CN105873111A - Soft and hard fault diagnosis and self restoration method suitable for health monitoring - Google Patents

Abstract

The invention discloses a method for diagnosing and self-repairing soft and hard faults suitable for health monitoring. The method forms a routing link. The first node of the link is a cluster head, and the second node is a backup cluster head. MA‑B is proposed .The method proposed by Abdo extracts two characteristic parameters for node fault detection and structural damage identification respectively from the original collected data, so as to avoid the interference of faulty nodes on structural damage. The self-healing wireless sensor network structure system plans to adopt a hybrid hierarchical topology. In addition to the original sensor nodes, it is considered to set up monitoring nodes in the network to monitor the network status. The method applies the cooperative communication mechanism to the multi-stream transmission problem of the wireless network to realize the maximum throughput of the network and ensure the stability and reliability of the network.

Description

A diagnosis and self-healing method for soft and hard faults suitable for health monitoring

technical field

The invention relates to a method for diagnosing and self-repairing soft and hard faults suitable for health monitoring, and belongs to the technical field of wireless sensor networks.

Background technique

The Structural Health Monitoring (SHM) system adopts the concept of intelligent material structure, uses the sensor/drive element network integrated in the structure to obtain information related to the health of the structure online in real time, and combines signals, information processing methods and The modeling method of material structure mechanics extracts characteristic parameters, identifies structural damage of materials (such as material cracks, holes, corrosion, etc.), and realizes self-diagnosis of structural health. The research on structural health monitoring is a cutting-edge research direction involving mechanics, machinery, communication, network and other scientific research fields. The research on structural health monitoring technology is involved in large-scale engineering structures such as aviation, bridges and engineering buildings. For example, with the support of USAF, the United States has carried out research on the application of structural health monitoring technology for F-18, F-22, JSF, DC-X2, X-33 and other aircraft. Hong Kong deployed 350 sensing channels on the Tsing Ma Bridge for bridge health monitoring.

Most of the traditional structural health monitoring is based on the collection of wired sensors. The wired monitoring systems and methods have problems such as many leads and large amount of information transmission, and the manpower and material resources required for maintenance are quite huge. In order to make up for the shortcomings, in recent years, domestic and foreign scholars have proposed a structural health monitoring system based on Wireless Sensor Networks (WSNs), which senses and collects relevant information of monitoring objects in real time through wireless sensor nodes deployed in the monitoring area. Collaborative processing and network delivery of information. WSNs has the advantages of rapid deployment, self-organization into a network, strong survivability and collaborative work ability, etc., and it is a hot field of research by scholars at home and abroad. The intelligent health monitoring system based on wireless sensor network has been paid more and more attention by domestic and foreign research institutions, and has achieved many application results in the fields of aviation, bridge, building and seismic structure monitoring.

When using wireless sensor networks to monitor and monitor large-scale engineering structures such as aviation, bridges, and engineering buildings, the system usually operates in a harsh environment, and its nodes will experience various failures due to exposure to the outside, directly causing errors in measured values. , and even cause the loss of certain functions of the WSN and even the paralysis of the entire network. In addition, in the specific field of health monitoring, the application of wireless sensor network has the following characteristics:

(1) The sensors are distributed in the key parts of the measured object. Once the nodes are deployed, they will not move, and there is no need to consider node mobility.

(2) Different from other application fields of WSN, the sensor nodes in structural health monitoring work at a fixed frequency, and the amount of raw data collected by each node is very large. Therefore, real-time and accurate acquisition of sensing signals is the key. The network needs to respond quickly in real time, and the communication is reliable, and the acquisition and forwarding can be completed in the shortest time.

(3) In order to complete health monitoring tasks, some key monitoring parts need to collect multiple sensing signals (such as Lamb wave, light, strain, displacement, acceleration, temperature, pressure, etc.), so WSN should have the function of multi-parameter collection.

(4) Damage identification in health monitoring must be completed through the cooperation of multiple sensor nodes around. Even if one of the nodes fails, the accuracy of damage identification will drop sharply.

