Key component identification method and system for multiple networked industrial software

Abstract

The invention discloses a key component identification method and system for multiple networked industrial software, and the method comprises the steps: firstly calculating the influence radius of all components on each network layer; calculating the importance degree # imgabs0 # of each component based on the dynamic influence range, then calculating the similarity between multiple feature vectors of each component on different relation network layers by using a cosine similarity algorithm; # imgabs1 #, calculating the association strength between different components on different relation network layers by using a hidden variable model; # imgabs2 #, and then calculating the importance degree LIalpha of each network layer G alpha; and finally, constructing a gravity centrality model based on the importance degree of the dynamic influence range # imgabs3 # and the importance degree LIalpha of the association strength # imgabs4 # network layer G alpha between different components, and calculating the importance degree of each component. And sorting is carried out according to the importance degree values, so that the key components on the multiple networked industrial software system are identified. The method disclosed by the invention is efficient and reliable, and the identification capability of the key component on the multi-networked industrial software system is enhanced.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent computing, and mainly relates to a key component identification method and system for multi-networked industrial software. Background Art

[0002] Key node identification is one of the most important research issues in network science. Key component identification on multi-networked industrial software systems refers to modeling different components into network nodes in an industrial software component network based on the association and coupling of multiple data stream types, and mining components that have an important impact on the operational stability of the multi-networked industrial software system based on the topological connection relationship between components. This result is crucial to exploring the stability of multi-networked industrial software systems. For example, operational errors in a few components with greater influence may lead to large-scale cascading failures of other components in a multi-networked industrial software system, and even cause the collapse of the entire industrial software system, seriously affecting industrial production efficiency. Therefore, accurate and efficient key component identification methods are of great significance to the stable operation of components on the entire industrial software system.

[0003] The core idea of ​​the traditional identification method for key components in industrial software systems is to consider the information of the component itself and other related components within its influence range, and calculate the importance of each component by modeling the interaction between them. In addition, most key component identification methods are based on single-layer industrial software systems for research. At present, industrial software systems show the characteristics of multiplicity of relationships between components, that is, different software components are simultaneously affected by multiple networks of associated coupling such as information flow, control flow, and business flow, corresponding to multiple different relationship networks. Abstracting such a complex industrial software system into a single-layer network will ignore the characteristics of different network layers themselves and their impact on the entire industrial software system. Therefore, the simple single-layer network method is no longer suitable for the identification of key components in industrial software systems. In order to effectively and reasonably describe industrial software with multiple relationships between different components, it is necessary to model the different relationships between components in the industrial software system, and calculate the importance of each component based on the importance of components under different relationships and the impact of different relationship networks on the entire industrial software system.

[0004] However, most of the existing studies on multiple networks sum up the key component identification results of each layer, and do not further consider the impact of different network relationships on key component identification. In addition, the existing methods are based on the multi-topological structure of the component for research, ignoring the use of multiple feature vectors of different components. For different industrial components, there are large differences in their scope of application, execution task category, carrying capacity, etc., and the same components on different network layers also have differences due to different characteristics such as task division and operating environment. Therefore, it is necessary to consider using different multiple feature vectors to represent industrial software components on different network layers. However, existing studies have ignored the use of component characteristics. At the same time, the greater the association strength between components, the higher the degree of attraction between them. When calculating the importance of components, it is necessary not only to consider the properties of the component itself, but also the association strength with other components. The greater the association strength, the higher the importance between components. For a specific component, components with higher association strength should be given higher weights when calculating importance. However, existing studies have ignored the modeling of this important phenomenon. Summary of the invention

[0005] The present invention aims at the problems existing in the prior art in identifying key components in multi-networked industrial software systems, and proposes a key component identification method and system for multi-networked industrial software. First, the influence radius of all components on each network layer is calculated, and the importance of each component based on the dynamic influence range is calculated. The cosine similarity algorithm is then used to calculate the similarity between the multiple feature vectors of each component at different relational network layers. Use latent variable models to calculate the strength of association between different components at different relationship network layers Then calculate each network layer G α Importance of LI α ; Finally, based on the importance of the dynamic impact range The strength of the associations between different components Network Layer G α The importance of LI α A gravity centrality model is constructed to calculate the importance of each component. The components are sorted according to the importance values, thereby identifying the key components on the multi-networked industrial software system. The method of the invention is efficient and reliable, and enhances the ability to identify key components on the multi-networked industrial software system.

