Military tent production management method and system based on performance prediction optimization

Abstract

The invention discloses a military tent production management method based on performance prediction optimization, and relates to the field of performance optimization, and the method comprises the steps: obtaining tent production information; constructing a finite element model according to the tent production information, and identifying a stress concentration area based on the finite element model; constructing an initial knowledge graph by using tent production information; calculating an initial confidence coefficient of the initial knowledge graph; distributing a key weight for the initial knowledge graph to obtain a target knowledge graph; the performance correlation parameters are optimized in a multi-objective mode, and the optimal production parameters are obtained. The performance of the military tent can be effectively optimized.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of performance optimization, and in particular to a production management method and system for military tents based on performance prediction optimization. Background Art

[0002] Military tents can provide effective wind protection and cold resistance functions for users, ensuring a good rest environment even under harsh weather conditions. Its importance is self-evident. Since the usage scenarios of military tents are mostly very harsh, the performance requirements for military tents are higher than those of ordinary tents. Therefore, it is necessary to further optimize the production process of tents.

[0003] Since there are many directions for optimizing the performance of military tents, in order to simplify the optimization process, traditional production optimization methods for military tents often only focus on performance indicators in a single dimension. However, there are conflicts and contradictions between different performances of military tents, such as wind resistance and portability. Therefore, only optimizing a single performance indicator may lead to performance imbalance of military tents, and thus may cause a decline in the overall performance of military tents. Summary of the Invention

[0004] The embodiments of the present application provide a production management method for military tents based on performance prediction optimization, which is used to solve the problem of insufficient overall performance of existing military tents.

[0005] To achieve the above object, the embodiments of the present application adopt the following technical solutions:

[0006] In a first aspect, a production management method for military tents based on performance prediction optimization is provided, and the method includes:

[0007] Obtain the tent production information of tent samples in the target factory;

[0008] Construct a finite element model of the tent sample according to the tent production information, complete the stress analysis of the tent sample based on the finite element model, and identify the stress concentration area of the tent sample according to the stress analysis result;

[0009] Generate multiple tent triples by using the tent production information, and construct an initial knowledge graph based on all tent triples;

[0010] Evaluate the fusion confidence of all tent triples based on the principle of translational invariance, calculate the initial confidence of the initial knowledge graph according to all fusion confidences, and complete the reliability verification of the initial knowledge graph according to the initial confidence;

[0011] If the reliability verification of the initial knowledge graph passes, then trace back the correlation factors of the stress concentration area in reverse through the initial knowledge graph and using the Bayesian theory, and assign key weights to the initial knowledge graph according to the correlation factors to obtain the target knowledge graph;

[0012] Screen out the performance-related parameters in the target knowledge graph according to the preset performance optimization objectives, and optimize the performance-related parameters with multiple objectives to obtain the optimal production parameters of the tent sample.

[0013] Optionally, the tent production information includes tent structure information, tent raw material information, tent process information, tent performance information, and production environment information.

[0014] Optionally, constructing a tent finite element model of the tent sample based on the tent production information, completing the stress analysis of the tent sample based on the finite element model, and identifying the stress concentration area of the tent sample according to the stress analysis result includes the following steps:

[0015] Construct a three-dimensional model of the tent sample according to the tent structure information;

[0016] Perform adaptive mesh division on the three-dimensional tent model to construct a tent finite element model;

[0017] Apply multiple types of loads to the tent finite element model, and calculate the stress distribution nephogram of the tent sample under the coupling action of multiple types of loads;

[0018] Complete the analysis of the stress distribution law of the tent sample according to the stress distribution diagram, and identify the stress concentration area of the tent sample according to the analysis result of the stress distribution law.

[0019] Optionally, generating multiple tent triples using the tent production information, and constructing an initial knowledge graph based on all tent triples includes the following steps:

[0020] Identify all key entities in the tent production information based on the entity naming rules;

[0021] Obtain the context relationships of all key entities in the tent production information, and identify the association relationships and entity attributes between all key entities according to the context relationships;

[0022] Generate multiple tent triples based on all key entities, association relationships, and entity attributes;

[0023] Construct an initial knowledge graph based on all tent triples. The nodes of the initial knowledge graph are used to represent key entities, and the edges of the initial knowledge graph are used to represent the association relationships between key entities.

[0024] Optionally, evaluate the fusion confidence of all tent triples based on the principle of translational invariance, calculate the initial confidence of the initial knowledge graph according to the fusion confidence, and complete the reliability verification of the initial knowledge graph according to the initial confidence includes the following steps:

[0025] For any tent triple, perform a vector translation projection between the head entity and the associated relationship in the tent triple to obtain a projection vector;

[0026] Calculate the vector distance between the tail entity in the tent triple and the projection vector, and calculate the local confidence of the tent triple based on the vector distance;

[0027] Take the head entity in the tent triple as the starting point;

[0028] Starting from the starting point, traverse the initial knowledge graph to search for all connected paths between the head entity and the tail entity, and merge all the connected paths to generate a local directed graph;

[0029] Assign an initial resource value to the head entity, and the initial resource value circulates to the tail entity through the local directed graph;

[0030] Calculate the tail resource value of the tail entity through the resource iteration formula and based on the initial resource value, and quantify the entity association degree between the tail entity and the head entity according to the tail resource value;

[0031] Determine the global confidence of the tent triple according to the entity association degree;

[0032] Fuse the local confidence and the global confidence of each tent triple respectively to obtain the fusion confidence of all tent triples;

[0033] Calculate the confidence mean of all the fusion confidences, and take the confidence mean as the initial confidence of the initial knowledge graph;

[0034] If the initial confidence is greater than the preset confidence threshold, it is determined that the reliability check of the initial knowledge graph passes;

[0035] If the initial confidence is less than or equal to the confidence threshold, it is determined that the reliability check of the initial knowledge graph fails.

[0036] Optionally, the resource iteration formula is as follows:

[0037]

[0038] Where, G(xy) is the tail resource value flowing from the head entity x to the tail entity y; μ is the resource circulation probability; N is the total number of nodes in the local directed graph; α is the node set between the head entity x and the tail entity y; z is the adjacent node adjacent to the tail entity y; G(zy) is the tail resource value flowing from the adjacent node z to the tail entity y; D z，y is the edge weight between the adjacent node z and the tail entity y; C z is the out-degree of the adjacent graph node z.

[0039] Optionally, based on the initial knowledge graph and using Bayesian theory to trace back the correlation factors of the stress concentration area, and assigning key weights to the initial knowledge graph according to the correlation factors, obtaining the target knowledge graph includes the following steps:

[0040] Assign initial weights to the edges and nodes of the initial knowledge graph based on the fusion confidence, and update the initial knowledge graph according to the initial weight assignment result;

[0041] Generate stress entities based on the stress concentration area, and add the stress entities to the initial knowledge graph after the update is completed to obtain an intermediate knowledge graph;

[0042] Map the intermediate knowledge graph to an intermediate knowledge network, trace back the correlation factors of the stress entities through the intermediate knowledge network, and assign key weights to the intermediate knowledge graph according to the tracing results of the correlation factors to obtain the target knowledge graph.

[0043] Optionally, mapping the intermediate knowledge graph to an intermediate knowledge network, tracing back the correlation factors of the stress entities through the intermediate knowledge network, and assigning key weights to the intermediate knowledge graph according to the tracing results of the correlation factors to obtain the target knowledge graph includes the following steps:

[0044] Traverse the intermediate knowledge graph and extract the direct and indirect dependency relationships between all intermediate nodes in the intermediate knowledge graph;

[0045] Construct multiple dependency matrices based on the direct dependency relationship, indirect dependency relationship, and initial confidence, and superimpose and merge all the dependency matrices to obtain a fusion relationship matrix;

[0046] Traverse the fusion relationship matrix, identify the conflict relationships in the fusion relationship matrix, and delete the fusion relationship matrix according to the conflict relationship identification result to obtain the target relationship matrix;

[0047] Generate a relationship directed graph based on the target relationship matrix, and use the loop detection algorithm to identify whether there is a loop structure in the relationship directed graph;

[0048] If there is a loop structure in the relationship directed graph, then delete the relationship edges in the relationship directed graph according to the fusion confidence until there is no loop structure in the relationship directed graph;

[0049] If there is no loop structure in the relationship directed graph, then use the relationship directed graph as the intermediate knowledge network;

[0050] Use the stress network node corresponding to the stress entity in the intermediate knowledge network as the starting node;

[0051] Starting from the starting node, trace back the parent nodes of the stress network nodes level by level until reaching the root node of the intermediate knowledge network, and extract the correlation factors of the stress entities according to the parent node tracing results;

[0052] Calculate the association probability between all association factors and stress entities using the Bayesian inference algorithm, and assign key weights to the intermediate nodes corresponding to each association factor according to the association probability to obtain the target knowledge graph.

[0053] Optionally, screen out the performance association parameters in the target knowledge graph according to the preset performance optimization objectives, and perform multi-objective optimization on the performance association parameters. The steps to obtain the optimal production parameters of the tent sample are as follows:

[0054] Infer the performance association path in the target knowledge graph according to the preset performance optimization objectives, and extract the performance association parameters corresponding to the performance optimization objectives based on the performance association path;

[0055] Define the performance optimization weight of the performance optimization objective according to the weight of the target knowledge graph;

[0056] Generate a multi-objective optimization function by combining the tent performance optimization objective and the performance optimization weight;

[0057] Based on the multi-objective optimization function, use the multi-objective optimization algorithm to iteratively optimize the performance association parameters until the preset maximum number of optimizations is reached, and output the optimal production parameters of the tent sample.

