A method and device for predicting spot welding nugget quality based on multi-factor mapping model

Through multi-factor mapping model combined with machine learning, the problem of relying on experience in determining traditional resistance spot welding process parameters is solved, and the welding quality and efficiency is improved, providing more accurate prediction of melting core quality and process parameter optimization is provided.

CN114693060BActive Publication Date: 2024-11-05HUAZHONG UNIV OF SCI & TECH +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210139462.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-02-15
Publication Date
2024-11-05
Estimated Expiration
2042-02-15

AI Technical Summary

Technical Problem

The determination of traditional resistance spot welding process parameters depends on experience, resulting in long definition of process parameters and wasted resources, and it is difficult to obtain multi-physical process information, affecting welding quality and efficiency.

Method used

Using the spot welding nucleus quality prediction method based on multi-factor mapping model, the process parameters and material parameters are combined with machine learning models through Z-score standardization, multi-process feature fusion, principal component analysis, feature selection and Bayesian optimization algorithm. Fusion, predict the mass of melting nucleus.

Benefits of technology

It improves the accuracy and efficiency of welding quality prediction, can design welding process and material combinations under quality constraints, and improves the quality and efficiency of welding processes.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114693060B_ABST
    Figure CN114693060B_ABST
Patent Text Reader

Abstract

The present invention provides a method and device for predicting the quality of spot welding nugget based on a multi-factor mapping model, the method comprising: using the Z-score method to standardize the original data set of the body-in-white to obtain pre-processed data; performing a multi-process feature fusion operation on the pre-processed data to obtain three types of data sets, and then using a feature selection method to uniformly express the three types of data sets, thereby obtaining a fused data set; cutting the fused data set into a training set and a validation set according to a retention method; putting the training set into iterative training of multiple machine learning models, and combining the Bayesian optimization algorithm to complete the parameter selection of the corresponding machine learning model, verifying on the validation set whether the trained corresponding machine learning model meets the selected index threshold, thereby determining the best machine learning model of multi-factor mapping, and then using the model to predict the quality index of the nugget. The present invention can effectively perform online feasibility analysis of body shaping in welding experiments, thereby reducing costs and improving efficiency.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of automobile body-in-white welding, and in particular to a method and device for predicting spot welding nugget quality based on a multi-factor mapping model. Background Art

[0002] The body-in-white (BIW) not only needs to provide passengers with a comfortable riding and driving space, but also needs to ensure the personal safety of passengers in emergencies, that is, when a collision, rollover or other road condition crisis occurs. Therefore, the body-in-white must give priority to impact collision in development and design. Among the four main processes of the body-in-white (stamping, welding, painting, and assembly), welding mainly involves heating and pressurizing materials such as castings, stamping plates, and profiles separately or simultaneously at points, lines, and surfaces to complete the complete body. Usually, a body-in-white contains about 6,000 welding points, so the strength and service life of the body-in-white depend to a large extent on the quality of the welding points, which is very important. Resistance spot welding is a welding method in which the workpiece to be welded is pressed between electrodes and an electric current is passed through it. The resistance heat generated by the current through the contact surface of the workpiece and the adjacent area is used to heat it to a molten or plastic state to form a metal bond. It includes four cycles of pre-pressing, energized welding, maintenance, and rest. Compared with other welding processes, resistance spot welding is the most widely used in automobile welding processes, and the body after spot welding has good strength.

[0003] In the traditional processing process, the process parameters and procedures of resistance welding are mainly based on the experience of the welding workers with some trial operations to define the welding industrial parameters, and then conduct post-welding destructive testing to test the determined process parameters and finally determine the corresponding process parameters. The whole process takes a long time, and because of the use of destructive experiments, resources are wasted seriously. In order to solve this problem, some manufacturers use the method of studying the welding mechanism and simulating to simulate the welding process, input different process parameters, simulate the welding process, and thus determine the best welding process parameters. At the same time, the welding quality is predicted based on the welding process parameters. However, in this method, it is difficult to obtain the required multi-physical process information. The welding process precision control of the automobile body in white is difficult, the process consistency requirements are high, the manufacturing cycle is short, and the factors affecting the welding quality and efficiency are complex, such as the complex physical and chemical changes of the materials. Therefore, it is necessary to explore new process research and development and quality control methods to meet its high-quality and efficient processing needs.

