Method for constructing laser stitch welding interface fusion width prediction model, laser stitch welding interface fusion width prediction method and system and electronic equipment

Abstract

The invention provides a method for constructing a fusion width prediction model of a laser stitch welding interface. The method comprises the following steps: collecting surface fusion width data and fusion depth data of a plurality of welding seams of laser stitch welding of a first workpiece and a second workpiece which are overlapped with each other, and interface fusion width data at an interface position between the first workpiece and the second workpiece; and constructing a data set by taking the collected surface fusion width data and fusion depth data as input data and taking the interface fusion width data as output data, and training the interface fusion width prediction model by using the data set. According to some embodiments of the invention, a laser stitch welding interface fusion width prediction model can be constructed, so that convenient and non-destructive interface fusion width prediction can be realized by using the prediction model in production, and the evaluation of the laser stitch welding quality is facilitated. The invention further provides a corresponding laser stitch welding interface fusion width prediction method and system, electronic equipment and a computer program product.

Description

Technical Field

[0001] The present invention relates to the field of laser processing and detection technology, and in particular to a method for constructing a laser stitch welding interface melt width prediction model, a laser stitch welding interface melt width prediction method, a laser stitch welding interface melt width prediction system, and electronic equipment and computer program products for laser stitch welding interface melt width prediction. Background Art

[0002] In laser welding, the characteristics of the weld, especially the dimensional characteristics of a certain part, can often reflect the quality of the weld. In particular, in laser stitch welding for batteries, the weld has electrical conductivity requirements. Accordingly, the weld width at the interface between the two workpieces being stitched is an important indicator for evaluating the electrical conductivity of the weld, that is, the quality of this type of weld. However, if destructive measures are not applied to the welded objects, current existing technologies are unable to obtain data on the weld width at the interface. Therefore, there is a need for a technology to non-destructively obtain weld width data for laser stitch welding. Summary of the Invention

[0003] The purpose of the present invention is to provide a method for constructing a laser stitch welding interface melt width prediction model, a laser stitch welding interface melt width prediction method, a laser stitch welding interface melt width prediction system, an electronic device for laser stitch welding interface melt width prediction and a computer program product, so as to at least partially solve the problems existing in the prior art.

[0004] According to a first aspect of the present invention, a method for constructing a weld width prediction model for a laser stitch welding interface is provided, the method comprising: Collecting surface melt width data, melt depth data, and intersection melt width data of a plurality of welds of a first workpiece and a second workpiece stacked on each other by laser welding under predetermined beam characteristic parameters, and intersection melt width data at an intersection position between the first workpiece and the second workpiece; and The collected surface weld width data and weld depth data are used as input data, and the interface weld width data are used as output data to construct a data set, and the interface weld width prediction model is trained using the data set.

[0005] According to the method for constructing a laser stitch welding interface melt width prediction model according to the first aspect of the present invention, a laser stitch welding interface melt width prediction model can be constructed, so that during or after welding, only two measurable data, the surface melt width and the melt depth of the weld, can be simply obtained, and the prediction model can be used to realize non-destructive prediction of the interface melt width, especially real-time prediction of the interface melt width, thereby facilitating the evaluation of the quality of laser stitch welding, especially in real time, without considering various relevant operating parameters of the laser stitch welding operation or the characteristics or parameters of the workpiece.

[0006] According to an optional embodiment of the present invention, the dataset is divided into a training dataset and a validation dataset. The training dataset is used to train the interface width prediction model and adjust its parameters. The trained interface width prediction model is then validated using the validation dataset until the deviation between the interface width data output by the trained interface width prediction model and the corresponding interface width data in the validation dataset is less than a preset threshold, thereby obtaining a trained interface width prediction model. This embodiment enables the interface width prediction model to effectively converge, enabling it to effectively predict the interface width.

[0007] According to an optional embodiment of the present invention, surface weld width data, weld depth data, and weld width data for multiple welds of the laser stitch welding process under different process control parameters are collected. This embodiment enables data on welds of different sizes to be obtained by varying laser power, welding speed, and other parameters while maintaining the same laser beam characteristics. This enriches the diversity of the collected data, enabling a better dataset to be constructed for training the weld width prediction model and preventing model overfitting.

