A Tool Intelligent Selection Method Based on k-Nearest Neighbor Entropy Estimation and Kriging Method

Abstract

This invention discloses an intelligent tool selection method based on k-nearest neighbor entropy estimation and kriging, comprising the following steps: S1, formulating a rule-based pre-screening method for cutting tools to obtain a set of candidate tools for machining; S2, performing tool performance reasoning and tool ranking based on the candidate tool set obtained in S1; S3, combining the analysis of tools and user behavior to obtain a comprehensive cutting tool recommendation method with process scenario adaptability. This invention employs the aforementioned intelligent tool selection method based on k-nearest neighbor entropy estimation and kriging, and by mining tool selection experience from cutting records, obtains a comprehensive cutting tool recommendation method with certain process scenario adaptability. This method can provide predicted machining results, possible user behaviors, and possible tool combinations based on user input when users formulate process manuals, thus achieving intelligent selection of cutting tools.

Description

Technical Field

[0001] This invention relates to the field of tool selection technology, and in particular to a smart tool selection method based on k-nearest neighbor entropy estimation and Kriging method. Background Technology

[0002] During process planning, the selected cutting tools significantly impact the surface quality of the machined parts and the machining cost. Currently, tool selection largely relies on the experience of workers or experts and consultation of relevant tool manuals, which introduces a degree of subjectivity and leads to the loss of experience due to personnel turnover. In recent years, intelligent tool selection has gained increasing attention, with a growing trend towards algorithm-based intelligent tool recommendations. Tool recommendation algorithms refer to the process of selecting the most suitable tool from a pool of alternatives during process planning, based on certain judgment criteria.

[0003] Existing tool recommendation algorithms are typically designed from two perspectives: instance-based reasoning and rule-based reasoning. The former lacks constraints, making it prone to inappropriate tool recommendations, and as the instance database expands, it struggles to build meaningful indexes to accelerate case searches. The latter, based on rule-based reasoning, relies heavily on the experience of process engineers, requiring continuous maintenance and improvement. Data fusion algorithms, leveraging rapidly developing intelligent technologies, combine tool and user factors, leveraging their strengths to achieve more intelligent tool recommendations. Summary of the Invention

[0004] The purpose of this invention is to provide a tool intelligent selection method based on k-nearest neighbor entropy estimation and Kriging method. This method can provide predicted machining results, possible user behaviors, and possible tool combinations based on user input when the user formulates a process manual, thereby achieving intelligent selection of cutting tools.

[0005] To achieve the above objectives, this invention provides a method for intelligent tool selection based on k-nearest neighbor entropy estimation and Kriging, comprising the following steps:

[0006] S1. Develop a rule-based method for pre-screening cutting tools to obtain a set of candidate tools for machining;

[0007] S2. Based on the candidate tool set obtained from the pre-screening in S1, perform tool performance reasoning and tool ranking;

[0008] S3. Combining the analysis of cutting tools and user behavior, a comprehensive recommendation method for cutting tools with adaptability to process scenarios is obtained.

[0009] Preferably, the specific process of S1 is as follows:

[0010] The user inputs the machining conditions into the inference engine, which combines expert knowledge to search for suitable tools in the tool database. The machining conditions include machine tool material and process information.

[0011] Pre-screening is performed based on user-inputted feature requirements, including tool parameters, tool material, and machining accuracy.

[0012] Based on function, topology, and geometry, most unsuitable tools are filtered out, and after several rounds of filtering, a pre-screened set of candidate tools is obtained.

[0013] Preferably, in S2, the tool performance reasoning includes using k-nearest neighbor entropy to estimate and analyze the information gain ratio and using Kriging interpolation to reason about the tool. By analyzing the tool parameters, various evaluation indicators are obtained; then the tools are ranked, and finally the ranking of the candidate tool set is obtained based on the comprehensive multi-dimensional indicators, which serves as the benchmark ranking.

