZTs are replaced by the circuits printed in the flexible interlayer to reduce the cable weight. A total of 16 PSLs which contain 48 PZT sensors numbered from 1 to 48 are placed on the inner surface of the composite wing box structure. PZTs 1–12 are connected to monitor 1 and PZTs 13–24 are connected to monitor 2, forming 12 sub-regions denoted as sub-region 1 to sub-region 12; the remaining 24 sensors have the similar placement. The size of every monitoring sub-region is 150 × 150 mm². A multi-radio sink is adopted to receive data from cluster 1 and cluster 2. By the software shown in Figure 9, the monitoring center can give out the impact localization results of the entire WSN and further inspection recommendation.

Figure 15.

Evaluation setup of the composite wing box structure.

Evaluation of the EWF-based localization method

To explain and verify the working mechanism of the proposed EWF-based localization method, an impact is conducted on sub-region 1 of the structure. Since the monitoring regions of monitor 1 and monitor 2 of cluster 1 are close, both the monitors are triggered to respond.

As a demonstration, DRs, IFREs, and the corresponding EWF values of all the 24 PZT sensors are listed in Table 1. The four sensors with the maximal EWF values of monitor 1 are PZTs 5, 1, 2, and 4, which are all in sub-region 1 and have the values of 2.92, 1.54, 0.83, and 0.73 ms. The four sensors of monitor 2 are PZTs 14, 13, 16, and 19, and their EWF values are 0.89, 0.57, 0.41, and 0.17 ms. Hence, after receiving data packets transmitted by the two monitors, management node 1 reserves the four EWF values of PZTs 5, 1, and 2 of monitor 1 and PZT 14 of monitor 2 and sets all the others to 0, since the four PZT sensors have the top four EWF values among all the 24 sensors. According to equation (3), the EWF^s values of the 12 sub-regions are calculated and shown in Figure 16. Sub-region 1 can be considered as the impact occurring sub-region because of the maximal value. The feasibility of the method is proved.

Table 1.

Feature parameters of the 24 PZT sensors.

PZT	DR (ms)	IFRE	EWF (ms)	PZT	DR (ms)	IFRE	EWF (ms)
1	3.08	2	1.54	13	1.13	2	0.57
2	2.49	3	0.83	14	0.89	1	0.89
3	2.37	6	0.40	15	0.60	5	0.12
4	2.93	4	0.73	16	1.22	3	0.41
5	2.92	1	2.92	17	0.76	6	0.13
6	2.49	9	0.28	18	0.78	9	0.09
7	2.36	5	0.47	19	0.66	4	0.17
8	2.04	7	0.29	20	0.48	10	0.05
9	1.99	11	0.18	21	0.47	8	0.06
10	1.46	12	0.12	22	0.82	7	0.12
11	1.85	8	0.23	23	0.61	11	0.06
12	2.08	10	0.21	24	0.52	12	0.04

PZT: piezoelectric transducer; DR: duration of the rise; IFRE: index of the first rising edge; EWF: energy-weighted factor.

Figure 16.

EWF^s values of the 12 sub-regions.

In addition, another validation is performed to further examine the reliability of the proposed method. Thirteen positions uniformly distributed in sub-region 1 are chosen to be applied impacts to verify whether the method can accurately give the localization results. The distribution of the 13 impact positions is shown in Figure 17. A total of 65 impacts are averagely performed on these positions; only position 6 has a false localization. Overall, the proposed EWF-based localization method has an acceptable reliability.

Figure 17.

Distribution of impact positions in sub-region 1.

Evaluation of the multi-radio sink–based WSN

Since the two monitors in each of the two clusters adopted in the network are all close, they both will be triggered when an impact occurs within the monitoring scope of the cluster that they belong to. In this validation, 12 sub-regions are randomly selected and a serial of 120 impacts are averagely applied on them. The statistical results are shown in Table 2; almost all the impact events applied on the structure are correctly recorded by the network and the accuracy turns out to be approximately 97%. Hence, the performance of the proposed impact monitoring WSN is well verified.

Table 2.

Evaluation results of the network.

Sub-region no.	1	2	4	5	9	11	14	17	18	19	22	24
Actual impact times	10	10	10	10	10	10	10	10	10	10	10	10
Correct localization times	10	9	10	10	10	10	9	9	10	10	9	10

Conclusion

Based on the multi-radio sink, this article puts forward an impact monitoring WSN design method to locate impact online for large-scale aircraft composite structures. Besides, a EWF-based localization method is also introduced to enable impact localization of the network. To verify the localization method and the impact monitoring WSN, validation experiments are implemented on a composite wing box structure. The evaluation results show an acceptable localization accuracy of 97% for engineering applications.

Footnotes

Declaration of Conflicting Interest

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the National Science Fund for Distinguished Young Scholars (grant no. 51225502), the 333 High-Level Personnel Training Project in Jiangsu Province (grant no. BRA2013190), the Qing Lan Project and the Prioritized Academic Program Development of Jiangsu Institutions of Higher Educations, the Foundation of Graduate Innovation Center in NUAA (grant no. kfjj130102), and the Fundamental Research Funds for the Central Universities.

