CN101035041A - Node invalidation and pre-alarming method of radio sensor network based on Bays method - Google Patents

Abstract

The node failure early warning method of wireless sensor network based on Bayesian method is a node failure monitoring method in the aspect of network performance monitoring in wireless sensor network. It is mainly used to solve the judgment and prediction of node failure in wireless sensor network applications. By sending query commands to the network, the data returned by these commands are analyzed, and the returned data is corrected through different partition and classification models, and these data Through the comprehensive processing of partitions and classifications, these indicators are synthesized into a parameter inspection with a comprehensive model, and then according to the failure probability of the historical failure nodes in this area and the Bayesian method, it is obtained whether the current query node is facing failure and timely notification to the user, to achieve The purpose of node failure warning. It can well monitor network failure nodes and give early warning to network failure nodes, which provides convenience for network performance monitoring.

Description

Node failure early warning method for wireless sensor network based on Bayesian method

technical field

The invention relates to a node failure monitoring method in the aspect of network performance monitoring in a wireless sensor network. It is mainly used to solve the judgment and prediction of node failure in wireless sensor network applications, and belongs to the field of wireless sensor network applications and software development.

Background technique

A wireless sensor network is composed of a large number of wireless sensor nodes, which can perceive, collect data, process data, and transmit effective data to the destination autonomous network through a multi-hop network. Because of these advantages of wireless sensor networks and traditional networks, it is widely used in military, medical, environmental monitoring, cultural relics protection, smart home and other aspects.

At present, the application of wireless sensor network is being studied, and the research on software, hardware and software of wireless sensor network has made great progress. With the continuous deepening of the research, the management part of the wireless sensor network is now beginning to be studied, including node energy management, collaborative management, protocol management, network performance monitoring management, etc. Among them, network performance management is a very critical performance indicator in the research, which marks the status of the entire network. Its function is to monitor various network performance indicators in real time, and to provide early warning for abnormal conditions such as node function failure and energy exhaustion. , so as to provide help for timely troubleshooting of network faults or additional deployment of nodes. Abroad, the Embedded Networked Sensor Research Center of the University of California, Los Angeles proposed a network performance monitoring system consisting of three levels of software tools, which can continuously scan data related to network performance indicators from the network and aggregate them for processing. In terms of node energy, there is a comparison chart of the energy level of each area in the background. By analyzing the value of the index parameters in the current area and comparing it with the energy of the node in the area at the beginning, the network performance of the area is warned and processed. In China, the research on wireless sensor networks is gradually heating up. Most of the research focuses on the hardware and application software of wireless sensor networks, while the management of wireless sensor networks is still in its infancy. Research results related to node failure monitoring have not yet been published.

Contents of the invention

Technical problem: The purpose of this invention is to provide a node failure early warning method for wireless sensor networks based on Bayesian method, which provides support for the management of wireless sensor network performance and guarantees for the maintenance of wireless sensor networks.

Technical solution: The present invention is an early warning method for failure monitoring. By obtaining indicators (such as original energy, instant energy, residual energy, and communication volume) that affect node failure from the network, and then according to the historical node failure of the relevant area of the node Probability, using the Bayesian method to realize the early warning function. These concepts are described below.

Raw energy: the electric energy value of each node, the electric energy value of the node at the beginning of work, usually expressed in mAh.

Instant energy: After each node starts normal operation, when monitoring network performance, when the node receives different query instructions, the current energy value of the node.

Remaining energy: the remaining energy value of the node so far. This value is equal to the instantaneous energy of the node in certain cases, but in the model of this method, it will be calculated first and then estimated.

Traffic: The amount of information a node processes within a certain period of time, including the average amount of data sent and received.