[0014] Preferably, in S2, the information gain ratio is estimated using k-nearest neighbor entropy to calculate the weight of the influence of tool parameters on the machining result, including the following two parts:

[0015] First, k-nearest neighbor entropy estimates the entropy of multidimensional evaluation indicators in a tool set. The input to this method is the tool parameters and corresponding evaluation indicators of the machining results from N machining records. In tool application scenarios, the evaluation of whether the tool selection is reasonable is usually analyzed from three perspectives: the surface roughness of the workpiece, the wear of the machining tool, and the machining cost. Therefore, k-nearest neighbor entropy estimation is calculated by the following formula:

[0016]

[0017] Where H(x) represents the entropy of the three-dimensional vector of processing quality, wear, and cost; Γ is the gamma function, ψ is the logarithmic derivative of the gamma function, N is the number of samples, k is the number of nearest neighbors selected, D is the dimension of x, and ε i For x i Euclidean distance to the k-th neighbor;

[0018] Secondly, the information gain ratio analyzes the correlation between tool parameters and indicators by examining changes in entropy, and analyzes the importance of tool parameters.

[0019] For a tool vector with m parameters and its corresponding D-dimensional evaluation index, select one parameter to divide the input set into z groups, grouping data with the same value into the same group, and calculate the information entropy H of the D-dimensional evaluation index in each group. j Then the information gain ratio of the tool parameter is: Information gain ratio G of all parameters is obtained. riThen, all ratios are normalized to obtain the tool parameter importance vector G = [G r1 G r2 ,…,G rm ];

[0020] For continuous attributes in tool parameters, the information gain of continuous attributes is calculated using the bisection method. For attributes with incomplete information in tool parameters, the expected value is used instead of the default value.

[0021] Preferably, in S2, the tool with m parameters and its corresponding machining result evaluation index constitute a multi-dimensional space. Kriging interpolation uses known points in the space to estimate unknown points, specifically:

[0022] Define the basic interpolation formula as follows: Where w i As the weight, y i Given the values ​​of points, assuming the spatial properties are uniform, then every point in space has the same variance σ. 2 If the unbiased estimation is satisfied, then: The weights are calculated using the following formula:

[0023]

[0024] Where, r ij The semivariance function between points i and j is fitted using a Gaussian function: Where r(0) = 0, c0 and c are fitting parameters, d represents the distance, and γ is the parameter of the Gaussian model. After solving all weights, the predicted value of the unobserved target in the index space can be obtained through the basic interpolation formula.

[0025] Kriging interpolation yields performance predictions for all tools in the candidate tool set, including surface roughness Ra, tool life, and machining cost. Based on these performance predictions, tools are ranked and recommended.

[0026] Preferably, in S3, based on the baseline ranking of the tools obtained in S2, a comprehensive ranking is generated using a process scenario-based fusion algorithm, including the following steps:

[0027] S31. Use collaborative filtering to retrieve user usage records, calculate tool reputation similarity, and predict user preferences for different tools by calculating the similarity matrix, thus obtaining the user preference ranking of the tool selection set.

[0028] S32. The Pagerank method is used to analyze the tool combination usage from the machining records, construct the adjacency matrix of the tool association graph, and use the principal eigenvector obtained by it to approximate the solution to obtain the tool combination ranking.

[0029] S33. Combine the baseline sorting, user preference sorting, and tool combination sorting, and obtain the comprehensive sorting result by weighted summation sorting.

[0030] Therefore, the present invention employs the above-mentioned intelligent tool selection method based on k-nearest neighbor entropy estimation and Kriging method, and the beneficial effects are as follows:

[0031] (1) The rule-based cutting tool pre-screening method of the present invention screens tools based on machine tool materials, process information, tool parameters, material and machining accuracy, etc., which can quickly eliminate a large number of unsuitable tools, reduce subsequent calculations, and improve the efficiency of the entire tool selection process.

[0032] (2) This invention uses k-nearest neighbor entropy estimation to analyze information gain ratio to determine tool parameter weights, and then combines Kriging interpolation to infer tool performance. It can accurately predict the performance of tools in machining, such as surface roughness Ra, tool life and machining cost, and make benchmark rankings accordingly, providing a scientific basis for the accurate selection of tools.