References

Bond

Underwood

Adams

. (2014) Structural health monitoring–based methodologies for managing uncertainty in aircraft structural life assessment. Structural Health Monitoring 13(6): 621–628.

Chen

Yuan

(2012) Impact source identification in finite isotropic plates using a time-reversal method: experimental study. Smart Materials and Structures 21(10): 105025.

Chen

Casciati

Faravelli

(2015) In-situ validation of a wireless data acquisition system by monitoring a pedestrian bridge. Advances in Structural Engineering 18(1): 97–106.

Diamanti

Soutis

(2010) Structural health monitoring techniques for aircraft composite structures. Progress in Aerospace Sciences 46(8): 342–352.

Dos Santos

Pirmez

Lemos

. (2014) A localized algorithm for Structural Health Monitoring using wireless sensor networks. Information Fusion 15: 114–129.

Dürager

Heinzelmann

Riederer

(2013) A wireless sensor system for structural health monitoring with guided ultrasonic waves and piezoelectric transducers. Structure and Infrastructure Engineering 9(11): 1177–1186.

Gibson

(2010) A review of recent research on mechanics of multifunctional composite materials and structures. Composite Structures 92(12): 2793–2810.

Incel

(2011) A survey on multi-channel communication in wireless sensor networks. Computer Networks 55(13): 3081–3099.

Incel

van Hoesel

Jansen

. (2011) MC-LMAC: a multi-channel MAC protocol for wireless sensor networks. Ad Hoc Networks 9(1): 73–94.

10.

Katnam

Da Silva

LFM

Young

(2013) Bonded repair of composite aircraft structures: a review of scientific challenges and opportunities. Progress in Aerospace Sciences 61: 26–42.

11.

Liu

Yuan

(2008) Active damage localization for plate-like structures using wireless sensors and a distributed algorithm. Smart Materials and Structures 17(5): 055022.

12.

Liu

Yuan

Qiu

(2012) Development of a PZT-based wireless digital monitor for composite impact monitoring. Smart Materials and Structures 21(3): 035018.

13.

Markmiller

JFC

Chang

(2009) Sensor network optimization for a passive sensing impact detection technique. Structural Health Monitoring 9(1): 25–39.

14.

Mohammad

Huang

(2011) An antenna sensor for crack detection and monitoring. Advances in Structural Engineering 14(1): 47–53.

15.

Nguyen

Chan

Thambiratnam

(2014) Effects of wireless sensor network uncertainties on output-only modal analysis employing merged data of multiple tests. Advances in Structural Engineering 17(3): 319–330.

16.

Qiu

Yuan

Shi

. (2012) Design of piezoelectric transducer layer with electromagnetic shielding and high connection reliability. Smart Materials and Structures 21(7): 075032.

17.

Soutis

(2005) Carbon fiber reinforced plastics in aircraft construction. Materials Science and Engineering: A 412(1): 171–176.

18.

Staszewski

Mahzan

Traynor

(2009) Health monitoring of aerospace composite structures—active and passive approach. Composites Science and Technology 69(11): 1678–1685.

19.

Sundaram

Ravisankar

Senthil

. (2013) Wireless sensors for structural health monitoring and damage detection techniques. Current Science 104(11): 1496–1505.

20.

Wang

Yan

Tian

. (2012) Investigation of wireless sensor networks for structural health monitoring. Journal of Sensors 2012: 156329.

21.

Wang

Lynch

Law

(2007) A wireless structural health monitoring system with multithreaded sensing devices: design and validation. Structure and Infrastructure Engineering 3(2): 103–120.

22.

Yuan

Zhou

. (2009) Design and evaluation of a wireless sensor network based aircraft strength testing system. Sensors 9(6): 4195–4210.

23.

Stankovic

. (2008) Realistic and efficient multi-channel communications in wireless sensor networks. In: The 27th conference on computer communications INFOCOM’08, Phoenix, AZ, 13–18 April, pp. 1867–1875. New York: IEEE.

24.

Rangwala

Chintalapudi

. (2004) A wireless sensor network for structural monitoring. In: Proceedings of the 2nd international conference on embedded networked sensor systems, Baltimore, MD, pp. 13–24. Available at: http://graphics.stanford.edu/courses/cs321-05-fall/Readings/structural.pdf

25.

Yuan

Liu

Qiu

(2012) A miniaturized composite impact monitor and its evaluation research. Sensors and Actuators A: Physical 184: 182–192.

26.

Yuan

Mei

Qiu

. (2014) On a digital wireless impact-monitoring network for large-scale composite structures. Smart Materials and Structures 23(8): 085007.

27.

Zhou

(2013) Recent developments on wireless sensor networks technology for bridge health monitoring. Mathematical Problems in Engineering 2013: 947867.

28.

Zhou

Zhang

. (2015) Energy-aware wireless sensor placement in structural health monitoring using hybrid discrete firefly algorithm. Structural Control and Health Monitoring 22(4): 648–666.