A WSN clustering routing method based on measurement data similarity

Abstract

The present invention discloses a WSN clustering routing method based on measurement data similarity, which specifically includes the following steps: (1) designing an adaptive clustering algorithm to cluster nodes according to measurement value similarity; (2) designing a cluster head replacement method according to residual energy and transmission energy consumption, and dynamically selecting a cluster head to transmit data; (3) selecting a routing path for the cluster head according to a transmission number factor and a relative distance factor. The present invention provides a WSN clustering routing method based on measurement data similarity. From the perspective of node measurement data similarity, an adaptive clustering algorithm is designed to group nodes with high measurement value similarity into a cluster, and a node is selected from each cluster to represent the entire cluster to transmit data, while other nodes are dormant to save energy. A cluster head selection and routing method based on energy and distance is designed to balance node energy consumption.

Description

Technical Field

[0001] The invention belongs to the field of wireless sensor network routing protocols, and in particular relates to a WSN clustering routing protocol based on measurement data similarity. Background Art

[0002] Wireless Sensor Network (WSN) consists of a large number of small, low-cost, energy-constrained sensor nodes, and is currently widely used in military, medical, and industrial fields. The nodes in the network are powered by batteries and the power supply is difficult to replenish. After working for a certain period of time, the nodes will die due to energy exhaustion. The limited node energy affects the life cycle of the network, so reducing node energy consumption and extending the network life cycle have always been research hotspots in WSN. In wireless sensor networks, the large number of densely distributed nodes makes the measurement values ​​of some nodes have a high similarity. If all nodes transmit data to the base station, it will lead to the transmission of redundant data and generate additional energy consumption. At the same time, due to the different distances between the nodes and the base station, the node energy consumption is uneven, causing some nodes to die prematurely due to excessive energy consumption. Therefore, effectively removing the transmission of redundant data, avoiding excessive communication energy consumption, and balancing the energy consumption of nodes are urgent problems to be solved in WSN. Summary of the invention

[0003] To solve the above problems, the present invention provides a WSN clustering routing method based on similarity of measured data. First, an adaptive clustering algorithm is designed. The historical data of node measurements and a predefined error threshold are used to perform clustering according to the similarity of node measurement data. Then, a cluster head node is selected from each cluster to transmit data on behalf of the entire cluster, and other nodes enter a sleep mode to save energy. At the same time, in order to solve the problem that uneven node energy consumption in the network leads to premature death of nodes, the cluster head is dynamically updated in each cluster according to the remaining energy and transmission energy consumption of the node, and the transmission number factor and relative distance factor are designed to find data forwarding nodes for the cluster head node to balance the node energy consumption.

[0004] In order to achieve the above object, the present invention provides a WSN clustering routing method based on measurement data similarity, comprising the following steps:

[0005] Step 1: Design an adaptive clustering algorithm, use the preset error threshold and the historical data of node measurement to perform clustering operations on nodes, and group nodes with similar measurement values ​​into one cluster;

[0006] Step 2: Dynamically select a node in each cluster to transmit data on behalf of the entire cluster based on the remaining energy and transmission energy consumption of the node, and other nodes in the cluster sleep to save energy;

[0007] Step 3: Generate a node evaluation function based on the transmission number factor and the relative distance factor, select a routing path for the cluster head, and balance the node energy consumption.

[0008] In step 1, the adaptive clustering algorithm includes three definitions such as equations (1), (2), and (3):

[0009] Definition 1: The clustering algorithm is based on the similarity of the node measurement data to cluster. The measurement data includes temperature or humidity. The similarity of the data is defined by the error. The error formula is defined as follows:

[0010]

[0011] Where M ir is the data value measured by node i in round r, M jr is the data value measured by node j in the rth round, R is the total number of measurement rounds; δ ij is the root mean square error of the measured data between nodes i and j in R rounds, δ ij The smaller the value of , the higher the similarity of the measured data between nodes i and j;

[0012] Definition 2: Assume that DST (Data Similarity Threshold) is a predefined error threshold; based on formula (1), nodes are judged to be in the same cluster by the following formula:

[0013]

[0014] Where T ij When the value is 1, it means that nodes i and j are eligible to be grouped into the same cluster; otherwise, it means that nodes i and j cannot be grouped into the same cluster. The value of DST should be pre-set according to actual needs.