[0058] In a second aspect, the present application provides a military tent production management system based on performance prediction optimization, which is characterized by including:

[0059] A memory configured to store instructions; and

[0060] A processor configured to call instructions from the memory and capable of implementing the method of military tent production management based on performance prediction optimization according to any one of the first aspect when executing the instructions.

[0061] Through the above technical solutions, by constructing a tent finite element model and performing stress analysis on the tent finite element model, the stress concentration areas in the tent sample can be accurately identified, which is convenient for subsequently determining the association factors that will affect the stress concentration areas. During the process of optimizing the tent performance, if there are contradictions and conflicts between the tent performances, the association factors can be used to balance the contradictions and conflicts between the tent performances, that is, preferentially optimize the performances associated with the association factors. In addition, in order to more accurately identify the stress concentration area association factors and comprehensively screen out the performance association parameters associated with the performance optimization objectives, the present application constructs a target knowledge graph using the target factory of the target factory, accurately identifies the stress concentration area association factors through the target knowledge graph, and comprehensively screens out the performance association parameters associated with the performance optimization objectives, so as to be able to comprehensively optimize the military tent from multiple different directions and improve the overall performance of the tent.

[0062] Other features and advantages of the embodiments of the present application will be described in detail in the following specific implementation section. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 It is a schematic flowchart of a production management method for military tents based on performance prediction optimization provided by an embodiment of the present application;

[0064] Figure 2 It is an example diagram of the initial knowledge graph provided by an embodiment of the present application;

[0065] Figure 3 It is a schematic flowchart of a stress concentration area identification method provided by an embodiment of the present application;

[0066] Figure 4 It is a schematic diagram of a fusion relationship matrix construction method provided by an embodiment of the present application;

[0067] Figure 5 It is an example diagram of the target relationship matrix provided by an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0068] To make the objectives, technical solutions, and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. It should be understood that the specific implementation manners described herein are only for explaining and illustrating the embodiments of the present application, and are not used to limit the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present application.

[0069] It should be noted that if there are directional indications (such as up, down, left, right, front, back,...) involved in the embodiments of the present application, the directional indications are only used to explain the relative positional relationship and movement conditions between components in a specific posture (as shown in the drawings). If the specific posture changes, the directional indications will also change accordingly.

[0070] In addition, if there are descriptions involving "first", "second", etc. in the embodiments of the present application, the descriptions of "first", "second", etc. are only for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include at least one of such features. In addition, the technical solutions between various embodiments can be combined with each other, but it must be based on the fact that those of ordinary skill in the art can implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection required by the present application.

[0071] Figure 1 Schematically shown is a schematic flow chart of a military tent production management method based on performance prediction optimization according to an embodiment of the present application. As Figure 1 shown, an embodiment of the present application provides a method for military tent production management based on performance prediction optimization, and the method may include the following steps:

[0072] S101. Obtain the tent production information of the tent samples in the target factory.

[0073] In this embodiment, the tent production information includes tent structure information, tent raw material information, tent process information, tent performance information, and production environment information. The tent structure information refers to the tent geometric structure diagram and geometric parameters of the tent samples produced by the target factory, such as the volume after folding is 0.02 m 3 , and the tent components that make up the tent samples, such as the inner tent, outer tent, tent poles, etc.; the tent raw material information refers to the raw material types used to produce the tent samples, such as nylon, polyester fiber, aluminum alloy, and PU coating, etc., and the raw material parameters corresponding to various raw material types, such as nylon: 50D (the lower the Denier value, the lighter the raw material quality), aluminum alloy: elastic modulus 70 GPa, PU coating: waterproof 4000 mm, etc.; the tent process information includes process types and process parameters. The process types include sewing processes (flat seam, hot melt adhesive strip), coating processes (siliconization, PU coating), assembly processes, etc. Different process types correspond to different process parameters, such as hot melt adhesive strip: width 20 mm, the hot melt adhesive strip is the process type, and the width 20 mm is the process parameter; the tent performance information includes performance types and performance parameters, such as temperature range: -50 °C to +55 °C, the temperature range is the performance type, -50 °C to +55 °C is the performance parameter, wind resistance level: greater than or equal to level 12 (wind speed 35 m / s), the wind resistance level is the performance type, greater than or equal to level 12 is the performance parameter, fire resistance performance: the oxygen index of the flame retardant fabric is greater than or equal to 32, etc., the fire resistance performance is the performance type, and the oxygen index of the flame retardant fabric is greater than or equal to 32 is the performance parameter; the production environment information includes information types and information parameters. The information type refers to information such as the temperature and humidity in the production workshop, and the information parameter refers to the parameters corresponding to information such as temperature and humidity, such as the information parameter of temperature is 25 degrees Celsius.

[0074] S102. Construct a finite element model of the tent samples based on the tent production information, complete the stress analysis of the tent samples based on the finite element model, and identify the stress concentration areas of the tent samples according to the stress analysis results.

[0075] In this embodiment, first, a 3D model of the tent of the tent sample is constructed using CAD or finite element software (such as ANSYS, Abaqus). Then, raw material properties are added to the 3D model of the tent according to the raw material information of the tent sample. Adaptive mesh division is performed on the 3D model of the tent, that is, key attention is paid to potential stress concentration areas such as the connection between the support rod and the tarpaulin, and the ground nail fixing points. For potential stress concentration areas, the mesh needs to be refined. In addition, boundary conditions also need to be defined for the 3D model of the tent, that is, the constraint conditions applied on the boundary or contact surface of the 3D model of the tent, which are used to restrict the degrees of freedom of the structure. After completing the above steps, a finite element model of the tent without applied load can be initially obtained.

[0076] Then, multiple types of loads are applied to the finite element model of the tent. The multiple types of loads include wind load, snow load, pre-tension, and human load, etc. The wind load, snow load, pre-tension, and human load, etc. can be randomly combined. The multiple types of load data are imported into the finite element software. Through the finite element software (such as ANSYS, SolidWorks Simulation), the stress distribution of the finite element model of the tent under the coupling action of multiple types of loads is simulated, and the finite element model of the tent is analyzed non-linearly or linearly by methods such as static analysis and transient analysis to obtain the stress distribution contour map of the finite element model of the tent. Observe the stress distribution map and identify the overall distribution trend of the stress, such as the positions of the high-stress area and the low-stress area. The area where the stress difference from the surrounding area is greater than the preset difference threshold is used as the stress concentration area.

[0077] S103. Generate multiple tent triples using the tent production information, and construct an initial knowledge graph based on all the tent triples.

[0078] In this embodiment, entity naming rules can be used to identify entity names in tent production information, such as polyester fiber, nylon, canvas, aluminum alloy, tent peg, bracket, etc. The identified entity names are used as entities that form triples, i.e., key entities. Then, syntactic analysis can be performed on sentences containing key entities to identify sentence structures such as subject-predicate and verb-object, so as to identify the context relationships of key entities in tent production information. Context windows can also be defined to extract the vocabulary around key entities to capture their context relationships in tent production information. Through context relationships, the association relationships and entity attributes between key entities can be obtained. The association relationship refers to the mutual relationship between key entities. The corresponding key entities, association relationships, and entity attributes are formed into tent triples, such as (polyester fiber, used for, tarpaulin), (waterproof layer, waterproof index, 4000mm). Tent triples with the same key entity are identified, merged, and stored in a pre-constructed graph database. Then, the key entities and entity attributes of all tent triples in the graph database are mapped to nodes, and the association relationships between key entities are mapped to edges to construct an initial knowledge graph. Finally, visualization tools such as Draw.io and Neo4j Bloom can be used to visually display the initial knowledge graph. Refer to Figure 2 。

[0079] S104. Evaluate the fusion confidence of all tent triples based on the principle of translational invariance, calculate the initial confidence of the initial knowledge graph according to all fusion confidences, and complete the reliability verification of the initial knowledge graph according to the initial confidence.

[0080] In this embodiment, using the principle of translational invariance, first map the head entity, association relationship, and tail entity in the tent triple to a low-dimensional space to obtain a head vector, a relationship vector, and a tail vector. Perform vector translation projection on the head entity and the association relationship in the tent triple, that is, project the head vector and the relationship vector onto the same plane and then approach the tail vector through translation operations to obtain the projection vector between the two. The smaller the vector distance between the projection vector obtained after the head entity and the association relationship complete vector translation projection and the tail vector, the higher the association degree between the head entity and the tail entity inside the tent triple. And the higher the association degree, the higher the confidence of the tent triple. According to this theory, the local confidence of the tent triple can be calculated based on the vector distance between the projection vector and the tail vector.

[0081] Since the initial knowledge graph has a large amount of information and contains a large number of nodes and edges, there are multiple paths between the head entity and the tail entity in the same tent triple. The total amount of resources transferred from the head entity to the tail entity through multiple paths can be used to quantify the association between the two entities. If the total amount of resources transferred to the tail entity is large, it means that the association between the head entity and the tail entity of the tent triple is also higher, which means that the global confidence of the tent triple in the initial knowledge graph is also higher. Specifically, the head entity in the tent triple is first used as the starting point. Starting from the starting point, it gradually expands in different directions along the association relationship in the initial knowledge graph until it reaches the tail entity in the same tent triple as the starting entity, and the connection paths that can connect the two are integrated to generate a local directed graph.