[0033] (3) This invention uses collaborative filtering and Pagerank methods to consider user preferences and tool combination usage respectively, and generates a comprehensive ranking result, so that tool recommendation is not only based on processing performance, but also fits user habits and optimizes tool combination, thereby improving the overall coordination and adaptability of the processing process.

[0034] (4) The data fusion algorithm in this invention utilizes the current rapidly developing intelligent technology to combine multiple aspects such as cutting tools and users, taking the best of each other and making up for the shortcomings, so as to achieve more intelligent cutting tool recommendation.

[0035] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0036] Figure 1 This is an overall technical block diagram of an embodiment of the intelligent tool selection method based on k-nearest neighbor entropy estimation and Kriging method of the present invention;

[0037] Figure 2 This is a logic diagram of a rule-based pre-screening method according to an embodiment of the present invention, which is a tool intelligent selection method based on k-nearest neighbor entropy estimation and Kriging method.

[0038] Figure 3 This is a tool performance reasoning logic diagram of an embodiment of the intelligent tool selection method based on k-nearest neighbor entropy estimation and Kriging method of the present invention.

[0039] Figure 4 This is a flowchart illustrating the tool benchmark ranking process of an embodiment of the intelligent tool selection method based on k-nearest neighbor entropy estimation and Kriging method of the present invention.

[0040] Figure 5 This is a schematic diagram illustrating the comprehensive sorting of an embodiment of the intelligent tool selection method based on k-nearest neighbor entropy estimation and Kriging method of the present invention. Detailed Implementation

[0041] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0042] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0043] like Figure 1 As shown, the technical framework of this invention, a tool intelligent selection method based on k-nearest neighbor entropy estimation and Kriging method, has a three-layer structure and includes the following steps:

[0044] S1. Develop a rule-based method for pre-screening cutting tools to obtain a set of candidate tools for machining;

[0045] S2. Based on the candidate tool set obtained from the pre-screening in S1, perform tool performance reasoning and tool ranking;

[0046] S3. Combining the analysis of cutting tools and user behavior, a comprehensive recommendation method for cutting tools with adaptability to process scenarios is obtained.

[0047] In S1, the rule-based tool pre-screening method aims to pre-screen tools that are not suitable for the current machining features, thereby reducing the computational load for tool selection. The specific process is as follows:

[0048] The user inputs the machining conditions into the inference engine, which then combines expert knowledge to search for suitable tools in the tool database. The machining conditions include machine tool material and process information.

[0049] Pre-screening is performed based on user-inputted feature requirements, including tool parameters, tool material, and machining accuracy as mentioned above. This embodiment uses square shoulder milling as an example to illustrate the pre-screening method. Square shoulder milling is a non-rotating square shoulder machining process, generally involving several key parameters such as width, length, depth of cut, angle, and fillet radius. In addition, material and machining accuracy also affect tool selection. The logic for pre-screening based on user-inputted feature requirements is as follows: Figure 2 As shown, when selecting a solid end mill and a replaceable end mill, the logic for querying the solid end mill and replaceable end mill data table is as follows:

[0050] Step 1: Determine the relevant information regarding the cutting diameter:

[0051] 1) If the cutting diameter is within the range of plane width * (0.4-0.6), or if it is not within the range of (0.4-0.6), proceed to the next step of judgment;

[0052] 2) If the cutting diameter is less than the plane width * 0.8, proceed to the next step of judgment;

[0053] Step 2: Determine the radius of the fillet:

[0054] 1) If the maximum fillet radius is equal to the minimum fillet radius, then the fillet radius is not less than the maximum fillet radius;

[0055] 2) If the maximum fillet radius is not equal to the minimum fillet radius, then the fillet radius is greater than the maximum fillet radius;

[0056] 3) If the maximum and minimum fillet radii are 0, determine whether the fillet radius falls within the range of (0.4-1.2). If it does, proceed to the next step; otherwise, no result is obtained.

[0057] Step 3: Relationship between the material being processed and the applicable processing materials

[0058] 1) If the material being processed belongs to the set of applicable processing materials, proceed to the next step of judgment; otherwise, there is no result.

[0059] 2) If the machining process is roughing, finishing, or semi-finishing, then the tool can be included in the pre-screened tool set; if it does not meet this requirement, then there is no result.