[0006] In order to achieve the above object, the technical solution adopted by the present invention is: a key component identification method for multi-networked industrial software, comprising the following steps:

[0007] S1: Input the components into a multi-networked industrial software system G = {G1, G2, …, G L}, each network layer in the multi-networked industrial software system includes N identical components;

[0008] S2: Calculate the influence radius of all components on each network layer, and determine the influence of the component together with the average influence radius of its neighboring components. Calculate the importance of each component based on the dynamic influence range.

[0009] S3: Calculate the similarity between multiple feature vectors of each component at different relational network layers, and calculate the similarity by using the cosine similarity algorithm get;

[0010] S4: Using latent variable models to calculate the strength of association between different components at different relational network layers The strength of the association Based on the similarity of step S3 Sure;

[0011] S5: Calculate each network layer G α The importance of LI α It is obtained by comparing the ratio of the number of edges in this network layer to the total number of edges in the multi-network;

[0012] S6: Components obtained according to step S2 Importance based on dynamic impact range The correlation strength between different components obtained in step S4 The network layer G obtained in step S5 α The importance of LI α , the gravity centrality model is constructed as follows:

[0013]

[0014] in, Representation Components and At the network layer The shortest path length on ;

[0015] The importance of each component is as follows:

[0016]

[0017] S7: Sort the importance values ​​of the components obtained in step S6 to identify key components on the multi-networked industrial software system.

[0018] As an improvement of the present invention, in step S1, the network layer G of the multi-networked industrial software system α Components on and The multiple topological structures between are expressed as:

[0019]

[0020] in, Represents the network layer G α The i-th component on Representation Components and At the network layer G α A connection exists on the

[0021] As an improvement of the present invention, step S2 specifically includes the following steps:

[0022] S21: Introducing the iterative k-shell algorithm and gravity centrality model, defining nodes For any other node The applied gravity centrality is:

[0023]

[0024]

[0025] in, Respectively represent nodes and The degree, express and The shortest distance between represents the gravity coefficient, Representation Node The iterated k-shell value, Iks max , Iks min Represents the network layer G α Maximum and minimum iterative k-shell values ​​at midpoints;

[0026] For nodes and The dividing point Define nodes separately and For Node Applied gravity centrality:

[0027]

[0028]

[0029] S22: According to the principle of force balance, the node and For Node The applied gravitational forces are equal, then:

[0030]

[0031] S23: Using the above equation Solve and get the node Influence radius The solution formula is:

[0032]

[0033] S24: For each component Calculate importance based on dynamic impact range

[0034]

[0035] in, Representation Components The number of neighbors.

[0036] In order to achieve the above-mentioned purpose, the technical solution adopted by the present invention is: a key component identification system for multi-networked industrial software at least includes a component impact range calculation module, an association strength calculation module, a network layer importance calculation module and a component importance comprehensive evaluation module;

[0037] The component impact range calculation module: based on the topological connection between components in the entire multi-networked industrial software system, calculates the impact range of each component and obtains the importance of each component based on the dynamic impact range;

[0038] The association strength calculation module combines multiple feature vectors and multiple topological structures of components, and solves the association strength of different relationship network layers between components by constructing a latent variable model;

[0039] The network layer importance calculation module is used to calculate the importance of the network layer to the entire industrial software system for each different relationship network by considering the impact of different relationship networks on the network structure of the entire industrial software system;

[0040] The component importance comprehensive evaluation module constructs a weight-based gravity centrality model, combines the calculated association strength, component centrality based on the scope of influence, and network layer importance to obtain the overall importance of each component, thereby identifying key components on multiple networked industrial software systems.

