A method and device for recommending process parameters for robotic drilling

Abstract

The robot drilling process parameter recommendation method and device provided in the application comprise the following steps: calculating a weighted cosine similarity between a target machining quality parameter and a historical machining scheme, and selecting a candidate machining scheme set based on the weighted cosine similarity; judging whether the candidate machining scheme meets a triggering condition of a rule in a preset rule library, to obtain a judgment result; determining an overall quality score of the candidate machining scheme set according to the judgment result and the weighted cosine similarity corresponding to each candidate machining scheme; determining an individual quality score of the candidate machining scheme; determining a confidence degree of the candidate machining scheme according to the individual quality score and the overall quality score; selecting a recommended machining scheme according to the confidence degree of the candidate machining scheme; when the recommended machining scheme meets a target rule, performing a correction operation on the machining parameter in the recommended machining scheme based on a correction operation indicated by the target rule, otherwise, keeping the original machining parameter, to obtain a corrected machining scheme; and recommending the corrected machining scheme to a user.

Description

Technical Field

[0001] This application relates to the field of robotic drilling and riveting technology, and in particular to a method and apparatus for recommending robotic drilling and riveting process parameters. Background Technology

[0002] Robotic drilling and riveting processes are widely used in modern manufacturing, and their processing quality directly affects the performance and reliability of the final product. The rational selection and adjustment of drilling and riveting process parameters are crucial for ensuring processing quality and improving production efficiency. With the development of digitalization and intelligentization in manufacturing processes, recommending process parameters based on historical processing data is gradually becoming an important means of improving process efficiency.

[0003] Existing methods for recommending process parameters typically rely on similarity metrics between historical data and target quality parameters, using historical processing schemes to help determine appropriate process parameters. However, in real-world production environments, processing quality parameters are multidimensional and complex, with significant differences in the relative importance of different parameters, making it difficult for a single similarity metric to provide accurate recommendations. Summary of the Invention

[0004] In view of this, this application provides a method and apparatus for recommending robot drilling and riveting process parameters, which is used to reliably and accurately recommend dual-robot drilling and riveting process parameters.

[0005] Specifically, this application is implemented through the following technical solution:

[0006] The first aspect of this application provides a method for recommending process parameters for robot drilling and riveting, the method comprising:

[0007] In response to the target processing quality parameters input by the user, the weighted cosine similarity between the target processing quality parameters and historical processing schemes in the pre-established dual-robot drilling and riveting process knowledge graph is calculated based on the weights pre-determined for each target processing quality parameter, and a set of alternative processing schemes is selected.

[0008] For each alternative processing scheme, determine whether the historical processing quality parameters of the alternative processing scheme meet the triggering conditions of the rules in the preset rule base, and obtain the judgment result;

[0009] Based on the judgment result and the weighted cosine similarity corresponding to each alternative processing scheme in the alternative processing scheme set, the overall quality score of the alternative processing scheme set is determined;

[0010] For each alternative processing scheme, based on the matching results between the alternative processing scheme and the rules in the preset rule base, the individual quality score of the alternative processing scheme is determined, and the confidence level of the alternative processing scheme is determined based on the individual quality score and the overall quality score.

[0011] Based on the weighted cosine similarity and confidence of each alternative processing scheme, a recommended processing scheme is selected from the recommended set of alternative processing schemes;

[0012] When the historical processing quality parameters in the recommended processing scheme meet the triggering conditions of the target rule in the preset rule base, the processing parameters in the recommended processing scheme are corrected based on the correction operation indicated by the target rule; otherwise, the original processing parameters are retained to obtain the corrected processing scheme.

[0013] The revised processing solution is recommended to the user.

[0014] A second aspect of this application provides a robot drilling and riveting process parameter recommendation device, the device comprising a screening module, a judgment module, a determination module, a correction module, and a recommendation module; wherein...

[0015] The filtering module is used to respond to the target processing quality parameters input by the user, calculate the weighted cosine similarity between the target processing quality parameters and the historical processing schemes in the pre-established dual-robot drilling and riveting process knowledge graph based on the weights determined in advance for each target processing quality parameter, and filter out a set of alternative processing schemes.

[0016] The judgment module is used to determine, for each alternative processing scheme, whether the historical processing quality parameters of the alternative processing scheme meet the triggering conditions of the rules in the preset rule base, and obtain the judgment result;

[0017] The determining module is used to determine the overall quality score of the set of alternative processing schemes based on the judgment result and the weighted cosine similarity corresponding to each alternative processing scheme in the set of alternative processing schemes.

[0018] The determining module is further configured to, for each alternative processing scheme, determine the individual quality score of the alternative processing scheme based on the matching situation between the alternative processing scheme and the rules in the preset rule base in the judgment result, and determine the confidence level of the alternative processing scheme based on the individual quality score and the overall quality score;

[0019] The determining module is further configured to select a recommended processing scheme from the recommended set of alternative processing schemes based on the weighted cosine similarity and confidence of each alternative processing scheme;

[0020] The correction module is used to correct the processing parameters in the recommended processing scheme based on the correction operation indicated by the target rule when the historical processing quality parameters in the recommended processing scheme meet the triggering conditions of the target rule in the preset rule base; otherwise, the original processing parameters are retained to obtain the corrected processing scheme.

[0021] The recommendation module is used to recommend the modified processing scheme to the user.

[0022] The robot drilling and riveting process parameter recommendation method and apparatus provided in this application have the following advantages: First, when screening a set of candidate processing schemes, by determining a weight for each target processing quality parameter and using these weights to calculate the weighted cosine similarity between the target processing quality parameter and historical processing schemes, historical processing schemes similar to the target processing requirements can be found more accurately, thereby screening a set of candidate processing schemes. Second, by matching the historical processing quality parameters in each candidate processing scheme with rules in a preset rule base, it is ensured that the candidate processing schemes not only resemble the target requirements but also strictly adhere to key rules such as process safety, quality, and efficiency, effectively filtering out schemes that do not meet process requirements and improving the overall quality level of the recommended schemes. Third, by combining the overall quality score and individual quality score to calculate the confidence level of each candidate processing scheme, a comprehensive and quantifiable evaluation index is provided, which not only enhances the interpretability of the recommendation results but also provides a more intuitive and reliable basis. Fourth, when the recommended processing scheme meets specific rule triggering conditions, the technology can automatically correct the parameters, ensuring that the processing scheme can flexibly respond to dynamic changes in actual production, greatly improving the adaptability and robustness of the processing process. In this way, through precise matching, rule constraints, comprehensive evaluation, dynamic adjustment and recommendation, the accuracy, flexibility and practicality of the recommended drilling and riveting process parameters for dual robots are significantly improved, and reliable assurance of robot drilling and riveting operations is achieved. Attached Figure Description

[0023] Figure 1 A flowchart of Embodiment 1 of the method for recommending robot drilling and riveting process parameters provided in this application;

[0024] Figure 2 A schematic diagram illustrating the construction of a dual-robot drilling and riveting process knowledge graph provided for an exemplary embodiment of this application;

[0025] Figure 3 A schematic diagram of a dual-robot drilling and riveting process knowledge graph pattern layer provided for an exemplary embodiment of this application;

[0026] Figure 4 This is a hardware structure diagram of the robot drilling and riveting process parameter recommendation device, which is the robot drilling and riveting process parameter recommendation device of this application.

