Multi-attribute target decision evaluation method for anti-corrosion performance enhancement design of bent bolt

Through the multi-attribute target decision evaluation method and support vector regression algorithm, the problem of insufficient corrosion resistance of wind power tower bend bolt coating in extreme environments is solved, and the efficient corrosion resistance evaluation and design of bend bolts is achieved, which extends the equipment Service life.

CN119066802AActive Publication Date: 2024-12-03CHINA PRODUCTIVITY CENT FOR MASCH

Patent Information

Application Number
CN202411052478.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-02
Publication Date
2024-12-03
Estimated Expiration
2044-08-02

AI Technical Summary

Technical Problem

The problem of aging and peeling of the coating of pipe bend bolts for wind power towers in extreme environments leads to insufficient corrosion resistance and the traditional evaluation methods lack objectivity and accuracy.

Method used

Using a multi-attribute target decision evaluation method, combined with support vector regression (SVR) algorithm, a coating corrosion resistance evaluation model is established through adhesion performance testing and corrosion resistance testing, and the weights of each evaluation index are automatically solved to achieve objective coating performance. , accurate assessment.

Benefits of technology

It improves the objectivity and accuracy of the corrosion resistance performance evaluation of bent bolts, guides the design of enhanced corrosion resistance, extends the service life of the bolts, and reduces the risk of failure.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure BDA0004975319720000032
    Figure BDA0004975319720000032
  • Figure BDA0004975319720000042
    Figure BDA0004975319720000042
  • Figure BDA0004975319720000043
    Figure BDA0004975319720000043
Patent Text Reader

Abstract

The invention provides a multi-attribute target decision evaluation method for the design of enhancing the corrosion resistance of a bent bolt, and the method is based on a multi-attribute decision analysis theory, and innovatively introduces a support vector regression (SVR) machine learning algorithm to solve the weight coefficient of each performance evaluation index. The objective and accurate performance comprehensive evaluation result is expected to be obtained. Compared with a traditional subjective weighting method, the method has the greatest advantages that index association rules and weight distribution characteristics contained in historical evaluation data can be fully explored, adaptive solution from data to weight is achieved through a machine learning algorithm, and the method has the advantages of being high in objectivity, good in adaptability, high in interpretability and the like. The method can also be widely applied to corrosion resistance evaluation of a novel coating system, and has important guiding significance for optimization iteration of a coating formula and process parameters.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of design and evaluation of enhanced anti-corrosion performance of pipeline bent bolts, and in particular to a multi-attribute target decision-making evaluation method for designing enhanced anti-corrosion performance of pipeline bent bolts. Background Art

[0002] Bent bolts play a key role in the fixation of wind turbine towers with their unique bending strength and deformation capacity. They can mitigate and absorb the external forces on the structure, making the structure safer and more solid. Compared with ordinary bolts, bent bolts have a larger cross-sectional area, which can improve the bearing capacity of the connection and make the structure more solid and durable. At the same time, they have certain bending properties, can play a certain role in preventing loosening, and have good anti-seismic and shock-absorbing properties.

[0003] The anti-corrosion technology of bent bolts for concrete wind turbine tower segments is currently in a stage of continuous progress and improvement. In order to meet the challenges of long-term operation of wind power facilities under harsh environmental conditions, especially in highly corrosive areas such as oceans and deserts, modern anti-corrosion technology has adopted a variety of advanced materials, processes and coating systems to enhance the corrosion resistance of bent bolts, thereby significantly extending their service life and reducing the risk of failure due to corrosion, thereby improving the stability and reliability of the equipment.

[0004] However, due to the harsh operating environment of the fastening bolts for wind turbine towers, the performance limitations of coating materials in extreme environments remain a major problem. These extreme environments may include high salinity, high humidity, highly corrosive gases, etc. These environmental factors will accelerate the aging, cracking and peeling of the coating, causing the bolts to lose effective protection, which can be solved to a certain extent through moderate oxidation. In addition, the adhesion between the coating and the bolt substrate is also a key issue. If the adhesion is insufficient, the coating is easy to peel off, thereby losing its anti-corrosion function. Bolts treated with Dacromet technology have extremely strong corrosion resistance, and the coating is not easy to fall off, but the Dacromet coating contains extremely harmful hexavalent chromium, and will cause contact corrosion with stainless steel parts.

[0005] How to effectively evaluate the effects of the above-mentioned solutions on corrosion resistance, so as to seek the best enhanced corrosion resistance design for pipe bend bolts for wind turbine towers under room conditions, is of great significance for the long-term, safe and effective use of pipe bend bolts. Summary of the invention

[0006] The present invention adopts a corrosion resistance evaluation method based on multi-attribute objective decision-making to comprehensively evaluate the adhesion and corrosion resistance of pipeline bent bolt samples after anti-corrosion design treatment, and finally obtains more comprehensive and effective evaluation results to guide the design of enhanced corrosion resistance of pipeline bent bolts.