[0004] Data mining technology is a widely used data processing and analysis technology, which has played a considerable role in many fields. Industrial manufacturing and processing enterprises have a system for collecting and storing relevant data, which can reflect the degree of informatization of enterprises to varying degrees. With the rapid development of computing hardware, the storage and collection capabilities for data have been greatly enhanced, but the processing and analysis capabilities for data are relatively lacking. Traditional industrial database methods can efficiently add, delete, modify and query data throughout the manufacturing process, but it is difficult to discover the information contained therein, and idle information is difficult to effectively use. Summary of the invention

[0005] A main problem solved by the present invention is: the present invention collects process parameters and process information of the welding process, and proposes an improved method for predicting the quality of electric welding nuggets using a multi-factor mapping model of body-in-white welding quality commonly used in digital workshops.

[0006] The present invention provides a spot welding nugget quality prediction method based on a multi-factor mapping model, the spot welding nugget quality prediction method comprising:

[0007] The Z-score method is used to standardize the original data set of the body-in-white to obtain preprocessed data;

[0008] Perform a multi-process feature fusion operation on the pre-processed data to obtain a data set X containing only process parameters. p and a dataset X containing process parameters and material parameters D , and use principal component analysis to analyze the data set X D , obtain the data set X of process parameters and material parameters fusion D’ ;

[0009] Then use the feature selection method to unify the three types of data sets:

[0010] Select R 2 The determination coefficient or RMSE mean square error is used as the optimization evaluation index to determine the optimal feature combination of the three types of data sets, and weights are assigned to the optimal feature combination to obtain the fused data set;

[0011] The fused data set is divided into a training set and a validation set according to a hold-out method;

[0012] The training set is put into iterative training of multiple machine learning models, and the parameter selection of the corresponding machine learning model is completed in combination with the Bayesian optimization algorithm, and whether the trained corresponding machine learning model meets the selected indicator threshold is verified on the validation set, so as to determine the best machine learning model for multi-factor mapping;

[0013] The optimal machine learning model is used to predict the quality indicators of the weld nugget.

[0014] Preferably, the mathematical expression of the fused data set is:

[0015] R=X γ,γ =(r1, r2, ..., r n ) T

[0016] Where R is the fused data set, X is the feature set of the preprocessed data, γ is the weight matrix, r is a vector of all 0s or all 1s, n represents the number of features, and T represents transpose.

[0017] Preferably, after determining the best machine learning model, the method further includes the following steps:

[0018] Use swarm intelligence algorithms to reversely predict process parameter values.

[0019] Preferably, before the Z-score method is used to perform standardization processing on the raw data of the body-in-white, the method further includes the following steps:

[0020] According to the design requirements of different parts of the body-in-white, the margin thresholds T of different combinations of welding materials are determined, and the values ​​that do not meet the margin threshold T in the original data are filtered out to obtain the original data set.

[0021] Preferably, the original data set of the body-in-white includes process data, material mechanical properties and chemical element content.

[0022] According to another aspect of the present invention, a multi-factor mapping model prediction device for spot welding nugget quality is also disclosed, the spot welding nugget quality prediction device includes a memory, a processor;

[0023] The memory stores a spot welding nugget quality prediction program based on a multi-factor mapping model that can be run on the processor;

[0024] When the spot welding nugget quality prediction program based on a multi-factor mapping model is executed by the processor, the steps of the spot welding nugget quality prediction method based on a multi-factor mapping model as described in any of the preceding items are implemented.

[0025] Automobile production lines are transforming into efficient, intelligent and flexible production. In such production lines, ensuring product quality by designing and controlling the production process is the most basic and effective method. At the same time, the cost of online non-destructive testing during body-in-white welding is too high and the speed is too slow, which greatly affects the production rhythm. Therefore, using data mining technology to analyze the factors affecting welding quality, study data mining algorithms, understand welding mechanisms, establish accurate mapping models of welding quality factors, estimate welding quality, or design related processes and processes and material combinations under the constraints of welding quality requirements are important methods to improve the quality and efficiency of welding processes.

