Welded joint health state assessment method based on credibility enhancement BRB

Abstract

The invention discloses a welded joint health state assessment method based on credibility enhancement BRB. The method comprises the steps that friction stir welding multi-source data are obtained, and after preprocessing is conducted, a welding joint health state evaluation result is obtained based on a credibility enhancement BRB model; the credibility enhancement BRB model comprises a plurality of confidence rules, for each confidence rule, according to the initial confidence of each health state and an external index, a credibility value is obtained, and the confidence of the corresponding health state is adjusted by using the credibility value; and fusing the adjusted confidence values of all the health states under the current confidence rule to obtain a health state evaluation result of the welding joint. According to the method, fusion of external knowledge and multi-source data is fully utilized, meanwhile, a more comprehensive and accurate model is provided for welding quality evaluation through self-supervised learning of label-free data, and particularly, a more accurate and reliable evaluation result can be provided when uncertainty and fuzzy information are processed.

Description

Technical Field

[0001] This invention relates to the field of welding quality assessment technology, and more specifically to a method for assessing the health status of welded joints based on reliability-enhanced BRB. Background Technology

[0002] Friction stir welding (FSW), as a highly efficient and environmentally friendly solid-state welding technology, relies on the frictional heat generated between workpieces by a rotating tool to achieve plastic flow and bonding of metals without melting the materials. It boasts advantages such as high joint strength, a small heat-affected zone, and no spatter, and has been widely applied in engineering fields with extremely high joint performance requirements, including aerospace, automotive manufacturing, rail transportation, and shipbuilding. The health status of the welded joint directly affects the service life and operational safety of the entire structure. Therefore, achieving online health assessment during the welding process is a crucial step in ensuring manufacturing quality and subsequent structural safety.

[0003] Current methods for assessing the health status of friction stir welded joints mainly fall into three categories: modeling methods based on sensor data, qualitative methods based on expert rules, and data-driven machine learning methods. Sensor methods utilize multi-source data such as temperature, force, vibration, and rotational speed to analyze the dynamic characteristics of the welding process; expert system methods achieve diagnostic reasoning through empirical rules; while data-driven methods, such as neural networks and support vector machines, rely on a large number of historical labeled samples for training and prediction.

[0004] While the aforementioned methods each have their advantages, they generally share the following common problems: the evaluation results lack a reliable quantification mechanism, the models have poor interpretability, and they are prone to misjudgment in complex operating conditions or scenarios with insufficient samples. Especially in highly uncertain environments, such as inconsistent materials, operational disturbances, or sensor noise, the reliability of health assessment systems becomes uncontrollable. Furthermore, many assessment models heavily rely on the sufficiency and consistency of labeled data, which is difficult to meet in engineering practice, further limiting their application and promotion.

[0005] Therefore, how to overcome the above problems is an issue that urgently needs to be addressed by those skilled in the art. Summary of the Invention

[0006] In view of the above problems, the present invention is proposed to provide a method for assessing the health status of welded joints based on reliability-enhanced BRB that overcomes or at least partially solves the above problems.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: This invention provides a method for assessing the health status of welded joints based on a reliability-enhanced BRB (Brand Reliability Building Block) system, comprising the following steps: Multi-source data of friction stir welding were acquired, preprocessed, and then input into the credibility-enhanced BRB model to obtain the health status assessment results of the welded joint. The confidence enhancement BRB model includes multiple confidence rules. For each confidence rule, a confidence value is obtained based on the initial confidence of each health state and external indicators. The confidence value is then used to adjust the confidence of the corresponding health state. The model also integrates the adjusted confidence values ​​of all states under the current confidence rule and determines the health status assessment result of the welded joint based on the integrated confidence value.

[0008] Preferably, the multi-source data for friction stir welding includes X-direction current, Y-direction current, Z-direction current, voltage, and pressure.

[0009] Preferably, the preprocessing steps include: The current, voltage, and pressure data in the X, Y, and Z directions were measured. Normalize the current data, voltage data, and pressure data respectively; The normalized current, voltage, and pressure data are weighted and fused to form the welding quality value.

[0010] Preferably, adjusting the confidence level of the corresponding health status using the confidence value includes: The confidence values ​​of each health state are weighted and fused based on the confidence level of the current confidence rule to obtain the confidence value of the current confidence rule. ;

[0011] In the formula, It is the first k The confidence level of a confidence rule. Is with the first i A confidence value related to a health status; These are the types of health status; The confidence level of each health state is adjusted using the confidence value of the current confidence rule.

[0012] Preferably, the confidence values ​​after state adjustment under the multiple confidence rules are integrated, including:

[0013] L Indicates the number of confidence rules. k This represents the index of the confidence rule sequence number. N Indicates the types of health status. i Represents a health status index. This indicates the adjusted confidence level.