The above characteristics show that in structural health monitoring, higher requirements are put forward for wireless sensor nodes and networks; that is, the system must be able to provide real-time, accurate, stable, and reliable data collection for health monitoring, and the occurrence of faults directly affects the monitoring results. . The requirement for multi-parameter acquisition of WSN nodes also increases the probability of failure. Therefore, in order to improve the reliability of structural health monitoring, timely and accurate diagnosis and repair of node and network faults are very necessary.

At present, there are three main categories of traditional sensor fault detection and diagnosis methods, physical redundancy method, model-based method and neural network-based method. Although the fault diagnosis of nodes in wireless sensor networks is limited by wireless communication and energy efficiency, the detection and diagnosis methods are slightly different, but the traditional sensor diagnosis technology can still be used for reference. Node faults in wireless sensor networks are divided into two categories: hard faults and soft faults. Hard failure refers to the failure of a certain module of the sensor node, so that it cannot communicate with other nodes (for example, due to the failure of the node communication module, the exhaustion of node energy, and the movement of the node out of the communication range of the entire network, etc.); Soft failure means that although the sensor node fails, it can still work and communicate with other nodes (the software and hardware of the communication module are normal and have routing information), but the data perceived or transmitted by the node is incorrect, or the sensor node occurs instantaneously. Communication failure. In the existing node fault diagnosis method of wireless sensor network, the literature proposes a node fault identification algorithm based on Bayesian network, which can distinguish different types of faults in the network. J.L.Gao et al. proposed a faulty node identification algorithm to judge the status of faulty nodes by comparing with the mean value of adjacent nodes; X.Luo et al. proposed an energy-efficient fault-tolerant technology for fault detection in wireless sensor networks ; Aiming at the emergence of faults in the network, the literature proposes an autonomous aggregation node confirmation algorithm to reduce the impact of node faults on the overall network performance. J.R.Chen et al. proposed a DFD algorithm for diagnosing node status by using adjacent nodes to test each other. The literature proposes a diagnostic method for attribute violation type errors in wireless sensor network programs.

In terms of structural health monitoring-oriented wireless sensor network fault diagnosis, domestic and foreign scholars have just started research in this area, and there are few literatures in this area. For vibration-based structural health monitoring, the literature gives a mathematical model of sensor node failure and structural damage, and discards the faulty node from the network to avoid the interference of node failure; X.Liu et al. proposed a WSN-based structural health monitoring Fault tolerance technology to realize node fault detection; the literature proposes the self-healing research of the wireless sensor system based on bridge structure health monitoring, and proposes that when the relay node fails, the node is discarded, and the network self-healing is realized in the form of a backup relay node Features.

To sum up, ① most of the existing research is aimed at the diagnosis of soft faults of nodes, but neither the treatment of hard faults nor the repair of faults is involved. Once a hard failure occurs, since the hardware of wireless sensor nodes is fixed, the nodes cannot be reconfigured unless the nodes are removed from the sensor network and the hardware and software systems of the nodes are redesigned. ②Furthermore, existing soft fault diagnosis algorithms are often based on data exchange between nodes, but in structural health monitoring, due to the large amount of data collected and exchanged between nodes, they cannot be applied. The idea of replacement is to perform fault diagnosis after data compression of nodes. In the above two situations, once the sensor network is deployed, the configuration can no longer be changed, and the existing nodes cannot perform self-repair and self-reconfiguration when they fail, so the failed nodes can only be discarded. Therefore, it is necessary to propose effective software and hardware fault diagnosis methods and self-repair mechanisms for the faults of sensor networks in this field.

(1) Research status of hard faults in wireless sensor networks in SHM

In the actual biological world, self-reconfiguration and self-repair functions are ubiquitous. If the wireless sensor network has bionic functions from network nodes to network structure systems, it has functions similar to those of living things (such as biological immunity), and can self-repair. It is of great significance to improve the robustness and security of wireless sensor networks. The research on fault diagnosis of wireless sensor network for structural health monitoring is of great practical significance. The research in this specific application field has just started, and the research on fault diagnosis in this field has great space and value.

