Laser welding process parameter prediction method, equipment and medium

Abstract

The invention provides a laser welding process parameter prediction method which comprises the following steps: collecting laser welding experimental data including process parameter combinations and corresponding welding quality indexes; constructing a forward prediction model for predicting welding quality indexes according to the process parameters; constructing a reverse prediction model for predicting process parameters according to the target welding quality index; the process parameters output by the reverse prediction model are input into the forward prediction model, and predicted welding quality is generated; the error between the predicted welding quality and the target welding quality is obtained, the error serves as a loss value to optimize the reverse prediction model, a closed-loop feedback mechanism is formed, and an optimal reverse prediction model is obtained; and predicting the laser welding process parameters based on the optimal reverse prediction model. According to the method, after the reverse process parameters are predicted, the reverse process parameters are input into the trained forward prediction model to calculate the actual expected welding quality, and then the actual expected welding quality is compared with the target welding quality, so that the technical blind area that the reverse result cannot be verified is broken through.

Description

Technical Field

[0001] This application relates to the field of laser welding process optimization technology, and in particular to a method, equipment and medium for predicting laser welding process parameters. Background Technology

[0002] Laser welding, as a high-energy-density, non-contact processing method, has been widely used in high-end manufacturing fields such as automobile manufacturing, shipbuilding, rail transportation, and construction machinery. Especially with the increasing demand for welding thick plate structural steel, ring-spot lasers, with their unique energy distribution advantages, can achieve more stable molten pool control and more uniform energy input by adjusting the distribution of laser power between the outer ring and the center, making them an important process in deep penetration welding of medium and thick plates.

[0003] To improve the automation and intelligence of welding processes, researchers have focused on utilizing deep learning and machine learning methods to establish interactive mapping models between process parameters and welding quality, thereby enabling welding quality prediction and process parameter optimization recommendations. Currently, the main techniques for welding process modeling and optimization use process parameters as input and welding quality as output, achieving positive prediction modeling of welding quality. However, research on reverse prediction of process parameters based on welding quality indicators is relatively limited. This research primarily includes constructing machine learning or deep learning regression models such as support vector machines and then backfitting process parameters based on target quality indicators to achieve parameter optimization recommendations. It also explores methods such as Bayesian optimization and genetic algorithms to attempt to optimize and predict process parameters by finding the best among process parameters through optimization strategies.

[0004] While these methods have achieved a certain degree of modeling the mapping between laser welding quality and parameters, backfit-based research schemes suffer from "one-to-many" or "non-unique solution" problems, making it difficult for the model to effectively distinguish and determine the optimal combination, leading to unstable prediction results. Similarly, optimization-based algorithms also face challenges such as low parameter search efficiency and high search difficulty. Therefore, a new method is urgently needed to achieve accurate prediction of process parameters based on target welding quality indicators. Summary of the Invention

[0005] This application aims to solve at least one of the technical problems existing in the prior art mentioned above, and proposes a method, equipment and medium for predicting laser welding process parameters. It establishes an effective reverse mapping mechanism between welding quality indicators and process parameters through a closed-loop prediction structure, and dynamically corrects the model by combining prediction error feedback, so as to realize the verification and adjustment of the reverse prediction results. This overcomes the problems of uncontrollable and difficult-to-verify prediction models in the prior art, and provides reliable technical support for laser welding process optimization and intelligent control.

[0006] In a first aspect, embodiments of this application provide a method for predicting laser welding process parameters, including:

[0007] Collect laser welding experimental data, including process parameter combinations and corresponding welding quality indicators;

[0008] A forward prediction model is constructed to predict welding quality indicators based on process parameters; a reverse prediction model is constructed to predict process parameters based on target welding quality indicators.

[0009] The process parameters output by the reverse prediction model are input into the forward prediction model to generate the predicted welding quality.

[0010] The error between the predicted welding quality and the target welding quality is obtained. This error is used as the loss value to optimize the reverse prediction model, forming a closed-loop feedback mechanism, and the optimal reverse prediction model is obtained.

