A critical link evaluation method in industrial wireless networks

Abstract

The present invention discloses a method for evaluating key links in an industrial wireless network, the method comprising: establishing a typical application network, measuring stability data through a stability measurement device and normalizing it; labeling the stability data according to three categories: link itself, between links, and between networks; establishing a deep autoencoding network to extract stability data features; establishing a deep neural network to give a correlation evaluation result based on the stability features of two links; calculating the average correlation of all links, and the largest one is the key link. The present invention implements the evaluation of key links in a multi-logic link industrial wireless network, and provides a method for improving a wireless network by stopping key links when the network is unstable.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial wireless networks, and in particular to a key link evaluation method in an industrial wireless network. Background Art

[0002] With the booming development of smart manufacturing and industrial Internet, industrial wireless networks have been widely used in the perception and control of various equipment. Due to the special requirements of industrial applications for stability, the stability of industrial wireless networks has always been a hot spot for technical research. Industrial wireless network stability technology can be divided into stability assessment and stability control according to the stage. Generally, after obtaining the stability assessment results, the network is controlled to improve stability. The current stability control method mainly improves network stability by improving the stability parameters of the logical link, such as extending the transmission time or reducing the transmission data. However, in the future industrial wireless network applications, there will be more and more logical links, and there may be a situation where no matter how the parameters are changed, the network cannot be improved. At this time, it is necessary to propose a method that can improve stability by stopping the least logical links. Summary of the invention

[0003] According to the deficiencies of the prior art, the present invention proposes a key link evaluation method in an industrial wireless network, which simplifies the logical links and effectively improves the stability of the network.

[0004] In order to solve the above technical problems, the technical solution of the present invention is:

[0005] A method for evaluating key links in an industrial wireless network comprises the following steps:

[0006] S100, link stability data measurement

[0007] A typical application network is established, and the stability data is measured and normalized using stability measurement equipment. The normalized stability data with a certain step length is used to characterize the original characteristics of a logical link.

[0008] S200, Data Labeling

[0009] Stability data is labeled into three categories: link itself, between links, and between networks;

[0010] S300, link feature extraction

[0011] Establish a deep autoencoder network to extract stability data features;

[0012] S400, link correlation coefficient calculation

[0013] S500, key link calculation

[0014] Calculate the average correlation of all links, and the largest one is the critical link.

[0015] Preferably, the stability data includes transmission time, update time, packet loss status, number of consecutive packet losses, and packet loss rate.

[0016] Preferably, the normalization processing method is expressed as follows:

[0017]

[0018] Among them, m t is the mask vector, A is the upper limit vector, B is the lower limit vector, l t The data of the transmission time, update time, packet loss status, number of consecutive packet losses, and packet loss rate are collected at time t.

[0019] Preferably, in step S300, the method for extracting stability data features is: compressing the original features of the logical link into stability features of a low-dimensional vector through a deep autoencoder network.

[0020] Preferably, in step S400, a fully connected neural network is established and an evaluation result of the correlation is given according to the stability characteristics of the two links.

[0021] Preferably, the first layer of the fully connected neural network is:

[0022] Among them, W0 and b0 are the network parameters of the first layer of the fully connected neural network, L′ i ,L′ j are the stability characteristics of link i and link j respectively, ReLU is the activation function, and its expression is: f(x)=max(0,x);

[0023] The intermediate layers of the fully connected neural network can be uniformly expressed as: k =ReLU(W k y k-1 +b k ),k∈{1,2,...,K}

[0024] Among them, W k and b k is the network parameter of the kth layer of the fully connected neural network;

[0025] The last layer of the fully connected neural network outputs the correlation coefficient, expressed as: r ij =σ(W K y K +b K )

[0026] Among them, σ is the Sigmoid activation function, which is in the form of:

[0027] Preferably, the last layer output of the fully connected neural network obtains a correlation coefficient in the range of 0-1.

[0028] Preferably, in step S500, the calculation method of the critical link is:

[0029]

[0030] Where J is the number of links, r max When it reaches the maximum value, it is a critical link.

[0031] The present invention has the following characteristics and beneficial effects:

[0032] By adopting the above technical solution, the key link is evaluated. Since the key link has the closest relationship with all other links, if the wireless network becomes unstable, the key link can be stopped first, thereby improving the stability of the wireless network to the greatest extent. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0034] Figure 1 The present invention provides a flow chart of the method.

[0035] Figure 2 A schematic diagram of the overall method provided in Embodiment 1 of the present invention.

[0036] Figure 3 The present invention provides a device description for the first embodiment.

[0037] Figure 4 This is a diagram of the feature extraction network architecture provided in Example 1 of the present invention.

[0038] Figure 5 This is a diagram of the correlation coefficient calculation network architecture provided in Example 1 of the present invention.

