A deep learning-based fault detection method for wireless sensor networks

Abstract

The present invention relates to the technical field of wireless sensor network fault detection, and specifically discloses a wireless sensor network fault detection method based on deep learning. The present invention can timely discover sensor faults, processor faults and communication faults corresponding to each node in the target wireless sensor network by monitoring and analyzing the hardware status corresponding to each node in the target wireless sensor network, so as to avoid further expansion of hardware faults of each node in the wireless sensor network and affect the normal operation of the wireless sensor network; by monitoring and analyzing the energy supply status corresponding to each node in the target wireless sensor network, the energy supply status of the node can be fully understood, and the problem of insufficient energy supply or battery failure can be discovered in time, so as to effectively ensure the reliability and stability of the operation of the wireless sensor network.

Description

Technical Field

[0001] The present invention relates to the technical field of wireless sensor network fault detection, and in particular to a wireless sensor network fault detection method based on deep learning. Background Art

[0002] Wireless sensor networks are network systems composed of a large number of micro sensor nodes. These sensor nodes are usually distributed in a specific monitoring area and are connected to each other through wireless communication to jointly complete the monitoring task of the target area. They are widely used in environmental monitoring, industrial control, smart home and other fields. During the operation of wireless sensor networks, various faults may occur due to various reasons (such as hardware failure, environmental interference, energy exhaustion, etc.), such as node failure, communication interruption, data anomaly, etc. These faults may cause the performance of wireless sensor networks to decline, the coverage range to shrink, or even fail to work properly. Therefore, realizing fault detection of wireless sensor networks is a very important step in the operation and maintenance of wireless sensor networks, and it is also a key step to ensure the reliability of wireless sensor networks.

[0003] However, the existing methods for detecting wireless sensor network faults are mostly based on traditional fault diagnosis methods, which have low detection efficiency and low accuracy. At the same time, in the actual detection of wireless sensor network faults, the existing methods cannot timely analyze the potential problems existing in the wireless sensor network, resulting in the inability to achieve comprehensive fault detection of the wireless sensor network, increasing the low operating efficiency of the wireless sensor network and the difficulty in effectively locating the fault.

[0004] In order to solve the above defects, a technical solution is now provided. Summary of the invention

[0005] The purpose of the present invention is to provide a wireless sensor network fault detection method based on deep learning to solve the problems raised in the above background.

[0006] The purpose of the present invention can be achieved by the following technical solution: A wireless sensor network fault detection method based on deep learning, comprising the following steps:

[0007] S1, data collection: collect the node hardware status parameters, node energy supply parameters and network communication parameters corresponding to each node in the target wireless sensor network, and perform node hardware status detection analysis, node energy supply detection analysis and network communication detection analysis respectively;

[0008] S2, node hardware status detection and analysis: monitor the node hardware status parameters corresponding to each node in the target wireless sensor network, and analyze the node hardware status evaluation index corresponding to each node in the target wireless sensor network, and obtain the node hardware status evaluation index corresponding to each node in the target wireless sensor network;

[0009] Preferably, the node hardware status evaluation index corresponding to each node in the target wireless sensor network is analyzed to obtain the node hardware status evaluation index corresponding to each node in the target wireless sensor network. The specific analysis method is as follows:

[0010] Real-time monitoring obtains the sensor output data corresponding to each node in the target wireless sensor network within a period of time, which is recorded as i represents the number of each node, i = 1, 2, ..., n, n represents the total number of node numbers, j represents the number of each sensor output data, j = 1, 2, ..., m, m represents the total number of sensor output data numbers, and the average value is calculated to obtain the average value of sensor output data corresponding to each node in the target wireless sensor network within a period of time, which is recorded as SJ i , according to the formula Calculate the standard deviation σ of the sensor output data corresponding to each node in the target wireless sensor network within a period of time i , and according to the formula Calculate the coefficient of variation ν of the sensor output data corresponding to each node in the target wireless sensor network within a period of time i , and obtain the number of sensor output data jumps corresponding to each node in the target wireless sensor network in real time within a period of time, recorded as N i And the three data are analyzed in a formulaic manner according to the set digital model: Thus, the stability index WP of the sensor output data corresponding to each node in the target wireless sensor network within a period of time is obtained. i ;

[0011] Among them, e represents a natural constant;

[0012] σ', ν', N' respectively represent the set reference output data standard deviation, reference output data variation coefficient, and reference output data jump times, a1, a2, and a3 respectively represent the set output data standard deviation, output data variation coefficient, and output data jump times corresponding to the weight of the sensor output data stability index;

[0013] Real-time monitoring obtains the peak CPU usage of each node in the target wireless sensor network within a period of time, recorded as At the same time, the memory occupancy rate of each node in the target wireless sensor network corresponding to each monitoring time point within a period of time is obtained in real time, and the maximum memory occupancy rate and the minimum memory occupancy rate are extracted from them, respectively, and recorded as At the same time, the monitoring time points corresponding to the maximum memory occupancy and the minimum memory occupancy are extracted, and then the monitoring interval duration corresponding to the maximum memory occupancy and the minimum memory occupancy is obtained, recorded as T, according to the formula Calculate the memory occupancy rate change rate of each node in the target wireless sensor network over a period of time i ;