[0027] Figure 5 This is a schematic diagram of the structure of Embodiment 1 of the robot drilling and riveting process parameter recommendation device provided in this application. Detailed Implementation

[0028] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application.

[0029] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used herein are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any and all possible combinations of one or more of the associated listed items.

[0030] It should be understood that although the terms first, second, third, etc., may be used in this application to describe various information, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0031] The following specific embodiments are given to illustrate the technical solution of this application in detail.

[0032] Figure 1 This is a flowchart of an embodiment of the method for recommending robot drilling and riveting process parameters provided in this application. Please refer to... Figure 1 The method provided in this embodiment may include:

[0033] S101. In response to the target processing quality parameters input by the user, based on the weights pre-determined for each target processing quality parameter, calculate the weighted cosine similarity between the target processing quality parameter and the historical processing schemes in the pre-established dual-robot drilling and riveting process knowledge graph, and select a set of alternative processing schemes based on the calculated weighted cosine similarity.

[0034] Specifically, the data types of the target machining quality parameters input by the user are pre-set according to actual needs, and this embodiment does not limit this. For example, in one embodiment, the target machining quality parameters required for drilling and riveting operations can be determined according to the robot model. For example, in one embodiment, the target machining quality parameters are specifically shown in Table 1:

[0035] Table 1 Target processing quality parameters

[0036]

[0037] Furthermore, referring to the examples shown in Table 1, for example, in one embodiment, the target processing quality parameters input by the user are: piercing head height 2.0±0.05mm, piercing head diameter 7.0±0.1mm, riveting interference 0.05±0.02mm, nail head flatness ≤0.03mm, hole diameter 5.0±0.02mm, and burr height ≤0.04mm.

[0038] It should be noted that each target processing quality corresponds to a weight, which reflects the importance of different processing quality parameters in the recommendation process.

[0039] Optionally, in one possible implementation, the process of determining the weights for each target processing quality parameter is pre-defined:

[0040] (1) Evaluate the relative importance of any two target processing quality parameters in the processing quality, and form a judgment matrix based on the evaluation results.

[0041] Specifically, each processing quality parameter in the target processing quality parameters is compared with another processing quality parameter, and the comparison result is placed in the corresponding position of the matrix to form a judgment matrix.

[0042] Determine the elements in the matrix This can represent the relative importance of the i-th target machining quality parameter with respect to the j-th target machining quality parameter, and the elements are given based on the 1-9 scaling method. The value of .

[0043] Specifically, the judgment matrix can be as follows:

[0044] ;

[0045] Where A is the judgment matrix; elements It can represent the relative importance of the i-th target processing quality parameter to the j-th target processing quality parameter.

[0046] Specifically, the elements in the judgment matrix can be given based on the 1-9 scale method. The value is then determined based on expert experience to assess the importance of each target processing quality parameter, and the relatively important parameters are adjusted to larger values.

[0047] In a specific implementation, in one embodiment, the judgment matrix corresponding to all target processing quality parameters formed based on relative importance is as follows:

[0048] ;

[0049] Where A1 is the judgment matrix corresponding to all target processing quality parameters.

[0050] (2) Use the normalization method to solve for the eigenvectors and the largest eigenvalue of the judgment matrix.

[0051] Specifically, the normalization method is used to perform eigenvalue decomposition on the judgment matrix to solve for the matrix's eigenvectors and corresponding eigenvalues. At this point, the eigenvector corresponding to the largest eigenvalue is the required weight vector.

[0052] In practice, the feature vector can be calculated based on the following formula:

[0053] ;

[0054] in, Let be the eigenvector corresponding to the i-th row, and n be the row number of the judgment matrix, with n ranging from 1 to 6.

[0055] Furthermore, the largest eigenvalue can be calculated based on the following formula:

[0056] ;

[0057] in, It is the largest eigenvalue. Let i be the feature vector corresponding to the i-th row. Let be the eigenvector corresponding to the j-th row.

[0058] (3) Use the feature vector as the initial weight vector for all target processing quality parameters.

[0059] In a specific implementation, for example, in one embodiment, the feature vector obtained after normalization processing... As the initial weight vector.

[0060] (4) Calculate the consistency index of the judgment matrix based on the largest eigenvalue.

[0061] In practice, the consistency index of the judgment matrix can be calculated based on the following formula:

[0062] ;

[0063] Where CI stands for consistency index of the judgment matrix. It is the largest eigenvalue.

[0064] (5) Calculate the consistency ratio of the judgment matrix based on the consistency index and random consistency index of the judgment matrix, and when the consistency ratio is less than a preset threshold, determine the initial weight vector as the weight vector corresponding to all target processing quality parameters; otherwise, evaluate the relative importance of any two target processing quality parameters in processing quality again, and form a judgment matrix based on the evaluation results until the consistency ratio is less than the preset threshold.

[0065] Specifically, the formula for calculating the consistency ratio is as follows: .

[0066] Specifically, the specific value of the preset threshold is set according to actual needs, and this embodiment does not limit it. For example, in one embodiment, the preset threshold can be 0.1.

[0067] In specific implementation, for example, in one embodiment, the number of rows n of the determination matrix is ​​6, then... Substituting the value into the above formula, we get CI = 0.029. When n is 6, the random consistency index RI = 1.24. When CR < 0.1, the judgment matrix is ​​considered to have satisfactory consistency, and the calculated consistency ratio CI = 0.023 indicates that the judgment matrix has satisfactory consistency. For example, in another embodiment, if CR > 0.1, the consistency of the judgment matrix is ​​considered poor. Therefore, the relative importance of each target processing quality parameter is reassessed, and a new consistency ratio is calculated.

[0068] Understandably, accurately determining the weight of each target machining quality parameter through the analytic hierarchy process (AHP) significantly improves the accuracy of process parameter recommendations. First, pairwise comparisons are performed on the target machining quality parameters to construct a judgment matrix. Then, the relative importance of each parameter is quantified based on expert evaluation results. Subsequently, the eigenvectors and maximum eigenvalues ​​of the judgment matrix are solved using a normalization method to obtain the initial weight vector. By calculating the consistency index and consistency ratio, the consistency of the judgment matrix is ​​ensured to meet a preset threshold, thus obtaining reliable target machining quality parameter weights. This weight determination method not only enhances the scientific rigor of parameter recommendations but also effectively improves the accuracy and practicality of the intelligent recommendation system for dual-robot drilling and riveting process parameters.

[0069] In this step, instead of directly calculating the cosine similarity, the weighted cosine similarity between the target processing quality parameters and the historical processing schemes in the pre-established dual-robot drilling and riveting process knowledge graph is calculated, and a set of alternative processing schemes is selected based on the calculated weighted cosine similarity.