[0007] To achieve the above object, the present invention provides a multi-attribute objective decision-making evaluation method for the design of enhanced corrosion resistance of bent bolts, characterized in that the evaluation method comprises:

[0008] Step 1: Prepare the bent bolts required for the test and evaluate and test the status of the bent bolts;

[0009] Step 2: Only blackening treatment or Dacromet coating is performed on some bent bolts, and some bent bolts are firstly blackened and then Dacromet coating is performed;

[0010] Step 3: Conduct an adhesion performance test on the bent bolt coating treated in step 2, and then score and record the corresponding data based on the test results;

[0011] Step 4: Conduct corrosion resistance test on the bent bolt coating treated in step 2, and then score and record the corresponding data according to the test results;

[0012] Step 5: Score the comprehensive anti-corrosion performance of the bent bolt coating treated in step 2 and record the corresponding data;

[0013] Step 6: Based on the support vector regression method, the data recorded in steps 3-5 are used to solve the weights of the evaluation indicators of each performance test in steps 3 and 4 that affect the evaluation of the anti-corrosion performance of the coating to obtain an evaluation model;

[0014] Step 7: Use the evaluation model to evaluate the enhanced corrosion resistance design of bent bolts.

[0015] Further preferably, the adhesion performance test includes a wiping method, a metallographic observation method and a physical assembly method.

[0016] Further preferably, the corrosion resistance test includes a salt spray corrosion method.

[0017] More preferably, the scores are in five levels of 1-5.

[0018] Further preferably, in step 6, an SVR algorithm is selected to perform association modeling on the recorded data and its corresponding performance labels. The SVR model adopts a Gaussian radial basis kernel function (RBF). By combining grid search with cross-validation, different combinations of penalty factor C and kernel function parameter g are traversed. The generalization performance on the validation set is evaluated based on the mean square error (MSE) indicator, and the optimal parameter combination is determined to train and generate the SVR model.

[0019] Further preferably, the data recorded in steps 3-5 include the following indicators: rubbing adhesion grade x 1, Metallographic observation method adhesion grade x 2 , physical assembly method adhesion level x 3 And salt spray test corrosion resistance grade x 4 , and the comprehensive anti-corrosion performance score of the coating y∈[1,5]; if the performance data of n coating samples are collected, it constitutes the original "feature-label" data set,

[0020] D={(x 1 ,y 1 ),(x 2 ,y 2 ),L,(x n ,y n )}

[0021] Where xi = [x i1 ,x i2 ,x i3 ,x i4 ] T is the feature vector of the i-th sample, y i Provide a comprehensive performance score for it;

[0022] The scoring results of each indicator are normalized and uniformly mapped to the interval [0,1] to eliminate the influence of different dimensions. Specifically, the maximum and minimum value normalization method is used to normalize the original feature matrix To convert:

[0023]

[0024] After normalization, the preprocessed feature matrix X∈[0,1] is obtained n×4 .

[0025] Further preferably, based on the preprocessed data set D={(x 1 ',y 1 ),(x 2 ',y 2 ),L,(x n ',y n )}, the SVR algorithm is used to build the model. SVR introduces the ε-insensitive loss function to seek an optimal regression hyperplane f(x)=w in the feature space. T φ(x)+b, so that the model outputs f(x i ) and the true label y i The error between them is controlled within ε, where w and b are model parameters, and φ(·) represents the transformation function that maps the original features to the high-dimensional space;

[0026] The Gaussian radial basis kernel function (RBF) κ(x i ,x j)=exp(-γ‖x i -x j ‖ 2 ) is used as the kernel function of the SVR model, where γ>0 is the kernel function width parameter, and the slack variable is introduced The objective optimization problem of SVR is expressed as:

[0027]

[0028] Among them, C>0 is the penalty coefficient, which is used to control the complexity of the model and the tolerance to outliers. By solving the dual form of the above convex quadratic programming problem, the optimal decision function of SVR is obtained:

[0029]

[0030] In the formula, is a Lagrange multiplier satisfying

[0031] In order to obtain the SVR model with the best generalization performance, the hyperparameters $(C,\gamma)$ of the SVR model are tuned. A method combining grid search and K-fold cross-validation is adopted to traverse different (C,γ) combinations in the parameter space. The optimal parameter configuration is determined by minimizing the mean square error (MSE) on the validation set to guide the training and generation of the SVR model.