[0014] Real-time monitoring is used to obtain the response time and execution time of each task corresponding to each node in the target wireless sensor network within a period of time, and the average value is calculated to obtain the average response time and average execution time of each task corresponding to each node in the target wireless sensor network within a period of time, which are recorded as According to the set formula Calculate the task execution efficiency XL of each node in the target wireless sensor network within a period of time i , T 响应 '、T 执行 'represents the average response time and average execution time of the reference task respectively, ΔT 响应 , ΔT 执行 They are respectively represented by the set reference task average response time difference and reference task average execution time difference, a4 and a5 are respectively represented by the set task average response time and task average execution time corresponding to the weight of task execution efficiency;

[0015] According to the formula Calculate the processor performance index CX of each node in the target wireless sensor network over a period of time i , where SY max ', ζ', XL' represent the set reference CPU usage peak value, reference memory usage change rate, and reference task execution efficiency, respectively; a6, a7, and a8 represent the set CPU usage peak value, memory usage change rate, and task execution efficiency corresponding to the weight of the processor performance index;

[0016] Real-time monitoring obtains the corresponding signal strength mean, signal-to-noise ratio and packet error rate of each node in the target wireless sensor network over a period of time, and marks them as XD respectively. i , XB i , SC i , according to the formula Calculate the communication signal quality index ZP corresponding to each node in the target wireless sensor network within a period of time i, XD', XB', SC' represent the set reference signal strength mean, reference signal-to-noise ratio, and reference data packet error rate, respectively; b1, b2, and b3 represent the weights of the set signal strength mean, signal-to-noise ratio, and data packet error rate corresponding to the communication signal quality index, respectively;

[0017] According to the formula Calculate the hardware status evaluation index φ of each node in the target wireless sensor network within a period of time i , b4, b5, and b6 represent the weights of the node hardware status evaluation index corresponding to the set sensor output data stability index, processor performance index, and communication signal quality index, respectively.

[0018] S3. Node energy supply detection and analysis: monitor the energy supply parameters corresponding to each node in the target wireless sensor network, and analyze the energy supply evaluation index corresponding to each node in the target wireless sensor network to obtain the energy supply evaluation index corresponding to each node in the target wireless sensor network.

[0019] Preferably, the energy supply evaluation index corresponding to each node in the target wireless sensor network is analyzed to obtain the energy supply evaluation index corresponding to each node in the target wireless sensor network. The specific analysis method is:

[0020] Real-time monitoring obtains the remaining battery power of the node at each time monitoring point within a period of time, and extracts the remaining battery power of the node at the first time monitoring point and the remaining battery power of the node at the last monitoring time point, which are recorded as At the same time, extract the first monitoring time point and the last monitoring time point, and then obtain the monitoring interval length corresponding to the first monitoring time point and the last monitoring time point, recorded as ΔT, and calculate according to the formula The rate of change of the remaining battery power of each node in the target wireless sensor network over a period of time

[0021] Real-time monitoring obtains the charging time, charging capacity, charging average temperature and charging average voltage of each battery charge corresponding to each node in the target wireless sensor network within a period of time, which are marked as f represents the number of each charge, f = 1, 2, ..., g, g represents the total number of charge numbers;

[0022] The charging power of each battery corresponding to each node in the target wireless sensor network is matched with the reference charging time corresponding to each charging power, and the reference charging time of each battery corresponding to each node in the target wireless sensor network is obtained, which is recorded as

[0023] According to the formula

[0024] Calculate the battery charging health index corresponding to each node in the target wireless sensor network over a period of time in, They represent the set reference charging power, reference charging average temperature, and reference charging average voltage, respectively; c1, c2, c3, and c4 represent the proportion weights of the battery charging health index corresponding to the set charging time, charging power, charging average temperature, and charging average voltage, respectively;

[0025] Obtain in real time the discharge duration, discharge capacity, discharge mean temperature and discharge mean voltage of each battery discharge corresponding to each node in the target wireless sensor network within a period of time, marked as k represents the number of each charge, k = 1, 2, ..., l, l represents the total number of discharges;

[0026] The discharge power of each battery corresponding to each node in the target wireless sensor network is matched with the reference discharge duration corresponding to each discharge power, and the reference discharge duration of each battery corresponding to each node in the target wireless sensor network is obtained, which is recorded as

[0027] According to the formula, the battery discharge corresponding to each node in the target wireless sensor network within a period of time is calculated

[0028] Calculate the battery discharge health index corresponding to each node in the target wireless sensor network over a period of time in, They represent the set reference discharge capacity, reference discharge mean temperature, and reference discharge mean voltage respectively; c5, c6, c7, and c8 represent the proportion weights of the battery charging health index corresponding to the set discharge time, discharge capacity, discharge mean temperature, and discharge mean voltage respectively;

[0029] Real-time monitoring obtains the number of low battery warnings corresponding to each node in the target wireless sensor network within a period of time, recorded as YN i , conduct a comprehensive analysis of the above data, based on the set data model: Thus, the energy supply evaluation index η corresponding to each node in the target wireless sensor network within a period of time is obtained. i , where d1, d2, d3, and d4 represent the weights of the energy supply evaluation index corresponding to the set battery remaining power decrease rate, battery charging health index, battery discharging health index, and battery low power warning times, respectively.