[0008] According to an optional embodiment of the present invention, surface weld width data is obtained by photographing and measuring the weld using a camera. This embodiment allows for convenient and non-destructive acquisition of surface weld width data. In particular, by photographing and measuring with a camera coaxial with a machined lens, surface weld width data can be acquired in real time.

[0009] According to an optional embodiment of the present invention, the penetration data is acquired by means of an OCT (Optical Coherence Tomography) sensor. With this embodiment, the penetration data of the weld can be conveniently acquired in a non-destructive manner.

[0010] According to an optional embodiment of the present invention, the interface weld width data is obtained by performing destructive processing on the weld to expose the interface position between the first workpiece and the second workpiece and measuring the weld width at the interface position.

[0011] According to an optional embodiment of the present invention, the interface weld width prediction model is a linear regression model or a neural network model.

[0012] According to a second aspect of the present invention, a method for predicting the weld width of a laser stitch welding interface is provided. The method is performed using a weld width prediction model for a laser stitch welding interface constructed by the method for constructing a weld width prediction model for a laser stitch welding interface according to the first aspect of the present invention. The method comprises: Acquiring surface melt width data and melt depth data of a weld seam of laser stitch welding of the first workpiece and the second workpiece stacked on each other under the predetermined beam characteristic parameters; and The interface weld width prediction model is used to predict interface weld width data at an interface position between the first workpiece and the second workpiece based on surface weld width data and weld depth data of the weld.

[0013] According to the laser stitch welding interface melt width prediction method of the second aspect of the present invention, when it is desired to obtain the interface melt width of the welding weld, the surface melt width and melt depth data of the weld can be obtained only through relevant sensor equipment, and then the constructed interface melt width prediction model is used to predict the interface melt width data based on these two measurable data, without having to use destructive means to destroy the material, and without having to consider various relevant operating parameters of the laser stitch welding operation or the characteristics or parameters of the workpiece.

[0014] According to a third aspect of the present invention, a laser stitch welding interface weld width prediction system is provided, comprising: a data acquisition module configured to acquire surface weld width data and weld depth data of a weld seam of laser stitch welding of a first workpiece and a second workpiece stacked on each other; and A prediction module is configured to call an interface melt width prediction model to predict interface melt width data at the interface position between the first workpiece and the second workpiece based on the surface melt width data and the melt depth data of the weld, wherein the interface melt width prediction model is constructed according to the method for constructing a laser stitch welding interface melt width prediction model according to the first aspect of the present invention and for the laser stitch welding of the first workpiece and the second workpiece.

[0015] By means of the laser stitch welding interface width prediction system according to the third aspect of the present invention, the interface width data of the weld seam can be predicted in a non-destructive manner.

[0016] According to a fourth aspect of the present invention, there is provided an electronic device for predicting the weld width of a laser stitch welding interface, comprising: at least one processor; and A memory, wherein program instructions that can be executed by the at least one processor are stored in the memory, and when the program instructions are executed by the at least one processor, the method according to the first aspect or the second aspect of the present invention is implemented.

[0017] According to a fifth aspect of the present invention, a computer program product, such as a computer-readable program carrier, is provided, which contains or stores computer program instructions, and when the computer program instructions are executed by a processor, at least assists in implementing the steps of the method according to the first aspect or the second aspect of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The present invention will be described in more detail below with reference to the accompanying drawings, so that the principles, features and advantages of the present invention can be better understood. The accompanying drawings include: Figure 1 A flowchart schematically illustrates the main steps of a method for constructing a weld width prediction model for a laser stitch welding interface according to an exemplary embodiment of the present invention; Figure 2 Schematically shows a weld formed by laser stitch welding a first workpiece and a second workpiece stacked on each other; Figure 3 A flowchart schematically illustrates the main steps of a method for predicting weld width at a laser stitch welding interface according to an exemplary embodiment of the present invention; Figure 4 A schematic block diagram of a laser stitch welding interface weld width prediction system according to an exemplary embodiment of the present invention is shown; Figure 5 A schematic block diagram of an electronic device for predicting weld width at a laser stitch welding interface according to an exemplary embodiment of the present invention is shown. DETAILED DESCRIPTION

[0019] In order to make the technical problems, technical solutions and beneficial technical effects to be solved by the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and multiple exemplary embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the scope of protection of the present invention.