[0070] It should be noted that the dual-robot drilling and riveting process knowledge graph is a domain knowledge graph used to store and organize knowledge related to the dual-robot drilling and riveting process. The dual-robot drilling and riveting process knowledge graph combines information such as process knowledge, material properties, and processing parameters, and is represented and stored in the form of a graph structure.

[0071] The dual-robot drilling and riveting process knowledge graph is constructed based on historical drilling and riveting process data related to robot drilling and riveting processes. This knowledge graph adopts an ontology structure with tools, processes, workpieces, and processing results as core elements, and achieves the association and organization between multiple elements through processing scheme entities. Optionally, in one possible implementation, Figure 2 This is a schematic diagram illustrating the implementation principle of constructing a dual-robot drilling and riveting process knowledge graph, provided as an exemplary embodiment of this application. Please refer to... Figure 2 In constructing the knowledge graph of dual-robot drilling and riveting processes, this paper adopts a combination of top-down and bottom-up construction methods. First, through the analysis of historical process data and under the guidance of experts, a schema layer analysis is performed on the knowledge graph of dual-robot drilling and riveting processes to establish a good conceptual hierarchy. Then, the data layer is constructed by extracting knowledge of entities, relations, and attributes. Finally, the schema layer and data layer are associated and mapped and imported into the Neo4j graph database to obtain the knowledge graph of dual-robot drilling and riveting processes.

[0072] (1) Pattern layer design.

[0073] The knowledge graph architecture mainly consists of two parts: the schema layer and the data layer. The schema layer, located above the data layer, includes conceptual levels of knowledge classes such as processing steps, processing parameters, and processing quality in the production process, as well as the constraints between layers. The data layer stores specific data information in units of fact triples. Each entity in a triple contains specific attributes and corresponding values.

[0074] Table 2 Knowledge Graph Structure

[0075] structure content Example Example Pattern layer Data source knowledge structure Concepts and relationships Entity-Relationship-Entity, Entity-Attribute-Value Data layer Specific knowledge information Fact Triad Workpiece - Process Steps - Riveting

[0076] (2) Ontology construction.

[0077] A complete dual-robot drilling and riveting process includes multiple steps such as drilling, countersinking, riveting, and lubrication. Each step contains several processing steps, and each step describes the processing information of one or more features. Therefore, considering the completeness and logic of the dual-robot process knowledge representation, the process knowledge involved in the dual-robot drilling and riveting process is divided into the following five levels.

[0078] (1) Tool layer: The tool layer mainly includes drill bits, robots that perform drilling and riveting processes, and robot combinations consisting of two robots. The reason for this arrangement is that in the drilling and riveting of aerospace panels, the mass of the workpiece varies greatly, and if the robot load is less than the mass of the workpiece, the processing requirements cannot be met.

[0079] (2) Process layer: The process layer mainly includes specific processing parameters, which are mainly divided into drilling process parameters and riveting process parameters. Among them, the drilling process parameters include drilling parameters, such as drilling speed, drilling feed, countersinking speed, countersinking feed, and drilling depth; the riveting process parameters include top iron pressure, riveting pressure, hammer riveting time, top clamping force, and clamping force.

[0080] (3) Workpiece layer: The workpiece layer mainly includes the shape, size, and material of the workpiece to be processed. The material, as a sub-entity of the workpiece, includes the material name, strength, and thermal conductivity.

[0081] (4) Feature layer: mainly the processing results of the corresponding processing technology, including hole diameter, head height, head diameter, nail head flatness, and rivet interference.

[0082] (5) Case Layer (Processing Scheme Layer): This layer primarily serves as a bridge connecting the various entity layers, including the use of robots, the use of tools, the presence of workpieces to be processed, the use of drilling technology, the use of riveting technology, and the presence of drilling and riveting quality. The process information contained in the four layers of the dual-robot drilling and riveting process and the relationships between them allow for a comprehensive overview of the process knowledge for robotic drilling and riveting of complex thin-walled parts, improving the completeness, relevance, and reasoning speed of the process. Each processing scheme represents the use of robots, the use of tools, the presence of workpieces to be processed, the use of drilling technology, the use of riveting technology, and the presence of drilling and riveting quality; that is, which combination of robots performs the operation, which drill bit is used, which process parameters are called, and which workpiece is processed; and the corresponding processing quality parameters.

[0083] (3) Data layer storage.

[0084] After the ontology model is built, the data layer is stored: the schema layer and the data layer can be associated and mapped, and imported into the Neo4j graph database, and then a knowledge graph of the dual-robot drilling and riveting process can be built in the database, which includes entities, relationships and attributes.

[0085] In practical implementation, the data can be decrypted and organized into tables, stored in XML format, and automatically read from the XML data using the panda library in Python and stored in the Neo4j database. Simultaneously, ontology relationships are constructed. Some of the constructed ternary tables are shown in Table 1, and Table 3 shows the ternary tables of the schema layer in an exemplary embodiment of this application.

[0086] Table 3

[0087]

[0088] It should be noted that, in this embodiment, the constructed dual-robot drilling and riveting process knowledge graph includes 292 entities and 711 relationships.

[0089] Specifically, in this step, the target machining quality parameters and the pre-established dual-robot drilling and riveting process knowledge graph can be input into the similarity calculation model to obtain the machining schemes in the dual-robot drilling and riveting process knowledge graph that have a high weighted cosine similarity to the target machining quality parameters.

[0090] In specific implementation, the step of calculating the weighted cosine similarity between the target processing quality parameter and historical processing schemes in a pre-established dual-robot drilling and riveting process knowledge graph, based on pre-determined weights for each target processing quality parameter, includes:

[0091] (1) Normalize the target processing quality parameter and the historical processing quality parameter in the historical processing scheme respectively to obtain the normalized value of the target processing quality parameter and the normalized value of the historical processing quality parameter.

[0092] Specifically, a min-max normalization method can be used to scale the value of each target processing quality parameter to the [0,1] range to eliminate the differences in dimensions and numerical ranges between different target processing quality parameters, making them comparable.

[0093] Furthermore, the same normalization method as the target processing quality parameter can be used to scale the values ​​of the historical processing quality parameters to the [0,1] range, thereby eliminating the differences in dimensions and numerical ranges and making the historical processing quality parameters comparable to the target processing quality parameters.

[0094] (2) Based on the normalized values ​​of the target machining quality parameters and the normalized values ​​of the historical machining quality parameters, calculate the weighted cosine similarity between the target machining quality parameters and the historical machining schemes in the pre-established dual-robot drilling and riveting process knowledge graph according to the following formula:

[0095] ;

[0096] Among them, the The weighted cosine similarity is denoted as n, where n is the number of dimensions of the target processing quality parameter. Let xi be the weights pre-determined for the i-th dimension target processing quality parameter, yi be the normalized value of the i-th dimension target processing parameter, and yi be the normalized value of the i-th dimension historical processing quality parameter.

[0097] It is understandable that the closer the weighted cosine similarity value is to 1, the more similar the target processing quality parameters are to the processing quality parameters of the historical processing scheme; the closer it is to -1, the less similar they are; and the closer it is to 0, the less related they are.