The present invention is based on the node failure warning method of the wireless sensor network of Bayesian method, comprises the following steps:

Step 1). Define the partition strategy and partition the wireless sensor network. The definition here is to divide according to the geographical location. First, define the number of columns of the entire network node. Each row and column corresponds to a node and assigns a position, and then several columns of nodes into the same area;

Step 2). Determine the energy level of the wireless sensor network node;

Step 3). When the node starts to work, the original energy of the node is returned to the control processing end;

Step 4). The control processing end periodically sends request commands to query the real-time energy and traffic of the nodes;

Step 5). After a node in the area fails, the neighbor node reports the node number of its failed node, the area where the failed node is located, and the latest communication time of the failed node;

Step 6). According to the ID of the node sent back by the neighbors of the node in the area, the last communication with the neighbor node before the node failed, the original energy sent back by the node before the failure, the instant energy, and the calculation of the average communication volume of the area node in the original energy , instant energy, and the failure probability of the node when the average communication volume is this value; if there are multiple node failures, the average value should be used;

Step 7). The original energy, real-time energy, and communication volume of the district nodes returned from the current query, use the energy correction model "Ei=E(E1, E2, t)+ei", where i represents the area number, and Ei is the revised Energy value, while ei is a correction factor, E(E1, E2, t) is a functional relational expression and traffic correction model about original energy E1, instant energy E2, reference time difference t "Ti=T(T1, n)+ ti", where i represents the area number, Ti is the corrected energy value, and ti is the correction factor, T(T1, n) is a functional relational expression about the traffic T1 and the average processing capacity n;

Step 8). Use the comprehensive index evaluation model "A=Ei, Ti, f" to calculate the comprehensive reference value, where i represents the area number, Ei is the energy value after correction, Ti is the traffic after correction, and f is the correction factor ;

Step 9). According to the Bayesian method, it is estimated whether the current node is invalid,

Step 10). Once the failure of the node is estimated, the user can be notified with prompt information to realize early warning of node failure.

The partition strategy is as follows:

1): Define the number of columns of network nodes. The principle of definition is based on the number of columns of nodes, that is, how many columns of nodes are there in the horizontal direction, then the number of columns of network nodes is how many columns;

2): Determine the number of areas to be divided into, and each area takes a size of 3 rows and 3 columns as an area, that is, the total number of areas is the quotient obtained by dividing the total number of nodes by 9;

3): The number of columns of network nodes is divided by the area size to obtain the number of areas in each row;

4): Divide from the first column of the first row, find the nodes belonging to the first area, if the searched nodes exceed the number of columns in the first area, start from the first column of the second row to find the nodes belonging to the first area The nodes in the area until the searched nodes exceed the number of node columns in the first area; then divide the next line in the same way until all the nodes belonging to the first line are divided;

5): Starting from the first row of the start column number of the second area, search for nodes belonging to the second area, if the searched node exceeds the column number of the node in the second area, start from the second row, Proceed according to the above principle until all rows are divided;

6): If the current number of regions has been divided, stop the execution. Otherwise, go to 7)

7): Starting from the first row and the first column of the next region, start to divide the nodes of the region until the first row is divided, then divide the second row and the third row until all the rows are divided; repeat Execute 6);

The Bayesian method is a method of probability statistics. The Bayesian method is based on the current data and combined with historical data to infer the probability that this situation will occur at present, that is, the probability of event B _j occurring when event A occurs. In the case of , the probability of event A occurring when event B _i occurs can be calculated.

The energy level can indicate the energy corresponding to each level according to the energy expression characteristics of the battery: level five: 0-440mAh; level four: 441-880mAh; level three: 881-1320mAh; level two: 1321-1760mAh; level one: 1761-2200mAh .

The correction model is: when evaluating a node, it is corrected according to the performance indicators of the node's original energy, real-time energy, and communication traffic, and the correction model is used when correcting a single performance indicator, where the energy correction model is expressed as: Ei =E(E1, E2, t)+ei, i represents the area number, Ei is the corrected energy value, and ei is the correction factor, E(E1, E2, t) is about the original energy E1, the instant energy E2, Refer to the functional relational expression of the time difference t; when correcting the communication volume, the correction model Ti=T(T1,n)+ti is used, i represents the area number, Ti is the energy value after correction, and ti is the correction factor , T(T1, n) is a functional relational expression about the communication amount T1 and the average processing amount n.

The comprehensive evaluation model is: when evaluating a node, in order to better use the Bayesian method, multiple indicators of the query are processed in one, and only the comprehensive value of multiple performance indicators is considered. The comprehensive evaluation model introduced here is A=(Ei, Ti, f), i represents the area number, Ei is the energy value after correction, Ti is the communication volume after correction, and f is the correction factor.