In order to realize that the wireless sensor network has bionic functions from the network nodes to the network structure system, and has the ability of biological immunity, the existing bionic hardware technology can be used for reference. The so-called bionic hardware (BHW, Bio-inspired Hardware) realizes the real-time self-reconfiguration of electronic circuits in the system through an evolutionary mechanism, so that it can have the characteristics of hardware adaptation, self-organization, and self-repair like a living thing. At present, the research on bionic hardware technology has been carried out for more than ten years. The basic idea is to use programmable reconfigurable devices such as field programmable gate arrays FPGAs and field programmable analog arrays FPAAs to simulate biological evolution mechanisms to realize bionic evolution hardware circuits. In recent years, scholars in the United States, Germany, the United Kingdom and other countries have studied the application of bionic hardware in wireless sensor networks, and the research on bionic sensor networks based on self-healing has gradually become an important research direction of sensor networks.

(2) Research Status of Data Compression in Wireless Sensor Networks in SHM

In structural health monitoring, traditional soft fault diagnosis methods are limited by the large amount of data exchange between nodes. The solution is to compress the data of the nodes. The current compression methods include: In 2006, Y.F.Zhang. and J.Li from the Department of Civil and Environmental Engineering of Lehigh University in the United States proposed a lossy data compression algorithm based on lifting wavelet changes. It is used to eliminate the time correlation collected by the node and obtain a good compression effect. In the same year, Y.F. Zhang and J. Li proposed a data compression method based on the ARX model (Auto-Regressive with eXogenous input model) in order to realize the processing and research of seismic response data; in 2006, K.K. Chintalapudi proposed a linear predictive coding (Linear Predictive Coding, LPC) data compression method, which is lossless data compression, using a prediction algorithm based on the autoregressive moving average (ARMA) model to perform lossless compression on the collected data. In 2007, Y.F.Zhang. and J.Li further improved on the basis of the ARX model, using auto-regressive (Auto-Regressive, AR) as the model structure, using the instrumental variables method (instrumental variables method, IV) to calculate the prediction parameters, and A compression algorithm based on linear prediction is proposed to achieve lossless compression of data. On the other hand, in the field of structural monitoring based on wireless sensor networks, N. Xu et al. also proposed a wavelet data compression method based on local nodes in wireless sensors to solve the problem of bandwidth limitation of wireless sensor data transmission in structural monitoring. J.P.Lynch et al studied the use of Huffman codes to reduce the data transmission volume of wireless sensors for structural health monitoring. However, the data compression methods mentioned above all belong to traditional compression algorithms, that is, to obtain complete collection data first, and then perform compression processing on the data. At present, D.Donoho proposes a new compression sampling technology called Compressive Sensing (CS), which can directly collect data in a compressed format. This method is very suitable for the compression of narrow-band signals in structure monitoring. Although the research in the field of wireless sensor networks for structure monitoring has just started, it has a good application prospect.

The above research status provides ideas for the present invention. Based on data compression and bionic hardware, we will provide effective software and hardware fault diagnosis methods and self-repair mechanisms for fault diagnosis of wireless sensor networks in the field of structural health monitoring.

The limitations of the existing research work are as mentioned in "Research Status". At present, there have been many fruitful researches on the fault diagnosis of wireless sensor networks, and some scholars have tried to apply the artificial immune mechanism to the fault diagnosis of wireless sensor networks. However, due to the particularity of the application field of structural health monitoring based on WSN, there are still many key problems in the fault diagnosis in this field that have not been systematically resolved, and further research is needed. The specific questions are as follows:

(1) Extraction of node exchange data in fault diagnosis. Existing fault diagnosis techniques, etc., use the spatial similarity of adjacent nodes to exchange data to realize node fault diagnosis. However, in WSN-based structural health monitoring, nodes collect a large number of original samples at a fixed frequency each time, and a large number of data exchanges will exhaust the energy of the network. An effective data compression method is the key problem to be solved. In addition, extracting characteristic parameters that can identify abnormal data and normal damage data from the original sample is the focus of research.