[0098] Understandably, by calculating weighted pre-similarity, the target machining quality parameters are accurately matched with historical machining schemes in the dual-robot drilling and riveting process knowledge graph. Normalization eliminates the differences in dimensions and numerical ranges between parameters, making different parameters comparable. The weighted cosine similarity calculation fully considers the importance of parameters in each dimension. By introducing pre-determined weights, it more accurately reflects the degree of similarity between the target machining quality parameters and historical machining schemes, effectively improving the accuracy of process parameter recommendations and providing a more scientific and reliable basis for parameter selection in dual-robot drilling and riveting processes.

[0099] Furthermore, after calculating the weighted cosine similarity values, the set of the top K processing schemes can be selected as the candidate processing scheme set according to the order of weighted cosine similarity from high to low.

[0100] Furthermore, the specific value of K is set according to actual needs, and this embodiment does not limit it. For example, in one embodiment, K is 20.

[0101] S102. For each alternative processing scheme, determine whether the historical processing quality parameters in the alternative processing scheme meet the triggering conditions of the rules in the preset rule base, and obtain the judgment result.

[0102] Specifically, the preset rule base contains multiple rules, each rule indicating that when preset processing parameters meet preset access conditions, the specified processing parameters should be corrected according to the specified correction operation (i.e., the recommended measures corresponding to the rule); the multiple rules include basic rules, safety rules, quality rules, and efficiency rules.

[0103] Specifically, the pre-built rule base can include the category to which the rule belongs, the number of each rule, the triggering conditions of the rule, and the corresponding recommended measures:

[0104] For example, in one possible implementation, Table 4 is a partial table of the preset rule base shown in an exemplary embodiment of this application:

[0105] Table 4

[0106]

[0107] For each alternative processing scheme It matches the triggering conditions in the preset rule base, and after determining the matching triggering conditions, records the recommended measures corresponding to the triggering rule.

[0108] S103. Based on the judgment result and the weighted cosine similarity corresponding to each alternative processing scheme in the alternative processing scheme set, determine the overall quality score of the alternative processing scheme set.

[0109] Specifically, the judgment result can determine whether the processing plan triggers the rules in the table, and which rule is triggered. If it is triggered, it can be determined that the processing plan has a defect, so its score can be appropriately reduced. At the same time, combined with the weighted cosine similarity, the score of the high similarity can be increased to obtain its overall quality score.

[0110] Optionally, in one possible implementation, the specific implementation process of this step may include:

[0111] (1) Based on the judgment result, determine the data size score of the alternative processing scheme set.

[0112] Specifically, the data scale score is determined based on the total number of alternative processing schemes, the number of valid processing schemes in the alternative processing scheme set (processing schemes that do not match any rule in the preset rule base are considered valid processing schemes), and whether there are invalid processing schemes in the alternative processing scheme set (processing schemes that match any basic rule in the preset rule base are considered invalid processing schemes; in addition, those that meet the triggering rule are considered matches). The data rule score is then used to evaluate the "effectiveness" and "degree of conflict with the rule base" of a set of alternative processing schemes.

[0113] In practice, the data size score can be determined using the following formula:

[0114] ;

[0115] Among them, the Scoring is given for the data size, the The number of valid alternative processing schemes in the alternative processing scheme set that do not match the rules in the preset rule base. The number of alternative processing schemes included in the alternative processing scheme set; This is an indicator function. When the alternative processing scheme matches the basic rule in the preset rule set, its value is 0; otherwise, its value is 1.

[0116] Referring to the preceding description, to evaluate the effectiveness and innovativeness of the alternative processing scheme set, a data size score is set to reflect the extent to which the scheme set is outside the coverage of the rule base. When any alternative processing scheme in the alternative processing scheme set matches the basic rule, the data size score of the alternative processing scheme set is 0; when every alternative processing scheme in the alternative processing scheme set does not match the basic rule, the score is 0. When the value is 1, the data rule score in the set of alternative processing schemes is equal to... Using the above method, if any scheme matches the basic rule, the entire group of schemes is considered invalid (score of 0). Under the premise that all schemes do not match the basic rule, a score is given based on the proportion of valid schemes, reflecting the data scale and quality. Through the above scoring method, repetitive designs can be effectively screened out, and the scientificity and rationality of the processing scheme evaluation can be improved.

[0117] (2) Calculate the average weighted cosine similarity of each alternative processing scheme in the alternative processing scheme set, and determine the average value as the data quality score of the alternative processing scheme set.

[0118] Specifically, by calculating the average weighted cosine similarity of the alternative processing schemes, we can reflect the overall similarity between the set of alternative processing schemes and the target processing quality parameters. In other words, the quality level of the data can be determined by the degree of similarity.

[0119] In practice, the data quality score can be determined using the following formula:

[0120] ;

[0121] in, Score the data quality. is the weighted cosine similarity of the i-th alternative processing scheme, and N is the number of alternative processing schemes.

[0122] (3) Calculate the standard deviation of the weighted cosine similarity of each alternative processing scheme in the alternative processing scheme set, and determine the data distribution score of the alternative processing scheme set based on the standard deviation and the matching of the alternative processing scheme with the safety class rules in the preset rule set.

[0123] Specifically, the data distribution score reflects the degree of concentration and security of the alternative processing scheme set in terms of similarity distribution.

[0124] Furthermore, the standard deviation of the weighted cosine similarity corresponding to each alternative processing scheme can measure the degree of dispersion of the case set in the similarity distribution.

[0125] In practice, the data distribution score can be determined using the following formula:

[0126] ;

[0127] Where D is the data distribution score, and σ is the standard deviation of the weighted cosine similarity corresponding to each alternative processing scheme. The conflict intensity coefficient is initially set to 1. When a security conflict is triggered (when at least one alternative processing scheme matches at least one security-class rule in the preset rule set, it is considered that a security conflict has been triggered), the value is set to 1.5 to improve the sensitivity of the discreteness penalty.

[0128] It should be noted that by setting a threshold of 0.3, the negative impact of high dispersion on the overall quality score is suppressed while retaining the advantage of low dispersion, thus enhancing the model's robustness to data noise.

[0129] In this embodiment, firstly, once a security rule is detected to be triggered (i.e., a scheme matching a preset security rule exists), δconflict is automatically increased to 1.5, and the value is reduced. Lowering the data distribution score means that in security-sensitive scenarios, the system more strictly constrains design dispersion and encourages a more focused and secure design direction. Using this formula to calculate the data distribution score not only quantitatively measures the consistency and security matching of the solution set, but also dynamically adapts to conflict risks during the design process, effectively improving the system's intelligent analysis and control capabilities in complex rule-based environments.

[0130] (4) Determine the overall quality score of the alternative processing scheme set based on the data size score, the data quality score and the data distribution score.

[0131] Specifically, the data size score, data quality score, and data distribution score can be summed to obtain the overall quality score of the candidate processing scheme set; alternatively, different weights can be set for the data size score, data quality score, and data distribution score, and the weighted result can be determined as the overall quality score of the candidate processing scheme set.