[0032] Further preferably, all support vector samples (xs1, ys1), (xs2, ys2), L, (xsl, ysl) and their corresponding Lagrange multipliers are extracted from the trained SVR model. Calculate the inner product of the support vector and the multiplier to get the original weight of the input feature β = [β 1 ,β 2 ,β 3 ,β 4 ] T :

[0033]

[0034] The maximum and minimum values ​​of β are normalized to obtain the relative importance weights of each evaluation index:

[0035]

[0036] At this point, the weight vector w = [w 1 ,w 2 ,w 3 ,w 4 ] T , the scoring results of each index of the coating to be evaluated are x = [x1 ,x 2 ,x 3 ,x 4 ] T The weighted sum of them and w is obtained to obtain the comprehensive evaluation model for the design of enhanced corrosion protection performance:

[0037]

[0038] Further preferably, the Dacromet coating is divided into acidic conditions and alkaline conditions.

[0039] The present invention proposes a novel coating anti-corrosion performance evaluation method, which is based on the theory of multi-attribute decision analysis and innovatively introduces the support vector regression (SVR) machine learning algorithm to solve the weight coefficients of various performance evaluation indicators, in order to obtain objective and accurate comprehensive performance evaluation results. Compared with the traditional subjective weighting method, the biggest advantage of this method is that it can fully explore the indicator association rules and weight distribution characteristics contained in the historical evaluation data, and realize the adaptive solution from data to weight through the machine learning algorithm, which has the characteristics of strong objectivity, good adaptability, and strong interpretability. This method can be widely used in the evaluation of the anti-corrosion performance of new coating systems, and has important guiding significance for the optimization and iteration of coating formulations and process parameters.

[0040] In the data preprocessing stage, the present invention adopts the maximum and minimum value normalization method to perform dimensionless processing on the feature data according to the dimensional differences of different evaluation indicators, and uniformly maps them to the [0, 1] interval to eliminate the dimensional influence.

[0041] When calculating the weights, the SVR algorithm is used to perform association modeling on the normalized feature data and the corresponding performance labels. Considering the nonlinear correlation characteristics between the coating anti-corrosion performance evaluation index and the comprehensive performance, the model adopts the Gaussian radial basis kernel function (RBF) to enhance the nonlinear mapping capability. At the same time, in order to seek the optimal parameter configuration of the model, a method combining grid search and cross-validation is used to traverse different combinations of penalty factor C and kernel function parameter g. The generalization performance on the validation set is evaluated based on the mean square error (MSE) indicator, and then the optimal parameter combination is determined to guide the training and generation of the SVR model. DETAILED DESCRIPTION

[0042] The implementation method of the present invention is specifically explained below through examples.

[0043] The multi-attribute objective decision-making evaluation method for the design of enhanced corrosion resistance of bent bolts in the present invention is characterized in that the evaluation method comprises:

[0044] Step 1: Prepare the bent bolts required for the test and evaluate and test the status of the bent bolts;

[0045] Step 2: Only blackening treatment or Dacromet coating is performed on some bent bolts, and some bent bolts are firstly blackened and then Dacromet coating is performed;

[0046] Step 3: Conduct an adhesion performance test on the bent bolt coating treated in step 2, and then score and record the corresponding data based on the test results;

[0047] Step 4: Conduct corrosion resistance test on the bent bolt coating treated in step 2, and then score and record the corresponding data according to the test results;

[0048] Step 5: Score the comprehensive anti-corrosion performance of the bent bolt coating treated in step 2 and record the corresponding data;

[0049] Step 6: Based on the support vector regression method, the data recorded in steps 3-5 are used to solve the weight of the evaluation index of each performance test in steps 3 and 4 that affects the evaluation of the anti-corrosion performance of the coating to obtain an evaluation model;

[0050] Step 7: Use the evaluation model to evaluate the enhanced corrosion resistance design of bent bolts.

[0051] In the specific experiment, step 1 prepares the bent bolt samples required for the test as follows:

[0052] Prepare a batch of 40CrNiMo bent bolts for testing, with thread size of M36, and conduct tensile tests on the bent bolts. The mechanical properties of the bolt specimens tested should meet the performance requirements of wind turbine tower bolts in the high-strength bolt connection pair for wind turbine towers in NB / T 31082-2016 to ensure that the test specimens can meet practical needs. Then divide the batch of samples into six groups: A, B, C, D, E, and F, and carry out subsequent steps respectively.

[0053] Step 2.1 Blackening of bent bolts

[0054] The three groups of bolts B, C and D are blackened. The specific process is as follows:

[0055] Clean and degrease the bent bolts → cold water cleaning → pickling → cold water cleaning → oxidation and blackening → cold water cleaning → saponification → cold water cleaning (oil immersion) → drying

[0056] 1) Cleaning and degreasing

[0057] In order to remove the oil stains attached to the surface of the bolt, the bent bolt needs to be cleaned in the degreasing tank for 20 to 30 minutes. During the specific degreasing process, the workpiece must be completely immersed in the degreasing liquid. The pH value of the degreasing liquid concentration is generally controlled at 12-14, and the processing time is usually 10-30 minutes. During the degreasing process, it is necessary to shake up and down several times every 3-5 minutes to ensure that the degreasing liquid can act evenly on the surface of the workpiece. When the concentration of the liquid is lower than pH12, degreasing powder needs to be added to maintain the activity of the degreasing liquid. Composition and ratio of degreasing solvent:

[0058] Table 1 Degreasing liquid composition

[0059] name Mass percentage Sodium hydroxide 1-3% Fatty alcohol polyoxyethylene ether 1-5% Sodium Nitrite 0.1-0.5%

[0060] 2) Cold water cleaning

[0061] After degreasing, the bent bolts are cleaned in cold water to remove the degreasing solvent (acid) on the surface.