[0082] An initial resource value (which can be 1) is assigned to the head entity. The initial resource value iterates and circulates continuously from the head entity until it reaches the tail entity through all nodes in the local directed graph. Then, the resource iteration formula is used to calculate the remaining resource value after the initial resource value reaches the tail entity, i.e., the tail resource value. The larger the tail resource value, the greater the association degree (entity association degree) between the tail entity and the head entity. The greater the association degree, the greater the global confidence of the tent triple containing the tail entity and the head entity. Therefore, the entity association degree between the tail entity and the head entity can be quantified based on the tail resource value, and the entity association degree can be normalized to the interval [0, 1] as the global confidence of the tent triple.

[0083] By fusing the local confidence and global confidence of each tent triple, a weighted average of the local confidence and the global confidence can be performed to obtain the fused confidence. The average value of all fused confidences, i.e., the confidence mean, is calculated, and the calculated average value is used as the overall confidence of the initial knowledge graph, i.e., the initial confidence. If the initial confidence is greater than the preset confidence threshold, it is determined that the reliability check of the initial knowledge graph has passed, indicating that the initial knowledge graph can be used for subsequent parameter optimization of the tent samples. If the initial confidence is less than or equal to the confidence threshold, it is determined that the reliability check of the initial knowledge graph has failed, indicating that there may be errors in the construction process of the initial knowledge graph, or there are errors in the tent production information used to construct the initial knowledge graph. Therefore, it is necessary to eliminate the tent triplets with fused confidence less than the preset fused confidence threshold, and then reconstruct the initial knowledge graph. If the reliability check of the reconstructed initial knowledge graph still fails, it means that there may be errors in the tent production information. The error output reminds the relevant staff of the target factory to check the tent production information.

[0084] Since the performance optimization of military tents is based on the tent production information and the initial knowledge graph constructed using the tent production information, once an error occurs in the construction process of the initial knowledge graph or there is incorrect information in the tent production information, it is very likely that the subsequent performance optimization of military tents will not succeed. Moreover, it is even possible that the optimal parameters obtained will not only fail to optimize the performance of military tents but will instead reduce the performance of military tents, running counter to the original optimization goal. Therefore, it is necessary to verify the reliability of the initial knowledge graph. Only when the reliability verification passes can the subsequent steps be continued.

[0085] S105. If the reliability verification of the initial knowledge graph passes, then trace back the correlation factors of the stress concentration area through the initial knowledge graph and use the Bayesian theory, and assign key weights to the initial knowledge graph according to the correlation factors to obtain the target knowledge graph.

[0086] In this embodiment, first assign initial weights to the edges and nodes of the initial knowledge graph according to the fusion confidence, and update the initial knowledge graph according to the initial weight assignment result. Determine the tent components corresponding to the stress concentration area (such as strut nodes, seams or ground nail anchor points), and find the key entity corresponding to the tent component in the initial knowledge graph as the area entity. At the same time, determine the stress direction, stress mean and stress peak of the stress concentration area according to the stress distribution cloud map, and add the stress direction, stress mean and stress peak as stress entities to the initial knowledge graph and connect them to the area entity to realize the secondary update of the initial knowledge graph and obtain the intermediate knowledge graph containing stress entities. Map the intermediate knowledge graph to an intermediate knowledge network, and the intermediate knowledge network is a Bayesian network. Trace back the correlation factors of the stress entities through the Bayesian network, and assign key weights to the key entities corresponding to the correlation factors in the intermediate knowledge graph to obtain the target knowledge graph. In the subsequent process of optimizing the tent performance, if there are conflicts and contradictions among the various performances of the military tent, higher optimization priorities will be given to the performances corresponding to the key entities and correlation relationships with higher weight values in the target knowledge graph. Of course, the higher optimization priorities here are still relative, and the priorities will not differ too much. Assigning optimization priorities is only to balance the conflicting performances as much as possible, so as to improve the overall performance of the military tent.

[0087] S106. Screen out the performance correlation parameters in the target knowledge graph according to the preset performance optimization goal, and perform multi-objective optimization on the performance correlation parameters to obtain the optimal production parameters of the tent sample.

[0088] In this embodiment, the performance optimization objective refers to the type of performance that needs to be optimized, such as the wind resistance level, fire resistance performance, etc. Then, the key entities corresponding to the performance optimization objective are found in the target knowledge graph, and the found key entities are marked as performance entities. Starting from the performance entities, all performance association paths are traversed in reverse using the graph traversal algorithm, that is, the paths having an association relationship with the performance entities, until a key entity with content being a parameter (such as 50D, elastic modulus of 70 GPa, etc.) or the length of the performance association path is greater than or equal to a preset length threshold is found. After all performance association paths are found, the performance association parameters (such as raw material thickness) in all performance association paths are extracted. At the same time, a performance optimization weight is assigned to the performance optimization objective according to the weight of the performance entity in the target knowledge graph, and the weight in the target knowledge graph refers to the initial weight or the final weight obtained after the initial weight is corrected by the key weight. A multi-objective optimization function is generated according to the tent performance optimization objective and the performance optimization weight.

[0089] In addition, while constructing the multi-objective optimization function, the following constraint conditions also need to be defined to ensure that the optimization result is within the actual feasible range. These constraint conditions include: restricting the optimization ranges of performance optimization objectives such as wind resistance level, fire resistance performance, tent weight, etc.; restricting parameters such as process parameters and raw material parameters to ensure that they are within the achievable range. Finally, based on the multi-objective optimization function, the performance association parameters are iteratively optimized using a multi-objective optimization algorithm until a preset maximum number of optimization times is reached, and the optimal production parameters of the tent sample are output. Commonly used multi-objective optimization algorithms include non-dominated sorting genetic algorithm, multi-objective particle swarm algorithm, etc.

[0090] In one implementation manner, the tent production information includes tent structure information, tent raw material information, tent process information, tent performance information, and production environment information.

[0091] In this embodiment, the tent structure information refers to the tent geometric structure diagram and geometric parameters of the tent sample produced by the target factory, such as the volume after folding of 0.02 m 3, and tent components that make up the tent sample, such as the inner tent, outer tent, tent poles, etc.; tent raw material information refers to the types of raw materials used to produce the tent sample, such as nylon, polyester fiber, aluminum alloy, and PU coating, etc., and the raw material parameters corresponding to each type of raw material, such as nylon: 50D (the lower the Denier value, the lighter the raw material quality), aluminum alloy: elastic modulus 70GPa, PU coating: waterproof 4000mm, etc.; tent process information includes process types and process parameters. Process types include sewing processes (flat seam, hot melt adhesive tape), coating processes (siliconization, PU coating), assembly processes, etc. Different process types correspond to different process parameters. For example, hot melt adhesive tape: width 20mm, where hot melt adhesive tape is the process type and width 20mm is the process parameter; tent performance information includes performance types and performance parameters. For example, temperature range: -50°C to +55°C, where temperature range is the performance type and -50°C to +55°C is the performance parameter, wind resistance level: greater than or equal to 12 levels (wind speed 35m / s), where wind resistance level is the performance type and greater than or equal to 12 levels is the performance parameter, fire resistance performance: oxygen index of flame-retardant fabric greater than or equal to 32, etc., where fire resistance performance is the performance type and oxygen index of flame-retardant fabric greater than or equal to 32 is the performance parameter; production environment information includes information types and information parameters. The information type refers to information such as the temperature and humidity in the production workshop, and the information parameter refers to the parameters corresponding to information such as temperature and humidity. For example, the information parameter for temperature is 25 degrees Celsius.

[0092] In one of the embodiments, referring to Figure 3 , constructing a finite element model of the tent sample based on the tent production information, completing the stress analysis of the tent sample based on the finite element model, and identifying the stress concentration areas of the tent sample according to the stress analysis results includes the following steps:

[0093] Constructing a three-dimensional model of the tent sample according to the tent structure information;

[0094] Performing adaptive mesh division on the three-dimensional model of the tent to construct a finite element model of the tent;

[0095] Applying multiple types of loads to the finite element model of the tent and calculating the stress distribution nephogram of the tent sample under the coupling action of multiple types of loads;

[0096] Completing the analysis of the stress distribution law of the tent sample according to the stress distribution diagram and identifying the stress concentration areas of the tent sample according to the analysis results of the stress distribution law.