[0014] Preferably, the credibility-enhanced BRB model employs an adaptive evolutionary strategy algorithm based on the projection covariance matrix for parameter optimization, the steps of which include: The objective function for constructing the credibility-enhancing BRB model is expressed as:

[0015] In the formula, It is a loss function used to measure the difference between the model's prediction and the actual health status; Indicates unlabeled samples ; These are the parameters of the confidence-enhanced BRB model, including rule weights and adjusted confidence levels. It is the number of unlabeled samples; This indicates that the credibility enhancement BRB model has improved parameters. The following sample The predicted output, j Indicates the sample number. This represents the target output generated based on pseudo-labels.

[0016] An optimization model is constructed to constrain the objective function. The optimization model is as follows:

[0017]

[0018] In the formula, The adjusted confidence level. Using the rule weights, the parameters in the confidence enhancement BRB model are updated through the optimization model to obtain a health status prediction model.

[0019] This invention provides a method for assessing the health status of welded joints based on reliability-enhanced BRB, aiming to achieve high-precision and high-reliability assessment of the health status of welded joints by introducing external welding quality and parameter optimization algorithms.

[0020] Specifically, compared with the prior art, the beneficial effects of the above-mentioned technical solutions provided by the embodiments of the present invention include at least the following: 1. By using multi-source data and introducing external knowledge for "confidence correction", the reliability and robustness of the assessment can be effectively improved, and the interpretability and user trust of the model can be enhanced.

[0021] 2. By utilizing unlabeled data and employing self-supervised learning, the internal parameters of the credibility-enhancing BRB model are automatically optimized. This not only makes full use of massive amounts of unlabeled data and reduces data acquisition costs, but also improves the accuracy of health status identification. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0023] Figure 1 The attached figure is a flowchart of a weld joint health status assessment method based on a credibility-enhanced BRB model provided by the present invention. Figure 2 The attached figure is a comparative analysis diagram of the BRB model for the weld joint health status assessment method based on the credibility-enhanced BRB model provided by the present invention. Figure 3 The attached figure is a comparative analysis of the weld joint health status assessment method based on the credibility-enhanced BRB model provided by this invention with other methods; Figure 4 The attached figure is a comparative analysis of the weld joint health status assessment method based on the credibility-enhanced BRB model provided by the present invention and the Fuzzy C-Means (FCM) algorithm. Figure 5 The attached figure is a comparative analysis diagram of the weld joint health status assessment method based on the credibility-enhanced BRB model provided by the present invention and a fuzzy system. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] This invention discloses a method for assessing the health status of welded joints based on a Belief Rule Base (BRB), comprising the following steps: Multi-source data of friction stir welding were acquired, preprocessed, and then input into the credibility-enhanced BRB model to obtain the health status assessment results of the welded joint. The confidence enhancement BRB model includes multiple confidence rules. For each confidence rule, a confidence value is obtained based on the initial confidence of each health state and external indicators. The confidence value is then used to adjust the confidence of the corresponding health state. The model also integrates the adjusted confidence values ​​of all states under the multiple confidence rules and determines the health status assessment result of the welded joint based on the integrated confidence value.

[0026] In one optional embodiment, the multi-source data of friction stir welding selects features that can comprehensively characterize the health status, including X-direction current, Y-direction current, Z-direction current, voltage, and pressure.

[0027] Furthermore, the above multi-source data is preprocessed to obtain the welding quality; In this embodiment, the aforementioned feature quantities can be obtained from data from multiple sensors, and the preprocessing steps include: 1) The measured current data in the X, Y, and Z directions are as follows: , , Voltage data is V Pressure data is P .

[0028] 2) Normalize the current, voltage, and pressure data respectively; this is used to convert the current, voltage, and pressure indicators to the range of [0,1] so that they can be weighted and fused under a unified scale.

[0029] In this embodiment, the current, voltage, and pressure data can be normalized using the following formula: The normalization of current data in the X, Y, and Z directions is shown in the following formula:

[0030]

[0031]

[0032] The voltage data normalization process is shown in the following formula:

[0033] The normalization process for pressure data is shown in the following formula:

[0034] 3) The normalized current, voltage, and pressure data are weighted and fused to form the welding quality value.

[0035] Assume the weights of the current in the three directions are as follows: 、 、 The weights for voltage and pressure are respectively... and The formula for calculating welding quality is as follows:

[0036] in: , , These are the normalized measurements of the current in the X, Y, and Z directions, respectively. It is the normalized value of the voltage; It is the normalized value of the pressure; , , , and These are the weight coefficients of each feature, satisfying: .