[0011] Prediction of laser welding process parameters based on the optimal reverse prediction model.

[0012] Furthermore, after collecting laser welding experimental data, the following is also included:

[0013] The process parameter combinations and corresponding welding quality indicators are structured and divided into training and testing sets according to a preset ratio.

[0014] Furthermore, after collecting laser welding experimental data, the following is also included:

[0015] The Z-score standardization method is used to unify the dimensions of the process parameter combinations and their corresponding welding quality indicators.

[0016] Furthermore, both the forward prediction model and the backward prediction model are constructed based on a multilayer perceptron, wherein:

[0017] The input layer of the reverse prediction model includes two nodes, which respectively input the weld penetration depth and weld width, two welding quality indicators; the output layer of the reverse prediction model is a single node, which outputs the predicted values ​​of three process parameters: outer ring laser power, center laser power, and welding speed.

[0018] The input layer of the forward prediction model includes three nodes, which respectively input three process parameters output by the reverse prediction model: outer ring laser power, center laser power, and welding speed; the output layer of the forward prediction model is a single node, which outputs predicted values ​​of two welding quality indicators: weld penetration and weld width.

[0019] Furthermore, both the forward prediction model and the reverse prediction model include a hidden layer based on a modified linear unit function.

[0020] Furthermore, the error between the predicted welding quality and the target welding quality is obtained, and this error is used as the loss value to optimize the inverse prediction model, forming a closed-loop feedback mechanism, including:

[0021] The mean square error between the predicted welding quality and the target welding quality is obtained, and the model weight parameters are adjusted through the backpropagation algorithm to minimize the loss function.

[0022] Furthermore, the error between the predicted welding quality and the target welding quality is obtained, and this error is used as the loss value to optimize the inverse prediction model, forming a closed-loop feedback mechanism, including:

[0023] The weights of the forward prediction model are frozen, and the backward prediction model is optimized by backpropagation using only the mean square error between the predicted welding quality and the target welding quality as the loss value.

[0024] Furthermore, the error between the predicted welding quality and the target welding quality is obtained, and this error is used as the loss value to optimize the inverse prediction model, forming a closed-loop feedback mechanism, including:

[0025] The steps “inputting the process parameters output by the inverse prediction model into the forward prediction model to generate the predicted welding quality” and “obtaining the error between the predicted welding quality and the target welding quality, and using this error as the loss value to optimize the inverse prediction model” are executed alternately and iteratively on the training set until the closed-loop error converges.

[0026] Secondly, embodiments of this application provide an electronic device, including: one or more processors;

[0027] A memory for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors are able to implement the steps in any of the preceding prediction methods.

[0028] Thirdly, embodiments of this application provide a computer-readable medium storing a computer program that, when executed by a processor, can implement the steps in any of the aforementioned prediction methods.

[0029] This application provides a laser welding process parameter prediction method. After reverse process parameter prediction, the reverse prediction parameters are input into a trained forward prediction model to calculate the actual expected welding quality, which is then compared with the target welding quality, thus breaking the technical blind spot of "reverse results cannot be verified". This application constructs a novel learning mechanism with closed-loop error as training loss, and automatically adjusts parameters through error feedback, reducing the prediction drift risk caused by the uncertainty of one-to-many mapping, and realizing parameter inference self-calibration. Attached Figure Description

[0030] Figure 1A core flowchart of a laser welding process parameter prediction method provided in this application embodiment;

[0031] Figure 2 A more detailed flowchart of a laser welding process parameter prediction method provided for embodiments of this application;

[0032] Figure 3 This is a structural block diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0033] To enable those skilled in the art to better understand the technical solutions of this application, exemplary embodiments of this application are described below with reference to the accompanying drawings, including various details of the embodiments of this application to aid understanding. These should be considered merely exemplary. Therefore, those skilled in the art should recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this application. Similarly, for clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description. Unless otherwise specified, the various embodiments of this application and the features within those embodiments can be combined with each other.