Beneficial effect: using this method has the following advantages:

1. Self-monitoring network node failure

According to the set time or frequency, the network can be queried irregularly or periodically, and the performance indicators of all nodes in different areas in the network can be queried. Based on these indicators and the Bayesian method, it is possible to analyze whether all nodes in the current network have failed to deal with them.

2. Save energy

The analysis of the performance indicators of the entire network node is carried out in the background, not in the wireless sensor nodes with limited resources, which makes the nodes do not need to consume a lot of energy to calculate, analyze, count, and predict the failure of nodes.

3. Small error

The implementation of this method is based on certain historical statistics, and the posterior probability is derived from these prior probabilities. In this regard, although the results of historical data are of reference to the present, due to many network factors, this reference value will have certain errors. Therefore, the error reduction of these data is carried out. That is, the average value is taken by counting multiple data, which reduces the error.

4. Flexibility

It can flexibly configure the way of network collection indicators, which can be collected periodically or irregularly. It can collect the real-time energy of network nodes, and it is also possible to collect the regional location of network nodes. At the same time, the regional The middle nodes can be flexibly set, that is, the performance indicators of all nodes in an area can be collected within a certain period, and the performance indicators of some nodes can be collected regularly; some or all nodes in different areas can be collected regularly or irregularly. Node performance metrics.

5. Scalability

With the deepening of the research, it may be necessary to add network performance indicators, such as the sending data rate. Therefore, an extensible interface is left here, as long as the performance indicators that need to be counted are added to the comprehensive evaluation model.

Description of drawings

Figure 1 is a node failure architecture diagram of a Bayesian-based wireless sensor network. In the figure, there are wireless sensor network node areas A, B, C, D, and E.

Fig. 2 is a schematic diagram of partition classification processing.

Fig. 3 is a schematic diagram of comprehensive processing of divisions and classifications.

FIG. 4 is a flow chart of a Bayesian-based node failure method for a wireless sensor network.

Detailed ways

Figure 1 is a node failure architecture diagram of a Bayesian-based wireless sensor network. In the figure, wireless sensor network node areas (area A, area B, area C, area D, and area E, where A, B, C, D, and E are only symbols), and Fig. 2 is a schematic diagram of partitioning and hierarchical processing. The figure shows the preprocessing process of the group processing on the performance of the node failure, and performs partition and classification preprocessing on the query results to provide more reliable data for the partition and classification comprehensive processing.

Fig. 3 is a schematic diagram of comprehensive processing of divisions and classifications. The performance indicators returned from the query are unified, and the performance indicators related to node failure after partitioning and grading processing are corrected. Here, the integrated processing model is used to integrate the performance of a node, so that a comprehensive indicator can be used to reflect the performance of a node, and then The early warning of node failure can be realized by using the Bayesian method.

FIG. 4 is a flow chart of a Bayesian-based node failure method for a wireless sensor network.

Bayesian method: Bayesian method is a kind of probability and statistics method. The Bayesian method is based on the current data, combined with historical data, to infer the current probability of this situation happening. It can be described mathematically as follows: Let the sample space of experiment E be S, A be the event of E, B ₁ , B ₂ ,..., Bn be a division of S, and P(A)>0, P(Bi) >0 (i=1, 2, . . . , n), $P\left(A\right)=\underset{j=1}{\overset{\mathrm{no}}{\mathrm{\Σ}}}P\left(B_j\right|A)P\left(B_j\right),$ P(B _i |A) represents the probability of event B _i occurring when event A occurs, P(A|B _i ) represents the probability of event A occurring when event B _i occurs, and P(B _j |A) represents the probability of event A occurring The probability of event B _j occurring, the Bayesian formula is expressed as:

$\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; B</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; A</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; A</span>}<span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; B</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; B</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>}{\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; B</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; A</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; B</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; ,</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1,2</span>}<span\; class="notranslate">\; ,</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span>$

Energy and traffic correction model: The node failure early warning method of wireless sensor network based on the Bayesian method is corrected according to the performance indicators (raw energy, instant energy, and traffic) of the node when evaluating the node. The correction model is used when the performance index is corrected. The energy correction model is expressed as (Ei=E(E1, E2, t)+ei), i represents the area number, Ei is the energy value after correction, and ei is the correction factor. E(E1, E2, t) is a functional relational expression about the original energy E1, the instant energy E2, and the reference time difference t; when correcting the traffic, a correction model (Ti=T(T1,n)+ti) is used , i represents the area number, Ti is the corrected energy value, and ti is the correction factor, T(T1, n) is a functional relational expression about the traffic T1 and the average processing capacity n.