[0062] 3) Pickling

[0063] The main purpose of pickling is to remove the oxide scale, rust and other impurities on the surface of the bolt, and activate the surface of the bent bolt so that it is easier for the bent bolt to form a dense oxide film in the blackening liquid. This step requires the bent bolt to be pickled in 20-30% hydrochloric acid for more than 30 seconds. (The acid solution is sprayed onto the metal surface at high pressure to remove the oxide scale and rust. This method is suitable for large and complex-shaped parts and can improve the pickling efficiency.)

[0064] 4) Oxidation and blackening

[0065] The main purpose of this step is to form a thin film of oxide or compound on the metal surface. The bent bolts that have been cleaned after pickling are immersed in an alkaline blackening solution and heated to 130°C. They are fully immersed until a uniform black oxide film is formed on the surface of the bent bolts.

[0066] Table 2 Blackening liquid composition and mass (volume) fraction:

[0067] name Mass percentage Copper sulfate 3% Silver nitrate 1% Glass acetic anhydride (volume fraction) 5% Boric acid 2% Skimmed acid 10% Sodium sulfite 1%

[0068] 5) Saponification

[0069] The main function of this process is to convert the iron in the micropores of the iron oxide film formed by the blackening of the bent bolts into iron stearate. This conversion makes the surface of the workpiece passivated, thereby improving its corrosion resistance. Soak the bent bolts after blackening in 2-3% boiling soapy water for 5 minutes.

[0070] 6) Drying

[0071] The saponified and cleaned bent bolts are dried at 150°C to remove surface moisture.

[0072] After the blackening treatment is completed, the bent bolts in group B are retained, and the bent bolts in groups C, D, E, and F are coated with Dacromet coating respectively.

[0073] Step 2.2 The specific process of Dacromet coating on bent bolts is as follows:

[0074] Dip coating → centrifugal → pre-baking → sintering → cooling

[0075] 1) Dip coating

[0076] The bent bolts after blackening treatment are immersed in a chromium-free Dacromet solution for 30 seconds to ensure that a complete and uniform chromium-free Dacromet coating is formed on the surface of the bolts.

[0077] The composition and mass percentage of chromium-free Dacromet solution. The role of silicon dioxide, zinc oxide and succinic acid is to adjust the pH value.

[0078] After the preliminary preparation of the Dacromet solution is completed, the Dacromet solution is divided into two groups, ① and ②, and the pH values ​​are adjusted to slightly alkaline and slightly alkaline respectively. The bent bolts of groups C and E are immersed in the slightly alkaline solution ①, and the bent bolts of groups D and F are immersed in the slightly alkaline solution ②.

[0079] The detailed composition and proportion of Dacromet solution are shown in the table below.

[0080] Table 3 Dacromet solution composition table

[0081] name Mass percentage Polyethylene glycol-400 19% Zinc powder 22% Aluminum powder 4.5% Silane coupling agent KH560 21% water 15% Ethanol 15% Sodium molybdate 1% NP-10 2% Silicon dioxide Less than 0.5% Zinc Oxide Less than 0.5% Succinic acid Less than 0.5%

[0082] 2) Centrifugation

[0083] When centrifuging, the machine should be rotated forward and reversed twice, each time for ten to twenty seconds, and the speed is generally 200-300 rpm. The purpose of this step is to ensure that there is no accumulation of liquid, hanging liquid, bubbles, or missing coating on the parts.

[0084] 3) Pre-baking

[0085] The workpiece of Dacromet wet film needs to be pre-baked at 120±20℃ for 10-15 minutes as soon as possible to evaporate the water in the coating and achieve uniform leveling.

[0086] 4) Sintering

[0087] The pre-baked bent bolts need to be sintered at a high temperature of about 300°C for 20-40 minutes. The purpose of this step is to form a good bonding force between the Dacromet coating and the oxide film to improve the wear resistance and corrosion resistance of the coating.

[0088] 5) Cooling

[0089] The sintered bolts are cooled and prepared for subsequent testing.

[0090] The treatment method for each group of bolts is shown in the following table.