[0034] As used herein, the term “and / or” includes any and all combinations of one or more of the associated enumerated entries. The terminology used herein is for describing particular embodiments only and is not intended to limit the application. As used herein, the singular forms “a” and “the” are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that when the terms “comprising” and / or “made of” are used in this specification, the presence of the stated features, integrals, steps, operations, elements, and / or components is specified, but the presence or addition of one or more other features, integrals, steps, operations, elements, components, and / or groups thereof is not excluded. Terms such as “connected” or “linked” are not limited to physical or mechanical connections but can include electrical connections, whether direct or indirect.

[0035] Unless otherwise specified, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art. It should also be understood that terms such as those defined in commonly used dictionaries should be interpreted as having the meaning consistent with their meaning in the context of the relevant art and this application, and will not be interpreted as having an idealized or overly formal meaning, unless expressly so defined herein.

[0036] In back-projection of welding process parameters by mapping weld penetration and weld width, existing methods often directly output parameter combinations, lacking effective constraints and evaluation methods for the back-projected results. This leads to multiple parameter solutions for the same quality objective, causing uncertainty in the prediction results and making it difficult to determine their practical feasibility. Since there is a typical one-to-many mapping relationship between welding quality indicators and process parameters, existing back-projection models cannot dynamically adjust based on model output errors and lack adaptive convergence capabilities. This makes it difficult for the model to automatically correct errors in practical applications, resulting in insufficient reliability and stability of the prediction results. Existing methods for back-projecting process parameters based on optimization algorithms (such as genetic algorithms and particle swarm optimization) typically require frequent calls to the forward fitting model to estimate welding quality, resulting in high computational costs and inefficient search processes due to the high dimensionality and wide range of the parameter space. Furthermore, these methods are highly dependent on the accuracy of the forward model; if the model has biases, the optimization results will be amplified, reducing the feasibility of back-projecting parameters. In view of this, this application proposes a method for predicting laser welding process parameters. By establishing a reverse prediction model, the reverse prediction parameters are fed back to the forward quality prediction model for verification. The training process of the reverse model is dynamically adjusted using the closed-loop prediction error as the driving signal, thereby significantly improving the accuracy of parameter back-calculation and providing a reliable means of verification and evaluation of the prediction results.

[0037] refer to Figure 1 and Figure 2 One embodiment of this application provides a method for predicting laser welding process parameters, which may specifically include the following steps.

[0038] Step 1: Collect laser welding experimental data, including process parameter combinations and corresponding welding quality indicators.

[0039] In some embodiments, annular spot laser welding experiments are conducted under different combinations of process parameters (center laser power, outer ring laser power, welding speed), and the corresponding weld penetration and weld width data are collected. In practice, welding experimental data from lasers with other spot types can also be collected; this description focuses on annular spot laser welding experiments. More specifically, it may include:

[0040] Experimental data on annular spot laser welding were collected under different combinations of process parameters, including outer ring laser power (0 W-4000 W), center laser power (500 W-4000 W), and welding speed (10 mm / s-100 mm / s). Welding quality indicators, including weld penetration and weld width, were obtained by taking typical weld sections and observing metallographic images using an optical microscope.

[0041] Each process parameter is determined independently within its own constraints and combined to form an input sample with three dimensions. The measured welding quality indicators form an output sample with two dimensions. The input and output indicators of each experiment correspond one-to-one, forming an experimental dataset.

[0042] The collected process parameters and corresponding welding quality index data were structured and divided into training and testing sets at an 8:2 ratio. Specifically, this included:

[0043] The process parameter data and welding quality index data of each group of experiments are uniformly recorded as a five-element array. The three input process parameters include the outer ring laser power, the center laser power, and the welding speed. The two output welding quality indexes include the weld penetration depth and the weld width.

