Industrial internet multi-dimensional data acquisition method and system based on graph signal processing

Abstract

The invention provides an industrial internet multi-dimensional data acquisition method and system based on graph signal processing, and belongs to the technical field of network data, and the method comprises the steps: obtaining a combined graph adjacency matrix capable of representing the internal incidence relation of multi-dimensional data through the existing spatial connection relation and in combination with the characteristics of the data; obtaining a Laplacian eigenvector matrix through a combined graph adjacent matrix, and constructing a mapping relation of multi-dimensional graph signals; according to the mapping relation of the multi-dimensional graph signals, a sampling distance is defined, an IGDR algorithm is expanded through the defined sampling distance, and the condition of undistorted reconstruction of the original signals is obtained; and collecting multi-dimensional data according to the condition of undistorted reconstruction of the original signal. According to the method, a reconstruction algorithm is redesigned, sufficient conditions are obtained by utilizing a combination graph structure based on a clustering scene, and the problem of multi-dimensional data acquisition is solved.

Description

Technical Field

[0001] The present invention belongs to the field of network data technology, and in particular relates to an industrial Internet multidimensional data collection method and system based on graph signal processing. Background Art

[0002] Edge computing-based Industrial Internet of Things (ECIIoT) consists of edge computing servers and a large number of sensors deployed on the same side of the industrial environment to provide monitoring services. Due to the bottleneck of data transmission range, the edge computing architecture can reduce network data traffic and decision latency. However, due to the large number of multidimensional data sets generated by the high monitoring frequency, new data compression acquisition models are needed to overcome the limitations of resource-limited Industrial Internet of Things networks, which promotes the development of graph sampling. Summary of the invention

[0003] In view of the above-mentioned deficiencies in the prior art, the present invention provides an industrial Internet multidimensional data acquisition method and system based on graph signal processing, which solves the problem of multidimensional data acquisition.

[0004] In order to achieve the above objectives, on the one hand, the technical solution adopted by the present invention is: a method for collecting multidimensional data of industrial Internet based on graph signal processing, comprising the following steps:

[0005] S1. Through the existing spatial connection relationship and combined with the characteristics of the data itself, a composite graph adjacency matrix that can characterize the intrinsic correlation of the multidimensional data of the Industrial Internet is obtained;

[0006] S2. Obtain the Laplace eigenvector matrix by combining the graph adjacency matrix and construct the mapping relationship of the multi-dimensional graph signal;

[0007] S3, according to the mapping relationship of the multi-dimensional image signal, define the sampling distance, and use the defined sampling distance to expand the IGDR algorithm to obtain the condition for distortion-free reconstruction of the original signal;

[0008] S4. Based on the conditions for distortion-free reconstruction of the original signal, the industrial Internet multi-dimensional data that meets the conditions is collected to complete the collection of the industrial Internet multi-dimensional data based on graph signal processing.

[0009] The beneficial effects of the present invention are as follows: the present invention designs a reconstruction algorithm, and obtains sufficient conditions by utilizing a combined graph structure (i.e., step S3) based on an even clustering scenario. The present invention proposes a combined graph structure construction method for connected different variable data sets. Based on the redesigned recovery algorithm, sufficient conditions for perfect reconstruction in an even clustering scenario are given.

[0010] Furthermore, the mapping relationship of the multi-dimensional graph signal is expressed as follows:

[0011]

[0012] Among them, y i,r represents the mapping relationship of multi-dimensional graph signals, P represents the sampling operator, and f i,r represents the graph signal of the rth variable on the i-th vertex, U k The feature vector representing the structure of the combined graph, Represents the projection of the spatial graph signal to the eigenvector of the spatial Laplacian matrix.

[0013] Beneficial effect of the above further solution: The mapping relationship of the multi-dimensional graph signal of the present invention can effectively obtain the sampled graph signal.