[0020] Figure 1 A flowchart schematically illustrates the main steps of a method for constructing a weld width prediction model for a laser stitch welding interface according to an exemplary embodiment of the present invention. For ease of explanation, Figure 2 The figure schematically shows the weld formed by laser stitch welding of a first workpiece W1 and a second workpiece W2 stacked on each other.

[0021] like Figure 1 As shown, the method for constructing a prediction model for weld width at the laser stitch welding interface according to an exemplary embodiment of the present invention mainly includes the following steps: S101: collecting surface weld width data Ws, weld depth data D, and interface weld width data Wi at an interface position P between the first workpiece W1 and the second workpiece W2 of a plurality of welds of a laser stitch welded together under predetermined beam characteristic parameters; and S102: Using the collected surface weld width data Ws and weld depth data D as input data and the interface weld width data Wi as output data to construct a data set, and using the data set to train an interface weld width prediction model.

[0022] In step S101, the beam characteristic parameters involve parameters related to the spatial characteristics and energy distribution of the laser beam, such as beam quality, spot characteristics, beam mode, focusing parameters, and polarization state. These parameters determine how the laser beam interacts with the material and, therefore, the weld formation mechanism. Using predetermined beam characteristic parameters for laser stitch welding can produce welds of the same type, ensuring that the mathematical relationship between the surface weld width data Ws, the weld depth data D, and the weld width data Wi remains the same, regardless of the specific size of the weld, which is determined by process control parameters (such as laser power, dwell time, and welding speed).

[0023] On the other hand, preferably, multiple laser stitch welding operations are performed under different process control parameters to obtain surface weld width data, weld depth data, and interface weld width data for multiple welds for each laser stitch welding operation. This allows data on welds of different sizes but with the same mathematical relationship of the relevant data to be obtained by changing the laser power, welding speed, etc. under the same beam characteristic parameters, thereby enriching the diversity of the collected data and enabling a better data set to be constructed in subsequent steps to train the interface weld width prediction model and avoid model overfitting. For example, for a first workpiece W1 and a second workpiece W2 made of aluminum alloy with a thickness of 3 mm, laser stitch welding is performed on the two workpieces using laser powers of 2000 W, 2500 W, 3000 W, 3500 W, and 4000 W, respectively, under predetermined beam characteristic parameters. Multiple welds of different sizes can be obtained, and the weld width data Ws, weld depth data D, and interface weld width data Wi of these welds are collected to construct a data set for training the model. Additionally or alternatively, multiple welds of different sizes can be obtained and data collected by changing the welding speed. In addition, those skilled in the art will appreciate that in these laser stitch welding processes, auxiliary parameters or variables such as shielding gas type / flow rate, defocus amount, workpiece clamping method, etc. are also the same.

[0024] The surface melt width data Ws of the corresponding weld can be directly obtained by taking photos and measuring the weld using a camera. In particular, the surface melt width data Ws of the weld can be measured in real time by taking real-time photos of the weld using a coaxial camera with a processed lens during the laser stack welding process. In addition, the penetration data D of each weld can be directly measured with the aid of an OCT sensor. Since the interface position P between the first workpiece W1 and the second workpiece W2 is located inside the welded workpiece, the interface position P can be exposed by destructively processing the weld, and then the interface melt width data Wi at the interface position P can be measured. Obviously, the surface melt width data Ws and penetration data D of the weld can also be measured after the weld is destructively processed to expose the interface position P. Various well-known and appropriate methods can be used for measurement.

[0025] After the data is collected, a data set can be constructed in step S102 to train the interface weld width prediction model. The collected surface weld width data and weld depth data are used as input data, and the corresponding interface weld width data is used as output data. The input data and output data are used as a set of data, and the multiple sets of data are divided into a training data set and a validation data set. For example, 80% of the data can be randomly selected from the multiple sets of data as the training data set, and the remaining 20% ​​as the validation data set. The training data set is then input into the interface weld width prediction model to train the interface weld width prediction model and adjust the parameters of the interface weld width prediction model. After training with the training data set, the trained interface weld width prediction model is validated using the validation data set. Specifically, the input data in the validation data set is input into the interface weld width prediction model to obtain the output data output by the model, and then the deviation between the output data output by the model and the corresponding output data in the validation data set is calculated. If the deviations are all less than or equal to a preset threshold, it can be considered that the trained interface weld width prediction model has converged, and the training of the interface weld width prediction model is completed. If there is a deviation greater than the preset threshold, the number of training rounds can be increased, that is, the training data set is input into the interface weld width prediction model again to train the model, and then the trained model is verified again using the verification data set until the deviation between the interface weld width data output by the trained interface weld width prediction model and the corresponding interface weld width data in the verification data set is less than the preset threshold, thereby obtaining a trained interface weld width prediction model.