[0044] All sample data were deduplicated, missing values ​​were filled, and outliers were removed. The data were then uniformly stored in a two-dimensional structured table format according to the physical magnitude and dimensional standards of the process parameters.

[0045] Using weld penetration and weld width as output samples, and the corresponding process parameters as independent inputs, three datasets are constructed for the three process parameters, which can be represented as outer ring laser power datasets. Central laser power dataset And welding speed dataset .in For the first One input sample, Indicates the size format of the input sample. , as well as Corresponding to the first The outer ring laser power, center laser power, and welding speed of each input sample. The total number of samples is denoted as . Based on the principle of random uniform sampling, all structured sample data are divided into training and test sets in an 8:2 ratio, with the test set reserved for evaluating the generalization performance of the prediction model.

[0046] Considering the dimensional differences between various process parameters and welding quality indicators, the Z-score standardization method is used to unify them. Specifically, this includes:

[0047] Considering the differences in physical units and orders of magnitude between different process parameters and welding quality indicators, the Z-score standardization method is used to unify the processing of all input process parameters and output welding quality indicators. For each output variable (weld penetration depth) in the dataset... With weld width ) and input variables (including outer ring laser power) Central laser power Welding speed ), calculate their mean in the dataset respectively. and standard deviation The specific formula is expressed as follows:

[0048] (1)

[0049] In the formula, For the first of the input or output variables A number, To correspond to the standardized results, the input and output variables of the corresponding standardized inverse prediction model can be expressed as follows: , and , , .

[0050] Step 2: Construct a forward prediction model to predict welding quality indicators based on process parameters; construct a backward prediction model to predict process parameters based on target welding quality indicators. Specifically, three independent backward prediction models and two independent forward prediction models are constructed based on a multilayer perceptron (MLP). The backward prediction model is used to fit the mapping relationship between welding quality indicators and process parameters, while the forward prediction model is used to fit the nonlinear relationship between process parameters and welding quality indicators.

[0051] To address the mapping relationship between welding quality indicators and process parameters, a regression model based on a Multi-Level Processing (MLP) (referring to a backpropagation model) is constructed. Three process parameters serve as outputs, corresponding to three independent regression models. Welding quality (weld penetration and weld width) serves as the input to each regression model. The MLP model is a feedforward neural network, consisting of an input layer, several hidden layers, and an output layer. The input layer contains two nodes, corresponding to two standardized welding quality parameters: weld penetration and weld width. and weld width The number of hidden layers and nodes can be adjusted based on experimental results, and an activation function can be used. Enhanced nonlinear representation. The output layer is a single node, outputting standardized predicted values ​​for the outer ring laser power, center laser power, and welding speed, respectively.

[0052] To establish the mapping relationship between process parameters and welding quality indicators, a regression model based on a Multi-Level Processing (MLP) (referring to the forward prediction model) is constructed. Two welding quality indicators are used as outputs, corresponding to two independent regression models, with three process parameters serving as inputs to each model. The MLP model is a feedforward neural network, consisting of an input layer, several hidden layers, and an output layer. The input layer contains three nodes, corresponding to three process parameters output from the aforementioned backward prediction model: outer ring laser power, center laser power, and welding speed. The number of hidden layers and nodes can be adjusted based on experimental results, and an activation function is used. Enhance the nonlinear representation effect. The output layer is a single node, outputting the standardized predicted values ​​of weld penetration depth and weld width respectively.

[0053] Taking two hidden layers as an example, each hidden layer has 64 nodes. Both the backpropagation model and the forward propagation model have the structure of input layer - hidden layer 1 - hidden layer 2 - output layer. The input layer of the backpropagation model can be described as:

[0054] (2)

[0055] At this point, x is the input welding quality index vector.

[0056] The input layer of a positive prediction model can be represented as:

[0057] (3)

[0058] at this time, This is the input vector of process parameters.