[0014] Furthermore, the expression of the sampling distance is as follows:

[0015] s v =E(|C i,u | i∈K,u∈S )

[0016] Among them, s v represents the sampling distance, S represents the sampling set, K represents the index set of the cluster list, E represents the expected operator, i represents the i-th cluster, C i,u Represents the number of vertices in the cluster, and u represents the u-th sampling.

[0017] Beneficial effect of the above further solution: The present invention constructs a sampling cluster through the above formula, which is convenient for multi-dimensional data sampling.

[0018] Furthermore, the IGDR algorithm is extended using the defined sampling distance, and its expression is as follows:

[0019]

[0020] Among them, f k+1 represents the k+1th reconstructed signal, P w Represents the graph frequency cutoff operator, F and F -1 Respectively represent the graph Fourier transform and inverse transform, I represents the identity operator, P T represents the subsampling operator, P m represents the local mean diffusion operator, P b represents the global deviation diffusion operator, f k represents the kth reconstructed signal, represents the kth reconstruction residual.

[0021] Beneficial effect of the above further solution: The present invention can obtain the upper limit of the sampling distance by expanding the IGDR algorithm by utilizing the defined sampling distance.

[0022] Furthermore, the expression of the condition for reconstructing the original signal without distortion is as follows:

[0023]

[0024] Among them, s v represents the sampling distance, w represents the cutoff frequency of the graph, and They represent the minimum and maximum quadratic values ​​of the graphic signal respectively.

[0025] The beneficial effect of the above further solution is: according to the condition of distortion-free reconstruction of the original signal, the multi-dimensional data meeting the condition is collected.

[0026] On the other hand, the present invention provides an industrial Internet multidimensional data acquisition system based on graph signal processing, wherein the industrial Internet multidimensional data acquisition system is applied to an industrial Internet multidimensional data acquisition method, comprising:

[0027] The first processing module is used to obtain a combined graph adjacency matrix that can characterize the intrinsic correlation relationship of the multi-dimensional data of the Industrial Internet through the existing spatial connection relationship and the characteristics of the data itself;

[0028] The second processing module is used to obtain the Laplace eigenvector matrix by combining the graph adjacency matrix and construct the mapping relationship of the multi-dimensional graph signal;

[0029] The third processing module is used to define a sampling distance according to the mapping relationship of the multi-dimensional image signal, and use the defined sampling distance to expand the IGDR algorithm to obtain the condition for distortion-free reconstruction of the original signal;

[0030] The fourth processing module is used to collect the industrial Internet multi-dimensional data that meets the conditions according to the conditions of distortion-free reconstruction of the original signal, and complete the collection of the industrial Internet multi-dimensional data based on graph signal processing.

[0031] The beneficial effects of the present invention are as follows: the present invention designs a reconstruction algorithm, and obtains sufficient conditions by utilizing a combined graph structure (i.e., the third processing module) based on an even clustering scenario. The present invention proposes a combined graph structure construction method for connected different variable data sets. Based on the redesigned recovery algorithm, sufficient conditions for perfect reconstruction in an even clustering scenario are given. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 The figure is a flow chart of the method of the present invention.

[0033] Figure 2 A schematic diagram of the simulation effect of the convergence performance of the Pearson correlation coefficient.

[0034] Figure 3 This is a schematic diagram of the convergence performance simulation effect under the same standard Gaussian random distribution.

[0035] Figure 4 Schematic diagram of the simulation effect of the convergence performance of a linear coefficient.