[0060] Based on function, topology, and geometry, most unsuitable tools are filtered out, and after several rounds of filtering, a pre-screened set of candidate tools is obtained.

[0061] like Figure 3As shown in Figure S2, the tool performance reasoning includes two steps: estimating and analyzing the information gain ratio using k-nearest neighbor entropy and reasoning about the tool using Kriging interpolation. By analyzing the tool parameters, various evaluation indicators are obtained; then, the tools are ranked, and finally, a ranking of the candidate tool set is obtained based on comprehensive multi-dimensional indicators, which serves as the benchmark ranking. Figure 4 As shown.

[0062] In S2, the following two key calculation methods are applied:

[0063] (1) The information gain ratio is estimated and analyzed using k-nearest neighbor entropy to calculate the weight of the influence of tool parameters on the machining results, laying the foundation for the Kriging method. This mainly includes the following two parts:

[0064] First, k-nearest neighbor entropy estimates the entropy of multidimensional evaluation indicators in a tool set. The input to this method is the tool parameters and corresponding evaluation indicators of the machining results from N machining records. In tool application scenarios, the evaluation of whether the tool selection is reasonable is usually analyzed from three perspectives: the surface roughness of the workpiece, the wear of the machining tool, and the machining cost. Therefore, k-nearest neighbor entropy estimation is calculated by the following formula:

[0065]

[0066] Where H(x) represents the entropy of the three-dimensional vector of processing quality, wear, and cost. Γ is the gamma function, ψ is the logarithmic derivative of the gamma function, N is the number of samples, k is the number of nearest neighbors selected, D is the dimension of x, and ε i For x i The Euclidean distance to the k-th neighboring point.

[0067] Secondly, the information gain ratio analyzes the correlation between tool parameters and indicators through entropy changes, thus revealing the importance of tool parameters. For an input tool vector with m parameters and its corresponding D-dimensional evaluation index, one parameter is selected to divide the input set into z groups, with data of the same value grouped into the same set. The information entropy H of the D-dimensional evaluation index in each set is then calculated. j Then the information gain ratio of the tool parameter is:

[0068]

[0069] Information gain ratio G of all parameters is obtained. ri Then, all ratios are normalized to obtain the tool parameter importance vector G = [G r1 G r2 ,…,G rm For continuous attributes in tool parameters, the information gain of continuous attributes is calculated using the bisection method. For attributes with incomplete information in tool parameters, the expected value is used instead of the default value.

[0070] 2) Kriging interpolation inference tool

[0071] In S2, Kriging is an interpolation method for multidimensional spaces, capable of modeling such spaces. A multidimensional space is formed between a tool with m parameters and its corresponding machining result evaluation index. Kriging interpolation uses known points in this space to estimate unknown points, thus enabling the inference of the approximate performance of pre-selected candidate tools based on existing examples.

[0072] Define the basic interpolation formula as follows: Where w i As the weight, y i Let be the value of a known point. Assuming the spatial properties are uniform, then every point in the space has the same variance σ. 2 If the unbiased estimation is satisfied, then: The weights are calculated using the following formula:

[0073]

[0074] Where, r ij The semivariance function between points i and j is fitted using a Gaussian function: Where r(0) = 0, c0 and c are fitting parameters, d represents the distance, and γ is the parameter of the Gaussian model. After solving for all weights, the predicted value of the unobserved target in the index space can be obtained through the basic interpolation formula.

[0075] Kriging interpolation can be used to obtain the performance prediction values ​​of all tools in the candidate tool set, including surface roughness Ra, tool life, and machining cost. Based on the performance prediction values ​​of all tools in the candidate tool set, the tools can be ranked and recommended.

[0076] In S3, based on the baseline tool ranking obtained in S2, to better reflect user preferences and the combined use of tools, a fusion algorithm based on process scenarios is used to generate a comprehensive ranking on the baseline tool ranking, including the following steps:

[0077] S31. Use collaborative filtering to retrieve user usage records, calculate tool reputation similarity, and predict user preferences for different tools by calculating the similarity matrix, thus obtaining a user preference ranking of the tool selection set.