[0091] Table 4 Treatment table of each group of bolts

[0092] A No treatment B Blackening treatment C Blackening treatment and Dacromet (weak alkaline) D Dacromet (weakly alkaline) E Blackening and Dacromet (alkaline) F Dacromet (alkaline)

[0093] Step 3 Bend bolt coating adhesion test

[0094] 1) Use clean white wiping paper to repeatedly and evenly wipe the surface coating of the bent bolt. If no other colors appear on the white paper, the coating quality is excellent; if a small amount of other colors appear on the white paper, the coating quality is average; if other colors appear obviously on the white paper, the coating quality is poor. Score the six groups of bolts based on the test results.

[0095] 2) Metallographic SEM organization observations were performed on the six groups of bolts to observe the microscopic state of the surface coating. If a dense coating was observed on the surface without damage, the coating had excellent adhesion; if damage to the Dacromet coating was observed but no bare metal surface was observed, the coating had good adhesion; if obvious damage to the coating was observed and a bare metal surface was seen, the coating quality was poor.

[0096] 3) Manually tighten and disassemble the six groups of bolts respectively, and visually observe whether the surface coating has fallen off or whether there are scratches on the surface. If there is no coating falling off or damage, it means that the coating quality is excellent; if the coating surface is damaged, or a small amount of coating falls off, but no bare metal surface is found, the coating quality is good; if obvious coating falling off occurs, and the metal surface is exposed or even damaged, it means that the coating quality is poor.

[0097] After the test, the adhesion tests of the six groups of bolts were scored and recorded according to the test results. The scoring results are as follows.

[0098] Table 5 Adhesion ability score

[0099] Serial number Wipe method Metallographic observation Physical assembly A 5.0 1.0 1.0 B 4.0 3.0 3.0 C 5.0 5.0 4.0 D 5.0 5.0 5.0 E 5.0 4.0 4.0 F 5.0 5.0 4.0

[0100] The numbers in the table indicate the coating adhesion level, where 1 indicates poor adhesion and 5 indicates excellent adhesion. Step 4 Corrosion resistance test of bent bolts

[0101] The six groups of bent bolts were placed in a salt spray test machine for salt spray corrosion for a certain period of time. The corrosion results were recorded at regular intervals. After the test, the corrosion resistance of the six groups of bent bolts was scored and recorded based on the salt spray test results, i.e., the corrosion time and corrosion degree.

[0102] Table 6 Corrosion resistance rating

[0103] Serial number Corrosion resistance rating A 1.0 B 2.0 C 5.0 D 4.0 E 5.0 F 3.0

[0104] The numbers in the table indicate the corrosion resistance level, 1 indicates poor corrosion resistance and 5 indicates excellent corrosion resistance.

[0105] Step 5: Comprehensive corrosion protection performance score of coating

[0106] Experts were asked to score and record the comprehensive corrosion resistance of the six groups of bent bolts based on historical experience. The specific scoring results are shown in Table 7 below.

[0107] Table 7 Comprehensive anti-corrosion performance score

[0108]

[0109]

[0110] The numbers in the table indicate the corrosion resistance level, 1 indicates poor comprehensive corrosion resistance level, and 5 indicates excellent comprehensive corrosion resistance level.

[0111] Step 6: Solution of coating anti-corrosion performance evaluation index weights based on support vector regression

[0112] 6.1 Data Preparation and Preprocessing

[0113] By testing the adhesion and corrosion resistance of anti-corrosion coatings processed under different process conditions, a set of performance evaluation data was obtained. The data includes the following indicators: Wiping method adhesion level x 1 , Metallographic observation method adhesion grade x 2 , physical assembly method adhesion level x 3 And salt spray test corrosion resistance grade x 4 , and the comprehensive anti-corrosion performance score of the coating y∈[1,5]. Assume that the performance data of n coating samples are collected to form the original "feature-label" data set:

[0114] D={(x 1 ,y 1 ),(x 2 ,y 2 ),L,(x n ,y n )}

[0115] Where xi = [x i1 ,x i2 ,x i3 ,x i4 ] T is the feature vector of the i-th sample, y i Give it a corresponding comprehensive performance score.

[0116] In the data preprocessing stage, the scoring results of each indicator are first normalized and uniformly mapped to the interval [0,1] to eliminate the influence of different dimensions. The maximum and minimum value normalization method is used to normalize the original feature matrix. To convert:

[0117]

[0118] After normalization, the preprocessed feature matrix X∈[0,1] is obtained n×4 .

[0119] 6.2 Construction and Training of SVR Model

[0120] Based on the preprocessed data set D = {(x 1 ',y 1 ),(x 2 ',y 2 ),L,(x n ',y n )}, the SVR algorithm is used to build a comprehensive anti-corrosion performance evaluation model for coatings. SVR introduces an ε-insensitive loss function to seek an optimal regression hyperplane f(x)=w in the feature space. T φ(x)+b, so that the model outputs f(x i ) and the true label y i The error between them is controlled within ε. Among them, w, b are model parameters, and φ(·) represents the transformation function that maps the original features to the high-dimensional space.