[0041] Compared with the prior art, the present invention has the following beneficial effects:

[0042] (1) The method of the present invention combines the multiple feature vectors and multiple topological structures of each component in a multi-networked industrial software system for the first time, and infers the correlation strength between different components in the multi-networked industrial software system by constructing a latent variable model.

[0043] (2) The present invention models the relationship types between industrial software components in a multi-networked environment, and quantifies the influence of different relationship types on the importance of industrial software components. In the process of key component identification, the influence of different relationship networks on the overall multi-networked industrial software system structure is considered, and the importance of each different relationship network to the multi-networked industrial software system is calculated, thereby enhancing the ability to distinguish different network layers in the component relationship network.

[0044] (3) The present invention constructs an improved weighted power model, which can more accurately reflect the degree of attraction between components based on the strength of association, and combines the strength of association between industrial software components and the importance of different relationship types to obtain a component sequence that has an important impact on the entire multi-networked industrial software system, thereby enhancing the ability to identify key components in the multi-networked industrial software system.

[0045] (4) The present invention uses multiple feature vectors and multiple topological structures of components to identify key components on multiple networked industrial software systems. The addition of multiple feature vectors makes the information considered in identifying key components more comprehensive, not only considering the structural information of the entire multiple industrial software network, but also considering the characteristics of different components themselves, further improving the effect of key component identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 It is the overall framework diagram of the method of the present invention;

[0047] Figure 2 This is a schematic diagram of the connections between components on a multi-networked industrial software system in Example 2 of the present invention. DETAILED DESCRIPTION

[0048] The present invention will be further explained below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the following specific embodiments are only used to illustrate the present invention and are not used to limit the scope of the present invention.

[0049] Example 1

[0050] For the key component identification method of multiple networked industrial software systems, the existing methods are all based on the multiple topological structures of components for research, ignoring the actual situation of different components on multiple industrial software networks, and not further considering the influence of different network relationships on the key component identification. There is almost no excessive research on the correlation strength that can accurately reflect the degree of attraction between components. Therefore, the present invention discloses a key component identification method suitable for multiple networked industrial software systems, such as Figure 1 As shown, the specific implementation process is as follows:

[0051] Step S1: Input a multi-networked industrial software system G containing L different relationships = {G1, G2, ..., G L}, each layer of the network contains N identical components. For a multi-networked industrial software system G, Indicates the network layer G α The i-th (1≤i≤N) component on (1≤α≤L). Network layer G α Components on and The multiple topological structures between can be expressed as:

[0052]

[0053] in, Representation Components and At the network layer G α A connection exists on the

[0054] Step S2: Calculate the influence radius of all components on each network layer.

[0055] The method of the present invention takes into account the differences in the impact ranges of different components and obtains different impact radii for each component on a multi-networked industrial software system, thereby accurately describing the heterogeneous characteristics of components in a complex network. Its impact range is expressed as This means that the component At the network layer G α The maximum number of Level neighbors. Assume that for any component There is a component outside its influence range in the same layer of relationship network At the same time, components and There are components in the overlapping scopes. More than just components of Level neighbors, and components of Obviously, the components Located in component and In this sense, components Can be considered as a component and The dividing point.

[0056] Step S21: Introduce iterative k-shell algorithm and gravity centrality model to define nodes For any other node The applied gravity centrality is:

[0057]

[0058]

[0059] in, Respectively represent nodes and The degree, express and The shortest distance between Represents the gravity coefficient, which is used to describe the degree of mutual attraction between nodes. Representation Node The iterated k-shell value, Iks max , Iks min Represents the network layer G α Maximum and minimum iterative k-shell values ​​for the midpoint nodes.

[0060] For nodes and The dividing point Define nodes separately and For Node Applied gravity centrality:

[0061]

[0062]

[0063] Step S22: According to the force balance principle, the node and For Node The applied gravitational forces are equal, then:

[0064]

[0065] Step S23: Apply the above equation to Solve and get the node Influence radius The solution formula is:

[0066]

[0067] Step S24: The method of the present invention defines that the influence of a component is determined by the influence radius of the component itself and the average influence radius of its neighboring components. After that, the method of the present invention is for each component Calculate importance based on dynamic impact range

[0068]

[0069] in, Representation Components The number of neighbors.