[0030] S4, network communication detection and analysis: monitoring the network communication parameters corresponding to the target wireless sensor network, and thereby analyzing the network communication evaluation index corresponding to the target wireless sensor network to obtain the network communication evaluation index corresponding to the target wireless sensor network;

[0031] Preferably, the network communication parameters corresponding to the target wireless sensor network are analyzed to obtain the network communication evaluation index corresponding to the target wireless sensor network. The specific analysis method is:

[0032] Real-time monitoring obtains the number of node connection changes, average path length change rate and clustering coefficient change of the target wireless sensor network within a period of time, and marks them as N respectively. 变化 , R 变化 , C 变化 , according to the formula Calculate the network topology stability index λ corresponding to the target wireless sensor network within a period of time, where e1, e2, and e3 represent the weights of the network topology stability index corresponding to the set number of node connection changes, average path length change rate, and clustering coefficient change, respectively;

[0033] Real-time monitoring obtains the abnormal values ​​of data transmission rate, data packet size, and data packet number corresponding to the target wireless sensor network within a period of time, and marks them as Y1, Y2, and Y3 respectively. According to the formula Calculate the network traffic evaluation index γ corresponding to the target wireless sensor network, e4, e5, and e6 represent the weights of the network traffic evaluation index corresponding to the set data transmission rate abnormal value, data packet size abnormal value, and data packet number abnormal value, respectively;

[0034] According to the formula Calculate the network communication evaluation index corresponding to the target wireless sensor network f1 and f2 represent the weights of the set network topology stability index and network traffic evaluation index corresponding to the network communication evaluation index respectively.

[0035] S5, fault analysis: analyzing the fault type corresponding to the target wireless sensor network based on the node hardware status evaluation index and the node energy supply evaluation index corresponding to each node in the target wireless sensor network and the network communication evaluation index corresponding to the target wireless sensor network, and obtaining the fault type corresponding to the target wireless sensor network;

[0036] Preferably, the fault type corresponding to the target wireless sensor network is analyzed based on the node hardware status evaluation index and the node energy supply evaluation index corresponding to each node in the target wireless sensor network and the network communication evaluation index corresponding to the target wireless sensor network. The specific analysis method is as follows:

[0037] The node hardware status evaluation index corresponding to each node in the target wireless sensor network is compared with the set node hardware status evaluation index threshold. If the node hardware status evaluation index corresponding to a node in the target wireless sensor network is less than the set node hardware status evaluation index threshold, the fault type corresponding to the node in the target wireless sensor network is determined to be a node hardware type.

[0038] The energy supply evaluation index corresponding to each node in the target wireless sensor network is compared with the set energy supply evaluation index threshold. If the energy supply corresponding to a node in the target wireless sensor network is less than the set energy supply evaluation index threshold, the fault type corresponding to the node in the target wireless sensor network is determined to be the node energy supply type.

[0039] The network communication evaluation index corresponding to the target wireless sensor network is compared with the set network communication evaluation index threshold. If the network communication evaluation index corresponding to the target wireless sensor network is less than the set network communication evaluation index threshold, the fault type corresponding to the target wireless sensor network is determined to be the network communication type.

[0040] S6. Fault processing: receiving the fault type corresponding to the target wireless sensor network, and performing corresponding processing based on the fault type corresponding to the target wireless sensor network.

[0041] Beneficial effects of the present invention:

[0042] The present invention monitors and analyzes the sensor output data stability index, processor performance index, and communication signal quality index corresponding to each node in the target wireless sensor network, and then evaluates and analyzes the node hardware status corresponding to each node. It can timely discover sensor failures, processor failures, and communication failures corresponding to each node in the target wireless sensor network, and avoid further expansion of hardware failures of each node in the wireless sensor network to affect the normal operation of the wireless sensor network.

[0043] The present invention monitors and analyzes the rate of change of battery remaining power decrease, battery charging health index, battery discharging health index and battery low power warning times corresponding to each node in the target wireless sensor network, and then evaluates and analyzes the node energy supply status corresponding to each node. It can fully understand the energy supply situation of the node, timely discover energy supply shortage or battery failure problems, and effectively ensure the reliability and stability of the operation of the wireless sensor network.

[0044] The present invention monitors and analyzes the network topology stability and network traffic status corresponding to the target wireless sensor network, and then evaluates and analyzes the network communication status corresponding to the target wireless sensor network, thereby avoiding further paralysis of the overall network communication failure of the target wireless sensor network and effectively improving the timeliness of the response to the network communication failure.

[0045] By displaying corresponding prompt information according to the fault type of the wireless sensor network, the fault type of the wireless sensor network can be accurately determined, which can help technicians quickly locate the problem and take corresponding repair measures, effectively improving the fault maintenance efficiency to a certain extent and providing a strong guarantee for the stable operation of the wireless sensor network. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The present invention will be further described below in conjunction with the accompanying drawings.