[0026] The advantages of the present invention mainly include two aspects: (1) taking into account both material factors and process factors, and extracting the optimal feature combination by fusing multiple algorithms. (2) proposing a Bayesian multi-machine learning comprehensive model, which can realize the welding quality prediction under certain material and process constraints, or the manufacturing parameter prediction under quality constraints, for the optimal feature combination.

[0027] In general, the present invention connects and links the process steps and material-related combined data of the body-in-white in the entire welding process, mines the information knowledge in these combined data, and uses these information to build a quality evaluation monitoring prediction model. It can effectively design welding process parameters and processes under the constraints of quality requirements, and can also predict welding quality under existing constraints, thereby improving the accuracy of welding quality prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The accompanying drawings, which constitute a part of the specification, illustrate embodiments of the invention and, together with the description, serve to explain the principles of the invention.

[0029] Figure 1 The present invention is a schematic flow chart of a method for predicting spot welding nugget quality based on a multi-factor mapping model in an embodiment of the present invention.

[0030] Figure 2 Schematic diagram of the feature fusion process in an embodiment of the present invention.

[0031] Figure 3 Schematic diagram of the hyperparameter optimization process in an embodiment of the present invention. DETAILED DESCRIPTION

[0032] Various exemplary embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings. It should be noted that unless otherwise specifically stated, the relative arrangement of components and steps, numerical expressions and numerical values ​​set forth in these embodiments do not limit the scope of the present invention.

[0033] At the same time, it should be understood that for the convenience of description, the sizes of the various parts shown in the drawings are not drawn according to the actual proportional relationship.

[0034] The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the invention, its application, or uses.

[0035] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments and with reference to the accompanying drawings.

[0036] Technologies, methods, and apparatus known to ordinary technicians in the relevant field may not be discussed in detail, but where appropriate, such technologies, methods, and apparatus should be considered part of the authorization specification.

[0037] In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limiting. Therefore, other examples of the exemplary embodiments may have different values.

[0038] It should be noted that like reference numerals and letters refer to similar items in the following figures, and therefore, once an item is defined in one figure, it need not be further discussed in subsequent figures.

[0039] Embodiment 1, as Figure 1 As shown, the present embodiment discloses a method for predicting the quality of spot welding nuggets based on a multi-factor mapping model, and the main steps include:

[0040] Step 1: Screening of raw data.

[0041] According to the design requirements of different parts, the margin threshold T of different combinations of welding materials is determined, and the values ​​that do not meet the margin threshold T in the original data of the body-in-white welding are filtered out to obtain the original data set.

[0042] At present, in actual production, a weld nugget larger than the designed weld nugget diameter is determined to meet the performance requirements of the body in white in all aspects. In order to ensure the qualified rate of the spot welding weld nugget, the production margin is usually increased, such as increasing the welding current, extending the welding time, etc.

[0043] Although too large residual nugget can meet the performance requirements, it also increases energy consumption and reduces efficiency. Too small residual nugget can meet the performance requirements and save energy, but it is difficult to control and obtain.

[0044] Therefore, this embodiment performs data mining by dynamically adjusting the margin threshold T before prediction, and selects appropriate data from the original data as basic data used in the spot welding nugget quality prediction method.

[0045] Step 2: Data standardization.

[0046] The Z-score method is used to standardize the original data set of the body in white to obtain preprocessed data. The process data, material mechanical properties and chemical element content in the original data set are standardized and preprocessed, and multiple indicators with different attributes, orders of magnitude and units are processed using a unified standardization method and formula to make them at the same order of magnitude, and pull these indicators to a unified baseline.

[0047] Specifically, the Z-score method is used for data standardization, and the specific formula is as follows:

[0048]

[0049] Among them, N i is the standardized result, x i is the characteristic of the dataset, is the mean value of feature x, and σ is the standard deviation of feature x.

[0050] Step 3: Data feature fusion.

[0051] Based on the feature fusion algorithm, the pre-processed data is subjected to a multi-process feature fusion operation, including common steps such as material feature dimensionality reduction and feature selection, combined with Figure 2 The process shown in the figure obtains a data set X containing only process parameters. p and a dataset X containing process parameters and material parameters D , and use principal component analysis to analyze the data set X D , obtain the data set X of process parameters and material parameters fusion D’ ;

[0052] The three types of data sets are then uniformly expressed using feature selection methods, such as the maximum mutual information coefficient, maximum correlation minimum redundancy coefficient, Boruta method, genetic algorithm, recursive feature elimination, etc. According to actual production, R is selected. 2 The determination coefficient or RMSE mean square error is used as an optimization evaluation indicator to determine the optimal feature combination of the three types of data sets, and weights are assigned to the optimal feature combination to obtain a fused data set.