[0097] In this embodiment, first, the tent sample is simplified into key components (such as tarpaulin, support poles, connection nodes, tent peg fixing points, etc.) according to the tent structure information, and secondary details (such as stitches, decorations) are ignored. Then, a 3D model of the tent for the tent sample is constructed using CAD or finite element software (such as ANSYS, Abaqus), and it is necessary to ensure that there are no gaps or overlaps in the 3D model of the tent to avoid affecting subsequent mesh generation. Next, according to the tent raw material information of the tent sample, raw material attributes are added to the 3D model of the tent, that is, the types of raw materials for the 3D model of the tent are defined, such as the tarpaulin being nylon or polyester fiber, etc., and raw material parameters, such as the elastic modulus, Poisson's ratio, tensile strength, etc. of the tarpaulin. Then, adaptive mesh generation is performed on the 3D model of the tent. By abandoning equidistant and uniform difference meshes and adopting curve meshes with uneven density that can automatically adapt to the structural characteristics of the tent, through adaptive mesh generation, it can be ensured that the calculated mesh coincides with the actual boundary at the boundary, and the density of mesh points can be arbitrarily adjusted in the 3D model of the tent, that is, key attention can be paid to potential stress concentration areas such as the connection between the support pole and the tarpaulin, tent peg fixing points, etc. For potential stress concentration areas, the mesh needs to be refined. In addition, when generating the mesh, information such as the aspect ratio and interior angle of the mesh needs to be concerned to avoid mesh distortion. In addition, boundary conditions also need to be defined for the 3D model of the tent, that is, the constraint conditions applied on the boundary or contact surface of the 3D model of the tent, which are used to restrict the degrees of freedom of the structure. For example, the tent pegs of the tent sample can be used as the boundary conditions of the 3D model of the tent to constrain the degrees of freedom of the 3D model of the tent. After completing the above steps, a finite element model of the tent without applied load can be initially obtained. The finite element model is a model established when using the finite element analysis method, and it is a combination of elements that are only connected at nodes, only transfer forces through nodes, and are only constrained at nodes. Finite element analysis refers to using mathematical approximation methods to simulate real physical systems (geometry and load conditions). The purpose of constructing the finite element model is to perform finite element analysis on the tent to find the stress concentration areas in the tent sample. As is well known, a uniform tensile stress distribution can ensure that the tarpaulin is taut, maintain the expected shape of the tent, and enhance the wind resistance. Stress concentration, however, may cause local tearing of the tent or instability of the support poles, and is likely to cause the collapse of the tent structure in strong winds or snow. Therefore, in order to further optimize the performance of military tents, it is necessary to identify the stress concentration areas in the tent sample and extract the associated factors that affect the stress concentration areas. For example, the sewing process around the stress concentration area, the welding accuracy of the tent support poles, etc. may all affect the occurrence and stress intensity of the stress concentration area. In the subsequent process of parameter optimization of the tent sample, more attention needs to be paid to the associated factors of the stress concentration area.

[0098] Next, multiple types of loads are applied to the finite element model of the tent. The multiple types of loads include wind load, snow load, pre-tension, and human load, etc. In terms of snow load, the weight of the accumulated snow needs to be considered, and the distribution may be uneven, especially at the top and edges of the tent. The snow load standards in different regions may vary, and it is necessary to select according to the actual situation. Since the military tent is a flexible structure, the wind load may trigger dynamic responses (such as vibration and flutter), so dynamic analysis (such as transient analysis or frequency domain analysis) needs to be adopted. CFD can be used to simulate the wind field distribution and export the pressure data to be loaded into the finite element model. At the same time, the environment where the military tent is used is relatively harsh, so the peak load caused by the instantaneous maximum wind speed (such as 3-second gust) is considered. In addition, the snow load may also change over time, such as the melting and refreezing of the accumulated snow, resulting in load changes. However, finite element analysis is usually static, so the most unfavorable situation also needs to be considered based on the usage environment of the military tent. Pre-tension is unique to the tent structure because the tent fabric will be pre-tensioned by support poles and guy ropes during erection to maintain the shape and stability. At the same time, the magnitude of the pre-tension will also affect the response of the structure under external loads. For example, excessive pre-tension may cause the material to tear, and too little may cause excessive deformation. Therefore, it is necessary to accurately simulate these pre-tensions during the finite element simulation analysis of the tent finite element model. In addition, since the military tent is for people, it is necessary to simulate the transient loads (such as short-term concentrated forces) when users touch or strongly pull the tent, and asymmetric loads caused by uneven down guy cable tension or support pole offset, etc. Since the usage environment of the military tent is relatively complex, in order to take into account different usage environments of the military tent, the wind load, snow load, pre-tension, and human load, etc. can be randomly combined, and the multi-type load data is imported into the finite element software. Through the finite element software (such as ANSYS, SolidWorks Simulation), the stress distribution of the tent finite element model under the coupling action of multi-type loads is simulated, and the tent finite element model is analyzed non-linearly or linearly by methods such as static analysis and transient analysis to obtain the stress distribution nephogram of the tent finite element model. Observe the stress distribution diagram and identify the overall distribution trend of the stress, such as the positions of the high-stress area and the low-stress area, and take the area where the stress difference from the surrounding area is greater than the preset difference threshold as the stress concentration area.

[0099] In one implementation, generating multiple tent triples using tent production information and constructing an initial knowledge graph based on all tent triples includes the following steps:

[0100] Identify all key entities in the tent production information based on the entity naming rules;

[0101] Obtain the context relationships of all key entities in the tent production information, and identify the association relationships and entity attributes between all key entities according to the context relationships;

[0102] Generate multiple tent triples based on all key entities, association relationships, and entity attributes;

[0103] Construct an initial knowledge graph based on all tent triples. The nodes of the initial knowledge graph are used to represent key entities, and the edges of the initial knowledge graph are used to represent the association relationships between key entities.

[0104] In this embodiment, the entity naming rule (NER) is a basic task in natural language processing technology (NLP). It can identify nominal referring terms from text, laying the foundation for tasks such as relation extraction. By using pre-constructed rule templates and selecting features including statistical information, punctuation marks, keywords, indicator words, direction words, position words (such as the last character), central words, etc., combined with a named entity library, weight assignment is performed for each rule, and then type judgment is carried out based on the matching situation between the entity and the rule, that is, the text is matched with the rule to identify named entities. The entity naming rule can identify the entity names in tent production information, such as polyester fiber, nylon, canvas, aluminum alloy, tent peg, bracket, etc., and use the identified entity names as the entities that make up the triples, that is, key entities. Then, syntactic analysis can be performed on the sentences containing key entities to identify the subject-predicate, verb-object, etc. structures of the sentences, thereby identifying the context relationships of key entities in tent production information. It is also possible to define a context window to extract the vocabulary around the key entity to capture its context relationship in tent production information. Through the context relationship, the association relationship and entity attributes between key entities can be obtained. The association relationship refers to the mutual relationship between key entities. For example, "polyester fiber is used to produce the tarpaulin of military tents", and through context analysis, the association relationship between polyester fiber and tarpaulin can be analyzed as being used for. The entity attribute refers to the attribute information of the key entity, including material attributes, performance attributes, and structural attributes, etc. For example, "the wind resistance level of the tarpaulin is greater than or equal to level 12", and through context relationship analysis, the entity attribute of the tarpaulin can be analyzed as having a wind resistance level greater than or equal to level 12.

[0105] Form tent triples with the corresponding key entities, association relationships, and entity attributes, such as (polyester fiber, used for, tarpaulin), (waterproof layer, waterproof index, 4000mm). Identify the tent triples with the same key entity, merge them, and store them in a pre-constructed graph database. Then, map the key entities and entity attributes of all tent triples in the graph database to nodes, and map the association relationships between key entities to edges to construct an initial knowledge graph. Visualization tools such as Draw.io and Neo4j Bloom can be used to visually display the initial knowledge graph.

[0106] The tent triple is the basic building block of the initial knowledge graph. The tent triple can effectively capture the relationships between key entities. Through these relationships, the interactions and dependencies between different key entities can be better understood, helping to construct a complex primary knowledge graph. At the same time, by constructing tent triples, the format of tent production information can be standardized, facilitating the subsequent construction of a unified primary knowledge graph. Moreover, during the subsequent production process of military tents using the optimal production parameters of tent samples, new tent production information will continue to be generated, and it may be necessary to update the target knowledge graph. The tent triple form allows the knowledge graph to be flexibly extended and updated without reconstructing the entire target knowledge graph.

[0107] In one implementation, based on the principle of translational invariance, the fusion confidence of all tent triples is evaluated, the initial confidence of the initial knowledge graph is calculated according to the fusion confidence, and the reliability verification of the initial knowledge graph is completed according to the initial confidence, including the following steps:

[0108] For any tent triple, complete the vector translation projection between the head entity and the associated relationship in the tent triple to obtain a projection vector;

[0109] Calculate the vector distance between the tail entity in the tent triple and the projection vector, and calculate the local confidence of the tent triple according to the vector distance;

[0110] Take the head entity in the tent triple as the starting point;

[0111] Starting from the starting point, traverse the initial knowledge graph to search for all connected paths between the head entity and the tail entity, and merge all connected paths to generate a local directed graph;

[0112] Assign an initial resource value to the head entity, and the initial resource value circulates to the tail entity through iterative circulation in the local directed graph;

[0113] Calculate the tail resource value of the tail entity through the resource iteration formula according to the initial resource value, and quantify the entity association degree between the tail entity and the head entity according to the tail resource value;

[0114] Determine the global confidence of the tent triple according to the entity association degree;

[0115] Fuse the local confidence and the global confidence of each tent triple respectively to obtain the fusion confidence of all tent triples;

[0116] Calculate the confidence mean of all fusion confidences, and take the confidence mean as the initial confidence of the initial knowledge graph;

[0117] If the initial confidence level is greater than the preset confidence threshold, it is determined that the reliability verification of the initial knowledge graph passes;

[0118] If the initial confidence level is less than or equal to the confidence threshold, it is determined that the reliability verification of the initial knowledge graph fails.