[0047] Figure 1 It is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0048] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0049] See also Figure 1 As shown, the present invention is a wireless sensor network fault detection method based on deep learning, comprising the following steps:

[0050] S1, data collection: collect the node hardware status parameters, node energy supply parameters and network communication parameters corresponding to each node in the target wireless sensor network, and perform node hardware status detection analysis, node energy supply detection analysis and network communication detection analysis respectively;

[0051] S2. Node hardware status detection and analysis: Monitor the node hardware status parameters corresponding to each node in the target wireless sensor network, and analyze the node hardware status evaluation index corresponding to each node in the target wireless sensor network to obtain the node hardware status evaluation index corresponding to each node in the target wireless sensor network. The specific analysis method is as follows:

[0052] Real-time monitoring obtains the sensor output data corresponding to each node in the target wireless sensor network within a period of time, which is recorded as i represents the number of each node, i = 1, 2, ..., n, n represents the total number of node numbers, j represents the number of each sensor output data, j = 1, 2, ..., m, m represents the total number of sensor output data numbers, and the average value is calculated to obtain the average value of sensor output data corresponding to each node in the target wireless sensor network within a period of time, which is recorded as SJ i , according to the formula Calculate the standard deviation σ of the sensor output data corresponding to each node in the target wireless sensor network within a period of time i , and according to the formula Calculate the coefficient of variation ν of the sensor output data corresponding to each node in the target wireless sensor network within a period of time i , and obtain the number of sensor output data jumps corresponding to each node in the target wireless sensor network in real time within a period of time, recorded as N i And the three data are analyzed in a formulaic manner according to the set digital model: Thus, the stability index WP of the sensor output data corresponding to each node in the target wireless sensor network within a period of time is obtained. i ;

[0053] Among them, e represents a natural constant;

[0054] σ', ν', N' respectively represent the set reference output data standard deviation, reference output data variation coefficient, and reference output data jump times, a1, a2, and a3 respectively represent the set output data standard deviation, output data variation coefficient, and output data jump times corresponding to the weight of the sensor output data stability index;

[0055] It should be noted that the standard deviation of the sensor output data reflects the degree of discreteness of the data. The larger the standard deviation, the greater the fluctuation of the data. The coefficient of variation of the sensor output data more intuitively reflects the relative degree of fluctuation of the data. The number of jumps in the sensor output data refers to the number of times the sensor output data changes from one value to another. Frequent data jumps may be a manifestation of sensor failure or interference.

[0056] Real-time monitoring obtains the peak CPU usage of each node in the target wireless sensor network within a period of time, recorded as At the same time, the memory occupancy rate of each node in the target wireless sensor network corresponding to each monitoring time point within a period of time is obtained in real time, and the maximum memory occupancy rate and the minimum memory occupancy rate are extracted from them, respectively, and recorded as At the same time, the monitoring time points corresponding to the maximum memory occupancy and the minimum memory occupancy are extracted, and then the monitoring interval duration corresponding to the maximum memory occupancy and the minimum memory occupancy is obtained, recorded as T, according to the formula Calculate the memory occupancy rate change rate of each node in the target wireless sensor network over a period of time i ;

[0057] Real-time monitoring is used to obtain the response time and execution time of each task corresponding to each node in the target wireless sensor network within a period of time, and the average value is calculated to obtain the average response time and average execution time of each task corresponding to each node in the target wireless sensor network within a period of time, which are recorded as According to the set formula Calculate the task execution efficiency XL of each node in the target wireless sensor network within a period of time i , T 响应 '、T 执行 'represents the average response time and average execution time of the reference task respectively, ΔT 响应 , ΔT 执行 They are respectively represented by the set reference task average response time difference and reference task average execution time difference, a4 and a5 are respectively represented by the set task average response time and task average execution time corresponding to the weight of task execution efficiency;

[0058] It should be noted that the task response time refers to the time interval from the time a node receives a communication request from other nodes or a sensor triggers an event to the time the processor starts processing it. A fast task response time indicates that the processor can process various inputs in a timely manner, thereby improving the overall performance and responsiveness of the system. The task execution time refers to the time required for the measuring node to read data from the sensor, process the data and send the results. A shorter task execution time indicates that the processor can respond and process important tasks quickly, with higher performance.

[0059] According to the formula Calculate the processor performance index CX of each node in the target wireless sensor network over a period of time i , where SY max ', ζ', XL' represent the set reference CPU usage peak value, reference memory usage change rate, and reference task execution efficiency, respectively; a6, a7, and a8 represent the set CPU usage peak value, memory usage change rate, and task execution efficiency corresponding to the weight of the processor performance index;

[0060] Real-time monitoring obtains the corresponding signal strength mean, signal-to-noise ratio and packet error rate of each node in the target wireless sensor network over a period of time, and marks them as XD respectively. i , XB i , SC i , according to the formula Calculate the communication signal quality index ZP corresponding to each node in the target wireless sensor network within a period of time i , XD', XB', SC' represent the set reference signal strength mean, reference signal-to-noise ratio, and reference data packet error rate, respectively; b1, b2, and b3 represent the weights of the set signal strength mean, signal-to-noise ratio, and data packet error rate corresponding to the communication signal quality index, respectively;

[0061] It should be noted that the signal-to-noise ratio refers to the ratio of the measured communication signal to the noise. The larger the signal-to-noise ratio, the less interference the signal has and the better the communication quality. The packet error rate refers to the proportion of errors in the packets received and sent by the node. The larger the packet error rate, the more communication failure the node has.

[0062] According to the formula Calculate the hardware status evaluation index φ of each node in the target wireless sensor network within a period of time i , b4, b5, and b6 represent the weights of the node hardware status evaluation index corresponding to the set sensor output data stability index, processor performance index, and communication signal quality index, respectively.