[0015] Definition 3: Based on formula (1), the merging formula between clusters is defined as follows:

[0016]

[0017] Where N i and N j are the number of nodes in cluster i and cluster j respectively, node p and node q are the nodes in cluster i and cluster j respectively; δ′ ij The smaller the value of is, the higher the similarity of the node measurement data in cluster i and cluster j is, and the greater the probability of merging is.

[0018] In step 1, based on formulas (1), (2), and (3), the execution steps of the adaptive clustering algorithm are as follows:

[0019] Step 1: Divide each node into an independent cluster; calculate the δ′ value between any two clusters and store it in the set δ′_set;

[0020] Step 2: Determine whether the set δ′_set is empty. If it is empty, jump directly to Step 5; otherwise, jump to Step 3;

[0021] Step 3: Find the minimum value δ′ from δ′_set ij ,judge Is the value of N i N j ; If yes, merge cluster i and cluster j and jump to Step 4; otherwise delete δ′ from δ′_set ij And jump to Step 2;

[0022] Step 4: Delete all δ′ values ​​related to cluster i or cluster j from δ′_set, calculate the δ′ value between the merged cluster and all other remaining clusters, store it in δ′_set, and jump to Step 2;

[0023] Step 5: Output the results.

[0024] In the step 2, the process of the cluster head selection method is as follows:

[0025] In the initialization phase after clustering is completed, the base station broadcasts a data packet to each node; after receiving the broadcast information, the node estimates its distance from the base station based on the received signal strength and direction, and then senses its own remaining energy and sends the energy and distance information to the base station; assuming that the size of the data packet transmitted by the node in each round is l bit, according to the remaining energy and distance information of the node, the base station performs the following calculations:

[0026]

[0027] Where E i represents the residual energy of node i, d iBS represents the distance between node i and the base station, E TX (l,d iBS ) represents the distance d from node i iBS The energy consumed by sending l bit of location data; E i With E TX (l,d iBS ) ratio ADTn i represents the maximum number of times node i can send data directly to the base station. The more remaining energy node i has, the less transmission energy consumption it consumes. i The larger the value, the more ADTn iis the data transmission amount of node i; assuming that node i is a node of cluster j, based on formula (4), the selection formula of cluster head node is defined as follows:

[0028]

[0029] Where Nj represents the number of nodes in cluster j, ADTc j represents the average data transmission amount of all nodes in cluster j, CHfit ij Indicates the fitness value of node i being selected as the cluster head node in cluster j; CHfit ij The higher the value, the greater the probability that node i is selected as the cluster head. The node with the largest CHfit value is selected as the cluster head node. If the maximum CHfit values ​​are equal, one of the nodes is randomly selected as the cluster head node. After determining the cluster head node of each cluster, the base station wakes up the node selected as the cluster head through broadcasting, and puts other nodes in the cluster into sleep mode to save energy. Then the base station transmits the ADTc value of its corresponding cluster to each cluster head node, and then enters the network operation phase. The cluster head node begins to transmit and forward data in rounds. At the beginning of each round, each cluster head node senses its own remaining energy and calculates its own data transmission volume based on formula (4), and then calculates the cluster head replacement coefficient:

[0030]

[0031] Where: CH i represents the cluster head node of cluster i, 0<λ1<1, CH_T i Represents the cluster head replacement coefficient. When CH_T i When the value of is 1, the cluster head node needs to be replaced. The cluster head node in this round sends its remaining energy information to the base station after the data packet, waiting for the base station to select a new cluster head node in the next round; otherwise, it transmits the data packet.

[0032] In step 3, the process of the cluster head routing method is as follows:

[0033] The cluster head node broadcasts its ADTn value to the network. After receiving information from other nearby cluster head nodes, each cluster head node selects a forwarding node according to the node evaluation function. Considering the remaining energy of the node, the distance from the base station, and the distance between nodes, the transmission number factor and the relative distance factor are designed to form the node evaluation function. The formula of the transmission number factor is defined as follows:

[0034]

[0035] In the formula and are the data transmission amounts of the cluster head nodes of cluster i and cluster j respectively; the smaller the data transmission amount of node i, the larger the data transmission amount of node j, and the greater the probability that node i chooses node j as a forwarding node; the relative distance factor is defined as follows:

[0036]