[0059] The mathematical formula for hidden layer 1 can be expressed as:

[0060] (4)

[0061] In the formula, This is the weight matrix of the network layer. This is the bias vector of the network layer. The activation function is denoted as . In some embodiments, the modified linear unit function (ReLU) is used as the activation function, and its expression is as follows:

[0062] (5)

[0063] When the input value is less than 0, the output value is always equal to 0. When the input is greater than or equal to 0, the output value is equal to the input value.

[0064] The mathematical formula for hidden layer 2 can be expressed as:

[0065] (6)

[0066] In the formula, This is the weight matrix of the network layer. This is the bias vector for this network layer.

[0067] Taking the inverse prediction model as an example, the mathematical formula for the output layer can be expressed as:

[0068] (7)

[0069] In the formula, This is the standardized predicted value of the outer ring laser power. This is the weight matrix of the network layer. This is the bias vector of the network layer. This is the output of hidden layer 2.

[0070] Similarly, an MLP model with the same structure can be constructed for the reverse prediction model to predict the center laser power and welding speed, predicting the output center laser power. and welding speed For the positive prediction model, an MLP model with the same structure can be constructed for predicting weld penetration depth and weld width, with the predicted output being penetration depth. and melt width .

[0071] This application independently constructs three inverse MLP models (predicting outer ring laser power, center laser power, and welding speed, respectively) and two forward MLP models (predicting weld penetration and weld width, respectively). The ReLU activation function is employed to enhance nonlinear fitting capabilities, thereby avoiding the repeated searches required by traditional optimization algorithms and reducing training time by more than 50% (compared to genetic algorithms). The MLP models adapt to complex nonlinear mappings, and the measured parameter prediction error rate is less than 3%.

[0072] Step 3: Input the process parameters output by the reverse prediction model into the forward prediction model to generate the predicted welding quality; obtain the error between the predicted welding quality and the target welding quality, use this error as the loss value to optimize the reverse prediction model, form a closed-loop feedback mechanism, and obtain the optimal reverse prediction model.

[0073] Specifically, based on the one-to-one correspondence between process parameters and welding quality indicators, forward prediction regression models for weld penetration and weld width are trained using training set data. Simultaneously, a backward prediction model is trained on the training set, and the three process parameters predicted by the backward prediction model are combined to form the input of the forward prediction model. The error between the obtained forward output penetration and weld width and the target input penetration and weld width is calculated, and this error is used as the loss value to optimize the backward prediction model.

[0074] Using the positive prediction models for weld penetration and weld width constructed in step 2, respectively, with process parameters from the training set... As input to the model, the corresponding weld penetration depth or weld width standardized value ( , The target is used as the output, and iterative training is performed round by round. The number of training cycles can be flexibly set according to the experimental results.

[0075] Using the inverse prediction models for outer ring laser power, center laser power, and welding speed constructed in step 2, respectively, welding quality indicators from the training set are used. As input to the model, the corresponding standardized values ​​of process parameters ( , , The target is used as the output, and iterative training is performed round by round. The number of training cycles can be flexibly set according to the experimental results.

[0076] In the training of the aforementioned forward and backward prediction models, the mean squared error (MSE) between the model's predicted output and the true value is calculated, and the model weight parameters are adjusted using the backpropagation algorithm to minimize the loss function. The mathematical expression for the mean squared error function is as follows:

[0077] (8)

[0078] In the formula, The total number of samples, and These represent the model's predicted output and the actual value, respectively. During backpropagation, different optimization algorithms, such as stochastic gradient descent (SGD) or the Adam optimizer, can be used to iteratively update the network weights and biases. The choice of optimization algorithm can be determined based on the actual results. Taking SGD as an example, the weights of each layer... With bias The update process can be described as follows:

[0079] (9)

[0080] In the formula, For learning rate, and The loss function is respectively paired with the first... The gradients of layer weights and biases are calculated. This process iterates until the loss function converges to a set threshold or the maximum number of training epochs is reached, thereby obtaining the optimal model parameters for welding quality prediction.