[0119] In this embodiment, the head entity refers to the key entity or entity attribute located at the head of the tent triple. Similarly, the tail entity is the key entity or entity attribute located at the tail of the tent triple. For example, in (polyester fiber, used for, tarpaulin), polyester fiber is the head entity and tarpaulin is the tail entity. For any tent triple, using the principle of translational invariance, first map the head entity, relationship, and tail entity in the tent triple to a low-dimensional space to obtain the head vector, relationship vector, and tail vector. Perform vector translation projection on the head entity and relationship in the tent triple, that is, project the head vector and relationship vector onto the same plane and then approach the tail vector through translation operations to obtain the projection vector between the two. The principle of translational invariance means that in geometric transformations, no matter how a graph is translated in space, its basic characteristics (such as shape, size, and direction) remain unchanged. Therefore, the smaller the vector distance between the projection vector obtained after the head entity and relationship complete vector translation projection and the tail vector, the higher the correlation degree between the head entity and the tail entity inside the tent triple. And the higher the correlation degree, the higher the confidence level of the tent triple. According to this theory, the local confidence level of the tent triple can be calculated based on the vector distance between the projection vector and the tail vector. The calculation formula is as follows:

[0120]

[0121] Where J represents the local confidence level of the tent triple, ||x + y - z|| represents the vector distance between the projection vector and the tail vector, x represents the head vector, y represents the relationship vector, and z represents the tail vector.

[0122] Due to the large amount of information in the initial knowledge graph, including a large number of nodes and edges, there are multiple paths between the head entity and the tail entity within the same tent triple. The total amount of resources passed from the head entity to the tail entity through multiple paths can be used to quantify the association degree between the two entities. If the total amount of resources passed to the tail entity is large, it indicates that the association degree between the head entity and the tail entity of the tent triple is also higher, and it also indicates that the global confidence of the tent triple in the initial knowledge graph is higher. Specifically, first take the head entity in the tent triple as the starting point, start from the starting point, and gradually expand in different directions along the association relationships in the initial knowledge graph until reaching the tail entity that is in the same tent triple as the starting entity, and integrate the connected paths that can connect the two to generate a local directed graph. The process of searching for connected paths can be implemented using depth-first search (DFS) or breadth-first search (BFS). Taking breadth-first search (BFS) as an example, start from the starting point and expand along any path in the initial knowledge graph. If the corresponding tail entity can be reached, then take this path as a connected path, and mark the nodes in this path as visited. Then continue to explore other paths from the starting point. Similarly, take the path that can reach the corresponding tail entity as a connected path, and mark the nodes in this connected path that have not been marked as visited as visited until no connected path between the head entity and the tail entity can be found. Then integrate all the nodes marked as visited to obtain a local directed graph with the direction from the head entity to the tail entity.

[0123] Assign an initial resource value (which can be 1) to the head entity. The initial resource value starts from the head entity and iteratively circulates until it reaches the tail entity through all the nodes in the local directed graph. Then use the resource iteration formula to calculate the remaining resource value of the initial resource value reaching the tail entity, that is, the tail resource value. The larger the tail resource value, the greater the association degree (entity association degree) between the tail entity and the head entity. The greater the association degree, the greater the global confidence of the tent triple containing the tail entity and the head entity. Therefore, the entity association degree between the tail entity and the head entity can be quantified according to the tail resource value, and the entity association degree is normalized to the interval [0, 1] as the global confidence of the tent triple. The resource iteration formula is as follows:

[0124]

[0125] Where, G(xy) is the tail resource value flowing from the head entity x to the tail entity y; μ is the resource circulation probability; N is the total number of nodes in the local directed graph; α is the set of nodes between the head entity x and the tail entity y; z is the adjacent node adjacent to the tail entity y; G(zy) is the tail resource value flowing from the adjacent node z to the tail entity y; D z，y is the edge weight between the adjacent node z and the tail entity y; Cz is the out-degree of the adjacent graph node z.

[0126] In addition, for the edge weights between adjacent nodes and tail entities, the local confidence of the tent triples can be used to assign temporary weights to the edges and nodes of the initial knowledge graph.

[0127] By fusing the local confidence and global confidence of each tent triple, a weighted average of the local confidence and the global confidence can be performed to obtain the fused confidence. The average value of all fused confidences, i.e., the confidence mean, is calculated, and the calculated average value is used as the overall confidence of the initial knowledge graph, i.e., the initial confidence. If the initial confidence is greater than the preset confidence threshold, it is determined that the reliability check of the initial knowledge graph has passed, indicating that the initial knowledge graph can be used for subsequent parameter optimization of the tent samples. If the initial confidence is less than or equal to the confidence threshold, it is determined that the reliability check of the initial knowledge graph has failed, indicating that there may be errors in the construction process of the initial knowledge graph, or there are errors in the tent production information used to construct the initial knowledge graph. Therefore, it is necessary to eliminate the tent triplets with fused confidence less than the preset fused confidence threshold, and then reconstruct the initial knowledge graph. If the reliability check of the reconstructed initial knowledge graph still fails, it means that there may be errors in the tent production information. The error output reminds the relevant staff of the target factory to check the tent production information.

[0128] Since the performance optimization of military tents is based on tent production information and the initial knowledge graph constructed using the tent production information, once an error occurs in the construction process of the initial knowledge graph, or there is erroneous information in the tent production information, then the subsequent performance optimization of the military tent will most likely not be successful. It is even possible that the optimal parameters obtained will not only fail to optimize the performance of the military tent, but will instead reduce the performance of the military tent, which runs counter to the original optimization goal. Therefore, it is necessary to verify the reliability of the initial knowledge graph. Only when the reliability verification is passed can the subsequent steps be continued.

[0129] In one embodiment, based on the initial knowledge graph and the Bayesian theory, the correlation factors of the stress concentration area are reversely traced, and key weights are assigned to the initial knowledge graph according to the correlation factors to obtain the target knowledge graph, which includes the following steps:

[0130] Assign initial weights to the edges and nodes of the initial knowledge graph based on the fusion confidence, and update the initial knowledge graph according to the initial weight assignment results;

[0131] Generate stress entities based on stress concentration areas, add the stress entities to the updated initial knowledge graph, and obtain an intermediate knowledge graph;

[0132] Map the intermediate knowledge graph to an intermediate knowledge network, trace the correlation factors of stress entities through the intermediate knowledge network, and assign key weights to the intermediate knowledge graph according to the tracing results of the correlation factors to obtain the target knowledge graph.

[0133] In this embodiment, first assign initial weights to the edges and nodes of the initial knowledge graph according to the fusion confidence, and update the initial knowledge graph according to the initial weight assignment results. For example, if the confidence of tent triple A is a, then the weights of the nodes and edges corresponding to the head entity, tail entity, and association relationship in tent triple A in the initial knowledge graph are also a. Determine the tent components corresponding to the stress concentration area (such as strut nodes, seams, or peg anchors), find the key entity corresponding to the tent component in the initial knowledge graph as the area entity, and at the same time determine the stress direction, stress mean, and stress peak of the stress concentration area according to the stress distribution cloud map, and add the stress direction, stress mean, and stress peak as stress entities to the initial knowledge graph and connect them to the area entity to realize the secondary update of the initial knowledge graph, obtain an intermediate knowledge graph containing stress entities, and set the fusion confidence between all stress entities and area entities to the maximum confidence (which can be 1).

[0134] Extract the direct and indirect dependency relationships between intermediate nodes through graph traversal algorithms. Commonly used graph traversal algorithms include depth-first search (DFS) and breadth-first search (BFS), etc. Construct multiple dependency matrices based on the direct dependency relationship, indirect dependency relationship, and initial confidence, and superimpose and merge all the dependency matrices to obtain a fused relationship matrix. Specifically, complete the type division of intermediate nodes according to the information type of tent production information to obtain multiple sets of intermediate nodes of different types, and evenly divide the set of intermediate nodes into multiple subsets of intermediate nodes. Construct a dependency matrix according to the direct dependency relationship, indirect dependency relationship between all nodes in the subset of intermediate nodes, and the fused confidence corresponding to the intermediate nodes. Expand and superimpose and merge all the dependency matrices corresponding to the same set of intermediate nodes to obtain a first-level dependency matrix. Then expand and superimpose and merge all the first-level dependency matrices corresponding to all sets of intermediate nodes to obtain a fused relationship matrix. Traverse the fused relationship matrix, identify and delete the conflict relationships in the fused relationship matrix to obtain the target relationship matrix. In addition, since the Bayesian network is a network without a cycle structure, it is necessary to use a loop detection algorithm to identify whether there is a cycle structure in the relationship directed graph. Commonly used loop detection algorithms include depth-first search (DFS), breadth-first search (BFS), strongly connected component algorithm, etc. If there is a cycle structure in the relationship directed graph, delete the relationship edges with a fused confidence less than the preset fused confidence threshold until there is no cycle structure in the relationship directed graph. If there is no cycle structure in the relationship directed graph, use the relationship directed graph as the intermediate knowledge network. Then, find the stress network node corresponding to the stress entity in the intermediate knowledge network as the starting node, and trace back the parent node of the stress network node and the parent node of its parent node (such as the intermediate node corresponding to the sewing process) in reverse from the starting node until reaching the root node of the intermediate knowledge network (such as the intermediate node corresponding to the process parameters). Take all the obtained parent nodes and root nodes as the associated factors of the stress entity, which can be considered as the associated factors of the stress concentration area. The associated factor refers to the factor that will affect the appearance of the stress concentration area and the stress direction, stress mean value, and stress peak value of the stress concentration area, such as raw material type, sewing process, and process parameters, etc. Finally, assign key weights according to the intermediate nodes corresponding to the associated factors, and correct the initial weights according to the key weight assignment results to obtain the target knowledge graph after two weight assignments. For example, if the initial weight of fire resistance performance is 0.5 and the assigned key weight is 0.8, then use 0.8 to correct 0.5, that is, calculate 0.8×0.7 + 0.5×0.3 = 0.71, that is, the final weight of fire resistance performance is 0.71.