[0037] In the formula is the distance between the cluster head node of cluster i and the cluster head node of cluster j, is the distance between the cluster head node of cluster i and the base station; the closer the distance between node i and node j, the less energy is consumed, and the greater the probability that node i chooses node j as a forwarding node; combined with the transmission number factor and the relative distance factor, the node evaluation function is defined as follows:

[0038]

[0039] Where 0<α, β<1, and α+β=1. The node evaluation function comprehensively considers the data transmission volume of the node and the relative distance between the nodes, making the energy consumption between the nodes more balanced; after the node evaluation function is defined, the forwarding node is selected for the cluster head node by defining the threshold, and the formula is as follows:

[0040]

[0041] Where 0<λ2<1; Based on formula (10), each cluster head node i selects The cluster head node j with the largest value is used as the data forwarding node. When the value of is 0, cluster head node i directly transmits data to the base station.

[0042] The present invention provides a WSN clustering routing method based on measurement data similarity, which has the following beneficial effects:

[0043] The present invention designs an adaptive clustering algorithm to cluster nodes according to the similarity of measurement values, selects one node from each cluster as a cluster head to transmit data, and other nodes are in sleep mode, thereby avoiding the transmission of redundant data and saving network energy; designs a cluster head replacement method according to the remaining energy and transmission energy consumption, dynamically selects the cluster head to transmit data, balances the node energy consumption, and avoids the node from dying prematurely due to energy exhaustion; designs a transmission number factor and a relative distance factor, selects a routing path for the cluster head, balances the node energy consumption, and avoids the node from dying prematurely due to energy exhaustion. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 A topological diagram of the WSN clustering routing method based on measurement data similarity provided by the present invention.

[0045] Figure 2The present invention provides a flow chart of a WSN clustering routing method based on measurement data similarity. DETAILED DESCRIPTION

[0046] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments, but it should be understood that the protection scope of the present invention is not limited to the specific implementation methods, and all technologies implemented by the present invention should be within the protection scope.

[0047] like Figure 1 , Figure 2 As shown, the WSN clustering routing method based on measurement data similarity provided by the present invention includes the following steps:

[0048] Step 1: Design an adaptive clustering algorithm, use the preset error threshold and the historical data of node measurements to perform clustering operations on nodes, and group nodes with similar measurement values ​​into a cluster:

[0049] Figure 1 (a) shows the network structure of WSN, including several sensor nodes and an independent base station. Based on this network structure, the base station uses the preset error threshold and the historical data of node measurements to cluster the nodes. To perform the clustering operation, three definitions must be set in advance:

[0050] Definition 1: The clustering algorithm is based on the similarity of the node measurement data to cluster. The measurement data includes temperature or humidity. The similarity of the data is defined by the error. The error formula is defined as follows:

[0051]

[0052] Where M ir is the data value measured by node i in round r, M jr is the data value measured by node j in the rth round, and R is the total number of measurement rounds; assuming that the nodes in the network measure data every 5 minutes, and the historical measurement data is set to 5 days (7200 minutes), then the total number of rounds R is 7200 / 5=1440; δ ij is the root mean square error of the measured data between nodes i and j in R rounds, δ ij The smaller the value of , the higher the similarity of the measured data between nodes i and j.

[0053] Definition 2: Assume that DST (Data Similarity Threshold) is a predefined error threshold; based on formula (1), nodes are judged to be in the same cluster by the following formula:

[0054]

[0055] Where T ijWhen the value is 1, it means that nodes i and j are eligible to be grouped into the same cluster; otherwise, it means that nodes i and j cannot be grouped into the same cluster. The value of DST should be pre-set according to actual needs.

[0056] Definition 3: Based on formula (1), the cluster merging formula is defined as follows:

[0057]

[0058] Where N i and N j are the number of nodes in cluster i and cluster j respectively, node p and node q are the nodes in cluster i and cluster j respectively; δ′ ij The smaller the value of is, the higher the similarity of the node measurement data in cluster i and cluster j is, and the greater the probability of merging is.