[0039] Figure 6 This is the calculation result of the 10-meter wireless network correlation coefficient provided in the first embodiment of the present invention.

[0040] Figure 7 This is the calculation result of the 100-meter wireless network correlation coefficient provided in the first embodiment of the present invention.

[0041] Figure 8 This is the calculation result of the 150-meter wireless network correlation coefficient provided in the first embodiment of the present invention. DETAILED DESCRIPTION

[0042] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0043] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and the like are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, features defined as "first", "second", and the like may explicitly or implicitly include one or more of the features. In the description of the present invention, unless otherwise specified, "multiple" means two or more.

[0044] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood by specific circumstances.

[0045] The present invention provides a method for evaluating key links in an industrial wireless network. Figure 1 As shown, the following steps are included:

[0046] S100, link stability data measurement

[0047] Establish a typical application network, measure stability data through stability measurement equipment and normalize it;

[0048] It can be understood that the collected stability data includes but is not limited to transmission time, update time, packet loss status, number of consecutive packet losses, and packet loss rate, and the collected data contains dirty data, so the upper and lower limit vectors A and B are set to filter abnormal data and normalize the data to the range of 0 to C. Assume that the above 5-dimensional data l is collected at time tt Then, the normalization processing method is expressed as follows:

[0049]

[0050] Among them, m t is the masking vector, which corresponds to l t For each dimension, if one dimension is abnormal data, then l t The corresponding dimension value in is 0, otherwise it is 1.

[0051] In this embodiment, the stability data change of a certain time step N is used as the original feature of each logical link, denoted as L i , L i =[l1,l2,...,l N ].

[0052] S200, Data Labeling

[0053] Stability data is labeled into three categories: link itself, between links, and between networks;

[0054] Specifically, the labeled values ​​are recorded as vectors α, β, and γ. Further, subscripts can be used to distinguish each internal value, for example, α1 represents the label value of the link and the link itself at the same time, and α2 represents the label value of the link and the link itself at different times.

[0055] S300, link feature extraction

[0056] Establish a deep autoencoder network to extract stability data features;

[0057] Specifically, the original features of the logical link are compressed into the stability features of a low-dimensional vector through a deep autoencoder network.

[0058] Furthermore, the deep autoencoder network structure is as follows Figure 4 As shown in Figure 1, the loss function is the mean square error between the input original features and the network output. Through this network, the original features of the link with a step size of 10 and a dimension of 5 are compressed to a 5-dimensional vector, thereby reducing the amount of calculation and reducing the solution space for the subsequent correlation calculation.

[0059] S400, link correlation coefficient calculation

[0060] By establishing a fully connected neural network and based on the stability characteristics of the two links, the correlation evaluation results are given.

[0061] Specifically, Figure 5 As shown, the first layer of the fully connected neural network is: y1 = ReLU (W0 [L ′ i ,L′ j ]+b0)

[0062] Among them, W0 and b0 are the network parameters of the first layer of the fully connected neural network, L i ′,L j ′ are the stability characteristics of link i and link j respectively, ReLU is the activation function, and its expression is: f(x) = max(0,x);

[0063] The intermediate layers of the fully connected neural network can be uniformly expressed as: k =ReLU(W k y k-1 +b k ),k∈{1,2,...,K}

[0064] Among them, W k and b k is the network parameter of the kth layer of the fully connected neural network;

[0065] The last layer of the fully connected neural network outputs the correlation coefficient, expressed as: r ij =σ(W K y K +b K )

[0066] Among them, σ is the Sigmoid activation function, which is in the form of:

[0067] The last layer output of the fully connected neural network obtains a correlation coefficient in the range of 0-1.

[0068] S500, key link calculation

[0069] Calculate the average correlation of all links, and the largest one is the critical link.

[0070] Specifically, the calculation method of the key link is:

[0071]

[0072] Where J is the number of links, r max When it is the maximum value, it is the critical link. max The link number can be obtained by using the link represented by . Since the key link has the closest relationship with all other links, if the wireless network becomes unstable, the key link can be stopped first. In particular, the order of the correlation coefficients of all links is also known in the calculation of the maximum average correlation. Therefore, if there is a first key link that cannot be stopped, the links ranked later can be considered in turn.