Comprehensive evaluation model: The node failure early warning method of wireless sensor network based on the Bayesian method. The composite value of the performance index. The comprehensive evaluation model introduced here is (A=(Ei, Ti, f)), where i represents the area number, Ei is the energy value after correction, Ti is the traffic volume after correction, and f is the correction factor.

Architecture: Figure 1 shows the architecture of a node failure early warning method for wireless sensor networks based on the Bayesian method. The entire wireless sensor network is logically divided into multiple regions, and the nodes in each region have similar characteristics, such as the equivalent traffic of the nodes, the geographical proximity of the nodes, and the similarity of the data obtained by the regions.

Wireless sensor network node area (area A, area B, area C, area D, area E, where A, B, C, D, E only represent symbols), request dispatch, dispatch reception, parameter configuration, partition classification processing, partition Hierarchical comprehensive processing, early warning of failed nodes, monitoring of failed nodes and historical database.

Parameter configuration: Complete the setting of the parameters required by this method, such as querying the network regularly or irregularly, if it is regular, how long is the regular time interval, whether to query for all areas or only for individual areas, is the query area All nodes or some nodes in the network, set the performance index of the query, whether to take the real-time energy and communication volume, or to take additional indicators, the time interval used when counting the communication volume, whether to monitor the number of sent data or the number of received nodes, Or consider both.

Request dispatch: responsible for assembling and sending the commands to be queried to the network. It involves the time interval of sending the request and the target node of the request query, etc.

Dispatch reception: Continuously receive query messages from the network, analyze the query results according to the set query indicators, and send the results to different processing modules according to different area numbers and real-time energy.

Partition hierarchical processing: each area has a corresponding processing model, according to the query message returned by the node in the current area, the energy level of the current query node is counted, and these parameter values are evaluated and corrected to get closer to the performance of the current node Condition. For example, by calculating the amount of information and predicting the energy of the node at the time of the current query and the average energy consumed by communication, the correction value of the current node energy is calculated. See Figure 2 for the processing flow chart.

Comprehensive processing of partition classification: the results of partition classification processing are handed over to partition classification comprehensive processing, and some indicators related to node performance are comprehensively inspected to obtain the comprehensive inspection value, and then according to the statistics of the partition classification history, the current node is facing failure. For the program, see Figure 3.

Failure node monitoring: Search the database at regular intervals to find nodes that have recently faced failure and nodes that have failed so far.

Failure node early warning: According to the results of failure node monitoring, it displays the nodes that are about to fail and the nodes that have failed.

Historical database: save some information related to nodes, area number, node number, original energy, real-time energy, traffic, comprehensive evaluation value, failure rate, original energy of regional failure nodes, real-time energy, comprehensive evaluation value and failure node The ratio in .

Method flow:

The method flow chart of the node failure early warning method for wireless sensor networks based on the Bayesian method is shown in the figure

4. The steps involved are as follows:

Step 1). Define the partition strategy and partition the wireless sensor network. The definition here is to divide according to the geographical location. First, define the number of columns of the entire network node, and assign a position to each row and column corresponding to the node, and then divide several columns of nodes into the same area;

Step 2). According to the energy value of the wireless sensor network node, it is divided into five energy levels (level five: 0-440mAh; level four: 441-880mAh; level three: 881-1320mAh; level two: 1321-1760mAh; level one : 1761-2200mAh);

Step 3). When the node starts to work, the original energy of the node is returned to the control processing end;

Step 4). The control processing end periodically sends request commands to query the real-time energy and traffic of the nodes;