[0081] After training the forward prediction model and the backward prediction model, the training data (weld penetration depth and weld width) is used to input the backward prediction model to generate the prediction output (predicted process parameters), which is then used as the input to the forward prediction model to obtain the predicted weld penetration depth and weld width. By calculating the mean square error between the predicted weld penetration depth and weld width (predicted welding quality) and the input weld penetration depth and weld width (target welding quality), the backward prediction model is optimized in reverse (keeping the forward prediction model fixed, i.e., freezing the weights of the forward prediction model, and using only the mean square error between the predicted welding quality and the target welding quality as the loss value for backpropagation optimization of the backward prediction model. This setting can achieve self-calibration of the prediction results and avoid the problem of "error propagation amplification"). This training process is consistent with the steps in the aforementioned formulas (8) and (9).

[0082] This application freezes the weights of the forward prediction model and optimizes only the backward prediction model. Combined with Z-score standardization and unified dimensionality, it can prevent the bias of the forward prediction model from being transmitted to the backward prediction model. The standardization process improves the robustness of the model to industrial data disturbances (the generalization error on the test set can be reduced by 40%).

[0083] The steps "inputting the process parameters output by the back prediction model into the forward prediction model to generate the predicted weld quality" and "obtaining the error between the predicted weld quality and the target weld quality, and using this error as the loss value to optimize the back prediction model" are executed alternately and iteratively on the training set until the closed-loop error converges. That is, after each round of back prediction model training, the forward prediction model is immediately used to verify and feedback the error. This setup allows for dynamic correction of the back prediction model's parameters, thereby improving the stability of the back prediction model's predictions.

[0084] Step 4: Predict laser welding process parameters based on the optimal reverse prediction model.

[0085] Input the target welding quality into the optimal reverse prediction model obtained in step 3, and output the prediction results of laser welding process parameters (outer ring laser power, center laser power and welding speed).

[0086] The aforementioned method for predicting laser welding process parameters can be applied to annular spot laser welding of medium-thick plate structural steel. For Q235 steel with a thickness of not less than 12 mm, the measured weld depth / weld width deviation after predicting the process parameters is less than 5%, thus verifying the reliability of the closed-loop mechanism in high-precision industrial welding.

[0087] This application proposes a laser welding process parameter prediction method that inputs the process parameters output by the inverse prediction model into a trained forward prediction model to generate a predicted welding quality. The predicted welding quality is then compared with the target welding quality to calculate the error. This method achieves real-time verification of the inverse prediction results for the first time, filling the technical gap of "inability to verify inverse results." By verifying the output of the inverse prediction model through the forward prediction model, it ensures that the combination of process parameters can achieve the target welding quality in actual welding. This application uses the closed-loop error (predicted welding quality vs. target welding quality) as the loss value to inversely optimize the inverse prediction model, dynamically constraining the model output and automatically converging to the optimal combination of process parameters. This eliminates the risk of non-unique solutions and reduces prediction drift through iterative calibration.

[0088] This application overcomes the four major technical bottlenecks in reverse prediction of process parameters in annular spot welding (one-to-many mapping leading to prediction uncertainty, lack of verification mechanism, low search efficiency, and uncontrollable error) through a closed-loop architecture of "reverse prediction → forward verification → error feedback → dynamic calibration", providing a highly reliable and efficient process optimization solution for intelligent welding systems.

[0089] Based on the same inventive concept, embodiments of this application also provide an electronic device. Figure 3 This is a structural block diagram of an electronic device provided in an embodiment of this application. Figure 3 As shown in the embodiments of this application, an electronic device includes: one or more processors 101, a memory 102, and one or more I / O interfaces 103. The memory 102 stores one or more programs, which, when executed by the one or more processors, cause the one or more processors to implement any of the laser welding process parameter prediction methods described in the above embodiments; the one or more I / O interfaces 103 are connected between the processor and the memory, configured to enable information interaction between the processor and the memory.