[0037] In an optional embodiment, the BRB model is used to obtain the weld joint health status assessment results based on the fusion weld quality; In this application, the confidence enhancement BRB model includes multiple confidence rules. For each confidence rule, a confidence value is obtained based on the initial confidence of each health state and external indicators. The confidence value is then used to adjust the confidence of the corresponding health state. Finally, the adjusted confidence values ​​of each state under all confidence rules are integrated to obtain the health status assessment result of the welded joint.

[0038] In one specific embodiment, the confidence enhancement BRB model consists of a series of confidence rules, wherein the first... k A confidence rule is expressed as: :IF ,

[0039] Then

[0040]

[0041] in: Indicates the first k Confidence rules, ; Indicates the first i The possible values ​​of the prerequisite attributes; Indicates the first k Rule number 1 i The possible values ​​of the preconditions; Indicates the first i Individual health status assessment values; Indicates the first k Each confidence rule applies to the results. confidence level ; Indicates the first k The rule weight of each rule; Indicates a relationship.

[0042] Furthermore, for each confidence rule, a confidence value is obtained based on the initial confidence level of each health state and external indicators; the expression is:

[0043] in, It is based on confidence level and external indicators (such as report quality). A function for calculating credibility. It is aimed at The confidence level. In this application, the confidence level... Reflects the assessment results The level of trust.

[0044] Calculate the confidence score The BRB system introduces a confidence value. To quantify the level of trust in the assessment results. k Confidence value of each confidence rule The calculation is based on the following formula:

[0045] in: It is the first k The confidence level of each confidence rule; Is with the first i A confidence value related to a health status; These are categories of health status.

[0046] For each confidence rule Its output The confidence level will be adjusted in conjunction with external confidence data. The adjustment process is as follows:

[0047] In the formula, It is the confidence level after confidence adjustment.

[0048] Furthermore, the confidence scores adjusted for each state under all confidence rules are integrated, including:

[0049] In the formula, L Indicates the number of confidence rules. k This represents the index of the confidence rule sequence number. N Indicates the types of health status. i Represents a health status index. This indicates the adjusted confidence level.

[0050] In a preferred embodiment, based on the welding quality valueq Different health status assessment values The assessment results for the health status of welded joints can be categorized into several levels to determine their effectiveness, including: =

[0051] In a preferred embodiment, a projection covariance matrix adaptive evolution strategy algorithm is used to optimize the parameters of the credibility-enhanced BRB model. By combining covariance matrix adaptation and projection operations, constrained optimization problems can be handled efficiently.

[0052] In this embodiment, the optimization steps include: The objective function for constructing the credibility-enhancing BRB model is expressed as:

[0053] In the formula, It is a loss function used to measure the difference between the model's prediction and the actual health status; Indicates unlabeled samples ; These are the parameters of the confidence-enhanced BRB model, including rule weights and adjusted confidence levels. It is the number of unlabeled samples; This indicates that the credibility enhancement BRB model has improved parameters. The following sample The predicted output, j Indicates the sample number. This represents the target output generated based on pseudo-labels.

[0054] An optimization model is constructed to constrain the objective function. The optimization model is as follows:

[0055]

[0056] In the formula, The adjusted confidence level. Using the rule weights, the parameters in the confidence enhancement BRB model are updated through the optimization model to obtain a health status prediction model.

[0057] This application enables the credibility-enhanced BRB model to provide the optimal assessment of the health status of welded joints on unlabeled samples by minimizing the loss function.

[0058] Furthermore, in one specific embodiment, the effectiveness and accuracy of the optimization results are verified.

[0059] The steps include: 1. In the health status assessment of friction stir welded joints, reference values ​​for characteristic attributes are set based on expert knowledge. X, Y, and Z are selected as low, medium, and high reference points, abbreviated as L, M, and H. This set of input data is matched with the premise parts of all rules in the rule base (i.e., the reference values ​​defined by L, M, and H) to determine the rule weights.

[0060] 2. The health status of friction stir welded joints is classified into three levels based on expert knowledge: good (H1), normal (H2), and moderately deteriorated (H3).

[0061] The input data is divided into training and testing sets. An adaptive evolutionary strategy algorithm based on the projection covariance matrix is ​​used to optimize the parameters of the confidence-enhancing BRB model and update the confidence score. After the input dataset is input into the system, the welding quality values ​​calculated in the previous steps are applied. Further calculate the health status assessment value. The system can output different health status levels: =

[0062] The method of this invention makes full use of the fusion of expert knowledge and sensor data, providing a more comprehensive and accurate model for welding quality assessment. In particular, it can provide more accurate and reliable assessment results when dealing with uncertainties and fuzzy information.