[0135] Due to conflicts and contradictions between various performance characteristics of military tents, such as waterproofness and breathability, increasing the thickness of the surface coating can improve waterproofness, but this will likely reduce breathability. Increasing the thickness of the tarpaulin to improve the structural strength of the tent will increase the weight, affecting its portability. Therefore, optimizing the performance of a military tent cannot focus solely on a single performance characteristic; rather, it is necessary to consider a wide range of performance characteristics to achieve overall performance optimization. Improving the overall performance of a military tent inevitably requires making trade-offs between different performance characteristics. Of course, these trade-offs are relative, and there is no such thing as drastically ignoring certain performance characteristics. The decision on how to make these trade-offs requires considering the weights of the key entities, entity attributes, and relationships corresponding to different performance characteristics in the target knowledge graph. Fusion confidence can better reflect the reliability of the information in the target knowledge graph. This information reliability is a prerequisite for subsequently selecting optimal parameters. Therefore, initial weights can be assigned to the target knowledge graph based on the fusion confidence. The stress concentration area of a military tent can affect many properties of the tent, such as its structural reliability, wind resistance and service life. For example, reducing the stress peak at the connection between tent poles or tarpaulins (such as optimizing the ply thickness or adopting a flexible transition design) can reduce the risk of material fatigue damage and extend the durability of the tent in repeated stress scenarios (such as long-term outdoor use). However, if the stress peak in the stress concentration area is increased, when located in a strong wind environment, it is very likely to cause local tearing or breakage of the military tent. Based on this, in the process of optimizing the performance of the tent, it is necessary to pay attention to the stress concentration area of the tent, that is, when optimizing the performance of the military tent, it is necessary to avoid the stress concentration area as much as possible or minimize the stress peak in the stress concentration area as much as possible. Therefore, it is necessary to use the target knowledge graph to extract the correlation factors that have direct and indirect relationships with the stress entity, and quantify the correlation between the correlation factor and the stress entity. The correlation degree is used as the key weight of the correlation factor for the node in the target knowledge graph. Therefore, in the subsequent process of optimizing the tent performance, the performance corresponding to the key entities and associations with higher weight values in the target knowledge graph will be given a higher optimization priority. Similarly, the higher optimization priority here is still relative, and the priority will not differ too much. The allocation of optimization priority is only to try to balance the conflicting performance, such as structural strength and portability. If the key entities and associations corresponding to structural strength have higher weights, then when optimizing these two performances, the portability of the military tent will be slightly sacrificed, and its structural strength will be prioritized. Similarly, there will still be excessive attention to its structural strength, and the portability of the military tent will still be taken into account, so that the overall performance of the military tent is improved.

[0136] In one implementation, mapping the intermediate knowledge graph to an intermediate knowledge network, tracing the correlation factors of stress entities through the intermediate knowledge network, and assigning key weights to the intermediate knowledge graph according to the tracing results of the correlation factors to obtain the target knowledge graph includes the following steps:

[0137] Traverse the intermediate knowledge graph and extract the direct and indirect dependency relationships between all intermediate nodes in the intermediate knowledge graph;

[0138] Construct multiple dependency matrices based on the direct dependency relationship, indirect dependency relationship, and initial confidence, and superimpose and merge all dependency matrices to obtain a fused relationship matrix;

[0139] Traverse the fused relationship matrix, identify the conflict relationships in the fused relationship matrix, and prune the fused relationship matrix according to the conflict relationship identification results to obtain the target relationship matrix;

[0140] Generate a relationship directed graph based on the target relationship matrix, and use a loop detection algorithm to identify whether there is a loop structure in the relationship directed graph;

[0141] If there is a loop structure in the relationship directed graph, prune the relationship edges in the relationship directed graph according to the fused confidence until there is no loop structure in the relationship directed graph;

[0142] If there is no loop structure in the relationship directed graph, use the relationship directed graph as the intermediate knowledge network;

[0143] Take the stress network node corresponding to the stress entity in the intermediate knowledge network as the starting node;

[0144] Starting from the starting node, trace the parent nodes of the stress network nodes level by level until reaching the root node of the intermediate knowledge network, and extract the correlation factors of the stress entities according to the parent node tracing results;

[0145] Use the Bayesian inference algorithm to calculate the correlation probabilities between all correlation factors and the stress entities, and assign key weights to the intermediate nodes corresponding to each correlation factor according to the correlation probabilities to obtain the target knowledge graph.

[0146] In this embodiment, the direct dependency means that a certain intermediate node directly affects other intermediate nodes. For example, if 50D (nylon parameter) can directly affect the tent weight (performance parameter), there is a direct dependency between 50D and the tent weight, that is, the intermediate node corresponding to 50D directly points to the intermediate node corresponding to the tent weight. An intermediate node refers to a node in the intermediate knowledge graph. The indirect dependency means that an intermediate node affects other intermediate nodes through several other intermediate nodes. For example, the raw material type can determine the raw material parameters, and the raw material parameters can directly affect the tent performance parameters. Therefore, there is an indirect dependency between the raw material type and the performance parameters. In addition, during the construction of the intermediate knowledge graph, and by analyzing the context relationship of key entities in the tent production information, the association relationships and entity attributes between all key entities are identified, and multiple tent triples are generated therefrom. Therefore, the direct and indirect dependencies between intermediate nodes can be directly extracted through graph traversal algorithms. Commonly used graph traversal algorithms include depth-first search (DFS) and breadth-first search (BFS), etc. Taking breadth-first search as an example, select an intermediate node as the starting node. Starting from the starting node, first visit all its adjacent nodes to obtain the direct dependencies between the starting node and other intermediate nodes, and then visit the adjacent nodes of these adjacent nodes in turn to obtain the direct dependencies between other intermediate nodes and the indirect dependencies between the starting node and other intermediate nodes, and so on, until all intermediate nodes are traversed to obtain the direct and indirect dependencies between all intermediate nodes. Since the extracted direct and indirect dependencies are local dependencies between intermediate nodes, and the reasons for the stress concentration area in the military tent and the increase of the stress peak value in the stress concentration area are relatively complex. Analyzing the intermediate nodes pairwise one by one easily ignores the hidden factors affecting the stress concentration area. The Bayesian network (intermediate knowledge network) can model through hidden variables to represent complex non-linear relationships, such as analyzing the combined influence of information parameters and material parameters on the stress concentration area. Therefore, it is necessary to construct a Bayesian network, that is, an intermediate knowledge network, based on the direct and indirect dependencies between intermediate nodes, and use the intermediate knowledge network to analyze the explicit and hidden factors affecting the stress concentration area from a global perspective, and finally integrate the explicit and hidden factors to obtain the correlation factors of the stress concentration area.

[0147] Refer to Figure 4 , constructing multiple dependency matrices based on the direct dependency, indirect dependency, and initial confidence, and superimposing and merging all the dependency matrices to obtain a fusion relationship matrix includes the following steps:

[0148] 1.1. Complete the type division of intermediate nodes according to the information types of tent production information to obtain multiple intermediate node sets of different types, such as a process node set, a performance node set, etc.;

[0149] 1.2. For any intermediate node set, evenly divide the intermediate node set into multiple intermediate node subsets;

[0150] 1.3. For any intermediate node subset, construct a dependency relationship matrix based on the direct dependency relationship, indirect dependency relationship among all nodes in the intermediate node subset, and the fusion confidence level corresponding to the intermediate nodes. Specifically, the elements of the dependency relationship matrix are the fusion confidence levels among all intermediate nodes. For example, if the fusion confidence level of the tent triple (nylon, used for, tarpaulin) is 0.8, then the fusion confidence level between the intermediate nodes corresponding to nylon and tarpaulin is 0.8, and the element at the corresponding position in the dependency relationship matrix constructed using the intermediate nodes corresponding to nylon and tarpaulin is 0.8;

[0151] 1.4. Expand and then superimpose and merge all the dependency relationship matrices corresponding to the same intermediate node set to obtain a first-level dependency relationship matrix. Specifically, first expand the number of rows and columns of each dependency relationship matrix to the number of global intermediate nodes. For example, if the global intermediate nodes are {1, 2, 3, 4} and the dependency relationship matrix is 2×2, then it needs to be expanded to 4×4, and fill zeros in the corresponding positions of the expanded rows and columns, that is, the elements in the first two rows / columns have values, and the elements in the remaining positions are zero. After all the dependency relationship matrices are expanded, perform superimposing and merging, which means adding the elements in the same positions of the dependency relationship matrices. Refer to Figure 3 and superimpose and merge the n dependency relationship matrices corresponding to the first intermediate node set to obtain the first-level dependency relationship matrix G′1 of the first intermediate node set, and superimpose and merge the n dependency relationship matrices corresponding to the second intermediate node set to obtain the first-level dependency relationship matrix G′2 of the second intermediate node set;

[0152] 1.5. Expand and then superimpose and merge all the first-level dependency relationship matrices corresponding to all intermediate node sets to obtain a fusion relationship matrix. For example, as Figure 3 shown, superimpose and merge the first-level dependency relationship matrix G′1 and the first-level dependency relationship matrix G′2 to obtain the fusion relationship matrix G M .

[0153] Divide all intermediate nodes into multiple intermediate node sets, and divide the intermediate node sets into multiple intermediate node subsets because there are many intermediate nodes. If a fusion relationship matrix is directly constructed, errors are likely to occur. To improve the reliability of the fusion relationship matrix, it is necessary to divide the intermediate nodes into subsets.