[0059] After the above three definitions are completed, based on formulas (1), (2), and (3), the base station uses the preset error threshold and the historical data of node measurements to cluster the nodes. The clustering steps are as follows:

[0060] Step 1: Divide each node into an independent cluster; calculate the δ′ value between any two clusters and store it in the set δ′_set;

[0061] Step 2: Determine whether the set δ′_set is empty. If it is empty, jump directly to Step 5; otherwise, jump to Step 3;

[0062] Step 3: Find the minimum value δ′ from δ′_set ij ,judge Is the value of N i N j ; If yes, merge cluster i and cluster j and jump to Step 4; otherwise delete δ′ from δ′_set ij And jump to Step 2;

[0063] Step 4: Delete all δ′ values ​​related to cluster i or cluster j from δ′_set, calculate the δ′ value between the merged cluster and all other remaining clusters, store it in δ′_set, and jump to Step 2;

[0064] Step 5: Output the results.

[0065] like Figure 1As shown in (b), after the clustering algorithm is completed, the clustering result is obtained. Then the base station selects a node from each cluster as the cluster head to transmit data, and puts other nodes in the cluster to sleep to save network energy. However, the process of selecting nodes needs to be reasonably designed. If a node with low energy and far away from the base station is selected to transmit data, the node will die prematurely due to low residual energy and excessive transmission energy consumption. At the same time, since the transmission energy consumption of each cluster head node is different, if data is transmitted in a single-hop manner, some nodes far away from the base station will die prematurely due to excessive energy consumption. Therefore, the present invention designs a cluster head node selection and routing method for balancing node energy consumption.

[0066] Step 2: Dynamically select a node in each cluster to transmit data on behalf of the entire cluster based on the remaining energy and transmission energy consumption of the node, and other nodes in the cluster sleep to save energy:

[0067] In the initialization phase after clustering is completed, the base station broadcasts a data packet to each node; after receiving the broadcast information, the node estimates its distance from the base station based on the received signal strength and direction, and then senses its own remaining energy and sends the energy and distance information to the base station; assuming that the size of the data packet transmitted by the node in each round is l bit, according to the remaining energy and distance information of the node, the base station performs the following calculations:

[0068]

[0069] Where E i represents the residual energy of node i, d iBS represents the distance between node i and the base station, E TX (l,d iBS ) represents the distance d from node i iBS The energy consumed by sending l bit of location data; E i With E TX (l,d iBS ) ratio ADTn i represents the maximum number of times node i can send data directly to the base station. The more remaining energy node i has, the less transmission energy consumption it consumes. i The larger the value, the more ADTn i is the data transmission amount of node i; assuming that node i is a node of cluster j, based on formula (4), the selection formula of cluster head node is defined as follows:

[0070]

[0071] Where N j represents the number of nodes in cluster j, ADTc j represents the average data transmission amount of all nodes in cluster j, CHfit ijIndicates the fitness value of node i being selected as the cluster head node in cluster j; CHfit ij The higher the value, the greater the probability that node i is selected as the cluster head. The node with the largest CHfit value is selected as the cluster head node. If the maximum CHfit values ​​are equal, one of the nodes is randomly selected as the cluster head node. After determining the cluster head node of each cluster, the base station wakes up the node selected as the cluster head through broadcasting, and puts other nodes in the cluster into sleep mode to save energy. Then the base station transmits the ADTc value of its corresponding cluster to each cluster head node, and then enters the network operation phase. The cluster head node begins to transmit and forward data in rounds. At the beginning of each round, each cluster head node senses its own remaining energy and calculates its own data transmission volume based on formula (4), and then calculates the cluster head replacement coefficient:

[0072]

[0073] In the formula: In the formula: CH i represents the cluster head node of cluster i, 0<λ1<1, CH_T i Represents the cluster head replacement coefficient. When CH_T i When the value of is 1, the cluster head node needs to be replaced. The cluster head node in this round sends its remaining energy information to the base station after the data packet, waiting for the base station to select a new cluster head node in the next round; otherwise, it transmits the data packet. The dynamic replacement of cluster heads can balance the energy consumption of nodes in the cluster and prevent nodes from dying due to premature energy exhaustion.