[0132] In practice, the overall quality score of the alternative processing scheme set can be determined according to the following formula:

[0133] P_total =

[0134] Where Ptotal represents the overall quality score.

[0135] S104. For each alternative processing scheme, based on the matching status of the alternative processing scheme with the rules in the preset rule base in the judgment result, determine the individual quality score of the alternative processing scheme, and determine the confidence level of the alternative processing scheme based on the individual quality score and the overall quality score.

[0136] Specifically, once a candidate processing scheme matches a rule, an individual quality score for that candidate processing scheme can be calculated to ensure that candidate processing schemes with more matching rules receive lower scores, while candidate processing schemes with fewer matching rules or no matching rules receive higher scores.

[0137] In one possible implementation, the individual quality score of the alternative processing scheme is determined according to the following formula:

[0138] ;

[0139] Among them, the The individual quality score is given, where m is the number of rules contained in the preset rule base. The penalty coefficient is a pre-set value for the i-th rule in the preset rule base. Let be the trigger coefficient of the i-th rule in the preset rule base. If the alternative processing scheme matches the i-th rule, then... Select 1 if the value is 1, otherwise select 0.

[0140] It should be noted that the penalty coefficient pre-set for the i-th rule in the preset rule base is set according to actual needs, and is not limited in this embodiment. For example, in one possible implementation,

[0141] Furthermore, in one possible implementation, determining the confidence level of the alternative processing scheme based on the individual quality score and the overall quality score includes:

[0142] (1) Based on the matching of the alternative processing scheme with the rules in the preset rule base in the judgment result, determine the first weight of the overall quality score and the second weight of the individual quality score.

[0143] In a specific implementation, in one possible approach, the first weight is determined according to the following formula, and the second weight is determined based on the first weight:

[0144] ;

[0145] Among them, the The first weight is m, where m is the number of rules contained in the preset rule base. The penalty coefficient is a pre-set value for the i-th rule in the preset rule base. Let be the trigger coefficient corresponding to the i-th rule in the preset rule base. If the alternative processing scheme matches the i-th rule, then... Select 1 if the value is 1, otherwise select 0.

[0146] Furthermore, the second weight is equal to the difference between 1 and the first weight.

[0147] (2) The individual quality score and the overall quality score are weighted based on the first weight and the second weight, and the weighted result is determined as the confidence level of the alternative processing scheme.

[0148] In practice, the second weight can be This means ensuring that the weighting of the first and second weights does not affect the reasonableness of the result.

[0149] In practice, the confidence level of the alternative processing schemes can be determined using the following formula:

[0150] ;

[0151] Where C is the confidence level. Scoring based on data size D represents the data quality score, and D represents the data distribution score. As the first weight, As the second weight, Assess individual quality.

[0152] As can be understood from the preceding description, and The specific value is preset according to actual needs, and this embodiment does not limit it. For example, in one embodiment, the penalty coefficient can be:

[0153] ;

[0154] The trigger coefficient can be:

[0155] ;

[0156] Understandably, this calculation method, which integrates data scale scoring, data quality scoring, and data distribution scoring, achieves a two-way interaction mechanism between knowledge and data. Rules serve both as filtering conditions for forward constraints and as weighting factors for backward correction, realizing a closed-loop feedback in confidence calculation and obtaining an accurate overall quality score. At the same time, by combining constraint truncation with dynamic weights, a balance is achieved between the safety, quality, and efficiency of recommendation parameters.

[0157] As can be understood from the preceding description, the introduction It can differentiate the impact of rules such as safety, quality, and efficiency, adapt to different scenario needs, and enhance the personalization and flexibility of the assessment. As long as a few rules are triggered, the first weight of the overall quality score will drop rapidly, meaning that the impact of the overall score will decrease rapidly, thus improving the sensitivity and response speed to rule-triggered events. Furthermore, the overall quality score plays a dominant role in the confidence level. The overall quality score emphasizes that the solution set has attributes such as innovation, sufficient quantity, concentrated similarity, and controllable noise. Rule conflicts affect individual quality scores. When a solution violates multiple rules, ωr approaches 0, and the confidence level of the alternative processing solution depends more on the individual quality score, strengthening the filtering ability of rule constraints on unsafe and unreasonable solutions.

[0158] S105. Based on the weighted cosine similarity and confidence of each alternative processing scheme, select a recommended processing scheme from the recommended set of alternative processing schemes.

[0159] Specifically, the two indicators, weighted cosine similarity and confidence level, are considered together. The weighted cosine similarity reflects the degree of matching between the alternative solution and the target processing quality parameters, while the confidence level reflects the reliability and stability of the alternative solution.

[0160] In practical implementation, methods such as weighted summation and ranking selection can be used to combine similarity and confidence for decision-making. For example, in one embodiment, different weights can be assigned to similarity and confidence, a comprehensive score can be calculated for each candidate, and then the candidate with the highest comprehensive score can be selected as the recommended processing plan.

[0161] S106. When the historical processing quality parameters in the recommended processing scheme meet the triggering conditions of the target rule in the preset rule base, the processing parameters in the recommended processing scheme are modified based on the modification operation indicated by the target rule; otherwise, the original processing parameters are retained to obtain the modified processing scheme.

[0162] Referring to the preceding description, it can be understood that the preset rule base includes multiple types of rules, each with a different priority; the historical processing quality parameters in the recommended processing scheme meet the triggering conditions of the target rule in the preset rule base, and the processing parameters in the recommended processing scheme are corrected based on the correction operation indicated by the target rule, including:

[0163] (1) If the historical processing quality parameters in the recommended processing scheme meet the triggering condition of at least one rule under the same type of rule, then the processing parameters in the recommended processing scheme are corrected in sequence based on the correction operations indicated by the multiple rules.

[0164] Specifically, if all triggered rules belong to the same category, then the corrections will be performed sequentially according to the correction operations corresponding to all satisfied trigger conditions.

[0165] In specific implementation, for example, in one embodiment, the machining parameter a triggers safety rules IC-01 and RH-01, both of which are safety rules. At this time, the correction operation indicated by IC-01 is first performed to replace the high-rigidity tool and correct the step-by-step drilling process. Then, the riveting force is reduced by 5% according to the correction operation indicated by RH-01.

[0166] (2) If the historical processing quality parameters in the recommended processing scheme meet the triggering conditions of multiple rules and the multiple rules belong to different types of rules, the target rule category with the highest priority is identified according to the rule category to which the multiple rules belong, and the processing parameters in the recommended processing scheme are corrected in sequence based on the correction operation corresponding to the target rule under the target rule category among the multiple rules.

[0167] Specifically, based on the category to which the rule belongs, the highest priority target rule category is identified, and the correction operation indicated by the target rule category is executed.

[0168] In specific implementation, for example, in one embodiment, the rule priority queue is set as follows: basic rules > safety rules > quality rules > efficiency rules. At this time, when IC-01 (safety rule) and IC-02 (quality rule) are triggered at the same time, the replacement of high-rigidity tool and step-by-step drilling process correction of IC-01 will be executed directly.