[0063] In a specific embodiment, the present invention monitors and analyzes the sensor output data stability index, processor performance index, and communication signal quality index corresponding to each node in the target wireless sensor network, and then evaluates and analyzes the node hardware status corresponding to each node, so as to timely discover sensor failures, processor failures, and communication failures corresponding to each node in the target wireless sensor network, and avoid further expansion of hardware failures of each node in the wireless sensor network and affect the normal operation of the wireless sensor network.

[0064] S3, node energy supply detection and analysis: monitor the energy supply parameters corresponding to each node in the target wireless sensor network, and analyze the energy supply evaluation index corresponding to each node in the target wireless sensor network, and obtain the energy supply evaluation index corresponding to each node in the target wireless sensor network. The specific analysis method is as follows:

[0065] Real-time monitoring obtains the remaining battery power of the node at each time monitoring point within a period of time, and extracts the remaining battery power of the node at the first time monitoring point and the remaining battery power of the node at the last monitoring time point, which are recorded as At the same time, extract the first monitoring time point and the last monitoring time point, and then obtain the monitoring interval length corresponding to the first monitoring time point and the last monitoring time point, recorded as ΔT, and calculate according to the formula The rate of change of the remaining battery power of each node in the target wireless sensor network over a period of time

[0066] Real-time monitoring obtains the charging time, charging capacity, charging average temperature and charging average voltage of each battery charge corresponding to each node in the target wireless sensor network within a period of time, which are marked as f represents the number of each charge, f = 1, 2, ..., g, g represents the total number of charge numbers;

[0067] The charging power of each battery corresponding to each node in the target wireless sensor network is matched with the reference charging time corresponding to each charging power, and the reference charging time of each battery corresponding to each node in the target wireless sensor network is obtained, which is recorded as

[0068] According to the formula

[0069] Calculate the battery charging health index corresponding to each node in the target wireless sensor network over a period of time in, They represent the set reference charging power, reference charging average temperature, and reference charging average voltage, respectively; c1, c2, c3, and c4 represent the proportion weights of the battery charging health index corresponding to the set charging time, charging power, charging average temperature, and charging average voltage, respectively;

[0070] Obtain in real time the discharge duration, discharge capacity, discharge mean temperature and discharge mean voltage of each battery discharge corresponding to each node in the target wireless sensor network within a period of time, marked as k represents the number of each charge, k = 1, 2, ..., l, l represents the total number of discharges;

[0071] The discharge power of each battery corresponding to each node in the target wireless sensor network is matched with the reference discharge duration corresponding to each discharge power, and the reference discharge duration of each battery corresponding to each node in the target wireless sensor network is obtained, which is recorded as

[0072] According to the formula, the battery discharge corresponding to each node in the target wireless sensor network within a period of time is calculated

[0073] Calculate the battery discharge health index corresponding to each node in the target wireless sensor network over a period of time in, They represent the set reference discharge capacity, reference discharge mean temperature, and reference discharge mean voltage respectively; c5, c6, c7, and c8 represent the proportion weights of the battery charging health index corresponding to the set discharge time, discharge capacity, discharge mean temperature, and discharge mean voltage respectively;

[0074] Real-time monitoring obtains the number of low battery warnings corresponding to each node in the target wireless sensor network within a period of time, recorded as YN i , conduct a comprehensive analysis of the above data, based on the set data model: Thus, the energy supply evaluation index η corresponding to each node in the target wireless sensor network within a period of time is obtained. i , where d1, d2, d3, and d4 represent the weights of the energy supply evaluation index corresponding to the set battery remaining power decrease rate, battery charging health index, battery discharging health index, and battery low power warning times, respectively.

[0075] In a specific embodiment, the present invention monitors and analyzes the rate of change of battery remaining power decrease, battery charging health index, battery discharge health index and battery low power warning times corresponding to each node in the target wireless sensor network, and then evaluates and analyzes the node energy supply status corresponding to each node. It can fully understand the energy supply situation of the node, timely discover energy supply shortage or battery failure problems, and effectively ensure the reliability and stability of the operation of the wireless sensor network.

[0076] S4, network communication detection and analysis: monitor the network communication parameters corresponding to the target wireless sensor network, and analyze the network communication evaluation index corresponding to the target wireless sensor network to obtain the network communication evaluation index corresponding to the target wireless sensor network. The specific analysis method is as follows:

[0077] Real-time monitoring obtains the number of node connection changes, average path length change rate and clustering coefficient change of the target wireless sensor network within a period of time, and marks them as N respectively. 变化 , R 变化 , C 变化 , according to the formula Calculate the network topology stability index λ corresponding to the target wireless sensor network within a period of time, where e1, e2, and e3 represent the weights of the network topology stability index corresponding to the set number of node connection changes, average path length change rate, and clustering coefficient change, respectively;

[0078] It should be noted that the number of node connection changes refers to the number of connection changes between all nodes and other nodes within a period of time. Frequent connection changes indicate unstable network topology. The average path length change rate refers to the change rate of the average path length between nodes in the wireless sensor network within a period of time. A larger average path length change rate means that the network topology has undergone significant changes. The clustering coefficient change refers to the difference between the clustering coefficient at the end of the monitoring time point and the clustering coefficient at the initial monitoring time point within a period of time. The clustering coefficient reflects the degree of clustering of nodes in the network. The larger the value of the clustering coefficient change, the greater the change in the network topology.