[0154] Traverse the fusion relationship matrix to identify the conflict relationships in the fusion relationship matrix to obtain a target relationship matrix. For example, refer to Figure 5, there are directed edges that are reverse to each other in the fusion relationship matrix, and there is a contradiction in the directed edges constructed by intermediate node A and intermediate node B. The directed edges C→B and C→A are normal directed edges. At the same time, the directed edges A→B and B→A appear, and the elements in their corresponding positions are not zero, then it is regarded as a contradiction. The directed edge A→B with a smaller element value can be set to zero, and the directed edge B→A with a larger element value is retained. Finally, the target relationship matrix is converted into a relationship directed graph, that is, the directed edges with non-zero elements are connected to generate a relationship directed graph. For example, the directed edges C→B and C→A are connected to generate a relationship directed graph in which both A and B are connected to C.

[0155] Since the Bayesian network is a network without a cycle structure, it is necessary to use a cycle detection algorithm to identify whether there is a cycle structure in the relationship directed graph. Commonly used cycle detection algorithms include depth-first search (DFS), breadth-first search (BFS), strongly connected component algorithm, etc. Taking the strongly connected component algorithm as an example, through the strongly connected component algorithm (such as Tarjan algorithm or Kosaraju algorithm), the relationship directed graph is divided into multiple strongly connected components. If the size of a strongly connected component is greater than 1, it means that there is a cycle structure in the relationship directed graph. On the contrary, if the strongly connected components are all less than 1, it means that there is no cycle structure in the relationship directed graph. If the relationship directed graph has a cycle structure, delete the relationship edges in the relationship directed graph whose fusion confidence is less than the preset fusion confidence threshold until there is no cycle structure in the relationship directed graph. If the relationship directed graph has no cycle structure, use the relationship directed graph as the intermediate knowledge network. Then, find the stress network node corresponding to the stress entity in the intermediate knowledge network as the starting node, and trace back the parent node of the stress network node and the parent node of its parent node (such as the intermediate node corresponding to the sewing process) in the reverse direction from the starting node until reaching the root node of the intermediate knowledge network (such as the intermediate node corresponding to the process parameters). All the obtained parent nodes and the root node are used as the correlation factors of the stress entity, which can be considered as the correlation factors of the stress concentration area.

[0156] The Bayesian inference algorithm refers to determining the influence degree of each correlation factor on the stress concentration area, that is, the correlation probability, through prior probability and conditional probability. The prior probability is the initial estimate of the probability of an event (the occurrence of a stress concentration area) based on historical experience or domain knowledge before new data or observation results are obtained. Stress analysis can be carried out on multiple tent samples with different performances, and the stress concentration areas of all tent samples can be identified through finite element simulation. The proportion of samples with stress concentration areas and the stress peak values in the stress concentration areas greater than the preset peak threshold among all tent samples is used as the prior probability. The conditional probability is the revised estimate of the probability of an event after obtaining new data or observation results, combined with the prior probability and current evidence. Specifically, the relevant parameters of the correlation factor in tent samples with different performances when the stress concentration area appears are statistically analyzed. For example, if the correlation factor is the sewing process, the relevant parameter is a width of 20 mm. The probability of the stress concentration area appearing when the relevant parameters of the correlation factor are the same is statistically analyzed as the conditional probability. After calculating the conditional probabilities of all correlation factors, the correlation probabilities of all correlation factors are calculated by combining the conditional probability and the prior probability. For example, the average value of the conditional probability and the prior probability can be calculated as the correlation probability of the correlation factor. Finally, according to the correlation probability, key weights are assigned to the corresponding intermediate nodes of each correlation factor, that is, the greater the correlation probability, the higher the assigned key weight, and vice versa, the smaller the correlation probability, the lower the assigned key weight, and finally the target knowledge graph with two weight assignments completed is obtained.

[0157] In one implementation, the performance correlation parameters in the target knowledge graph are screened according to the preset performance optimization objectives, and the performance correlation parameters are optimized for multiple objectives. The steps to obtain the optimal production parameters of the tent samples are as follows:

[0158] Infer the performance correlation path in the target knowledge graph according to the preset performance optimization objective, and extract the performance correlation parameters corresponding to the performance optimization objective based on the performance correlation path;

[0159] Define the performance optimization weight of the performance optimization objective according to the weight of the target knowledge graph;

[0160] Generate a multi-objective optimization function by combining the tent performance optimization objective and the performance optimization weight;

[0161] Based on the multi-objective optimization function, use the multi-objective optimization algorithm to iteratively optimize the performance correlation parameters until the preset maximum number of optimizations is reached, and output the optimal production parameters of the tent samples.

[0162] In this embodiment, the performance optimization objective refers to the type of performance that needs to be optimized, such as the wind resistance level, fire resistance performance, etc. Then, the key entities corresponding to the performance optimization objective are found in the target knowledge graph, and the found key entities are marked as performance entities. Starting from the performance entities, use the graph traversal algorithm to traverse all performance association paths in reverse, that is, the paths that have an association relationship with the performance entities, until a key entity with content as a parameter (such as 50D, elastic modulus 70GPa, etc.) or the length of the performance association path is greater than or equal to the preset length threshold is found. After finding all the performance association paths, extract the performance association parameters (such as raw material thickness) in all the performance association paths. At the same time, allocate a performance optimization weight to the performance optimization objective according to the weight of the performance entity in the target knowledge graph. The weight in the target knowledge graph refers to the initial weight or the final weight obtained after the key weight modifies the initial weight. For example, the weight of the wind resistance level in the target knowledge graph is 0.6, and the sum of the weights of all performance entities in the target knowledge graph is 2. Calculate 0.6 divided by 2 equals 0.3, then the performance optimization weight of the wind resistance level is 0.3. Generate a multi-objective optimization function according to the tent performance optimization objective and the performance optimization weight. Taking the performance optimization objective as maximizing the wind resistance level and fire resistance performance and minimizing the tent weight as an example, the multi-objective optimization function is as follows:

[0163] E = μ1f(x) + μ2f(y) - μ3f(z)

[0164] Among them, μ1, μ2, and μ3 are the performance optimization weights of the wind resistance level f(x), fire resistance performance f(y), and tent weight f(z), respectively.

[0165] In addition, while constructing the multi-objective optimization function, the following constraint conditions also need to be defined to ensure that the optimization results are within the actual feasible range. These constraint conditions include: restricting the optimization ranges of performance optimization objectives such as the wind resistance level, fire resistance performance, and tent weight; restricting parameters such as process parameters and raw material parameters to ensure that they are within the achievable range.

[0166] Finally, based on the multi-objective optimization function, use the multi-objective optimization algorithm to iteratively optimize the performance association parameters until the preset maximum number of optimizations is reached, and output the optimal production parameters of the tent sample. Commonly used multi-objective optimization algorithms include non-dominated sorting genetic algorithm, multi-objective particle swarm algorithm, etc. Taking the non-dominated sorting genetic algorithm as an example, its optimization steps include:

[0167] 2.1. Encode all performance association parameters as real number vectors, and set the parameter feasible region, for example (such as raw material thickness ∈ [0.5, 2.0] mm);

[0168] 2.2. Randomly generate an initial population containing N (N can be 150) individuals, and each individual corresponds to a combination of performance association parameters;

[0169] 2.3. Pairwise comparison is performed on all individuals. If individual A is not inferior to individual B in the multi-objective optimization function, then A dominates B. Based on this, the entire population is divided into multiple non-dominated levels. The first level is the current optimal solution set, and the subsequent levels are sub-optimal solutions in turn.

[0170] 2.4. For individuals within the same non-dominated level, calculate their crowding distance (Euclidean distance between adjacent individuals) in the multi-objective optimization function space. The larger the distance, the higher the diversity of the solution set. Priority is given to retaining boundary solutions and solutions in sparse regions.

[0171] 2.5. The tournament selection method is adopted. Priority is given to selecting individuals with a lower non-dominated level and a higher crowding degree as parents. Simulated binary crossover (SBX) is used to generate offspring, and the parameters of the parents are exchanged (such as the crossover rate of 0.8). Random perturbations are imposed on the parameters of the offspring (such as the mutation rate of 0.05) to explore new regions of the parameter space.

[0172] 2.6. The parent and offspring populations are merged to form a temporary population of size 2N. The merged population is re-sorted, and the first N optimal and highly diverse individuals are retained as the next-generation population.

[0173] 2.7. Repeat the above iterative optimization steps until the preset maximum number of optimization times, and output the optimal solution as the optimal production parameters of the tent sample.

[0174] This application also provides a military tent production management system based on performance prediction optimization, which is characterized by including:

[0175] A memory configured to store instructions; and

[0176] A processor configured to call instructions from the memory and capable of implementing the method for military tent production management based on performance prediction optimization according to any one of the above when executing the instructions.

[0177] Among them, the processor can adopt a central processing unit (CPU). Of course, according to the actual usage situation, other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. can also be adopted. The general-purpose processor can adopt a microprocessor or any conventional processor, etc. This application does not make any restrictions on this.

[0178] Among them, the memory can be an internal storage unit of the computer device, for example, the hard disk or memory of the computer device, or an external storage device of the computer device, for example, a plug-in hard disk, a smart media card (SMC), a secure digital card (SD), or a flash card (FC) equipped on the computer device. Moreover, the memory can also be a combination of the internal storage unit and the external storage device of the computer device. The memory is used to store computer programs and other programs and data required by the computer device. The memory can also be used to temporarily store data that has been output or will be output. This application does not limit this.