[0026] The above data partitioning and training methods are merely exemplary. For example, if the model fails to converge after one round of training, the validation dataset can be partially merged back into the training dataset, and a new validation dataset can be repartitioned. A new round of training and validation can then be performed on the model until the trained model converges. Other well-known methods are also feasible.

[0027] Note that after step S101, a data pre-cleaning step may also be included, that is, by checking the data and identifying obvious deviation data or erroneous data based on experience or calculation, deleting them or re-measuring and collecting data for the corresponding welds to ensure the reliability of the data set constructed in the subsequent step S102, thereby improving the training efficiency and ensuring the reliability of the trained interface weld width prediction model.

[0028] In a preferred embodiment, the interface weld width prediction model is a linear regression model. The linear regression model has low algorithm complexity, fast training speed, and is suitable for large-scale data. The strong constraints of the linear assumption reduce the risk of overfitting, especially when the feature dimension is much smaller than the sample size. For the predetermined first workpiece and second workpiece, the interface weld width of the lap weld is approximately linearly related to the surface weld width and the weld depth. Therefore, selecting an appropriate linear regression model can meet the requirements for interface weld width prediction. Of course, a neural network model, such as a shallow neural network model, can also be used as the interface weld width prediction model.

[0029] The above describes the process of constructing a model for predicting the weld width of the laser stitch welding interface. The following describes a method for predicting the weld width of the laser stitch welding interface using the model.

[0030] Figure 3 The flowchart schematically shows the main steps of the method for predicting the weld width of the laser stitch welding interface according to an exemplary embodiment of the present invention. Figure 3 As shown, a method for predicting the weld width of a laser stitch welding interface according to an exemplary embodiment of the present invention mainly includes the following steps: S301: Acquiring surface weld width data Ws and weld depth data D of a laser stitch weld of a first workpiece W1 and a second workpiece W2 stacked on each other; and S302: Using the interface weld width prediction model constructed according to the model construction method described above, predict the interface weld width data Wi at the interface position P between the first workpiece W1 and the second workpiece W2 based on the surface weld width data Ws and the weld depth data D.

[0031] Similarly, in step S301, the surface weld width data Ws of the corresponding weld can be directly obtained by photographing and measuring the weld using a camera. In particular, during the laser stitch welding process, the surface weld width data Ws can be obtained in real time by photographing and measuring the weld using a coaxial camera with a processed lens. Furthermore, the penetration depth data D of each weld can be directly measured using an OCT sensor. In this way, the two input data can be obtained non-destructively, and in particular, in real time.

[0032] In step S302, since the interface width prediction model has been trained, after inputting the surface width data Ws and the penetration data D, the interface width prediction model can predict and output the interface width data Wi of the corresponding weld. This allows the weld interface width data Wi to be obtained without damaging the weld object. In particular, the weld interface width data Wi can be obtained in real time, thereby better assisting the production process.

[0033] Figure 4FIG. 1 is a schematic block diagram of a laser stitch welding interface weld width prediction system 1 according to an exemplary embodiment of the present invention. Figure 4 As shown, a laser stitch welding interface weld width prediction system 1 according to an exemplary embodiment of the present invention includes: A data acquisition module 11 configured to acquire surface weld width data and weld depth data of a weld seam of laser stack welding of a first workpiece and a second workpiece stacked on each other; and The prediction module 12 is configured to: call the interface melt width prediction model constructed for the laser stitch welding of the first workpiece and the second workpiece by the above-mentioned method of constructing the laser stitch welding interface melt width prediction model, and predict the interface melt width data at the interface position between the first workpiece and the second workpiece based on the surface melt width data and the melt depth data of the weld.