[0121] Considering the possible complex nonlinear relationship between various performance indicators and the comprehensive score, this paper uses the Gaussian radial basis kernel function (RBF) κ(x i ,x j )=exp(-γ‖x i -x j ‖ 2 ) is used as the kernel function of the SVR model, where γ>0 is the kernel function width parameter. Introducing the slack variable The objective optimization problem of SVR can be expressed as:

[0122]

[0123] Among them, C>0 is the penalty coefficient, which is used to control the complexity of the model and the tolerance to outliers. By solving the dual form of the above convex quadratic programming problem, the optimal decision function of SVR can be obtained:

[0124]

[0125] In the formula, is a Lagrange multiplier satisfying

[0126] In order to obtain the SVR model with the best generalization performance, it is necessary to tune the model hyperparameters $(C,\gamma)$. A method combining grid search and K-fold cross validation is used to traverse different (C,γ) combinations in the parameter space, and the optimal parameter configuration is determined by minimizing the mean square error (MSE) on the validation set to guide the training and generation of the model.

[0127] 6.3 Solution of evaluation index weights

[0128] Extract all support vector samples (xs1, ys1), (xs2, ys2), L, (xsl, ysl) and their corresponding Lagrange multipliers from the trained SVR model Calculate the inner product of the support vector and the multiplier to get the original weight of the input feature β = [β 1 ,β 2 ,β 3 ,β 4 ] T :

[0129]

[0130] In order to make the weights comparable, the maximum and minimum values ​​of β are normalized to obtain the relative importance weights of each evaluation index:

[0131]

[0132] At this point, the weight vector w = [w 1 ,w 2 ,w 3 ,w 4 ] T , the scoring results of each index of the coating to be evaluated are x = [x 1 ,x 2 ,x 3 ,x 4 ] T By taking the weighted sum of each with w, the comprehensive score of the coating's corrosion resistance can be obtained:

[0133]

[0134] Step 7 Summary and evaluation of the performance of anti-corrosion coating on bent bolts

[0135] Data on the adhesion grade by wiping method, the adhesion grade by metallographic observation method, the adhesion grade by physical assembly method, and the corrosion resistance performance grade. To comprehensively evaluate the corrosion resistance performance of each group of bolts, this paper further comprehensively adopts a multi-attribute decision-making method based on TOPSIS (Technique for Order Preference by Similarity to an Ideal Solution), weights and integrates the scoring results of different indicators to obtain the comprehensive score of the corrosion resistance performance of the stud coating, providing a quantitative basis for the optimal selection of the coating scheme.

[0136] In this scheme, adhesion tests and salt spray corrosion tests were carried out on bolt samples treated by different processes, and data on the adhesion grade by wiping method, the adhesion grade by metallographic observation method, the adhesion grade by physical assembly method, and the corrosion resistance performance grade of six groups of bolts were obtained. To comprehensively evaluate the corrosion resistance performance of each group of bolts, this paper uses a multi-attribute decision-making method based on TOPSIS (Technique for Order Preference by Similarity to an Ideal Solution), weights and integrates the scoring results of different indicators to obtain the comprehensive score of the corrosion resistance performance of the bolt coating, providing a quantitative basis for the optimal selection of the coating scheme.

[0137] The TOPSIS method is a commonly used multi-attribute decision-making method. Its basic idea is to calculate the distances between each alternative solution and the positive and negative ideal solutions, establish a relative closeness index, and take the solution with the smallest distance from the positive ideal solution and the largest distance from the negative ideal solution as the optimal choice. In the problem of evaluating the corrosion resistance performance of coatings, the highest grade of each evaluation index can be regarded as the positive ideal solution, and the lowest grade as the negative ideal solution. By calculating the weighted Euclidean distance between each coating scheme and the ideal solution, the degree of closeness to the optimal performance can be obtained.

[0138] Suppose there are m groups of coating schemes to be evaluated, and each group of schemes has n evaluation indicators. Let the The implementation steps of the TOPSIS method are as follows:

[0139] Step 7.1: Construct a standardized scoring matrix R = [r ij m×n .

[0140] Step 7.2: Calculate the weighted standardized scoring matrix {V} = [v {ij} {m×n} , where

[0141] v ij = w j ·r ij , i = 1, 2, L, m; j = 1, 2, L, n ​​

[0142] Step 7.3: Determine the positive and negative ideal solutions:

[0143]

[0144] in,

[0145]

[0146] Step 7.4: Calculate the weighted Euclidean distance between the i-th group of solutions and the positive and negative ideal solutions:

[0147]

[0148] Step 7.5: Calculate the relative closeness of the i-th group of solutions:

[0149]

[0150] Where 0≤c i ≤1, and c i The closer it is to 1, the better the overall performance of the i-th group of solutions.