[0073] Furthermore, the above technical solution is applied to provide the following specific embodiments:

[0074] In this embodiment, the method of the present invention is described by evaluating the stability of three wireless networks, where: Figure 2 and Figure 3 As shown, each wireless network has three wireless devices, and the distances between the wireless devices are 10 meters, 100 meters, and 150 meters respectively. The devices are transmitted via Bluetooth 4.0. The data collection interval is 0.131 milliseconds. Each network obtains 10,000 pieces of data. In this embodiment, first, 10,000 data in each link are divided into 9990 groups according to a step size of 10 units, and 9 sequences containing 9990 groups are obtained, among which sequence numbers 0-2 are 10-meter grouping sequences, sequence numbers 3-5 are 100-meter grouping sequences, and sequence numbers 6-8 are 150-meter grouping sequences. After the grouping is completed, all the data are stored in the array S in order, and the three sequences of the same network category are combined in pairs. In this embodiment, the combination steps mainly include:

[0075] 1. Set a combination upper limit, that is, the maximum number of random combinations generated between two sequences. In this embodiment, the combination upper limit is set to 1000;

[0076] 2. Select two sequences to be combined, and set their combination index to the sequence start value of the first sequence (for example, the sequence start value of sequence number 0 is 0, and the sequence start value of sequence number 1 is 9990);

[0077] 3. The combination number is increased by a random step between 0 and 20 each time to obtain a new combination number, which corresponds to the subscript of the first data in the combination in array S, and the array subscript of the second data comes from the position of the corresponding combination number of another sequence (for example, sequence 0 is combined with sequence 2, and the combination number is 10, then the data s[0*9990+10], s[2*9990+10] are used as the combination);

[0078] 4. Repeat step 3 until the number of generated combinations reaches the upper limit of combinations set in step 1;

[0079] After pairing the same network data sequences in pairs according to the above steps, 9,000 records can be obtained for each of the three categories of 10 meters, 100 meters, and 150 meters, totaling 27,000 records. Then the 9,000 records in each category are randomly divided into 7,200 training set data for training the model and 1,800 test data for testing the model effect. The training set data of the three categories totals 21,600 as a complete model training set. After completing the model training, for each category, the model is used to test its 1,800 test set data to evaluate the actual effect of the model.

[0080] In the model training, the two networks set the same training parameters, namely, learning rate 0.001, batch size 100, and training round number 1000. The correlation coefficient calculation results of the three networks are as follows Figure 6-Figure 8 As shown. Further calculation of the average correlation coefficient shows that if the network becomes unstable, link 3 in network 1, link 1 in network 2, and link 3 in network 3 can be stopped first to improve the stability of the wireless network to the greatest extent.

[0081] The embodiments of the present invention are described in detail above with reference to the accompanying drawings, but the present invention is not limited to the described embodiments. For those skilled in the art, various changes, modifications, substitutions and variations of these embodiments including components are made without departing from the principles and spirit of the present invention, and still fall within the scope of protection of the present invention.

Claims

1. A method for evaluating key links in an industrial wireless network, characterized in that: The steps include: S100, link stability data measurement Establish a typical application network, measure and normalize the stability data using stability measurement equipment, and use the normalized stability data with a certain step length to characterize the original characteristics of a logical link; S200, Data Labeling Stability data is labeled into three categories: link itself, between links, and between networks; S300, link feature extraction Establish a deep autoencoder network to extract stability data features; S400, link correlation coefficient calculation By building a fully connected neural network and based on the stability characteristics of the two links, the correlation evaluation results are given; The first layer of the fully connected neural network is: Among them, W0 and b0 are the network parameters of the first layer of the fully connected neural network, L i ′,L j ′ are the stability characteristics of link i and link j respectively, ReLU is the activation function, and its expression is: f(x) = max(0,x); The intermediate layers of the fully connected neural network can be uniformly expressed as: k =ReLU(W k y k-1 +b k ),k∈{1,2,...,K}where W k and b k is the network parameter of the kth layer of the fully connected neural network; The last layer of the fully connected neural network outputs the correlation coefficient, expressed as: r ij =σ(W K y K +b K ) where σ is the Sigmoid activation function, which is in the form of: S500, key link calculation Calculate the average correlation of all links. The largest one is the critical link. The calculation method of the critical link is: Where J is the number of links, r max When it reaches the maximum value, it is a critical link.

2. The critical link evaluation method in an industrial wireless network according to claim 1, characterized in that: The stability data includes transmission time, update time, packet loss status, number of consecutive packet losses, and packet loss rate.

3. The critical link evaluation method in an industrial wireless network according to claim 2 is characterized in that: The normalization processing method is expressed as follows: l t =[1-diag -1 (B-A)(x t -A)⊙(1-C)]⊙m t Among them, m t is the mask vector, A is the upper limit vector, B is the lower limit vector, l t The data of the transmission time, update time, packet loss status, number of consecutive packet losses, and packet loss rate are collected at time t.

4. The critical link evaluation method in an industrial wireless network according to claim 2, characterized in that: In step S300, the method for extracting stability data features is: compressing the original features of the logical link into stability features of a low-dimensional vector through a deep autoencoder network.

5. The critical link evaluation method in an industrial wireless network according to claim 4, characterized in that: The last layer output of the fully connected neural network obtains a correlation coefficient in the range of 0-1.

Images