[0079] Real-time monitoring obtains the abnormal values ​​of data transmission rate, data packet size, and data packet number corresponding to the target wireless sensor network within a period of time, and marks them as Y1, Y2, and Y3 respectively. According to the formula Calculate the network traffic evaluation index γ corresponding to the target wireless sensor network, e4, e5, and e6 represent the weights of the network traffic evaluation index corresponding to the set data transmission rate abnormal value, data packet size abnormal value, and data packet number abnormal value, respectively;

[0080] It should be noted that the data transmission rate anomaly refers to the ratio of the number of abnormal data transmission rate occurrences corresponding to the target wireless sensor network in a period of time to the total number of occurrences; the data packet size anomaly refers to the ratio of the number of abnormal data packet size occurrences corresponding to the target wireless sensor network in a period of time to the total number of occurrences; the data packet number anomaly refers to the ratio of the number of abnormal data packets corresponding to the target wireless sensor network in a period of time to the total number of occurrences;

[0081] It should also be pointed out that the acquisition of data transmission rate anomalies, data packet size anomalies, and data packet number anomalies can all be collected through network traffic monitors, and feature extraction and analysis can be performed using deep learning technology. Deep learning mainly refers to the use of neural network models to detect and identify network traffic data anomalies. Deep learning is a branch of machine learning that uses multi-layer neural network models to simulate the human nervous system to achieve data learning and processing. In the network traffic data anomaly detection system, the ability to detect and identify network traffic data anomalies is improved by learning and training a large amount of network traffic data. Generally speaking, the deep learning process includes: data preparation, model construction, model training, model evaluation, and model application;

[0082] According to the formula Calculate the network communication evaluation index corresponding to the target wireless sensor network f1 and f2 represent the weights of the set network topology stability index and network traffic evaluation index corresponding to the network communication evaluation index respectively.

[0083] In a specific embodiment, the present invention monitors and analyzes the network topology stability and network traffic status corresponding to the target wireless sensor network, and then evaluates and analyzes the network communication status corresponding to the target wireless sensor network, thereby avoiding further paralysis of the overall network communication failure of the target wireless sensor network and effectively improving the timeliness of the response to the network communication failure.

[0084] S5. Fault analysis: Based on the node hardware status evaluation index and node energy supply evaluation index corresponding to each node in the target wireless sensor network and the network communication evaluation index corresponding to the target wireless sensor network, the fault type corresponding to the target wireless sensor network is analyzed to obtain the fault type corresponding to the target wireless sensor network. The specific analysis method is as follows:

[0085] The node hardware status evaluation index corresponding to each node in the target wireless sensor network is compared with the set node hardware status evaluation index threshold. If the node hardware status evaluation index corresponding to a node in the target wireless sensor network is less than the set node hardware status evaluation index threshold, the fault type corresponding to the node in the target wireless sensor network is determined to be a node hardware type.

[0086] The energy supply evaluation index corresponding to each node in the target wireless sensor network is compared with the set energy supply evaluation index threshold. If the energy supply corresponding to a node in the target wireless sensor network is less than the set energy supply evaluation index threshold, the fault type corresponding to the node in the target wireless sensor network is determined to be the node energy supply type.

[0087] The network communication evaluation index corresponding to the target wireless sensor network is compared with the set network communication evaluation index threshold. If the network communication evaluation index corresponding to the target wireless sensor network is less than the set network communication evaluation index threshold, the fault type corresponding to the target wireless sensor network is determined to be the network communication type.

[0088] S6. Fault processing: receiving the fault type corresponding to the target wireless sensor network, and performing corresponding processing based on the fault type corresponding to the target wireless sensor network.

[0089] In a specific embodiment, if the fault type corresponding to a node in the target wireless sensor network is a node hardware type, the node hardware corresponding to the node in the target wireless sensor network is abnormally displayed through the display in the fault handling terminal, specifically: "The node hardware corresponding to a node in the target wireless sensor network is defective, please repair it", and at the same time, the location of the node is extracted from the node location stored in the system repository and displayed accordingly.

[0090] If the fault type corresponding to a certain node in the target wireless sensor network is a node energy supply type, the node energy supply corresponding to the node in the target wireless sensor network will be abnormally displayed through the display in the fault handling terminal, specifically: "There is a defect in the node energy supply corresponding to a certain node in the target wireless sensor network, please repair it", and at the same time, the position of the node is extracted from the node position stored in the system repository and displayed accordingly.

[0091] If the fault type corresponding to the target wireless sensor network is a network communication type, the display in the fault handling terminal will display the abnormality of the network communication corresponding to the target wireless sensor network, specifically: "The network communication corresponding to the target wireless sensor network is defective, please repair it."

[0092] In a specific embodiment, corresponding prompt information is displayed according to the fault type of the wireless sensor network, which can accurately determine the fault type of the wireless sensor network, help technicians quickly locate the problem and take corresponding repair measures, effectively improve the fault maintenance efficiency to a certain extent, and provide strong guarantee for the stable operation of the wireless sensor network.