[0151] Example

[0152] This scheme provides a method for strengthening the corrosion resistance of bent bolts and evaluating the corrosion resistance based on multi-attribute objective decision-making. The specific operation process is as follows:

[0153] (1) Calculation of indicator weights for multi-attribute target evaluation

[0154] i. Historical data collection and preprocessing

[0155] By collecting and screening the historical evaluation data of 6 typical anti-corrosion coatings in a certain engineering field, the performance score data of 10 valid samples of each coating formula were obtained to support the subsequent data analysis and modeling. The original data includes the following attributes: adhesion grade of wiping method (1-5 levels), adhesion grade of metallographic observation method (1-5 levels), adhesion grade of physical assembly method (1-5 levels), corrosion resistance grade of salt spray test (1-5 levels) and comprehensive anti-corrosion performance score (1-5 points), and the corresponding data matrix is ​​shown below.

[0156] Adhesion rating matrix:

[0157]

[0158] Corrosion resistance rating vector: [1 2 5 4 5 3]

[0160] Comprehensive performance score vector: [1.5 2.6 4.8 4.5 4.6 3.8]

[0162] The adhesion grade matrix and the corrosion resistance grade vector are normalized to their maximum and minimum values, respectively, and the heterogeneous data are uniformly mapped to the [0,1] interval to eliminate the dimension difference. The normalized feature matrix and label vector are as follows:

[0163] Feature Matrix:

[0164]

[0165] Label vector: [0.09 0.33 0.97 0.91 0.94 0.70]

[0167] ii.SVR model training and parameter optimization

[0168] Based on the normalized feature matrix and label vector, a "feature-label" dataset for training the SVR model is constructed. Considering the small sample size, the RBF kernel function is selected to achieve nonlinear transformation of the feature space.

[0169] In order to obtain the SVR model with the best performance, the penalty factor C and the kernel function parameter g are used as optimization variables, and different combinations of C and g are traversed through the grid search method. Among them, the candidate value of C is set to 10, and the candidate value of g is set to [0.01, 0.1, 1]. The data set is randomly divided into 5 mutually exclusive subsets using a 5-fold cross-validation method. Four of the subsets are selected as training sets in turn, and the remaining subset is used as a validation set. The generalization performance of the model under each parameter combination is evaluated, and the prediction error on the validation set is measured by the mean square error (MSE).

[0170] By comparing the average cross-validation errors of different parameter combinations, the optimal parameters were finally determined to be C = 10 and g = 0.1. Based on this parameter combination, the SVR model was retrained using all sample data to obtain the final coating anti-corrosion performance evaluation model.

[0171] iii. Solving the evaluation index weights

[0172] All support vector samples and their corresponding Lagrange multipliers are extracted from the trained SVR model, and the original weight vector of the input feature is obtained through matrix product operation. The original weight vector is normalized to the maximum and minimum values, so that the weight component is scaled to the [0,1] interval. The normalized weight vector is the relative importance measure of each evaluation index.

[0173] The normalized indicator weight vector is:

[0174] Wipe weight: 0.16

[0175] Metallographic observation weight: 0.21

[0176] Physical assembly method weight: 0.06

[0177] Corrosion resistance weight: 0.57

[0178] (2) Performance evaluation of anti-corrosion coating on bent bolts based on multi-attribute weights

[0179] According to the test results of the above-mentioned adhesion test and salt spray test, the multi-attribute target decision-making TOPSIS method was comprehensively adopted to conduct the final comprehensive anti-corrosion performance evaluation according to the following weight ratio.

[0180] Table 8 Weight ratio

[0181] Test methods Wipe method Metallographic observation Physical assembly Salt spray corrosion Weight 0.16 0.21 0.06 0.57

Claims

1. A multi-attribute objective decision-making evaluation method for the design of enhanced corrosion resistance of bent bolts, characterized in that: The evaluation methods include: Step 1: Prepare the bent bolts required for the test and evaluate and test the status of the bent bolts; Step 2: Only blackening treatment or Dacromet coating is performed on some bent bolts, and some bent bolts are firstly blackened and then Dacromet coating is performed; Step 3: Conduct an adhesion performance test on the bent bolt coating treated in step 2, and then score and record the corresponding data based on the test results; Step 4: Conduct corrosion resistance test on the bent bolt coating treated in step 2, and then score and record the corresponding data according to the test results; Step 5: Score the comprehensive anti-corrosion performance of the bent bolt coating treated in step 2 and record the corresponding data; Step 6: Based on the support vector regression method, the data recorded in steps 3-5 are used to solve the weights of the evaluation indicators of each performance test in steps 3 and 4 that affect the evaluation of the anti-corrosion performance of the coating to obtain an evaluation model; Step 7: Use the evaluation model to evaluate the enhanced corrosion resistance design of bent bolts.