[0070] Step S3: Calculate the similarity between multiple feature vectors of each component at different relational network layers.

[0071] The method of the present invention defines components The multiple eigenvectors of can be expressed as: Among them C i1 ,C i2 ,…,C in It is a component Based on this vector, the cosine similarity algorithm is used to calculate the different components and At the network layer G α Similarity between multiple eigenvectors on :

[0072]

[0073] Where n represents the number of elements contained in the multiple feature vector used.

[0074] Step S4: Use the latent variable model to calculate the association strength between different components at different relationship network layers.

[0075] The latent variable model is based on two observations: (1) for the two components and Given a specific relational network layer G α , the strength of association By Component and At the network layer G α The similarity between multiple feature vectors on OK. (2) and At the relational network layer Gα The strength of the association Affect them in G α Multiple topologies on

[0076] The latent variable model represents the possible causal relationships between all variables by modeling conditional dependencies. Based on these dependencies, the joint distribution is decomposed as follows:

[0077]

[0078] in express and At the relational network layer G α The strength of the association.

[0079] The method of the present invention adopts the widely used Gaussian distribution to simulate the conditional probability and The dependencies are as follows:

[0080]

[0081] where w α is the weight to be estimated, and v is the variance in the Gaussian model. In the same way, and The dependencies between them are modeled as follows:

[0082]

[0083] where β α is the parameter to be learned. and In the relational network layer C α The similarity of multiple eigenvectors on Multiple topologies between components Association Strength The joint probability can be expressed as:

[0084]

[0085] Where λ1 and λ2 are preset parameters, G is the network layer α The set of the above component pairs.

[0086] Since the above joint probabilities are independent, the method of the present invention can decompose them into L independent joint probabilities and derive the solutions of each relational network layer respectively. Therefore, it is only necessary to derive the solution of the joint probability of the αth relational network layer. Take the logarithm of the above formula and list only the logarithmic likelihood of the αth relational network layer:

[0087]

[0088] Because of the function is a concave function. This method uses a gradient-based strategy to optimize the parameters. The coordinate ascent optimization method will iteratively update w α , β α , Until convergence. For different parameters, the gradients in the coordinate direction are:

[0089]

[0090]

[0091]

[0092] In each iteration, the method of the present invention uses the Newton-Raphson method to update the above parameters:

[0093]

[0094]

[0095]

[0096] Among them, the second-order derivative of each parameter is obtained by the following formula:

[0097]

[0098]

[0099]

[0100] After convergence, the final Component and In the component relationship network layer G of multi-networked industrial software systems α The strength of the association.

[0101] Similarly, the method of the present invention can obtain the association strength between components on different network layers.

[0102] Step S5: Calculate each network layer G α importance.

[0103] Network Layer G α The importance in a multi-layer network is represented by LI α , is calculated by comparing the ratio of the number of edges in this network layer to the total number of edges in the multinetwork, which can be given by:

[0104]

[0105] in It is the network layer G α The number of edges, is the total number of edges in the entire multinetwork.

[0106] LI α It is possible to quantify the contribution of each network layer to the overall multi-layer network structure. Layers with high importance indicate that they play a pivotal role in the multi-layer network, while layers with low importance contribute less to the overall network structure.

[0107] Step S6: Construct a gravity centrality model with weights.

[0108] As mentioned above, the components Importance based on dynamic impact range The strength of the associations between different components Network Layer G α The importance of LI α Based on the above results, the gravity centrality model is constructed as follows:

[0109]

[0110] in, Representation Components and At the network layer The shortest path length on the network layer, if there is no direct path between the two components on this network layer, then The value of is negative infinity.

[0111] Based on the above results, the importance of each component is as follows:

[0112]

[0113] Step S7: Sort the components according to their importance values ​​to identify key components on multiple networked industrial software systems, provide guidance for collaboration and decision-making between components, and improve the overall production and operation efficiency of industrial software.