[0093] The above contents are merely examples and explanations of the structure of the present invention. The technicians in this technical field may make various modifications or additions to the specific embodiments described or replace them in a similar manner. As long as they do not deviate from the structure of the invention or exceed the scope defined by the claims, they should all fall within the protection scope of the present invention.

Claims

1. A wireless sensor network fault detection method based on deep learning, characterized in that: The following steps are involved: S1, data collection: collect the node hardware status parameters, node energy supply parameters and network communication parameters corresponding to each node in the target wireless sensor network, and perform node hardware status detection analysis, node energy supply detection analysis and network communication detection analysis respectively; S2, node hardware status detection and analysis: monitor the node hardware status parameters corresponding to each node in the target wireless sensor network, and analyze the node hardware status evaluation index corresponding to each node in the target wireless sensor network, and obtain the node hardware status evaluation index corresponding to each node in the target wireless sensor network; S3, node energy supply detection and analysis: monitor the energy supply parameters corresponding to each node in the target wireless sensor network, and analyze the energy supply evaluation index corresponding to each node in the target wireless sensor network, so as to obtain the energy supply evaluation index corresponding to each node in the target wireless sensor network; S4, network communication detection and analysis: monitoring the network communication parameters corresponding to the target wireless sensor network, and thereby analyzing the network communication evaluation index corresponding to the target wireless sensor network to obtain the network communication evaluation index corresponding to the target wireless sensor network; S5, fault analysis: analyzing the fault type corresponding to the target wireless sensor network based on the node hardware status evaluation index and the node energy supply evaluation index corresponding to each node in the target wireless sensor network and the network communication evaluation index corresponding to the target wireless sensor network, and obtaining the fault type corresponding to the target wireless sensor network; S6. Fault processing: receiving the fault type corresponding to the target wireless sensor network, and performing corresponding processing based on the fault type corresponding to the target wireless sensor network.

2. A wireless sensor network fault detection method based on deep learning according to claim 1, characterized in that: The node hardware status evaluation index corresponding to each node in the target wireless sensor network is analyzed in the following specific analysis method: Real-time monitoring obtains the sensor output data corresponding to each node in the target wireless sensor network within a period of time, which is recorded as i represents the number of each node, i = 1, 2, ..., n, n represents the total number of node numbers, j represents the number of each sensor output data, j = 1, 2, ..., m, m represents the total number of sensor output data numbers, and the average value is calculated to obtain the average value of sensor output data corresponding to each node in the target wireless sensor network within a period of time, which is recorded as SJ i , according to the formula Calculate the standard deviation σ of the sensor output data corresponding to each node in the target wireless sensor network within a period of time i , and according to the formula Calculate the coefficient of variation ν of the sensor output data corresponding to each node in the target wireless sensor network within a period of time i , and obtain the number of sensor output data jumps corresponding to each node in the target wireless sensor network in real time within a period of time, recorded as N i And the three data are analyzed in a formulaic manner according to the set digital model: Thus, the stability index WP of the sensor output data corresponding to each node in the target wireless sensor network within a period of time is obtained. i , where e represents a natural constant, σ', ν', and N' represent the set reference output data standard deviation, reference output data variation coefficient, and reference output data jump times, respectively; a1, a2, and a3 represent the set output data standard deviation, output data variation coefficient, and output data jump times, respectively, corresponding to the weight of the sensor output data stability index; Real-time monitoring obtains the peak CPU usage of each node in the target wireless sensor network within a period of time, recorded as At the same time, the memory occupancy rate of each node in the target wireless sensor network corresponding to each monitoring time point within a period of time is obtained in real time, and the maximum memory occupancy rate and the minimum memory occupancy rate are extracted from them, respectively, and recorded as At the same time, the monitoring time points corresponding to the maximum memory occupancy and the minimum memory occupancy are extracted, and then the monitoring interval duration corresponding to the maximum memory occupancy and the minimum memory occupancy is obtained, recorded as T, according to the formula Calculate the memory occupancy rate change rate of each node in the target wireless sensor network over a period of time i ; Real-time monitoring is used to obtain the response time and execution time of each task corresponding to each node in the target wireless sensor network within a period of time, and the average value is calculated to obtain the average response time and average execution time of each task corresponding to each node in the target wireless sensor network within a period of time, which are recorded as According to the set formula Calculate the task execution efficiency XL of each node in the target wireless sensor network within a period of time i , T 响应 '、T 执行 'represents the average response time and average execution time of the reference task respectively, ΔT 响应 , ΔT 执行 They are respectively represented by the set reference task average response time difference and reference task average execution time difference, a4 and a5 are respectively represented by the set task average response time and task average execution time corresponding to the weight of task execution efficiency; According to the formula Calculate the processor performance index CX of each node in the target wireless sensor network over a period of time i , where SY max ', ζ', XL' represent the set reference CPU usage peak value, reference memory usage change rate, and reference task execution efficiency, respectively; a6, a7, and a8 represent the set CPU usage peak value, memory usage change rate, and task execution efficiency corresponding to the weight of the processor performance index; Real-time monitoring obtains the corresponding signal strength mean, signal-to-noise ratio and packet error rate of each node in the target wireless sensor network over a period of time, and marks them as XD respectively. i , XB i , SC i , according to the formula Calculate the communication signal quality index ZP corresponding to each node in the target wireless sensor network within a period of time i , XD', XB', SC' represent the set reference signal strength mean, reference signal-to-noise ratio, and reference data packet error rate, respectively; b1, b2, and b3 represent the weights of the set signal strength mean, signal-to-noise ratio, and data packet error rate corresponding to the communication signal quality index, respectively; According to the formula Calculate the hardware status evaluation index φ of each node in the target wireless sensor network within a period of time i , b4, b5, and b6 represent the weights of the set sensor output data stability index, processor performance index, and communication signal quality index corresponding to the node hardware status evaluation index, respectively.