[0074] Step 3: Design the transmission number factor and relative distance factor to generate the path evaluation function, select the routing path for the cluster head, and balance the node energy consumption:

[0075] The cluster head node broadcasts its ADTn value to the network. After receiving information from other nearby cluster head nodes, each cluster head node selects a forwarding node according to the node evaluation function. Considering the remaining energy of the node, the distance from the base station, and the distance between nodes, the transmission number factor and the relative distance factor are designed to form the node evaluation function. The formula of the transmission number factor is defined as follows:

[0076]

[0077] In the formula and are the data transmission amounts of the cluster head nodes of cluster i and cluster j respectively; the smaller the data transmission amount of node i, the larger the data transmission amount of node j, and the greater the probability that node i chooses node j as a forwarding node; the relative distance factor is defined as follows:

[0078]

[0079] In the formula is the distance between the cluster head node of cluster i and the cluster head node of cluster j, is the distance between the cluster head node of cluster i and the base station; the closer the distance between node i and node j, the less energy is consumed, and the greater the probability that node i chooses node j as a forwarding node; combined with the transmission number factor and the relative distance factor, the node evaluation function is defined as follows:

[0080]

[0081] Where 0<α, β<1, and α+β=1. The node evaluation function comprehensively considers the data transmission volume of the node and the relative distance between the nodes, making the energy consumption between the nodes more balanced; after the node evaluation function is defined, the forwarding node is selected for the cluster head node by defining the threshold, and the formula is as follows:

[0082]

[0083] Where 0<λ2<1; Based on formula (10), each cluster head node i selects The cluster head node j with the largest value is used as the data forwarding node. When the value of is 0, cluster head node i directly transmits data to the base station, such as Figure 1 (d) shows that cluster head nodes generate routing paths based on the number of data transmissions and relative distances, balancing the energy consumption between different cluster nodes and achieving energy balance for the entire network. Figure 2 The whole process of the invention is summarized. Figure 2 The base station first performs clustering operations on the nodes using the preset error threshold DST and the historical data of node measurements, then collects the energy and distance information of the nodes, and selects the appropriate node in each cluster to become the cluster head. After that, the base station wakes up the cluster head node and puts the non-cluster head nodes into a dormant state. Each cluster head node transmits data in rounds. At the beginning of each round, the cluster head node first determines whether the cluster head needs to be replaced, and then confirms its own data forwarding node by receiving broadcast information, and then forwards the data. If not all nodes are dead, the base station determines which cluster head needs to be replaced and selects a new cluster head; if all nodes are dead, it ends.

[0084] The above is only an example application of the present invention and cannot limit the application scope of the present invention. Any modification or replacement based on the present invention should be included in the protection scope of the present invention.