(2) Integration of fault diagnosis and self-healing methods. Most of the existing fault diagnosis methods are aimed at diagnosing soft faults of a single node, but not dealing with hard faults. Due to the characteristics of its field, the fault diagnosis of wireless sensor networks for structural health monitoring is particularly urgent for the diagnosis and self-repair of software and hardware faults. It is necessary to further deepen the research on the integrated method of fault diagnosis and repair.

(3) The ability to detect simultaneous failures of multiple sensors. Different from other application fields, damage identification in health monitoring must be completed through the cooperation of multiple surrounding sensor nodes. Even if one of the nodes fails, the accuracy of damage identification will drop sharply. The diagnosis and repair of multi-sensor faults or even partial network faults need further research.

To sum up, the present invention will aim at the specific application field of structural health monitoring of wireless sensor networks. Diagnosis and self-healing theory and method of soft and hard faults in sensor networks. This research is of great significance for improving the self-healing and fault-tolerant capabilities of wireless sensor networks in response to abnormal events such as network attacks and node failures, and improving the robustness and security of wireless sensor networks in SHM.

Contents of the invention

Purpose of the invention: (1) Solve the problem of identifying fault information under the superposition of multiple abnormal data: this is the premise of realizing fault diagnosis. The purpose of fault diagnosis is to accurately extract the damage data of structural materials to judge the health status of materials. The measurer of the node has a great influence on the identification of material damage, so it is necessary to analyze the essential difference and connection of the two superimposed measurement values, and find a method to solve the interference and identification of abnormal data on normal damage data.

(2) Solve the compression problem of data exchange of sensor nodes under network faults: this is an effective method to expand traditional soft fault diagnosis. Effective data compression and acquisition methods can extend traditional soft fault diagnosis methods to the field of structural health monitoring. Therefore, it is necessary to analyze and study a feasible data compression acquisition method.

(3) Solve the problem of constructing self-healing trigger conditions under multi-node failure: this is a powerful way to realize fault diagnosis and self-healing. When multiple sensor nodes fail at the same time, different triggering conditions should implement different self-healing mechanisms. That is, the determination of trigger conditions (critical thresholds) for node self-repair, network self-repair, and network rerouting.

The invention studies the problem of fault detection in wireless sensor networks under structural health monitoring. After summarizing the latest research results in this field at home and abroad, combined with the research objectives, the research content of the present invention is shown in Figure 1.

1: Research on efficient real-time routing mechanism in SHM: In structural health monitoring, wireless sensor networks must be able to provide real-time, accurate, stable and reliable data collection for health monitoring. Therefore, it is necessary to study the efficient and real-time routing mechanism in SHM. Specific studies include:

(1) Research on multi-stream transmission in SHM. In structural health monitoring based on wireless sensor networks, the reliability and stability of network transmission is the basis of various applications. However, due to the limitation of bandwidth and transmission power in wireless transmission, and the influence of signal fading (Signal Fading) effect, the performance of wireless system will be greatly reduced, including system capacity, transmission efficiency, service quality and energy efficiency. The solution can be to use the cooperative communication mechanism of multi-stream transmission, which can realize the maximum throughput of the network and ensure the stability and reliability of the network. The problem of multi-stream transmission in SHM is one of the basic contents of the research of the present invention.

(2) Research on clustering routing algorithm in structural health monitoring. Study the alternative method of standby cluster head when the cluster head node fails;

(3) The performance analysis of the routing mechanism to the structural health monitoring system and the fault diagnosis of the network.

2: The interference analysis of the abnormal data of the nodes in SHM to the measured value of structural damage: as mentioned in the "project establishment basis", the structural damage of the material refers to the changes in the cracks and holes of the material, and the identification and location of the damage must This is an important feature of structural health monitoring; when a node produces a measurement error due to a sensor module failure, it will drastically affect the accuracy of structural damage identification and location. At this time, the faulty node should be temporarily removed to avoid interference, and then damage identification and location should be performed, so timely detection of the faulty node becomes the key.