[0169] It should be noted that when there are conflicts between the correction operations corresponding to multiple target rules, the correction operations for the conflicting parts of the rules with lower priority can be canceled, and only the correction operations for the conflicting parts of the rules with higher priority can be executed.

[0170] Understandably, when the historical processing quality parameters in the recommended processing scheme meet the triggering conditions of the target rules in the preset rule base, the system can correct the processing parameters in the recommended processing scheme according to the correction operation indicated by the target rule. If the triggering conditions of at least one rule under the same type of rule are met, the correction is performed sequentially according to the correction operation indicated by that type of rule. If the triggering conditions of multiple rules under different types of rules are met, the target rule category with the highest priority is identified according to the priority of the rules, and the correction is performed based on the correction operation corresponding to the target rule under that type of rule. This rule base-based correction mechanism ensures that the recommended processing scheme can meet the requirements of safety, quality and efficiency, and improves the accuracy and reliability of drilling and riveting process parameters.

[0171] In this embodiment, a preset rule base is introduced into the entire process of processing parameter recommendation. The rules not only participate in the calculation of confidence as filtering conditions in the early stage, but also participate in the correction of processing parameters in the later stage. This mechanism avoids the risk of misjudgment caused by simply relying on data similarity and improves the controllability and engineering applicability of the recommendation results.

[0172] Furthermore, this embodiment introduces a dual evaluation model of individual quality scores and overall quality scores, respectively characterizing the quality features of individual solutions and the solution set as a whole. Different weight coefficients are dynamically assigned to the scores based on rule matching, thereby calculating the solution confidence level. This closed-loop feedback structure enables the system to automatically identify and suppress highly similar solutions that match key rules, thus improving the robustness and accuracy of the recommendation system.

[0173] Furthermore, the rule base is subdivided into multiple categories, and a priority determination mechanism is used to distinguish the importance of rules. When recommendation parameters trigger multiple rules, the system can prioritize the correction strategies in the high-priority rule set and cascade the parameter correction operations indicated by multiple rules, thereby improving the accuracy and relevance of parameter correction and ensuring a high degree of consistency between the final recommendation result and the user's goals.

[0174] S107. Recommend the modified processing scheme to the user.

[0175] In practice, the revised processing plan will be recommended to the user, so that the user can perform dual-robot drilling and riveting operations according to the recommended plan.

[0176] The recommended method for robot drilling and riveting process parameters has the following beneficial effects:

[0177] (1) Improve the relevance and accuracy of the recommended solutions.

[0178] By introducing weighted cosine similarity calculation between target processing quality parameters and historical processing schemes in the knowledge graph, the correlation between the current task and past experience can be accurately measured, thereby selecting the historical scheme that best matches the current processing target and improving the practicality and effectiveness of the recommended parameters.

[0179] (2) Integrating rule knowledge to enhance the rationality of decision-making

[0180] By combining a pre-defined rule base to assess the compliance of quality parameters of alternative processing schemes, the applicability of processing parameters within the scope of process rules is ensured. Furthermore, the rule-triggered mechanism can filter out potentially non-compliant or low-quality schemes, thereby enhancing the reliability of recommended schemes.

[0181] (3) Multidimensional scoring mechanism improves screening accuracy

[0182] By incorporating individual quality scores and overall quality scores into the evaluation of alternative processing schemes, and combining this with confidence level calculations, a quantitative expression of the multi-dimensional quality of processing schemes is achieved, thereby improving the accuracy and reliability of the final recommendation results.

[0183] (4) Supports intelligent correction and optimization of processing parameters

[0184] Based on the recommended processing scheme, if the historical processing quality parameters meet the triggering conditions of the target rules, the parameters can be modified and optimized according to the rule opinions, thereby improving the processing quality and efficiency in a targeted manner and enhancing the system's adaptability and intelligence.

[0185] (5) Enhance the process intelligence decision-making capability of the dual-robot drilling and riveting system

[0186] This method relies on knowledge graph structure and rule reasoning mechanism to structure experiential knowledge and intelligently recommend parameters, effectively improving the adaptive decision-making ability of dual-robot drilling and riveting systems for process parameters in complex manufacturing scenarios.

[0187] (6) Enhance user interaction experience and decision support capabilities

[0188] This method responds to the target processing quality requirements set by the user, and provides the user with transparent, explainable and quality-assured processing parameter solutions through intelligent analysis and recommendation throughout the entire process, thereby improving the system's decision support capabilities and interactive experience for engineers.

[0189] The robot drilling and riveting process parameter recommendation method and apparatus provided in this embodiment have the following advantages: First, when screening the set of candidate processing schemes, by determining the weight of each target processing quality parameter and using these weights to calculate the weighted cosine similarity between the target processing quality parameter and historical processing schemes, historical processing schemes similar to the target processing requirements can be found more accurately, thereby screening the set of candidate processing schemes. Second, by matching the historical processing quality parameters in each candidate processing scheme with the rules in the preset rule base, it is ensured that the candidate processing schemes are not only similar to the target requirements, but also strictly follow key rules such as process safety, quality, and efficiency, effectively filtering out schemes that do not meet the process requirements and improving the overall quality level of the recommended schemes. Third, by combining the overall quality score and the individual quality score to calculate the confidence level of each candidate processing scheme, a comprehensive and quantifiable evaluation index is provided, which not only enhances the interpretability of the recommendation results, but also provides a more intuitive and reliable basis. Fourth, when the recommended processing scheme meets the specific rule triggering conditions, the technology can automatically correct the parameters, ensuring that the processing scheme can flexibly respond to dynamic changes in actual production, greatly improving the adaptability and robustness of the processing process. In this way, through precise matching, rule constraints, comprehensive evaluation, dynamic adjustment and recommendation, the accuracy, flexibility and practicality of the recommended drilling and riveting process parameters for dual robots are significantly improved, and reliable assurance of robot drilling and riveting operations is achieved.

[0190] Corresponding to the aforementioned embodiment of a method for recommending robot drilling and riveting process parameters, this application also provides corresponding verification experiments.

[0191] Specifically, a dual-robot drilling and riveting process parameter recommendation device is built using the Flask framework in Python, aiming to achieve structured management of process knowledge and intelligent parameter decision-making in complex manufacturing scenarios. This device uses a knowledge graph and rule-based reasoning fusion method as its core, integrating a process parameter recommendation module through a lightweight web service architecture to form a decision-making system of "data input - knowledge reasoning - parameter inference".

[0192] Furthermore, to verify the knowledge graph node retrieval capabilities of the recommendation device, the following functional test was conducted: After the user triggers the node search function on the front-end interface and specifies the entity type as "processing result," they select the specific instance "drilling and riveting quality 21." The device's back-end executes a predefined Cypher retrieval statement through the graph database query engine: MATCH(n:Action{id:'drilling and riveting execution 21'})RETURNn. This statement locates the target node using the entity tag "Action" and the unique identifier "drilling and riveting execution 21," achieving accurate retrieval of specific drilling and riveting execution process entities. The query results are serialized and returned to the front-end, where a visualization component based on D3.js dynamically renders the target node and its relationships in a graph structure, including the robot combination, tools used, drilling process, riveting process, and drilling and riveting quality associated with the execution process.