[0063] To verify the overall effectiveness of the method of this invention, the model was validated using the collected data, and the results are as follows: Figures 2 to 5 As shown. To more intuitively illustrate the superiority of the models, the mean squared error (MSE) is mainly used to measure the performance of each model in the welding quality assessment task. This is achieved by comparing the ordinary BRB (Normal Belief Rule Base, N-BRB) model and the credibility-enhanced BRB (C-BRB) model with other common algorithms (such as Convolutional Neural Network (CNN), FCM, and Backpropagation Neural Network (BPNN)). Among them, by Figure 2 As can be seen, the MSE of the C-BRB model is significantly lower than that of the N-BRB model, indicating that after introducing the confidence level mechanism, C-BRB can more accurately assess the health status of the welded joint.

[0064] Figure 3The comparison shows that the C-BRB model significantly outperforms the other three models, especially when welding health conditions change significantly. The C-BRB model can stably adjust the evaluation results and effectively track changes in health conditions. In contrast, the evaluation results of the CNN, FCM, and BP models show greater fluctuations, particularly in areas with large changes in health conditions, exhibiting an unstable trend.

[0065] Figure 4 As can be seen, the health status assessment results of the C-BRB model are relatively stable and more consistent, especially when the health status of the welded joint changes significantly, accurately tracking these changes. In contrast, the assessment results of the FCM model show greater fluctuations, particularly when the health status changes drastically, demonstrating significant instability.

[0066] Figure 5 The results show that the C-BRB model performs the most stably in health status assessment, accurately tracking changes in health status, especially when significant changes occur in the weld joint, the C-BRB model can smoothly adjust the assessment results. Other models (N-BRB, CNN, FCM, BP) exhibit greater fluctuations, particularly when health status changes significantly, their assessment results are less stable.

[0067] Furthermore, the MSE comparison results are shown in the table below:

[0068] In this invention, the MSE value of the C-BRB model is significantly lower than that of other models. In particular, when the welding health status changes significantly, it can accurately track the changes in health status and performs superiorly in the health status assessment task.

[0069] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0070] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for assessing the health state of a welded joint based on a credibility enhancement BRB, characterized in that, The method comprises the following steps: The multi-source data of friction stir welding is acquired, preprocessed, and input into the credibility-enhanced BRB model to obtain the evaluation result of the health state of the welded joint; The credibility-enhanced BRB model comprises a plurality of confidence rules, for each confidence rule, a credibility value is obtained according to the initial credibility of each health state and external indexes, and the credibility of the corresponding health state is adjusted by using the credibility value; All adjusted credibility values of the health states under the plurality of confidence rules are fused, and the evaluation result of the health state of the welded joint is determined according to the fused credibility values.

2. The method of evaluating the health state of a welded joint according to claim 1, characterized by, The multi-source data of friction stir welding comprises X-direction current, Y-direction current, Z-direction current, voltage, and pressure.

3. The method of evaluating the health state of a welded joint according to claim 2, characterized by, The preprocessing comprises the following steps: The current data, voltage data, and pressure data in the X, Y, and Z directions are normalized respectively; The normalized current data, voltage data, and pressure data are weighted and fused to obtain a welding quality value.

4. The method of evaluating the health of a weld joint according to claim 1, wherein, Adjusting the credibility of the corresponding health state by using the credibility value comprises: The credibility values of the health states are fused by weighting based on the confidence degree of the current confidence rule to obtain a credibility value of the current confidence rule ; wherein is the confidence of the k th confidence rule, is the confidence value associated with the i th health state; is the health state category; Adjusting the credibility of each health state by using the credibility value of the current confidence rule.

5. The method of evaluating the health of a weld joint according to claim 1, wherein, Fusing all adjusted credibility values of the health states under the plurality of confidence rules comprises: wherein L represents the number of confidence rules, k represents the confidence rule sequence number index, N represents the health status category, i represents the health status index, represents the adjusted confidence.

6. The method of evaluating the health of a weld joint according to claim 1, wherein, The credibility-enhanced BRB model adopts a projection covariance matrix adaptive evolution strategy algorithm for parameter optimization, and the steps comprise: The objective function of the credibility-enhanced BRB model is constructed and expressed as: In the formula, is a loss function used to measure the difference between the model prediction result and the true health status; represents an unmarked sample ; is the parameter of the credibility-enhanced BRB model, including the rule weight, the adjusted confidence; is the number of unmarked samples; represents the prediction output of the credibility-enhanced BRB model under the parameters of the sample , j represents the sample serial number, represents the target output generated based on the pseudo label.

7. The method of evaluating the health state of a welded joint according to claim 6, characterized by, An optimization model is constructed to constrain the objective function, and the optimization model is: In the formula, is the adjusted confidence, is the rule weight, the parameters in the BRB model are updated by the optimization model to update the confidence, and a health state prediction model is obtained.

Images