Most of the existing WSN node fault detection methods need to exchange and collect data between nodes to detect faults. However, in SHM, as mentioned in "Feature 2" mentioned in "Basis for Establishment", nodes will collect a large number of original samples at a fixed frequency each time, and a large number of data exchanges will exhaust the energy of the network. Therefore, for the failure of sensing components, the existing WSN node failure detection methods are not fully applicable in structural health monitoring.

Therefore, it is a prerequisite for this study to extract characteristic parameters that can identify abnormal data and normal damage data from the original collected samples to realize node fault detection.

3: Soft fault diagnosis of wireless sensor nodes based on compressed sensing in SHM: In structural health monitoring based on wireless sensor networks, although nodes are fixed, there is no need to consider their mobility, and even manual routing can be configured. However, considering the flexibility and robustness of node fault diagnosis in this field, the use of cluster routing mechanism can make fault management distributed to the respective cluster areas. Therefore, a soft fault diagnosis method for nodes (including cluster heads) based on compressed sensing is proposed, which is suitable for structural health monitoring. Specific studies include:

(1) Research on the compression of data collected by nodes; analysis of the impact of compression algorithm reconstruction performance on structural health monitoring and network fault diagnosis;

(2) Node fault diagnosis method. The fault diagnosis method and routing algorithm of ordinary nodes are studied when transient communication soft faults occur in nodes or no damage occurs in structural materials. It intends to refer to and improve the existing neighboring node cooperation processing method to adapt to the specific field of the present invention, and the key point is to improve the type of data exchanged during the neighboring node cooperation.

4: The hard fault diagnosis mechanism of sensor nodes in SHM and the design of bionic self-repair nodes: according to the characteristics of wireless sensor networks and the specific needs of structural health monitoring, the application of bionic hardware to self-repair sensor nodes must meet the requirements of low cost. , small size and other characteristics. Therefore, how to use the advantages of bionic hardware to propose a self-repair mechanism suitable for hard failures of wireless sensor nodes is one of the focuses of this study. Specific studies include:

(1) Self-healing node software and hardware architecture;

(2) The selection and setting method of the self-healing module is studied for the strain, Lamb wave and other sensing components of the SHM;

(3) Design and implementation of self-healing sensor nodes, hard fault diagnosis and self-reconfiguration process of nodes;

(4) Performance testing of self-repairing sensor nodes, including: node power consumption, analog-to-digital conversion, transmission distance, signal spectrum and other performance.

Technical scheme of the present invention is as follows:

Aiming at the characteristics of the specific application field of structural health monitoring, the invention provides a theory and method for diagnosing and self-repairing of soft and hard faults suitable for health monitoring. On the basis of studying the efficient real-time routing mechanism, aiming at the diagnosis of soft faults, with the support of compressed sensing theory, a method of neighbor cooperation and cluster routing is proposed to detect soft faults on ordinary nodes and cluster head nodes; for hard faults, According to the theory of bionic self-repair, the theory and method of network fault diagnosis and self-repair of wireless sensor network under the condition of node failure, partial network failure and external node intrusion are studied. Through the above research, the self-healing and fault-tolerant capabilities of wireless sensor networks in response to abnormal events such as network attacks and node failures are improved.

The method for diagnosing and self-repairing soft and hard faults of health monitoring includes the following specific steps:

Initial scene setup:

Step 1) Efficient real-time routing in SHM: Based on the cooperative communication mechanism of multi-stream transmission, a distributed cluster routing algorithm is designed to form routing links. The first node of the link is the cluster head, and the second node is the backup cluster head, etc. .

Step 2) Interference processing of abnormal data in SHM: Combining with the feature extraction method of structural monitoring vibration signals studied by the invention group in the previous period, the method proposed by M.A-B.Abdo is proposed to extract from the original collected data for node fault detection and the two characteristic parameters of structural damage identification, after obtaining the feature extraction of the natural frequency, according to the local and global theories of the two measured values, it is proposed to use the sample statistical algorithm or the traditional node cooperative fault diagnosis algorithm to realize the node fault abnormality Distinguishing and extracting values and structural damage measurements.