[0053] The mathematical expression of the fused data set is:

[0054] R=X γ,γ =(r1, r2, ..., r n ) T

[0055] Where R is the fused data set, X is the feature set of the preprocessed data, γ is the weight matrix, r is a vector of all 0s or all 1s, n represents the number of features, and T represents transpose.

[0056] Step 4: Machine learning hyperparameter optimization.

[0057] like Figure 3 As shown, firstly, the data set is divided. The fused data set is cut into a training set and a test set (i.e., a validation set) according to the holdout method;

[0058] The training set is then put into multiple machine learning models for iterative training. In order to avoid overfitting, the Bayesian optimization algorithm is used to complete the parameter selection of the corresponding machine learning model, thereby reducing the variance and improving the performance of the model.

[0059] Verify on the test set whether the trained machine learning model meets the selected indicator threshold, for example, select R 2 The coefficient of determination is used to evaluate the degree of fit of the model to the data; RMSE and MAE are selected to evaluate the fitting error of the model; the average absolute relative deviation AARD is selected as the arithmetic mean of the absolute value of the relative error to judge the accuracy; PAP is selected as the percentage of accuracy-precision, representing reliability; the explainable error EV is selected to indicate the proportion of data that can be explained by the data model.

[0060] Taking into account the various indicators selected in the previous steps, the trained machine learning model is evaluated. If the evaluation criteria are met, the optimal machine learning model is determined. If the evaluation criteria are not met, new hyperparameters are searched again for the training step.

[0061] The mathematical expression for hyperparameter optimization is expressed as:

[0062]

[0063] x * represents the parameter to be optimized, X * represents the set of parameters to be optimized, and f(x) represents the objective function.

[0064] Step 5: Quality prediction and particle swarm reverse prediction.

[0065] The quality index of the weld nugget is predicted by using the determined optimal machine learning model. The selected optimal machine learning model can complete the prediction of the weld nugget quality index under the process constraints, or further use the swarm intelligence algorithm, such as the particle swarm algorithm, to reversely predict the process parameter values.

[0066] The original data in this embodiment are laboratory welding measurement data and historical database accumulation data. The retention method is used to divide the training set and the test set. Each piece of data is the process parameters and quality indicators of a weld nugget. Dynamic fusion of each feature is performed to extract the data set with the best performance of the machine learning model, and adjust the model hyperparameters to the optimal. Experiments have shown that the machine learning model that integrates material properties has improved the data fitting ability by 3%, which can effectively reduce the excess of welding materials. The thousands of welds on the body-in-white have numerous heterogeneous plate overlap combinations. In particular, new models require a large number of welding experiments for feasibility analysis of body shaping. In the mass production process, the accumulated historical data can be used with the help of the prediction method framework in the present invention to effectively perform online feasibility analysis, thereby reducing costs and improving efficiency.

[0067] The feature fusion processing part of the present invention includes unified expression of data preprocessing, feature dimensionality reduction, feature selection, etc. Through these steps, the data features such as welding process parameters, process parameters, material parameters, etc. are optimally selected, and finally the optimal feature data set is comprehensively and uniformly expressed, and the training convergence of the machine learning model is accelerated using a hyperparameter optimization algorithm.

[0068] The heat generated by the welder clamping the material directly affects the spot weld nugget formation process, which is subject to Joule's law:

[0069]

[0070] Wherein, I is welding current, ρ is resistivity, l is plate length, s is plate cross-sectional area, t is time, and R is resistance. Among them, current and resistance determined by the material have the greatest influence on the generation of heat. Therefore, adding the material properties of body-in-white welding to the method of the present invention can effectively improve the performance of the machine learning model.

[0071] According to the second aspect of the present application, the present invention also provides a multi-factor mapping model prediction device for spot welding nugget quality, especially a white body spot welding nugget quality prediction device used in a digital workshop, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the computer program is executed by the processor, the spot welding nugget quality prediction method based on the multi-factor mapping model in the embodiment of the first aspect is implemented.