[0114] Example 2

[0115] The new technology proposed in this patent deals with multiple networked industrial software components. Since the components can be independently developed, deployed, run, upgraded and interconnected, the components can adaptively switch between service combination modes and dynamically adjust service granularity to form multi-granular component services. In a real multi-networked industrial software system scenario, industrial components are jointly affected by multiple related coupling modes such as data flow, business flow and information flow. For example, in a simple multi-networked industrial software system, the connection between components is as follows: Figure 2 As shown, different layers G1 to G1 represent different coupling relationships of the components. α Industrial software components Based on the actual business process of the industrial software system, it is connected with other components. In order to realize accurate identification and efficient scheduling of components in complex coupled industrial production scenarios such as multiple products, multiple batches, and multiple workshops, solve the problems of green efficiency improvement, elastic service, and intelligent deployment of industrial software components, and provide strong technical support for the improvement of efficiency, robustness, and scalability of industrial software systems in multiple network environments, a key component identification system based on association strength on a multiple networked industrial software system implementing the method described in Example 1 is proposed, which includes at least four different functional modules: a component impact range calculation module, an association strength calculation module, a network layer importance calculation module, and a component importance comprehensive evaluation module.

[0116] The component influence range calculation module calculates different influence radii for each component in the multi-networked industrial software system based on the different characteristics of the influence ranges of different components, and obtains the importance of each component based on the dynamic influence range on this basis; the association strength calculation module combines the multiple characteristic vectors and multiple topological structures of the components at the same time, and solves the association strength of different relationship network layers between components by constructing a latent variable model; the network layer importance calculation module calculates the importance of each different relationship network to the entire multi-networked industrial software system by considering the impact of different relationship networks on the overall multi-networked industrial software system structure; the component importance comprehensive evaluation module obtains the overall importance of each component by constructing a weighted gravity centrality model, combining the calculated association strength, component centrality based on the influence range and network layer importance, thereby identifying the key components in the multi-networked industrial software system.

[0117] In summary, the method of the present invention is based on the combination of multiple eigenvectors and multiple topological structures. The gravity centrality method (combining Intrinsic and Topological Features to calculate Relationship Strength for Gravity Centrality, ITFRSGC) proposes a key component identification method and system on a multi-networked industrial software system based on association strength. For the first time, multiple eigenvectors and multiple topological structures of components are used simultaneously to identify key components, and the association strength between different components on different relationship networks is modeled based on the above information. At the same time, the importance of different relationship networks to the identification of key components is considered, and the key components on the multi-networked industrial software system are more efficiently identified. For example, relatively more critical components are monitored, and more resources are allocated to these components to prevent their failures from causing large-scale cascading failures of other components in the multi-networked industrial software system, provide guidance for collaboration and decision-making between components, and improve overall production and operation efficiency.

[0118] It should be noted that the above content only illustrates the technical idea of ​​the present invention and cannot be used to limit the protection scope of the present invention. For ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications all fall within the protection scope of the claims of the present invention.

Claims

1. A method for identifying key components of multi-networked industrial software, characterized in that: The steps include: S1: Input the components into a multi-networked industrial software system G = {G1, G2, …, G L }, each network layer in the multi-networked industrial software system includes N identical components; S2: Calculate the influence radius of all components on each network layer, and determine the influence of the component together with the average influence radius of its neighboring components. Calculate the importance of each component based on the dynamic influence range. S3: Calculate the similarity between multiple feature vectors of each component at different relational network layers, and calculate the similarity by using the cosine similarity algorithm get; S4: Using latent variable models to calculate the strength of association between different components at different relational network layers The strength of the association Based on the similarity of step S3 Sure; S5: Calculate each network layer G α The importance of LI α It is obtained by comparing the ratio of the number of edges in this network layer to the total number of edges in the multi-network; S6: Components obtained according to step S2 Importance based on dynamic impact range The correlation strength between different components obtained in step S4 The network layer G obtained in step S5 α The importance of LI α , the gravity centrality model is constructed as follows: in, Representation Components and At the network layer The shortest path length on ; The importance of each component is as follows: S7: Sort the importance values ​​of the components obtained in step S6 to identify key components on the multi-networked industrial software system.