The monitoring vibration signal obtained from the sensor uses Gabor order tracking and Viterbi maximum likelihood decoding algorithm to extract natural frequency, and uses sample statistical algorithm or traditional node cooperative fault diagnosis algorithm to identify node fault value and structural damage value.

Step 3) Diagnosis mechanism in SHM: In structural health monitoring, compressive sampling based on compressed sensing is used for node sampling to adapt to data exchange between nodes. For the measured value of the sensor node, a matrix Φ∈R ^{M × N} (M<<N) uncorrelated with the orthogonal base Ψ∈R ^{N × N} is used to project the high-dimensional signal onto a low-dimensional space to realize the node Compression of sampled signals.

Step 4) Hardware architecture: In the self-repairing wireless sensor node hardware architecture, programmable devices are selected as embedded solutions, and the main functional modules are connected through bionic hardware FPAAs as modules. The bionic hardware and field programmable analog array FPAAs are used to realize the signal link of the sensor module in the node, and the fault diagnosis function and self-repair control function of the sensor link are designed in the signal processing module of the node.

Step 5) software architecture: the data acquisition driver program converts the binary data in the A/D conversion register into decimal data and compares it with the threshold value set by the signal abnormality diagnosis program to give the diagnosis result; the central control program according to the diagnosis result Send a drive command to the FPAA driver; the program first reads the FPAA configuration file stored in the external Flash memory according to the drive command, and then dynamically configures the FPAA to complete the reconfiguration of the sensor and its redundancy layer signal link structure.

Step 6) Performance optimization: the self-repairing wireless sensor network structure system adopts a hybrid hierarchical topology structure, and a monitoring node for monitoring the network status is set in the network.

Beneficial effect

Compared with the existing research on fault diagnosis of wireless sensor networks, the characteristics and innovations are as follows:

1) In structural health monitoring, diagnose and repair soft and hard faults of wireless sensor networks, identify abnormal data and normal damage data of faulty nodes, and improve the robustness and accuracy of system monitoring.

2) Applying bionic hardware to the sensing components of sensor nodes to realize node fault diagnosis and self-repair, considering the actual engineering requirements of SHM, can solve the key problem of network failure caused by simultaneous failure of multiple sensors.

3) The mechanism of compressed sampling and multi-stream transmission is adopted to improve the performance of fault diagnosis and data transmission.

Description of drawings

Fig. 1 research content of the present invention.

Figure 2 shows the general technical route to be taken by the invention.

Figure 3 shows the soft fault diagnosis mechanism of wireless sensor nodes

Figure 4 proposes the self-repairing wireless sensor node hardware architecture.

Figure 5 proposes the self-healing wireless sensor node software architecture.

detailed description

The technical solution of the present invention will be described in detail below in conjunction with the accompanying drawings.

The invention is aimed at diagnosing and repairing software and hardware faults of wireless sensor networks in the field of structural health monitoring, focusing on the design of self-repairing nodes of bionic hardware, the distinction between abnormal data and normal data of fault nodes, and soft fault diagnosis based on compressed sensing , node and network fault diagnosis and repair mechanism, etc., as shown in Figure 1, is the research content of the present invention. The invention adopts a research plan combining theoretical analysis and experimental verification. Firstly, the fault type of sensor nodes or networks in SHM is obtained, and the number of faulty nodes is compared with the critical threshold, so as to repair the nodes or part of the network respectively. The entire research route is shown in the figure 2, where the parameter ψ is used to determine the duration of node soft faults identified as hard faults, φ1 represents the critical threshold of the number of faulty nodes that need to be re-clustered in SHM, and φ2 is the number of faulty nodes that have network faults in SHM critical threshold.

(1): Research on efficient real-time routing mechanism in SHM

In view of the fact that the sensor nodes of SHM are fixed and do not need to consider mobility, the routing mechanism and soft fault diagnosis algorithm of SHM should have the characteristics of stability, real-time, and small data transmission volume. Therefore, multi-stream transmission is considered in the routing mechanism.