[0090] The processor 101 is a device with data processing capabilities, including but not limited to a central processing unit (CPU); the memory 102 is a device with data storage capabilities, including but not limited to random access memory (RAM, more specifically SDRAM, DDR, etc.), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), and flash memory (FLASH); the I / O interface (read / write interface) 103 is connected between the processor 101 and the memory 102, and can realize information interaction between the processor 101 and the memory 102, including but not limited to a data bus (Bus).

[0091] In some embodiments, the processor 101, memory 102, and I / O interface 103 are interconnected via bus 104, and thus connected to other components of the computing device.

[0092] In some embodiments, the one or more processors 101 include a field-programmable gate array.

[0093] This application also provides a computer-readable medium. The computer-readable medium stores a computer program, which, when executed by a processor, implements the steps in any of the laser welding process parameter prediction methods described in the above embodiments. The computer-readable storage medium can be volatile or non-volatile.

[0094] This application also provides a computer program product, including computer-readable code, or a non-volatile computer-readable storage medium carrying computer-readable code. When the computer-readable code is run in the processor of an electronic device, the processor in the electronic device executes the above-described laser welding process parameter prediction method.

[0095] Those skilled in the art will understand that all or some of the steps, systems, and apparatuses disclosed above, and their functional modules / units, can be implemented as software, firmware, hardware, or suitable combinations thereof. In hardware implementations, the division between functional modules / units mentioned above does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be performed collaboratively by several physical components. Some or all physical components may be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit (ASIC). Such software can be distributed on a computer-readable storage medium, which may include computer storage media (or non-transitory media) and communication media (or transient media).

[0096] As is known to those skilled in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information, such as computer-readable program instructions, data structures, program modules, or other data. Computer storage media includes, but is not limited to, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), static random access memory (SRAM), flash memory or other memory technologies, portable compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical disc storage, magnetic cartridges, magnetic tape, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and is accessible to a computer. Furthermore, it is known to those skilled in the art that communication media typically contain computer-readable program instructions, data structures, program modules, or other data in modulated data signals such as carrier waves or other transmission mechanisms, and may include any information delivery medium.

[0097] The computer-readable program instructions described herein can be downloaded from computer-readable storage media to various computing / processing devices, or downloaded via a network, such as the Internet, local area network, wide area network, and / or wireless network, to an external computer or external storage device. The network may include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. A network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards them to the computer-readable storage media in the respective computing / processing device.

[0098] The computer program instructions used to perform the operations of this application may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, status setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk, C++, etc., and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The computer-readable program instructions may be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or may be connected to an external computer (e.g., via the Internet using an Internet service provider). In some embodiments, electronic circuits, such as programmable logic circuits, field-programmable gate arrays (FPGAs), or programmable logic arrays (PLAs), are personalized by utilizing the status information of the computer-readable program instructions. These electronic circuits can execute the computer-readable program instructions to implement various aspects of this application.

[0099] The computer program product described herein can be implemented specifically through hardware, software, or a combination thereof. In one alternative embodiment, the computer program product is specifically embodied in a computer storage medium; in another alternative embodiment, the computer program product is specifically embodied in a software product, such as a software development kit (SDK), etc.

[0100] Various aspects of this application are described herein with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer-readable program instructions.

[0101] These computer-readable program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that, when executed by the processor of the computer or other programmable data processing apparatus, they create means for implementing the functions / actions specified in one or more blocks of the flowchart and / or block diagram. These computer-readable program instructions can also be stored in a computer-readable storage medium that causes a computer, programmable data processing apparatus, and / or other device to operate in a particular manner; thus, the computer-readable medium storing the instructions comprises an article of manufacture that includes instructions for implementing aspects of the functions / actions specified in one or more blocks of the flowchart and / or block diagram.

[0102] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to perform the functions / actions specified in one or more boxes of a flowchart and / or block diagram.