3. A wireless sensor network fault detection method based on deep learning according to claim 1, characterized in that: The energy supply evaluation index corresponding to each node in the target wireless sensor network is analyzed in the following specific analysis method: Real-time monitoring obtains the remaining battery power of the nodes at each time monitoring point within a period of time, and extracts the remaining battery power of the nodes at the first time monitoring point and the remaining battery power of the nodes at the last monitoring time point respectively. At the same time, the first time monitoring point and the last monitoring time point are extracted, and then the monitoring interval duration corresponding to the first time monitoring point and the last monitoring time point is obtained. A comprehensive analysis is performed to obtain the rate of change of the decrease speed of the remaining battery power of each node in the target wireless sensor network within a period of time. Real-time monitoring obtains the charging time, charging capacity, charging average temperature and charging average voltage of each battery charge corresponding to each node in the target wireless sensor network within a period of time, which are marked as f represents the number of each charge, f = 1, 2, ..., g, g represents the total number of charge numbers; The charging power of each battery corresponding to each node in the target wireless sensor network is matched with the reference charging time corresponding to each charging power, and the reference charging time of each battery corresponding to each node in the target wireless sensor network is obtained, which is recorded as According to the formula Calculate the battery charging health index corresponding to each node in the target wireless sensor network over a period of time in, They represent the set reference charging power, reference charging average temperature, and reference charging average voltage, respectively; c1, c2, c3, and c4 represent the proportion weights of the battery charging health index corresponding to the set charging time, charging power, charging average temperature, and charging average voltage, respectively; Similarly, the battery discharge health index corresponding to each node in the target wireless sensor network within a period of time is obtained by analysis Real-time monitoring obtains the number of low battery warnings corresponding to each node in the target wireless sensor network within a period of time, recorded as YN i , conduct a comprehensive analysis of the above data, based on the set data model: Thus, the energy supply evaluation index η corresponding to each node in the target wireless sensor network within a period of time is obtained. i , where d1, d2, d3, and d4 represent the weights of the energy supply evaluation index corresponding to the set battery remaining power decrease rate, battery charging health index, battery discharging health index, and battery low power warning times, respectively.

4. A wireless sensor network fault detection method based on deep learning according to claim 1, characterized in that: The network communication parameters corresponding to the target wireless sensor network are analyzed in the following specific manner: Real-time monitoring obtains the number of node connection changes, average path length change rate and clustering coefficient change of the target wireless sensor network within a period of time, and marks them as N respectively. 变化 , R 变化 , C 变化 , according to the formula Calculate the network topology stability index λ corresponding to the target wireless sensor network within a period of time, where e1, e2, and e3 represent the weights of the network topology stability index corresponding to the set number of node connection changes, average path length change rate, and clustering coefficient change, respectively; Real-time monitoring obtains the abnormal values ​​of data transmission rate, data packet size, and data packet number corresponding to the target wireless sensor network within a period of time, and marks them as Y1, Y2, and Y3 respectively. According to the formula Calculate the network traffic evaluation index γ corresponding to the target wireless sensor network, e4, e5, and e6 represent the weights of the network traffic evaluation index corresponding to the set data transmission rate abnormal value, data packet size abnormal value, and data packet number abnormal value, respectively; According to the formula Calculate the network communication evaluation index corresponding to the target wireless sensor network f1 and f2 represent the weights of the set network topology stability index and network traffic evaluation index corresponding to the network communication evaluation index respectively.

5. A wireless sensor network fault detection method based on deep learning according to claim 1, characterized in that: The fault type corresponding to the target wireless sensor network is analyzed based on the node hardware status evaluation index and the node energy supply evaluation index corresponding to each node in the target wireless sensor network and the network communication evaluation index corresponding to the target wireless sensor network. The specific analysis method is as follows: The node hardware status evaluation index corresponding to each node in the target wireless sensor network is compared with the set node hardware status evaluation index threshold. If the node hardware status evaluation index corresponding to a node in the target wireless sensor network is less than the set node hardware status evaluation index threshold, the fault type corresponding to the node in the target wireless sensor network is determined to be a node hardware type. The energy supply evaluation index corresponding to each node in the target wireless sensor network is compared with the set energy supply evaluation index threshold. If the energy supply corresponding to a node in the target wireless sensor network is less than the set energy supply evaluation index threshold, the fault type corresponding to the node in the target wireless sensor network is determined to be the node energy supply type. The network communication evaluation index corresponding to the target wireless sensor network is compared with the set network communication evaluation index threshold. If the network communication evaluation index corresponding to the target wireless sensor network is less than the set network communication evaluation index threshold, the fault type corresponding to the target wireless sensor network is determined to be the network communication type.

Images