[0072] The spot welding nugget quality prediction method and device based on the multi-factor mapping model can be run on computing devices such as desktop computers, notebooks, PDAs, and cloud servers. The spot welding nugget quality prediction method and device based on the multi-factor mapping model can include, but is not limited to, a processor and a memory.

[0073] Those skilled in the art can understand that the examples are merely examples of a spot welding nugget quality prediction device based on machine learning multi-factor mapping model, and do not constitute a limitation on a spot welding nugget quality prediction device based on a multi-factor mapping model, and may include more or less components than the examples, or combine certain components, or different components, for example, may also include input and output devices, network access devices, buses, etc. The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may be any conventional processor, etc. The processor is the control center of the spot welding nugget quality prediction method device based on the multi-factor mapping model, and uses various interfaces and lines to connect the various parts of the entire spot welding nugget quality prediction method device based on the multi-factor mapping model. The memory can be used to store the computer program and / or module, and the processor realizes various functions of the spot welding nugget quality prediction method device based on the multi-factor mapping model by running or executing the computer program and / or module stored in the memory, and calling the data stored in the memory. In addition, the memory may include a high-speed random access memory, and may also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Sma rt-Media-Card, SMC), a secure digital (Secure-Digital, SD) card, a flash card (Flash-Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0074] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

[0075] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

Claims

1. A method for predicting spot welding nugget quality based on a multi-factor mapping model, characterized in that: The spot welding nugget quality prediction method comprises: The Z-score method is used to standardize the original data set of the body-in-white to obtain preprocessed data; Perform a multi-process feature fusion operation on the pre-processed data to obtain a data set X containing only process parameters. p With the data set X containing process parameters and material parameters D , and use principal component analysis to analyze the data set X D , obtain the data set X of process parameters and material parameters fusion D’ ; The feature selection method is used to uniformly express the three types of data sets: Select R 2 The determination coefficient or RMSE mean square error is used as the optimization evaluation index to determine the optimal feature combination of the three types of data sets, and weights are assigned to the optimal feature combination to obtain the fused data set; The fused data set is divided into a training set and a validation set according to a hold-out method; The training set is put into iterative training of multiple machine learning models, and the parameter selection of the corresponding machine learning model is completed in combination with the Bayesian optimization algorithm, and whether the trained corresponding machine learning model meets the selected indicator threshold is verified on the validation set, so as to determine the best machine learning model for multi-factor mapping; The optimal machine learning model is used to predict the quality indicators of the weld nugget.

2. A method for predicting spot welding nugget quality based on a multi-factor mapping model as claimed in claim 1, characterized in that: The mathematical expression of the fused data set is: R=Xγ,γ=(r1,r2,...,r n ) T Where R is the fused data set, X is the feature set of the preprocessed data, γ is the weight matrix, r is a vector of all 0s or all 1s, n represents the number of features, and T represents transpose.

3. A method for predicting spot welding nugget quality based on a multi-factor mapping model as claimed in claim 1, characterized in that: Once the best machine learning model is determined, the following steps are also included: Use swarm intelligence algorithms to reversely predict process parameter values.

4. A method for predicting spot welding nugget quality based on a multi-factor mapping model as claimed in claim 1, characterized in that: Before the Z-score method is used to standardize the raw data of the body-in-white, the following steps are also included: According to the design requirements of different parts of the body-in-white, the margin thresholds T of different combinations of welding materials are determined, and the values ​​that do not meet the margin threshold T in the original data are filtered out to obtain the original data set.

5. The method for predicting spot welding nugget quality based on a multi-factor mapping model as claimed in claim 1, characterized in that: The original data set of the body in white includes process data, material mechanical properties and chemical element content.

6. A spot welding nugget quality prediction device based on a multi-factor mapping model, characterized in that: The spot welding nugget quality prediction device includes a memory and a processor; The memory stores a spot welding nugget quality prediction program based on a multi-factor mapping model that can be run on the processor; When the spot welding nugget quality prediction program based on a multi-factor mapping model is executed by the processor, the steps of the spot welding nugget quality prediction method based on a multi-factor mapping model as claimed in any one of claims 1 to 5 are implemented.