[0103] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of an instruction containing one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0104] Example embodiments have been disclosed herein, and while specific terminology has been used, it is for general illustrative purposes only and should not be construed as limiting. In some embodiments, it will be apparent to those skilled in the art that features, characteristics, and / or elements described in connection with particular embodiments may be used alone, or in combination with features, characteristics, and / or elements described in connection with other embodiments, unless otherwise expressly indicated. Therefore, those skilled in the art will understand that various changes in form and detail may be made without departing from the scope of this application as set forth by the appended claims.

Claims

1. A method for predicting laser welding process parameters, characterized in that, include: Collect laser welding experimental data, including process parameter combinations and corresponding welding quality indicators; Construct a positive prediction model to predict welding quality indicators based on process parameters; Construct a reverse prediction model to predict process parameters based on target welding quality indicators; The process parameters output by the reverse prediction model are input into the forward prediction model to generate the predicted welding quality. The error between the predicted welding quality and the target welding quality is obtained. This error is used as the loss value to optimize the reverse prediction model, forming a closed-loop feedback mechanism, and the optimal reverse prediction model is obtained. Prediction of laser welding process parameters based on the optimal reverse prediction model.

2. The prediction method according to claim 1, characterized in that, After collecting the laser welding experimental data, the following is also included: The process parameter combinations and corresponding welding quality indicators are structured and divided into training and testing sets according to a preset ratio.

3. The prediction method according to claim 1, characterized in that, After collecting the laser welding experimental data, the following is also included: The Z-score standardization method is used to unify the dimensions of the process parameter combinations and their corresponding welding quality indicators.

4. The prediction method according to claim 1, characterized in that, Both the forward prediction model and the backward prediction model are built based on a multilayer perceptron, wherein: The input layer of the reverse prediction model includes two nodes, which respectively input the weld penetration depth and weld width, two welding quality indicators; the output layer of the reverse prediction model is a single node, which outputs the predicted values ​​of three process parameters: outer ring laser power, center laser power, and welding speed. The input layer of the forward prediction model includes three nodes, which respectively input three process parameters output by the reverse prediction model: outer ring laser power, center laser power, and welding speed; the output layer of the forward prediction model is a single node, which outputs predicted values ​​of two welding quality indicators: weld penetration and weld width.

5. The prediction method according to claim 4, characterized in that, Both the forward prediction model and the reverse prediction model include a hidden layer based on a modified linear unit function.

6. The prediction method according to claim 1, characterized in that, Obtain the error between the predicted welding quality and the target welding quality, use this error as the loss value to optimize the inverse prediction model, and form a closed-loop feedback mechanism, including: The mean square error between the predicted welding quality and the target welding quality is obtained, and the model weight parameters are adjusted through the backpropagation algorithm to minimize the loss function.

7. The prediction method according to claim 6, characterized in that, Obtain the error between the predicted welding quality and the target welding quality, use this error as the loss value to optimize the inverse prediction model, and form a closed-loop feedback mechanism, including: The weights of the forward prediction model are frozen, and the backward prediction model is optimized by backpropagation using only the mean square error between the predicted welding quality and the target welding quality as the loss value.

8. The prediction method according to claim 7, characterized in that, Obtain the error between the predicted welding quality and the target welding quality, use this error as the loss value to optimize the inverse prediction model, and form a closed-loop feedback mechanism, including: The steps "inputting the process parameters output by the inverse prediction model into the forward prediction model to generate the predicted welding quality" and "obtaining the error between the predicted welding quality and the target welding quality, and using this error as the loss value to optimize the inverse prediction model" are executed alternately and iteratively on the training set until the closed-loop error converges.

9. An electronic device, characterized in that, include: One or more processors; Memory, used to store one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors are enabled to implement the steps in the prediction method as described in any one of claims 1 to 8.

10. A computer-readable medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it can perform the steps in the prediction method as described in any one of claims 1 to 8.

Images