[0034] Figure 5 FIG. 2 shows a schematic block diagram of an electronic device 2 for predicting weld width at the laser stitch welding interface according to an exemplary embodiment of the present invention. Figure 5 As shown, an electronic device 2 for predicting the weld width of a laser stitch welding interface according to an exemplary embodiment of the present invention includes: at least one processor 21; and A memory 22, in which program instructions that can be executed by the at least one processor 21 are stored, and when the program instructions are executed by the at least one processor 21, a method for constructing a laser stitch welding interface melt width prediction model according to an exemplary embodiment of the present invention or a laser stitch welding interface melt width prediction method according to an exemplary embodiment of the present invention is implemented.

[0035] Although specific embodiments of the present invention are described in detail herein, they are provided for illustrative purposes only and should not be considered to limit the scope of the present invention. Various substitutions, changes, and modifications may be conceived without departing from the spirit and scope of the present invention.

Claims

1. A method for constructing a prediction model for weld width at a laser stitch welding interface, comprising: Collecting surface weld width data and weld depth data of a plurality of welds of a first workpiece and a second workpiece stacked on each other by laser welding under predetermined beam characteristic parameters, and weld width data of an intersection surface at an intersection surface position between the first workpiece and the second workpiece; and The collected surface weld width data and weld depth data are used as input data, and the interface weld width data are used as output data to construct a data set, and the interface weld width prediction model is trained using the data set.

2. The method for constructing a prediction model for weld width at the laser stitch welding interface according to claim 1, wherein: The data set is divided into a training data set and a validation data set. The training data set is used to train the interface melt width prediction model and adjust the parameters of the interface melt width prediction model. The trained interface melt width prediction model is verified using the validation data set until the deviation between the interface melt width data output by the trained interface melt width prediction model and the corresponding interface melt width data in the validation data set is less than a preset threshold, thereby obtaining a trained interface melt width prediction model.

3. The method for constructing a prediction model for weld width at the laser stitch welding interface according to claim 1 or 2, wherein: The surface melt width data, the melt depth data and the interface melt width data of the plurality of welds of the laser stitch welding under different process control parameters are collected.

4. The method for constructing a prediction model for weld width at the laser stitch welding interface according to claim 1 or 2, wherein: Obtain surface weld width data by taking photographs of the weld using a camera; and / or The penetration depth data is acquired with the help of OCT sensors.

5. The method for constructing a prediction model for weld width at the laser stitch welding interface according to claim 1 or 2, wherein: The interface weld width data is obtained by performing destructive processing on the weld to expose the interface position between the first workpiece and the second workpiece and measuring the weld width at the interface position.

6. The method for constructing a prediction model for weld width at the laser stitch welding interface according to claim 1 or 2, wherein: The interface weld width prediction model is a linear regression model or a neural network model.

7. A method for predicting the weld width of a laser stitch welding interface, the method being performed using a weld width prediction model constructed by the method for constructing a weld width prediction model for a laser stitch welding interface according to any one of claims 1 to 6, the method comprising: Acquiring surface melt width data and melt depth data of a weld seam of the first workpiece and the second workpiece stacked on each other by laser welding under the predetermined beam characteristic parameters; and The interface weld width prediction model is used to predict interface weld width data at an interface position between the first workpiece and the second workpiece based on surface weld width data and weld depth data of the weld.

8. A laser stitch welding interface weld width prediction system (1), comprising: A data acquisition module (11) configured to acquire surface weld width data and weld depth data of a laser stack weld of a first workpiece and a second workpiece stacked on each other; and A prediction module (12) is configured to call an interface melt width prediction model to predict interface melt width data at an interface position between the first workpiece and the second workpiece based on surface melt width data and melt depth data of the weld, wherein the interface melt width prediction model is constructed by the method for constructing a laser stitch welding interface melt width prediction model according to any one of claims 1 to 6 and for the laser stitch welding of the first workpiece and the second workpiece.

9. An electronic device (2) for predicting the weld width of a laser stitch welding interface, comprising: at least one processor (21); and A memory (22), wherein program instructions that can be executed by the at least one processor (21) are stored in the memory (22), and when the program instructions are executed by the at least one processor (21), the method according to any one of claims 1 to 7 is implemented. 10 . A computer program product, such as a computer-readable program carrier, comprising or storing computer program instructions, which, when executed by a processor, at least assist in implementing the steps of the method according to claim 1 .

Images