[0179] An embodiment of the present application also provides a machine-readable storage medium, on which instructions are stored, and the instructions are used to cause a machine to execute the above method for optimizing military tent production management based on performance prediction.

[0180] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0181] The present application is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram can be implemented by computer program instructions, and the combination of flows and / or blocks in the flowchart and / or block diagram can also be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate for implementing in the process Figure 1 a process or multiple processes and / or blocks Figure 1 a block or multiple blocks the device with the functions specified therein.

[0182] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including an instruction device, and the instruction device implements in the process Figure 1 a process or multiple processes and / or blocks Figure 1 a block or multiple blocks the functions specified therein.

[0183] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one process Figure 1 one process or multiple processes and / or blocks Figure 1 steps of the functions specified in one block or multiple blocks.

[0184] In a typical configuration, a computing device includes one or more processors (CPUs), an input / output interface, a network interface, and memory.

[0185] The memory may include non-permanent memory in the form of computer-readable media, random access memory (RAM), and / or non-volatile memory such as read-only memory (ROM) or flash memory (flash RAM). The memory is an example of computer-readable media.

[0186] Computer-readable media includes permanent and non-permanent, removable and non-removable media and can store information by any method or technology. The information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette tapes, magnetic disk storage or other magnetic storage devices, or any other non-transmission media that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transitory media such as modulated data signals and carrier waves.

[0187] It should also be noted that the term "comprising", "including" or any other variation thereof is intended to cover non-exclusive inclusion, so that a process, method, commodity or device comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, commodity or device. Without further limitation, an element defined by the statement "comprising an..." does not exclude the existence of additional identical elements in the process, method, commodity or device comprising the element.

[0188] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, various changes and modifications can be made to the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included within the scope of the claims of the present application.

Claims

1. A method for production management of military tents based on performance prediction and optimization, characterized in that: The method comprises the following steps: Obtain tent production information for tent samples in target factories; Construct a tent finite element model of the tent sample based on tent production information, complete stress analysis of the tent sample based on the finite element model, and identify stress concentration areas of the tent sample based on the stress analysis results; Use tent production information to generate multiple tent triplets, and build an initial knowledge graph based on all tent triplets; Based on the principle of translation invariance, the fusion confidence of all tent triplets is evaluated, the initial confidence of the initial knowledge graph is calculated based on all the fusion confidences, and the reliability of the initial knowledge graph is verified based on the initial confidence. If the reliability check of the initial knowledge graph passes, the correlation factors of the stress concentration area are traced back through the initial knowledge graph using Bayesian theory, and key weights are assigned to the initial knowledge graph according to the correlation factors to obtain the target knowledge graph; According to the preset performance optimization goals, the performance-related parameters in the target knowledge graph are screened out, and the performance-related parameters are optimized with multiple objectives to obtain the optimal production parameters of the tent sample.

2. The method according to claim 1, characterized in that: The tent production information includes tent structure information, tent raw material information, tent process information, tent performance information and production environment information.

3. The method according to claim 2, wherein The steps of constructing a tent finite element model of the tent sample based on the tent production information, performing stress analysis of the tent sample based on the finite element model, and identifying stress concentration areas of the tent sample based on the stress analysis results include the following steps: Constructing a three-dimensional tent model of the tent sample according to the tent structure information; Adaptively mesh the three-dimensional model of the tent and construct a finite element model of the tent; Apply multiple types of loads to the tent finite element model and calculate the stress distribution cloud map of the tent sample under the coupling of multiple types of loads; The stress distribution law analysis of the tent sample is completed according to the stress distribution diagram, and the stress concentration area of the tent sample is identified according to the stress distribution law analysis results.

4. The method according to claim 1, characterized in that: The method of generating multiple tent triples using tent production information and constructing an initial knowledge graph based on all tent triples includes the following steps: Identify all key entities in tent production information based on entity naming rules; Obtain the contextual relationships of all key entities in the tent production information, and identify the association relationships and entity attributes between all key entities based on the contextual relationships; Generate multiple tent triples based on all key entities, relationships and entity attributes; An initial knowledge graph is constructed based on all tent triples. The nodes of the initial knowledge graph are used to represent key entities, and the edges of the initial knowledge graph are used to represent the association relationships between key entities.

5. The method according to claim 1, characterized in that The method of evaluating the fusion confidence of all tent triplets based on the translation invariance principle, calculating the initial confidence of the initial knowledge graph based on the fusion confidence, and completing the reliability verification of the initial knowledge graph based on the initial confidence includes the following steps: For any tent triplet, complete the vector translation projection between the head entity and the association relationship in the tent triplet to obtain the projection vector; Calculate the vector distance between the tail entity and the projection vector in the tent triplet, and calculate the local confidence of the tent triplet based on the vector distance; Take the head entity in the tent triplet as the starting point; Starting from the starting point, traverse the initial knowledge graph to search for all connected paths between the head entity and the tail entity, and merge all connected paths to generate a local directed graph; Assign an initial resource value to the head entity, and the initial resource value is iteratively circulated to the tail entity through the local directed graph; The tail resource value of the tail entity is calculated based on the resource iteration formula and the initial resource value, and the entity association degree between the tail entity and the head entity is quantified based on the tail resource value; Determine the global confidence of the tent triple based on the entity association; The local confidence and global confidence of each tent triple are fused separately to obtain the fused confidence of all tent triplets; Calculate the confidence mean of all fused confidences and use the confidence mean as the initial confidence of the initial knowledge graph; If the initial confidence is greater than the preset confidence threshold, the reliability check of the initial knowledge graph is determined to have passed; If the initial confidence is less than or equal to the confidence threshold, it is determined that the reliability check of the initial knowledge graph has failed.

6. The method according to claim 5, characterized in that, The resource iteration formula is as follows: Among them, G(xy) is the value of the tail resource flowing from the head entity X to the tail entity y; μ is the resource flow probability; N is the total number of nodes in the local directed graph; α is the set of nodes between the head entity x and the tail entity y; z is the adjacent node adjacent to the tail entity y; G(zy) is the value of the tail resource flowing from the adjacent node z to the tail entity y; D z，y is the edge weight between the adjacent node z and the tail entity y; C z is the out-degree of the adjacent graph node z.

7. The method according to claim 1, characterized in that The method of reversely tracing the correlation factors of the stress concentration area based on the initial knowledge graph and the Bayesian theory, and assigning key weights to the initial knowledge graph according to the correlation factors to obtain the target knowledge graph includes the following steps: Assign initial weights to the edges and nodes of the initial knowledge graph based on the fusion confidence, and update the initial knowledge graph according to the initial weight assignment results; Generate stress entities based on stress concentration areas, add the stress entities to the updated initial knowledge graph, and obtain an intermediate knowledge graph; The intermediate knowledge graph is mapped into an intermediate knowledge network, the association factors of the stress entities are traced through the intermediate knowledge network, and key weights are assigned to the intermediate knowledge graph according to the tracing results of the association factors to obtain the target knowledge graph.

8. The method according to claim 7, wherein The process of mapping the intermediate knowledge graph into an intermediate knowledge network, tracing the correlation factors of stress entities through the intermediate knowledge network, and assigning key weights to the intermediate knowledge graph based on the tracing results of the correlation factors to obtain the target knowledge graph includes the following steps: Traverse the intermediate knowledge graph and extract the direct and indirect dependencies between all intermediate nodes in the intermediate knowledge graph; Based on direct dependency, indirect dependency and initial confidence, multiple dependency matrices are constructed, and all dependency matrices are superimposed and merged to obtain a fusion relationship matrix; Traverse the fusion relationship matrix, identify the conflict relationship in the fusion relationship matrix, and delete the fusion relationship matrix according to the conflict relationship identification result to obtain the target relationship matrix; Generate a directed relation graph based on the target relation matrix, and use a loop detection algorithm to identify whether the directed relation graph has a loop structure; If the relationship directed graph has a ring structure, the relationship edges in the relationship directed graph are deleted according to the fusion confidence until the relationship directed graph no longer has a ring structure; If the relationship directed graph does not have a ring structure, the relationship directed graph is used as an intermediate knowledge network; The stress network node corresponding to the stress entity in the intermediate knowledge network is used as the starting node; Starting from the starting node, the parent nodes of the stress network nodes are traced step by step until the root node of the intermediate knowledge network is reached, and the correlation factors of the stress entities are extracted based on the parent node tracing results; The Bayesian inference algorithm is used to calculate the association probability between all association factors and stress entities, and key weights are assigned to the intermediate nodes corresponding to each association factor according to the association probability to obtain the target knowledge graph.

9. The method according to claim 1, characterized in that: The method of screening the performance-related parameters in the target knowledge graph according to the preset performance optimization target, optimizing the performance-related parameters with multiple targets, and obtaining the optimal production parameters of the tent sample includes the following steps: Infer the performance association path in the target knowledge graph based on the preset performance optimization goal, and extract the performance association parameters corresponding to the performance optimization goal based on the performance association path; Define the performance optimization weight of the performance optimization target based on the weight of the target knowledge graph; Generate a multi-objective optimization function by combining the tent performance optimization goal and performance optimization weight; Based on the multi-objective optimization function, the performance-related parameters are iteratively optimized using a multi-objective optimization algorithm until the preset maximum number of optimizations is reached, and the optimal production parameters of the tent sample are output.

10. A military tent production management system based on performance prediction optimization, characterized in that, include: a memory configured to store instructions; as well as A processor is configured to call instructions from a memory and implement the method for military tent production management based on performance prediction optimization according to any one of claims 1 to 9 when executing the instructions.

Images