Our idea is to apply the cooperative communication mechanism to the problem of multi-stream transmission in wireless networks to improve its transmission performance. The technical route to be adopted is: use linear programming, integer programming and other methods to formally describe the problem; use the method of specification to prove that some problems are NP-hard; use bipartite graph saturation matching, Dijkstra method, branch and bound method, dynamic Planning and other techniques to design effective algorithms and protocols, and use mathematical methods such as binomial correlation properties to analyze the theoretical performance of algorithms.

On the basis of the cooperative communication mechanism of multi-stream transmission mentioned above, considering the actual needs of structural health monitoring, and inspired by the idea of distributed hierarchical aggregation clustering method, the scheme we intend to adopt is: design a distributed clustering routing algorithm, and form a routing link, the first node of the link is the cluster head, the second node is the backup cluster head, etc.

(2): Interference processing of the abnormal data of fault nodes in SHM to the measured value of structural damage

The interference of node faults on structural damage directly affects the accuracy of damage identification, and normal structural damage measurements often interfere with the fault diagnosis of sensor networks and cause confusion. Therefore, extracting characteristic parameters that can identify abnormal data and normal damage data from the original collected samples is the key to fault diagnosis.

Inspired by the mathematical model of node faults and structural damage in SHM, combined with the feature extraction method of structural monitoring vibration signals studied by the invention group in the previous period, the invention intends to use the method proposed by M.A-B. Based on the two characteristic parameters of node fault detection and structural damage identification, the interference of fault nodes on structural damage is avoided. The technical ideas to be adopted are as follows:

In structural health monitoring, node fault measurement values and structural damage measurement values belong to weak signals, and interfere and confuse each other. It is not conducive to the fault diagnosis of the network, but also affects the damage identification in the structural health monitoring. According to the theory in the literature: "The main means of damage identification and location in structural health monitoring is realized through the structural vibration characteristics (natural frequency) of the signal, and there is an obvious difference between the natural frequency of the measured value of the fault node and the natural frequency of the measured value of the healthy node." difference; the former is local, the latter is global."

The above theories provide an important basis for distinguishing faulty node measurements from normal damage measurements and realizing fault diagnosis. Therefore, we intend to use the Gabor coefficient C _m,n representing the time-frequency characteristics of the structural vibration signal, and use the Gabor order tracking method to obtain a time-frequency grid surface composed of discrete time points and discrete frequency points, and use the Viterbi algorithm to find the time-frequency grid surface. The optimal frequency path between the points, so as to realize the frequency feature extraction of the sensor measurement signal.

After obtaining the feature extraction of the natural frequency, according to the local and global theories of the two measured values, it is planned to use the sample statistical algorithm or the traditional node collaborative fault diagnosis algorithm to realize the distinction and extraction of node fault abnormal values and structural damage measurement values . The preliminary plan is shown in Figure 3.

(3): Soft fault diagnosis mechanism of wireless sensor nodes based on compressed sensing in SHM

In the soft fault diagnosis algorithm, the weighted median fault diagnosis method studied by the invention group in the previous period is combined to realize the fault detection of common nodes. This scheme is mainly aimed at the case of transient communication soft faults or no damage in structural materials, once the routing link is generated, it remains fixed until the trigger of the critical threshold of φ1. The technical ideas to be adopted are as follows:

Considering that sensor nodes have temporal and spatial correlations in the network, for the measured value x _i of a certain sensor node, its fault diagnosis can be compared with the perceived values of surrounding neighboring nodes. For a node to be diagnosed with M neighbor nodes. Its weighted median can be defined as:

in, is the weighted median value, x _j (j=1,2...M) is the perception measurement value of the neighbor node, and λ _j (j=1,2...M) is the weight value of each neighbor node. The identification and diagnosis of node soft faults (abnormal perception measurement values) can be realized through the fault diagnosis function as follows.