[0036] Figure 5 It is a schematic diagram of the system structure of the present invention. DETAILED DESCRIPTION

[0037] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0038] Example 1

[0039] The edge computing-based Industrial Internet of Things (ECIIoT) consists of edge computing servers and a large number of sensors deployed on the same aspect of the industrial environment to provide monitoring services. Due to the bottleneck of data transmission range, the edge computing architecture can reduce network data traffic and decision delays. However, due to the high monitoring frequency, a large number of multivariate data sets are generated, and a new data compression acquisition model is needed to overcome the limitations of the resource-limited Industrial Internet of Things network. This invention proposes an industrial Internet multidimensional data acquisition method based on graph signal processing. The invention redesigns the reconstruction algorithm and obtains sufficient conditions by utilizing a combined graph structure based on even clustering scenarios. Figure 1 As shown, the present invention provides an industrial Internet multi-dimensional data collection method based on graph signal processing, and its implementation method is as follows:

[0040] S1. Through the existing spatial connection relationship and combined with the characteristics of the data itself, a composite graph adjacency matrix that can characterize the intrinsic correlation of the multidimensional data of the Industrial Internet is obtained;

[0041] In this embodiment, it is assumed that the spatial domain parameters are the key influencing factors for the formation of the graph structure. Therefore, different variable data sets have the same graph structure. Then, the edges where the combined graph structure appears are connected to the vertices of the different graph structures located at the same spatial position associated with the different variable data sets. The combined adjacency matrices A2, A3, and A4 associated with the two, three, and four variable data sets are shown below:

[0042]

[0043] Where A represents the adjacency matrix.

[0044] S2. Obtain the Laplace eigenvector matrix by combining the graph adjacency matrix and construct the mapping relationship of the multi-dimensional graph signal;

[0045] In this embodiment, the mapping relationship of the multi-dimensional graph signal is expressed as follows:

[0046]

[0047] Among them, y i,r represents the mapping relationship of multi-dimensional graph signals, P represents the sampling operator, and f i,r represents the graph signal of the rth variable on the i-th vertex, U k The feature vector representing the structure of the combined graph, Represents the projection of the spatial graph signal to the eigenvector of the spatial Laplacian matrix.

[0048] S3, according to the mapping relationship of the multi-dimensional image signal, define the sampling distance, and use the defined sampling distance to expand the IGDR algorithm to obtain the condition for distortion-free reconstruction of the original signal;

[0049] In this embodiment, the expression of the sampling distance is as follows:

[0050] s v =E(|C i,u | i∈K,u∈S )

[0051] Among them, s v represents the sampling distance, S represents the sampling set, K represents the index set of the cluster list, E represents the expected operator, i represents the i-th cluster, C i,u Represents the number of vertices in the cluster, and u represents the u-th sampling.

[0052] In this embodiment, the IGDR algorithm is extended using the defined sampling distance, and its expression is as follows:

[0053]

[0054] Among them, f k+1 represents the k+1th reconstructed signal, P w Represents the graph frequency cutoff operator, F and F -1 Respectively represent the graph Fourier transform and inverse transform, I represents the identity operator, P T represents the subsampling operator, P m represents the local mean diffusion operator, P b represents the global deviation diffusion operator, f k represents the kth reconstructed signal, represents the kth reconstruction residual.

[0055] In this embodiment, the condition for reconstructing the original signal without distortion is expressed as follows:

[0056]

[0057] Among them, sv represents the sampling distance, w represents the cutoff frequency of the graph, and They represent the minimum and maximum quadratic values ​​of the graphic signal respectively.

[0058] S4. Collect multivariate data based on the condition of distortion-free reconstruction of the original signal.

[0059] In this embodiment, Figure 2 As shown, the correlation error RE and the Pearson correlation coefficient PCC are used to test the predictive ability of the model. The Pearson correlation coefficient is usually used to measure the linear relationship (or linear association) between two random variables. Its principle is as follows:

[0060]

[0061]

[0062] Among them, R represents the predicted value, P represents the predicted value, and PCC represents the correlation between data under the same node. The Pearson correlation coefficient, also known as the product-moment correlation coefficient, is a common statistical measure of the association between variable sets and is used to analyze sampling performance. A coefficient of 1 indicates a perfect positive correlation or direct relationship, and a correlation coefficient of 0 indicates that there is no relationship between the two variable sets. Therefore, consider the Pearson correlation coefficient changing from 0 to 1 with a step size of 0.1. The sampling volume is set to 3 and the graph cutoff frequency is set to 0.25. The simulation results are shown in Figure 2 As shown, the left ordinate is the relative error, the right ordinate is the standard deviation, and the horizontal axis is the Pearson correlation coefficient. It can be seen that the three sampling sets have the same convergence performance, and as the Pearson correlation coefficient value increases, the relative error decreases.