[0078] The formula for calculating the similarity of reputation between knives is:

[0079]

[0080] Where, N a N represents the number of process engineers who have selected tool a. b This represents the number of process engineers who have used tool b.

[0081] After calculating the similarity of all existing tools in the dataset, the similarity matrix W can be obtained. ij Based on the similarity matrix, the interest of user1 in tool q can be calculated using the following formula:

[0082] P user1c =∑ i∈N(user1)∩S(c,k) W ic R user1i ;

[0083] P user1q =∑ i∈N(user1)∩S(q,l) W iq R user1i ;

[0084] Among them, P user1q W represents the predicted value of user1's preference for tool q, N(user1) represents the set of tools used by user1, S(q,l) represents l tools with the most similar reputation to tool q, and W iq The similarity between tool i and tool q is obtained from the similarity matrix, R. user1i This represents user1's preference for tool i. Based on the predicted user preferences for different tools obtained by this method, the user preference ranking of the tool selection set can be obtained.

[0085] S32. The Pagerank method is used to analyze the tool combination usage from the machining records, construct the adjacency matrix of the tool association graph, and use the principal eigenvectors obtained from it to approximate the tool combination ranking. The specific process is as follows:

[0086] First, construct the adjacency matrix of the tool relationship graph: For a set of tools T selected according to rules, assume there is a tool set of 5 tools: T = {t1, t2, t3, t4, t5}. These tools have all processed the "Material 1 and Feature 1" process, while t5 has never participated in the processing. Among them, {t1, t3} have been processed together 3 times, {t1, t2} have been processed together 5 times, {t3, t4} have been processed together 4 times, and {t2, t4} have been processed together 2 times. Then the tool hyperlink matrix is:

[0087]

[0088] For the suspended node t5 (not connected to any other tool), pad all rows in the corresponding row to 1 / 5 to obtain HM′. Further add random noise to obtain the improved adjacency matrix: Where E represents the identity matrix, α is a variable scalar, typically taken as 0.85-0.99, and n is the number of tools.

[0089] Since the actual set of candidate tools may contain hundreds of tools, directly solving high-order equations would consume a lot of computational resources. Therefore, the power method is used to solve for the eigenvectors of the principal eigenvalues. The calculation method is as follows: π (u) =π (u-1) "HM" can eventually obtain an approximate solution through continuous iteration, and the approximate solution of the principal eigenvector obtained by it can be used to obtain the tool combination sorting.

[0090] S33, such as Figure 5 As shown, the baseline sort, user preference sort, and tool combination sort are combined and summed to obtain the comprehensive sorting result through weighted summation:

[0091]

[0092] Where, β i Score is the weight value. i For the ranking of the three sorting methods, β i =β0+pos, where pos is a custom tendency and β0 is an adjustable initial value.

[0093] This ranking result takes into account three aspects: predicted machining results, possible user behavior, and possible tool combinations. It is a general and complete tool recommendation method.

[0094] Therefore, this invention adopts the above-mentioned intelligent tool selection method based on k-nearest neighbor entropy estimation and Kriging method, and constructs a three-layer technical framework. Through a series of steps such as rule-based pre-screening, tool performance reasoning and ranking, and comprehensive ranking combined with process scenarios, it can accurately screen and recommend the most suitable tool for a specific machining task from a large number of tools, effectively improving the intelligence level, accuracy and efficiency of tool selection. At the same time, it takes into account user preferences and tool combination usage, and can provide a comprehensive, reasonable and efficient tool selection scheme for cutting machining, significantly improving machining quality and economic benefits.