However, in structural health monitoring, nodes collect a large number of original samples each time, the above-mentioned weighted median fault discrimination cannot adapt to a large amount of data exchange between nodes. The proposed solution is to use compressed sampling based on compressed sensing to adapt to data exchange between nodes when sampling nodes. For the measured value of the sensor node, a matrix Φ∈R ^{M × N} (M<<N) uncorrelated with the orthogonal base Ψ∈R ^{N × N} is used to project the high-dimensional signal onto a low-dimensional space to realize the node Compression of sampled signals. Its core formula is:

y=Φx=ΦΨα=Θα (4)

Among them, Φ is an M×N matrix, called a measurement matrix; Θ=ΦΨ is an M×N matrix, called an observation matrix. Finally, the reconstruction of the compressed signal is realized by solving the method of minimizing the l _p norm.

(4): Hard fault diagnosis mechanism of wireless sensor nodes in SHM and design of self-healing nodes

1) Self-healing wireless sensor node hardware architecture

The hardware architecture of the self-repairing wireless sensor node to be used in the research is shown in Figure 4.

In view of the characteristics of multi-parameter acquisition in structural health monitoring, without loss of generality, we intend to study the self-healing function of typical strain wireless sensor nodes, and the design principles of other types of sensor nodes are the same. Considering the application requirements of the actual engineering of structural health monitoring, the nodes of self-healing wireless sensors must meet the characteristics of low power consumption, low cost, and small size. A low-cost programmable device can be used as an embedded solution. The main functional modules change the way they are connected to each other in the past, and connect them as module cells through bionic hardware FPAAs. The strain wireless sensor node shown in Figure 4 includes three parts: sensing module, signal processing module, and wireless transceiver module. The bionic hardware and field programmable analog array FPAAs are used to realize the signal link of the sensor module in the node, and the fault diagnosis function and self-repair control function of the sensor link are designed in the signal processing module of the node. Wherein, the redundant part in the self-repair module can be realized by the same hardware circuit as the original sensor module, or can be realized by a functional unit composed of a programmable array.

2) Self-healing wireless sensor node software architecture

The software architecture of the self-repairing wireless sensor node to be adopted in the study is shown in Figure 5. The data acquisition driver program converts the binary data in the A/D conversion register into decimal data, and then the signal abnormality diagnosis program compares it with the set threshold to give the diagnosis result; the central control program sends a drive to the FPAA driver according to the diagnosis result command; the program first reads the FPAA configuration file stored in the external Flash memory according to the driver command, and then dynamically configures the FPAA to complete the reconstruction of the sensor and its redundancy layer signal link.

3) Performance considerations of self-healing wireless sensor networks

The self-healing wireless sensor network structure system plans to adopt a hybrid hierarchical topology. In addition to the original sensor nodes, it is considered to set up monitoring nodes in the network to monitor the network status. The system parameters that need to be considered during restoration include various parameters such as network function, reliability, power consumption, and restoration time.

Claims (1)

1. A soft and hard fault diagnosis and self-repair method suitable for health monitoring is characterized by comprising the following steps:

1) efficient real-time routing in SHM: based on a multi-stream transmission cooperative communication mechanism, a distributed clustering routing algorithm is designed to form a routing link, a first node of the link is a cluster head, a second node is a standby cluster head and the like;

2) interference processing of abnormal data in SHM: the method is characterized in that two characteristic parameters which are respectively used for node fault detection and structural damage identification are extracted from original collected data by combining a characteristic extraction method of a structural monitoring vibration signal, and after the characteristic extraction of natural frequency is obtained, a sample statistical algorithm or a traditional node cooperation fault diagnosis algorithm is adopted to realize the distinguishing and extraction of a node fault abnormal value and a structural damage measured value;

3) diagnostic mechanisms in SHM: in the structural health monitoring, compressed sampling based on compressed sensing is adopted during node sampling so as to adapt to data exchange among nodes;

4) hardware and software architecture: in a self-repairing wireless sensor node hardware architecture, a programmable device is selected as an embedded solution, and is connected by taking bionic hardware FPAAs as a module; in the self-repairing wireless sensor node software architecture, a program firstly reads an FPAA configuration file stored in an external Flash memory according to a driving command, and then dynamically configures the FPAA so as to complete the reconstruction of a sensing layer signal link and a redundant layer signal link thereof.