Step 5). After a node in the area fails, the neighbor node reports the node number of its failed node, the area where the failed node is located, and the latest communication time of the failed node;

Step 6). According to the ID of the node sent back by the neighbors of the node in the area, the last communication with the neighbor node before the node failed, the original energy sent back by the node before the failure, the instant energy, and the calculation of the average communication volume of the area node in the original energy , instant energy, and the failure probability of the node when the average communication volume is this value; if there are multiple node failures, the average value should be used;

Step 7). The original energy, real-time energy, and communication volume of the district nodes returned from the current query, use the energy correction model "Ei=E(E1, E2, t)+ei", where i represents the area number, and Ei is the revised Energy value, while ei is a correction factor, E(E1, E2, t) is a functional relational expression and traffic correction model about original energy E1, instant energy E2, reference time difference t "Ti=T(T1, n)+ ti", where i represents the area number, Ti is the corrected energy value, and ti is the correction factor, T(T1, n) is a functional relational expression about the traffic T1 and the average processing capacity n;

Step 8). Use the comprehensive index evaluation model "A=Ei, Ti, f" to calculate the comprehensive reference value, where i represents the area number, Ei is the energy value after correction, Ti is the traffic after correction, and f is the correction factor ;

Step 9). According to the Bayesian method, it is estimated whether the current node is invalid;

Step 10). Notify the user of the estimated failure node with prompt information to realize the node failure early warning function.

For the convenience of description, we assume two indicators of node energy and traffic, and take the topology in Figure 1 as an example. Now we focus on the failure of nodes in Area A in Figure 1.

The specific implementation method is:

(1) Divide the wireless sensor network into six areas: A, B, C, D, E, and F;

(2) According to the energy value of wireless sensor network nodes, it is divided into five energy levels (level five: 0-440mAh; level four: 441-880mAh; level three: 881-1320mAh; level two: 1321-1760mAh;

(3) Level 1: 1761-2200mAh);

(4) When the node starts working, the original energy (889mAh) of the node is returned to the control processing end;

(5) The control processing end sends a request command to the nodes in the A area every 20 seconds to obtain the instant energy and communication volume of the nodes in the A area;

(6) According to the ID of the node sent back by the neighbor of node 5 in area A, the last communication with the neighbor node before the failure of the node, the original energy (889mAh) and the instant energy (881mAh, 870mAh, 790mAh) sent back by node 5 before the failure ...), the calculation of the average traffic (20p/usec) node failure probability (1.2% ); here is only the failure of one node 5, if there are multiple nodes failure, the average value should be used;

(7) The original energy (887mAh), real-time energy (870mAh), and communication volume (20p/usec) of node 3 in area A currently queried, using the energy correction model (Ei=E(887,870,t)+ei) And traffic correction model (Ti=T(20,n)+ti);

(8) Calculate the comprehensive reference value according to the comprehensive index evaluation model (A=(Ei, Ti, f));

(9) Utilize the result of (7), according to Bayesian method, estimate the failure probability of current node 5;

(10) Display and prompt the estimated failure nodes to the user, so as to achieve early warning of node failure.

Claims (6)