[0095] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for intelligent tool selection based on k-nearest neighbor entropy estimation and Kriging, characterized in that, Includes the following steps: S1. Develop a rule-based method for pre-screening cutting tools to obtain a set of candidate tools for machining; S2. Based on the candidate tool set obtained from the pre-screening in S1, perform tool performance reasoning and tool ranking; S3. Combining the analysis of cutting tools and user behavior, a comprehensive recommendation method for cutting tools with process scenario adaptability is obtained; In S2, tool performance reasoning includes using k-nearest neighbor entropy to estimate and analyze the information gain ratio and using Kriging interpolation to reason about the tool. By analyzing the tool parameters, various evaluation indicators are obtained. Then, the tools are ranked, and finally, the ranking of the candidate tool set is obtained based on the comprehensive multi-dimensional indicators, which serves as the benchmark ranking. In S2, the k-nearest neighbor entropy estimation analysis is used to calculate the weight of the influence of tool parameters on the machining results, which includes the following two parts: First, k-nearest neighbor entropy estimates the entropy of multidimensional evaluation indicators in a tool set. The input to this method is the tool parameters and corresponding evaluation indicators of the machining results from N machining records. In tool application scenarios, the evaluation of whether the tool selection is reasonable is usually analyzed from three perspectives: the surface roughness of the workpiece, the wear of the machining tool, and the machining cost. Therefore, k-nearest neighbor entropy estimation is calculated by the following formula: ； in, It represents the entropy of the three-dimensional vector of processing quality, wear, and cost; For gamma function, It is the derivative of the logarithm of the gamma function. For the sample size, Select the number of nearest neighbors. for Dimensions for With the Euclidean distance between neighboring points; Secondly, the information gain ratio analyzes the correlation between tool parameters and indicators through changes in entropy, and analyzes the importance of tool parameters. For input with Tool vectors with various parameters and their corresponding parameters For dimensional evaluation metrics, selecting a parameter to divide the input set into... Data with the same value are grouped into the same set, and the result in each set is calculated separately. Information entropy of dimensional evaluation index Then the information gain ratio of the tool parameter is: After obtaining the information gain ratio of all parameters Then, all ratios are normalized to obtain the tool parameter importance vector. ; For continuous attributes in tool parameters, the information gain of continuous attributes is calculated using the bisection method. For attributes with incomplete information in tool parameters, the expected value is used instead of the default value.

2. The intelligent tool selection method based on k-nearest neighbor entropy estimation and Kriging method according to claim 1, characterized in that, The specific process of S1 is as follows: The user inputs the machining conditions into the inference engine, which combines expert knowledge to search for suitable tools in the tool database. The machining conditions include machine tool material and process information. Pre-screening is performed based on user-inputted feature requirements, including tool parameters, tool material, and machining accuracy. Based on function, topology, and geometry, most unsuitable tools are filtered out, and after several rounds of filtering, a pre-screened set of candidate tools is obtained.

3. The intelligent tool selection method based on k-nearest neighbor entropy estimation and Kriging method according to claim 2, characterized in that, In S2, a multidimensional space is formed between a tool with m parameters and its corresponding machining result evaluation index. Kriging interpolation uses known points in this space to estimate unknown points, specifically: Define the basic interpolation formula as follows: ,in As weight, Given the values ​​of points, assuming the spatial properties are uniform, then every point in space has the same variance. If the unbiased estimation is satisfied, then: The weights are calculated using the following formula: ； in, For point and points The semivariance function between them is fitted using a Gaussian function: ,in , and For the fitting parameters, Indicates distance, These are the parameters of the Gaussian model. After solving for all weights, the predicted values ​​of the unobserved targets in the index space are obtained through the basic interpolation formula. Kriging interpolation yields performance predictions for all tools in the candidate tool set, including surface roughness Ra, tool life, and machining cost. Based on these performance predictions, tools are ranked and recommended.

4. The intelligent tool selection method based on k-nearest neighbor entropy estimation and Kriging method according to claim 3, characterized in that, In S3, based on the baseline ranking of tools obtained in S2, a comprehensive ranking is generated using a process scenario-based fusion algorithm, including the following steps: S31. Use collaborative filtering to retrieve user usage records, calculate tool reputation similarity, and predict user preferences for different tools by calculating the similarity matrix, thus obtaining the user preference ranking of the tool selection set. S32. The Pagerank method is used to analyze the tool combination usage from the machining records, construct the adjacency matrix of the tool association graph, and use the principal eigenvector obtained by it to approximate the solution to obtain the tool combination ranking. S33. Combine the baseline sorting, user preference sorting, and tool combination sorting, and obtain the comprehensive sorting result by weighted summation sorting.

Images