2. The method for identifying key components of multi-networked industrial software according to claim 1, characterized in that: In step S1, the network layer G of the multi-networked industrial software system α Components on and The multiple topological structures between are expressed as: in, Indicates the network layer G α The i-th component on Representation Components and At the network layer G α A connection exists on the 3. The method for identifying key components of multi-networked industrial software according to claim 2, characterized in that: The step S2 specifically includes the following steps: S21: Introducing the iterative k-shell algorithm and gravity centrality model, defining nodes For any other node The applied gravity centrality is: in, Respectively represent nodes and The degree, express and The shortest distance between represents the gravity coefficient, Representation Node The iterated k-shell value, Iks max , Iks min Represents the network layer G α Maximum and minimum iterative k-shell values ​​at midpoints; For nodes and The dividing point Define nodes separately and For Node Applied gravity centrality: S22: According to the principle of force balance, the node and For Node The applied gravitational forces are equal, then: S23: Using the equation of step S22 Solve and get the node Influence radius The solution formula is: S24: For each component Calculate importance based on dynamic impact range in, Representation Components The number of neighbors.

4. The method for identifying key components of multi-networked industrial software according to claim 1, characterized in that: In step S3, the components The multiple eigenvectors of are expressed as: Among them C i1 ,C i2 ,…,C in It is a component Different features represent different components and At the network layer G α The similarity between multiple feature vectors on is specifically: Where n represents the number of elements contained in the multiple feature vector used.

5. The method for identifying key components of multi-networked industrial software according to claim 4, characterized in that: The step S4 specifically includes the following steps: S41: The latent variable model models the conditional dependencies. Based on the dependencies, the joint distribution is decomposed as follows: in, express and At the relational network layer G α The strength of the association on S42: Using Gaussian distribution to simulate conditional probability and The dependencies are as follows: Among them, w α is the weight to be estimated, v is the variance in the Gaussian model; and The dependencies between them are modeled as follows: where β α is the parameter to be learned; S43: Given and At the relational network layer G α The similarity of multiple eigenvectors on Multiple topologies between components Association Strength The joint probability can be expressed as: Where λ1 and λ2 are preset parameters, G is the network layer α The set of above component pairs; S44: Decompose the joint probability of step S43 into L independent joint probabilities, and derive the solution of each relationship network layer respectively. Only the solution of the joint probability of the αth relationship network layer needs to be derived. Take the logarithm of the above formula and list only the logarithmic likelihood of the αth relationship network layer: Because of the function is a concave function; S45: Using the gradient-based strategy to optimize the parameters, the coordinate ascent optimization method will iteratively update w α , β α , Until convergence, for different parameters, the gradients in the coordinate direction are: In each iteration, the above parameters are updated using the Newton-Raphson method: Among them, the second-order derivative of each parameter is obtained by the following formula: S46: After convergence, the final For components and In the component relationship network layer G of multi-networked industrial software systems α The strength of the association.

6. The method for identifying key components of multi-networked industrial software according to claim 1, characterized in that: The importance LI of each network layer in step S5 α The specific calculation method is: in It is the network layer G α The number of edges, is the total number of edges in the entire multinetwork.

7. A key component identification system for multi-networked industrial software using the method of claim 1, characterized in that: At least including a component impact range calculation module, an association strength calculation module, a network layer importance calculation module and a component importance comprehensive evaluation module; The component impact range calculation module: based on the topological connection between components in the entire multi-networked industrial software system, calculates the impact range of each component and obtains the importance of each component based on the dynamic impact range; The association strength calculation module combines multiple feature vectors and multiple topological structures of components, and solves the association strength of different relationship network layers between components by constructing a latent variable model; The network layer importance calculation module is used to calculate the importance of the network layer to the entire industrial software system for each different relationship network by considering the impact of different relationship networks on the network structure of the entire industrial software system; The component importance comprehensive evaluation module constructs a weight-based gravity centrality model, combines the calculated association strength, component centrality based on the scope of influence, and network layer importance to obtain the overall importance of each component, thereby identifying key components on multiple networked industrial software systems.

Images