[0063] In this embodiment, Figure 3 As shown in , it is assumed that both variable data sets follow a standard Gaussian random distribution. The sampling volume is set to 4 and the graph cutoff frequency is set to 0.25. The simulation results are shown in Figure 2 As shown, the left ordinate is the relative error, the right ordinate is the standard deviation, and the horizontal axis is the number of iterations. It can be found that all four sampling sets can reconstruct the graph signal and show similar convergence performance. The standard deviation decreases with the increase of the number of iterations. This means that based on the combined graph structure, when the two variable data sets follow the same Gaussian random distribution, the redesigned reconstruction algorithm can perfectly reconstruct the sampled graph signal.

[0064] In this embodiment, Figure 4 As shown in Figure 1, consider the linear coefficient between two variable data sets. The linear coefficient varies from 0.1 to 1 with a step size of 0.1. The sampling volume is set to 3 and the graph cutoff frequency is set to 0.25. The simulation results are shown in Figure 1. Figure 4As shown, the left ordinate is the relative reconstruction error, the right ordinate is the standard deviation, and the abscissa is the linear coefficient. It can be found that the three sampling groups can all restore the original graphic signal, and the higher the linear coefficient value, the smaller the relative error value.

[0065] Example 2

[0066] like Figure 5 As shown, the present invention provides an industrial Internet multidimensional data acquisition system based on graph signal processing, and the industrial Internet multidimensional data acquisition system is applied to the industrial Internet multidimensional data acquisition method, including:

[0067] The first processing module is used to obtain a combined graph adjacency matrix that can characterize the intrinsic correlation relationship of the multi-dimensional data of the Industrial Internet through the existing spatial connection relationship and the characteristics of the data itself;

[0068] The second processing module is used to obtain the Laplace eigenvector matrix by combining the graph adjacency matrix and construct the mapping relationship of the multi-dimensional graph signal;

[0069] The third processing module is used to define a sampling distance according to the mapping relationship of the multi-dimensional image signal, and use the defined sampling distance to expand the IGDR algorithm to obtain the condition for distortion-free reconstruction of the original signal;

[0070] The fourth processing module is used to collect the industrial Internet multi-dimensional data that meets the conditions according to the conditions of distortion-free reconstruction of the original signal, and complete the collection of the industrial Internet multi-dimensional data based on graph signal processing.

[0071] like Figure 5 The industrial Internet multi-dimensional data acquisition system provided by the illustrated embodiment can execute the technical solution shown in the industrial Internet multi-dimensional data acquisition method in the above-mentioned method embodiment 1. Its implementation principle and beneficial effects are similar and will not be repeated here.

[0072] In an embodiment of the present invention, the present application can divide the functional units according to the multi-dimensional data collection method of the industrial Internet. For example, each function can be divided into each functional unit, or two or more functions can be integrated into one processing unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of a software functional unit. It should be noted that the division of units in the present invention is schematic and is only a logical division. There may be other division methods in actual implementation.

[0073] In the embodiment of the present invention, in order to realize the principles and beneficial effects of the multi-dimensional data collection method of the industrial Internet, the industrial Internet multi-dimensional data collection system includes hardware structures and / or software modules corresponding to the execution of each function. Those skilled in the art should easily realize that, in combination with the various schematic units and algorithm steps described in the embodiments disclosed in the present invention, the present invention can be implemented in the form of hardware and / or a combination of hardware and computer software. Whether a function is executed in a hardware or computer software driven manner depends on the specific application and design constraints of the technical solution. Different methods can be used for each specific application to implement the described function, but such implementation should not be considered to exceed the scope of this application.