2. The evaluation method according to claim 1, characterized in that: The adhesion performance test includes a wiping method, a metallographic observation method and a physical assembly method.

3. The evaluation method according to claim 1, characterized in that: The corrosion resistance test includes a salt spray corrosion method.

4. The evaluation method according to claim 1, characterized in that: The ratings are all on a five-level scale of 1-5.

5. The evaluation method according to claim 1, characterized in that: In step 6, the SVR algorithm is selected to perform association modeling on the recorded data and its corresponding performance labels. The SVR model uses a Gaussian radial basis kernel function (RBF). By combining grid search with cross-validation, different combinations of penalty factor C and kernel function parameter g are traversed. The generalization performance on the validation set is evaluated based on the mean square error (MSE) indicator, and then the optimal parameter combination is determined to train and generate the SVR model.

6. The evaluation method according to claim 5, characterized in that: The data recorded in steps 3-5 include the following indicators: adhesion level x1 by wiping method, adhesion level x2 by metallographic observation method, adhesion level x3 by physical assembly method, corrosion resistance level x4 by salt spray test, and comprehensive corrosion resistance score y∈[1,5] of the coating. If the performance data of n coating samples are collected, the original "feature-label" data set is formed. D={(x1,y1),(x2,y2),L,(x n ,y n )} Where xi = [x i1 ,x i2 ,x i3 ,x i4 ] T is the feature vector of the i-th sample, y i Provide a comprehensive performance score for it; The scoring results of each indicator are normalized and uniformly mapped to the interval [0,1] to eliminate the influence of different dimensions. Specifically, the maximum and minimum value normalization method is used to normalize the original feature matrix To convert: After normalization, the preprocessed feature matrix X∈[0,1] is obtained n×4 .

7. The evaluation method according to claim 6, characterized in that: Based on the preprocessed data set D = {(x1',y1),(x2',y2),L,(x n ',y n )}, the SVR algorithm is used to build the model. SVR introduces the ε-insensitive loss function to seek an optimal regression hyperplane f(x)=w in the feature space. T φ(x)+b, so that the model outputs f(x i ) and the true label y i The error between them is controlled within ε, where w, b are model parameters, and φ(·) represents the transformation function that maps the original features to the high-dimensional space; The Gaussian radial basis kernel function (RBF) κ(x i ,x j )=exp(-γ‖x i -x j ‖ 2 ) is used as the kernel function of the SVR model, where γ>0 is the kernel function width parameter, and the slack variable ξ is introduced. i , The objective optimization problem of SVR is expressed as: s.t.f(x i )-y i ≤ε+ξ Among them, C>0 is the penalty coefficient, which is used to control the complexity of the model and the tolerance to outliers. By solving the dual form of the above convex quadratic programming problem, the optimal decision function of SVR is obtained: In the formula, α i , is a Lagrange multiplier satisfying In order to obtain the SVR model with the best generalization performance, the hyperparameters (C, γ) of the SVR model are tuned. A method combining grid search and K-fold cross-validation is used to traverse different (C, γ) combinations in the parameter space. The optimal parameter configuration is determined by minimizing the mean square error (MSE) on the validation set to guide the training and generation of the SVR model.

8. The evaluation method according to claim 7, characterized in that: Extract all support vector samples (xs1, ys1), (xs2, ys2), L, (xsl, ysl) and their corresponding Lagrange multipliers from the trained SVR model Calculate the inner product of the support vector and the multiplier to get the original weight of the input feature β = [β1, β2, β3, β4] T : The maximum and minimum values ​​of β are normalized to obtain the relative importance weights of each evaluation index: At this point, the weight vector w = [w1, w2, w3, w4] of the coating anti-corrosion performance evaluation index is obtained based on the SVR model. T , the scoring results of the various indicators of the coating to be evaluated are x = [x1, x2, x3, x4] T The weighted sum of them and w is obtained to obtain the comprehensive evaluation model for the design of enhanced corrosion protection performance:

9. The evaluation method according to claim 1, characterized in that: The Dacromet coating is divided into acidic conditions and alkaline conditions.

Citation Information

Patent Citations

  • Product concept performance evaluation method based on performance prediction

    CN105740984A

  • Railway bolt

    CN107760075A

  • Method and device for comprehensively evaluating coating protection performance

    CN108510166A

  • Support vector regression model based on iterative aggregation grid search algorithm

    CN112330044A

  • Concrete reinforcement corrosion state evaluation method and system, terminal and storage medium

    CN113344219A

Cited By

  • Optical fiber coating detection equipment and method based on low-temperature extremely-cold condition

    CN121090382A

  • Multi-bolt tightening method based on rolling balance control and related device

    CN121156725A