[0193] Furthermore, taking the drilling and riveting process requirements of a certain enterprise with dual robots as an example, the following process quality requirements are proposed, and Table 5 shows the process quality requirements to be processed:

[0194] Table 5

[0195]

[0196] Input the above process quality into the recommended device. The Top-5 similar cases recommended by the device are shown in Table 6. Table 6 shows similar cases of robot drilling and riveting processing parameters:

[0197] Table 6

[0198]

[0199] Using the original cosine similarity algorithm as the baseline method, and comparing it with the improved weighted cosine similarity case recommendation algorithm, the recommended process parameters of both are shown in Table 7. Table 7 shows the recommended processing parameters:

[0200] Table 7

[0201]

[0202] The overall confidence level of the above recommended parameters is 0.93, which meets the processing parameter requirements.

[0203] To verify the effectiveness of rule-based reasoning, new process quality requirements are introduced as shown in Table 8. Table 8 presents another set of process quality requirements to be processed.

[0204] Table 8

[0205]

[0206] Among the parameters above, the recommended process parameters before and after the correction of safety rule RH-01 and efficiency rule MA-01 are shown in Table 9. Table 9 shows the recommended processing parameters for rule constraints:

[0207] Table 9

[0208]

[0209] Among them, no similar cases were found in the baseline method, and the overall confidence of the recommended parameters without rule correction was 0.81, while the overall confidence of the parameters after rule correction was 0.84, which was better than the parameters before correction, proving the effectiveness of the rule constraints.

[0210] Corresponding to the aforementioned embodiment of a method for recommending robot drilling and riveting process parameters, this application also provides an embodiment of a device for recommending robot drilling and riveting process parameters.

[0211] An embodiment of the robot drilling and riveting process parameter recommendation device disclosed in this application can be applied to robot drilling and riveting process parameter recommendation equipment. The device embodiment can be implemented through software, hardware, or a combination of both. Taking software implementation as an example, as a logical device, it is formed by the processor of the robot drilling and riveting process parameter recommendation equipment loading the corresponding computer program instructions from non-volatile memory into memory for execution. From a hardware perspective, such as... Figure 4 The diagram shown is a hardware structure diagram of the robot drilling and riveting process parameter recommendation device, which is part of the robot drilling and riveting process parameter recommendation equipment of this application. Except for... Figure 4 In addition to the processor, memory, network interface, and non-volatile memory shown, the robot drilling and riveting process parameter recommendation device in the embodiment may also include other hardware depending on the actual function of the robot drilling and riveting process parameter recommendation device, which will not be described in detail here.

[0212] Figure 5 This is a schematic diagram of the structure of Embodiment 1 of the robot drilling and riveting process parameter recommendation device provided in this application. Please refer to... Figure 5 The apparatus provided in this embodiment includes a screening module 510, a judgment module 520, a determination module 530, a correction module 540, and a recommendation module 550; wherein,

[0213] The filtering module 510 is used to respond to the target processing quality parameters input by the user, calculate the weighted cosine similarity between the target processing quality parameters and the historical processing schemes in the pre-established dual-robot drilling and riveting process knowledge graph based on the weights determined in advance for each target processing quality parameter, and filter out a set of alternative processing schemes.

[0214] The judgment module 520 is used to determine, for each alternative processing scheme, whether the historical processing quality parameters of the alternative processing scheme meet the triggering conditions of the rules in the preset rule base, and obtain the judgment result.

[0215] The determining module 530 is used to determine the overall quality score of the set of alternative processing schemes based on the judgment result and the weighted cosine similarity corresponding to each alternative processing scheme in the set of alternative processing schemes.

[0216] The determining module 530 is further configured to, for each alternative processing scheme, determine the individual quality score of the alternative processing scheme based on the matching situation between the alternative processing scheme and the rules in the preset rule base in the judgment result, and determine the confidence level of the alternative processing scheme based on the individual quality score and the overall quality score;

[0217] The determining module 530 is further configured to select a recommended processing scheme from the recommended set of alternative processing schemes based on the weighted cosine similarity and confidence of each alternative processing scheme;

[0218] The correction module 540 is used to correct the processing parameters in the recommended processing scheme based on the correction operation indicated by the target rule when the historical processing quality parameters in the recommended processing scheme meet the triggering conditions of the target rule in the preset rule base; otherwise, the original processing parameters are retained to obtain the corrected processing scheme.

[0219] The recommendation module 550 is used to recommend the modified processing scheme to the user.

[0220] The apparatus of this embodiment can be used to perform... Figure 1 The steps of the method embodiment shown are similar in principle and process, and will not be repeated here.

[0221] Please continue to refer to Figure 4 This application also provides a robot drilling and riveting process parameter recommendation device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps of any of the methods provided in the first aspect of this application.

[0222] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the methods provided in this application.

[0223] The specific implementation process of the functions and roles of each unit in the above device can be found in the implementation process of the corresponding steps in the above method, and will not be repeated here.

[0224] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this application according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0225] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for recommending process parameters for robot drilling and riveting, characterized in that, The method includes: In response to the target processing quality parameters input by the user, the weighted cosine similarity between the target processing quality parameters and the historical processing schemes in the pre-established dual-robot drilling and riveting process knowledge graph is calculated based on the weights pre-determined for each target processing quality parameter. Based on the calculated weighted cosine similarity, a set of alternative processing schemes is selected. For each alternative processing scheme, determine whether the historical processing quality parameters of the alternative processing scheme meet the triggering conditions of the rules in the preset rule base, and obtain the judgment result; Based on the judgment result and the weighted cosine similarity corresponding to each alternative processing scheme in the alternative processing scheme set, the overall quality score of the alternative processing scheme set is determined; For each alternative processing plan, based on the matching results between the alternative processing plan and the rules in the preset rule base, an individual quality score is determined for the alternative processing plan, and the confidence level of the alternative processing plan is determined based on the individual quality score and the overall quality score; the determination of the individual quality score based on the matching results between the alternative processing plan and the rules in the preset rule base includes: determining the individual quality score of the alternative processing plan according to the following formula: ; wherein, the The individual quality score is given, where m is the number of rules contained in the preset rule base. The penalty coefficient is a pre-set value for the i-th rule in the preset rule base. Let be the trigger coefficient corresponding to the i-th rule in the preset rule base. If the alternative processing scheme matches the i-th rule, then... Select 1 if the value is 1, otherwise select 0. Based on the weighted cosine similarity and confidence of each alternative processing scheme, a recommended processing scheme is selected from the recommended set of alternative processing schemes; When the historical processing quality parameters in the recommended processing scheme meet the triggering conditions of the target rule in the preset rule base, the processing parameters in the recommended processing scheme are corrected based on the correction operation indicated by the target rule; otherwise, the original processing parameters are retained to obtain the corrected processing scheme. The revised processing solution is recommended to the user.