1. A node failure early warning method of a wireless sensor network based on a Bayesian method, characterized in that the method process comprises the steps: Step 1). Define the partition strategy and partition the wireless sensor network. The definition here is to divide according to the geographical location. First, define the number of columns of the entire network node. Each row and column corresponds to a node and assigns a position, and then several columns of nodes into the same area; Step 2). Determine the energy level of the wireless sensor network node; Step 3). When the node starts to work, the original energy of the node is returned to the control processing end; Step 4). The control processing end periodically sends request commands to query the real-time energy and traffic of the nodes; Step 5). After a node in the area fails, the neighbor node reports the node number of its failed node, the area where the failed node is located, and the latest communication time of the failed node; Step 6). According to the ID of the node sent back by the neighbors of the zone node, the last communication with the neighbor node before the node fails, the original energy sent back by the node before the failure, the instant energy, and the calculation of the average communication volume of the area node in the original energy , instant energy, and the failure probability of the node when the average communication volume is this value; if there are multiple node failures, the average value should be used; Step 7). The original energy, real-time energy, and communication volume of the district nodes returned from the current query, use the energy correction model "Ei=E(E1, E2, t)+ei", where i represents the area number, and Ei is the revised Energy value, while ei is a correction factor, E(E1, E2, t) is a functional relational expression and traffic correction model about original energy E1, instant energy E2, reference time difference t "Ti=T(T1, n)+ ti", where i represents the area number, Ti is the corrected energy value, and ti is the correction factor, T(T1, n) is a functional relational expression about the traffic T1 and the average processing capacity n; Step 8). Use the comprehensive index evaluation model "A=Ei, Ti, f" to calculate the comprehensive reference value, where i represents the area number, Ei is the energy value after correction, Ti is the traffic after correction, and f is the correction factor ; Step 9). According to the Bayesian method, it is estimated whether the current node is invalid, Step 10). Once the failure of the node is estimated, the user can be notified with prompt information to realize early warning of node failure. 2. the node failure early warning method of the wireless sensor network based on Bayesian method according to claim 1, it is characterized in that partition strategy is as follows: 1): Define the number of columns of network nodes. The principle of definition is based on the number of columns of nodes, that is, how many columns of nodes are there in the horizontal direction, then the number of columns of network nodes is how many columns; 2): Determine the number of areas to be divided into, and each area takes a size of 3 rows and 3 columns as an area, that is, the total number of areas is the quotient obtained by dividing the total number of nodes by 9; 3): The number of columns of network nodes is divided by the area size to obtain the number of areas in each row; 4): Divide from the first column of the first row, find the nodes belonging to the first area, if the searched nodes exceed the number of columns in the first area, start from the first column of the second row to find the nodes belonging to the first area The nodes in the area until the searched nodes exceed the number of node columns in the first area; then divide the next line in the same way until all the nodes belonging to the first line are divided; 5): Starting from the first row of the starting column number of the second area, search for nodes belonging to the second area, if the searched node exceeds the number of columns of nodes in the second area, start from the second row, Proceed according to the above principle until all rows are divided; 6): If the current number of regions has been divided, stop the execution. Otherwise, go to 7) 7): Starting from the first row and the first column of the next region, start to divide the nodes of the region until the first row is divided, then divide the second row and the third row until all the rows are divided; repeat Execute 6). 3. the node failure early warning method of the wireless sensor network based on Bayesian method according to claim 1, it is characterized in that Bayesian method is a kind of probability statistics method, and Bayesian method is exactly according to current data, in conjunction with history Data, infer the probability that this situation will occur at present, that is, when the probability of event B _j occurring when event A occurs, the probability of event A occurring when event B _i occurs can be calculated. 4. The node failure early warning method of wireless sensor network based on Bayesian method according to claim 1, characterized in that the energy level can represent the energy corresponding to each level according to the energy representation characteristics of the battery: five levels: 0-440mAh; Level 4: 441-880mAh; Level 3: 881-1320mAh; Level 2: 1321-1760mAh; Level 1: 1761-2200mAh. 5. the node failure early warning method of the wireless sensor network based on Bayesian method according to claim 1, it is characterized in that revision model is: when evaluating node, according to the original energy of node, instant energy, traffic The performance index has been corrected, and a correction model is used when correcting a single performance index, where the energy correction model is expressed as: Ei=E(E1, E2, t)+ei, i represents the area number, and Ei is the corrected energy value, while ei is the correction factor, E(E1, E2, t) is the functional relational expression about the original energy E1, the instant energy E2, and the reference time difference t; when correcting the traffic, the correction model Ti=T is used (T1, n)+ti, i represents the area number, Ti is the corrected energy value, and ti is the correction factor, and T(T1, n) is the functional relationship between the traffic T1 and the average processing capacity n. 6. the node failure warning method of the wireless sensor network based on Bayesian method according to claim 1, it is characterized in that comprehensive evaluation model is: when node is evaluated, in order to use Bayesian method better, to The multiple indicators of the query are processed together, and only the comprehensive value of multiple performance indicators is considered. The comprehensive evaluation model introduced here is A=(Ei, Ti, f), i represents the area number, and Ei is the energy value after correction. Ti is the revised traffic, and f is the correction factor.

Images