Claims

1. A WSN clustering routing method based on measurement data similarity, characterized by: Follow these steps in order: Step 1: Design an adaptive clustering algorithm, use the preset error threshold and the historical data of node measurement to perform clustering operations on nodes, and group nodes with similar measurement values ​​into one cluster; the execution steps of the adaptive clustering algorithm are as follows: Step 1: Divide each node into an independent cluster; calculate the δ′ value between any two clusters and store it in the set δ′_set; Step 2: Determine whether the set δ′_set is empty. If it is empty, jump directly to Step 5; otherwise, jump to Step 3; Step 3: Find the minimum value δ′ from δ′_set ij ,judge Is the value of N i N j ; If yes, merge cluster i and cluster j and jump to Step 4; On the contrary, delete δ′ from δ′_set ij And jump to Step 2; Step 4: Delete all δ′ values ​​related to cluster i or cluster j from δ′_set, calculate the δ′ value between the merged cluster and all other remaining clusters, store it in δ′_set, and jump to Step 2; Step 5: Output the results; Step 2: Dynamically select a node in each cluster as the cluster head based on the node's remaining energy and transmission energy consumption to transmit data on behalf of the entire cluster, while other nodes in the cluster sleep to save energy; the cluster head selection process is as follows: In the initialization phase after clustering is completed, the base station broadcasts a data packet to each node; after receiving the broadcast information, the node estimates its distance from the base station based on the received signal strength and direction, and then senses its own remaining energy and sends the energy and distance information to the base station; assuming that the size of the data packet transmitted by the node in each round is l bit, according to the remaining energy and distance information of the node, the base station performs the following calculations: Where E i represents the residual energy of node i, d iBS represents the distance between node i and the base station, E TX (l,d iBS ) represents the distance d from node i iBS The energy consumed by sending l bit of location data; ADTn i is the data transmission amount of node i, which represents the maximum number of times node i can send data directly to the base station. The more residual energy of node i, the less transmission energy consumption. i The larger the value is; assuming that node i is the node of cluster j, the selection formula of the cluster head node is defined as follows: Where N j represents the number of nodes in cluster j, ADTc j represents the average data transmission amount of all nodes in cluster j, CHfit ij Indicates the fitness value of node i being selected as the cluster head node in cluster j; CHfit ij The higher the value, the greater the probability that node i is selected as the cluster head. The node with the largest CHfit value is selected as the cluster head node. If the maximum CHfit values ​​are equal, one of the nodes is randomly selected as the cluster head node. After determining the cluster head node of each cluster, the base station wakes up the node selected as the cluster head through broadcasting and puts other nodes in the cluster into sleep mode. Then the base station transmits the ADTc value of its corresponding cluster to each cluster head node, and then enters the network operation phase. The cluster head node begins to transmit and forward data in rounds. At the beginning of each round, each cluster head node senses its own remaining energy and calculates its own data transmission volume based on formula (1), and then calculates the cluster head replacement coefficient: Where: CH i represents the cluster head node of cluster i, 0<λ1<1, CH_T i Represents the cluster head replacement coefficient. When CH_T i When the value of is 1, the cluster head node needs to be replaced. The cluster head node in this round sends its remaining energy information to the base station after the data packet, waiting for the base station to select a new cluster head node in the next round; otherwise, it transmits the data packet; Step 3: Generate a node evaluation function based on the transmission number factor and the relative distance factor, select a routing path for the cluster head, and balance the node energy consumption; the process of cluster head routing selection is as follows: The cluster head node broadcasts its ADTn value to the network. After receiving information from other nearby cluster head nodes, each cluster head node selects a forwarding node according to the node evaluation function. Considering the remaining energy of the node, the distance from the base station, and the distance between nodes, the transmission number factor and the relative distance factor are designed to form the node evaluation function. The formula of the transmission number factor is defined as follows: In the formula and are the data transmission amounts of the cluster head nodes of cluster i and cluster j respectively; the smaller the data transmission amount of node i, the larger the data transmission amount of node j, and the greater the probability that node i chooses node j as a forwarding node; The relative distance factor is defined as follows: In the formula is the distance between the cluster head node of cluster i and the cluster head node of cluster j, is the distance between the cluster head node of cluster i and the base station; the closer the distance between node i and node j, the less energy is consumed, and the greater the probability that node i chooses node j as a forwarding node; combined with the transmission number factor and the relative distance factor, the node evaluation function is defined as follows: Where 0<α, β<1, and α+β=1; the node evaluation function comprehensively considers the data transmission volume of the node and the relative distance between the nodes, so that the energy consumption between the nodes is more balanced; after the node evaluation function is defined, the forwarding node is selected for the cluster head node by defining the threshold, and the formula is as follows: Where 0<λ2<1; Based on formula (7), each cluster head node i selects The cluster head node j with the largest value is used as the data forwarding node. When the value of is 0, cluster head node i directly transmits data to the base station.

2. The WSN clustering routing method based on measurement data similarity according to claim 1, characterized in that: In the step 1, the adaptive clustering algorithm includes three definitions such as formulas (8), (9), and (10); Definition 1: The clustering algorithm is based on the similarity of the node measurement data to cluster. The measurement data is temperature or humidity. The similarity of the data is defined by the error. The error formula is defined as follows: Where M ir is the data value measured by node i in round r, M jr is the data value measured by node j in the rth round, and R is the total number of measurement rounds; δ ij is the root mean square error of the measured data between nodes i and j in R rounds, δ ij The smaller the value, the higher the similarity of the measured data between nodes i and j; Definition 2: Assume that DST (Data Similarity Threshold) is a predefined error threshold; based on formula (8), nodes are judged whether they are in the same cluster by the following formula: Where T ij When the value is 1, it means that nodes i and j are eligible to be grouped into the same cluster; otherwise, it means that nodes i and j cannot be grouped into the same cluster. The value of DST should be pre-set according to actual needs. Definition 3: Based on formula (8), the cluster merging formula is defined as follows: Where N i and N j are the number of nodes in cluster i and cluster j respectively, node p and node q are the nodes in cluster i and cluster j respectively; δ′ ij The smaller the value of is, the higher the similarity of the node measurement data in cluster i and cluster j is, and the greater the probability of merging is.

Images