2. The method according to claim 1, characterized in that, The step of determining the overall quality score of the candidate processing scheme set based on the judgment result and the weighted cosine similarity corresponding to each candidate processing scheme in the candidate processing scheme set includes: Based on the judgment result, the data size score of the alternative processing scheme set is determined; Calculate the average weighted cosine similarity of each alternative processing scheme in the alternative processing scheme set, and determine the average value as the data quality score of the alternative processing scheme set; Calculate the standard deviation of the weighted cosine similarity corresponding to each alternative processing scheme in the alternative processing scheme set, and determine the data distribution score of the alternative processing scheme set based on the standard deviation and the matching of the alternative processing scheme with the safety class rules in the preset rule set; The overall quality score of the alternative processing scheme set is determined based on the data size score, the data quality score, and the data distribution score.

3. The method according to claim 1, characterized in that, Based on the individual quality score and the overall quality score, the confidence level of the alternative processing plan is determined, including: Based on the matching results of the alternative processing scheme with the rules in the preset rule base, the first weight of the overall quality score and the second weight of the individual quality score are determined. The individual quality score and the overall quality score are weighted based on the first weight and the second weight, and the weighted result is determined as the confidence level of the alternative processing scheme.

4. The method according to claim 3, characterized in that, The determination of the first weight of the overall quality score and the second weight of the individual quality score based on the matching of the alternative processing scheme with the rules in the preset rule base in the judgment result includes: The first weight is determined according to the following formula, and the second weight is determined based on the first weight: ； Among them, the The first weight is m, where m is the number of rules contained in the preset rule base. The penalty coefficient is a pre-set value for the i-th rule in the preset rule base. The trigger coefficient is the i-th rule in the preset rule base.

5. The method according to claim 2, characterized in that, The step of determining the data size score of the alternative processing scheme set based on the judgment result includes: Determine the data size score using the following formula: ； Among them, the Scoring is given to the data size, the The number of valid alternative processing schemes in the alternative processing scheme set that do not match the rules in the preset rule base. The number of alternative processing schemes included in the alternative processing scheme set; This is an indicator function. When the alternative processing scheme in the set of alternative processing schemes matches the basic rule in the preset rule set, its value is 0; otherwise, its value is 1.

6. The method according to claim 1, characterized in that, The step of calculating the weighted cosine similarity between the target machining quality parameter and historical machining schemes in a pre-established dual-robot drilling and riveting process knowledge graph, based on pre-determined weights for each target machining quality parameter, includes: The target processing quality parameters and the historical processing quality parameters in the historical processing schemes are normalized respectively to obtain the normalized values ​​of the target processing quality parameters and the normalized values ​​of the historical processing quality parameters. Based on the normalized values ​​of the target machining quality parameters and the historical machining quality parameters, the weighted cosine similarity between the target machining quality parameters and the historical machining schemes in the pre-established dual-robot drilling and riveting process knowledge graph is calculated using the following formula: ； Among them, the The weighted cosine similarity is denoted as n, where n is the number of dimensions of the target processing quality parameter. Let xi be the weights pre-determined for the i-th dimension target processing quality parameter, yi be the normalized value of the i-th dimension target processing parameter, and yi be the normalized value of the i-th dimension historical processing quality parameter.

7. The method according to claim 1, characterized in that, The preset rule base includes multiple rule categories, each with a different priority; the historical processing quality parameters in the recommended processing scheme meet the triggering conditions of the target rule in the preset rule base, and the processing parameters in the recommended processing scheme are corrected based on the correction operation indicated by the target rule, including: If the historical processing quality parameters in the recommended processing scheme meet the triggering condition of at least one rule under the same type of rule, then the processing parameters in the recommended processing scheme are corrected sequentially based on the correction operations indicated by the multiple rules. If the historical processing quality parameters in the recommended processing scheme meet the triggering conditions of multiple rules, and the multiple rules belong to different rule categories, the target rule category with the highest priority is identified according to the rule category to which the multiple rules belong, and the processing parameters in the recommended processing scheme are corrected in sequence based on the correction operation corresponding to the target rule under the target rule category among the multiple rules.

8. The method according to claim 1, characterized in that, The method for determining the weights pre-determined for each target processing quality parameter includes: The relative importance of any two target processing quality parameters in the processing quality is evaluated, and a judgment matrix is ​​formed based on the evaluation results; The eigenvectors and the largest eigenvalue of the judgment matrix are obtained using a normalization method. The feature vector is used as the initial weight vector for all target processing quality parameters; Based on the largest eigenvalue, calculate the consistency index of the judgment matrix of the initial weight vector; The consistency ratio of the judgment matrix is ​​calculated based on the consistency index and random consistency index of the judgment matrix. When the consistency ratio is less than a preset threshold, the initial weight vector is determined as the weight vector corresponding to all target processing quality parameters. Otherwise, the relative importance of any two target processing quality parameters in the processing quality is evaluated again, and a judgment matrix is ​​formed based on the evaluation results until the consistency ratio is less than the preset threshold.

9. A device for recommending process parameters for robot drilling and riveting, characterized in that, The device includes a screening module, a judgment module, a determination module, a correction module, and a recommendation module; wherein, The filtering module is used to respond to the target processing quality parameters input by the user, calculate the weighted cosine similarity between the target processing quality parameters and the historical processing schemes in the pre-established dual-robot drilling and riveting process knowledge graph based on the weights determined in advance for each target processing quality parameter, and filter out a set of alternative processing schemes. The judgment module is used to determine, for each alternative processing scheme, whether the historical processing quality parameters of the alternative processing scheme meet the triggering conditions of the rules in the preset rule base, and obtain the judgment result; The determining module is used to determine the overall quality score of the set of alternative processing schemes based on the judgment result and the weighted cosine similarity corresponding to each alternative processing scheme in the set of alternative processing schemes. The determining module is further configured to, for each candidate processing scheme, determine an individual quality score for the candidate processing scheme based on the matching situation between the candidate processing scheme and the rules in the preset rule base in the judgment result, and determine the confidence level of the candidate processing scheme based on the individual quality score and the overall quality score; the step of determining the individual quality score of the candidate processing scheme based on the matching situation between the candidate processing scheme and the rules in the preset rule base in the judgment result includes: determining the individual quality score of the candidate processing scheme according to the following formula: ; wherein, the The individual quality score is given, where m is the number of rules contained in the preset rule base. The penalty coefficient is a pre-set value for the i-th rule in the preset rule base. Let be the trigger coefficient corresponding to the i-th rule in the preset rule base. If the alternative processing scheme matches the i-th rule, then... Select 1 if the value is 1, otherwise select 0. The determining module is further configured to select a recommended processing scheme from the recommended set of alternative processing schemes based on the weighted cosine similarity and confidence of each alternative processing scheme; The correction module is used to correct the processing parameters in the recommended processing scheme based on the correction operation indicated by the target rule when the historical processing quality parameters in the recommended processing scheme meet the triggering conditions of the target rule in the preset rule base; otherwise, the original processing parameters are retained to obtain the corrected processing scheme. The recommendation module is used to recommend the modified processing scheme to the user.

Images