Claims

1. A method for collecting multidimensional data of industrial Internet based on graph signal processing, characterized in that: The following steps are involved: S1. Through the existing spatial connection relationship and combined with the characteristics of the data itself, a composite graph adjacency matrix that can characterize the intrinsic correlation of the multidimensional data of the Industrial Internet is obtained; S2. Obtain the Laplace eigenvector matrix by combining the graph adjacency matrix and construct the mapping relationship of the multi-dimensional graph signal; S3, according to the mapping relationship of the multi-dimensional image signal, define the sampling distance, and use the defined sampling distance to expand the IGDR algorithm to obtain the condition for distortion-free reconstruction of the original signal; S4. Based on the conditions for distortion-free reconstruction of the original signal, the industrial Internet multi-dimensional data that meets the conditions is collected to complete the collection of the industrial Internet multi-dimensional data based on graph signal processing.

2. The method for collecting multidimensional data of industrial Internet based on graph signal processing according to claim 1 is characterized in that: The expression of the mapping relationship of the multi-dimensional graph signal is as follows: Among them, y i,r represents the mapping relationship of multi-dimensional graph signals, P represents the sampling operator, and f i,r represents the graph signal of the rth variable on the i-th vertex, U k The feature vector representing the structure of the combined graph, Represents the projection of the spatial graph signal to the eigenvector of the spatial Laplacian matrix.

3. The method for collecting multidimensional data of industrial Internet based on graph signal processing according to claim 1 is characterized in that: The expression of the sampling distance is as follows: s v =E(|C i,u | i∈K,u∈S ) Among them, s v represents the sampling distance, S represents the sampling set, K represents the index set of the cluster list, E represents the expected operator, i represents the i-th cluster, C i,u Represents the number of vertices in the cluster, and u represents the u-th sampling.

4. The method for collecting multidimensional data of industrial Internet based on graph signal processing according to claim 1 is characterized in that: The IGDR algorithm is extended using the defined sampling distance, and its expression is as follows: Among them, f k+1 represents the k+1th reconstructed signal, P w Represents the graph frequency cutoff operator, F and F -1 Respectively represent the graph Fourier transform and inverse transform, I represents the identity operator, P T represents the subsampling operator, P m represents the local mean diffusion operator, P b represents the global deviation diffusion operator, f k represents the kth reconstructed signal, represents the kth reconstruction residual.

5. The method for collecting multidimensional data of industrial Internet based on graph signal processing according to claim 1 is characterized in that: The expression of the condition for the reconstruction of the original signal without distortion is as follows: Among them, s v represents the sampling distance, w represents the cutoff frequency of the graph, and They represent the minimum and maximum quadratic values ​​of the graphic signal respectively.

6. An industrial Internet multidimensional data acquisition system based on graph signal processing, characterized in that: The industrial Internet multi-dimensional data collection system is applied to the industrial Internet multi-dimensional data collection method, including: The first processing module is used to obtain a combined graph adjacency matrix that can characterize the intrinsic correlation relationship of the multi-dimensional data of the Industrial Internet through the existing spatial connection relationship and the characteristics of the data itself; The second processing module is used to obtain the Laplace eigenvector matrix by combining the graph adjacency matrix and construct the mapping relationship of the multi-dimensional graph signal; The third processing module is used to define a sampling distance according to the mapping relationship of the multi-dimensional image signal, and use the defined sampling distance to expand the IGDR algorithm to obtain the condition for distortion-free reconstruction of the original signal; The fourth processing module is used to collect the industrial Internet multi-dimensional data that meets the conditions according to the conditions of distortion-free reconstruction of the original signal, and complete the collection of the industrial Internet multi-dimensional data based on graph signal processing.

Images