Method, computer equipment and storage medium for welding quality diagnosis under multiple welding tasks

Abstract

The purpose of this application is to provide a method and device for welding quality diagnosis under multiple welding tasks, the method comprising: obtaining first real-time welding data during the welding process, wherein the first real-time welding data matches a first welding task, and the first welding task belongs to one of multiple welding tasks; based on a target welding quality diagnosis model, determining first welding quality diagnosis information corresponding to the first real-time welding data, wherein the target welding quality diagnosis model is used to perform welding quality diagnosis for multiple welding tasks. This solution can perform quality diagnosis on different welding production activities by training a welding quality diagnosis model for multiple welding tasks. When executing different welding quality diagnosis tasks, there is no need to redeploy the model, saving resources, improving diagnosis efficiency, and then timely adjusting welding parameters during welding to ensure welding production efficiency and quality.

Description

Technical Field

[0001] The present application relates to the field of welding, and in particular to a technology for performing welding quality diagnosis under multiple welding tasks. Background Art

[0002] Welding, as industrial tailoring, is a very important processing method in industrial production. The quality of welding directly affects the overall performance of the welding product. Therefore, welding quality diagnosis is also an important part of welding processing. For welding production under different working conditions, welding quality diagnosis models suitable for each working condition are usually trained separately for corresponding quality diagnosis. These models are usually applicable in a certain working condition, but are no longer applicable when the working condition changes or other working conditions. Summary of the invention

[0003] One object of the present application is to provide a method and device for performing welding quality diagnosis under multiple welding tasks.

[0004] According to one aspect of the present application, a method for welding quality diagnosis under multiple welding tasks is provided, the method comprising:

[0005] Acquire first real-time welding data during the welding process, wherein the first real-time welding data matches a first welding task, and the first welding task belongs to one of a plurality of welding tasks;

[0006] Based on a target welding quality diagnosis model, first welding quality diagnosis information corresponding to the first real-time welding data is determined, wherein the target welding quality diagnosis model is used to perform welding quality diagnosis for the multiple welding tasks.

[0007] According to one aspect of the present application, a computer device for performing welding quality diagnosis under multiple welding tasks is provided, comprising a memory, a processor, and a computer program stored in the memory, characterized in that the processor executes the computer program to implement the steps of any of the methods described above.

[0008] According to one aspect of the present application, a computer-readable storage medium is provided, on which a computer program is stored, characterized in that when the computer program is executed by a processor, the steps of any of the methods described above are implemented.

[0009] According to one aspect of the present application, a computer program product is provided, including a computer program, characterized in that when the computer program is executed by a processor, the steps of any of the methods described above are implemented.

[0010] According to one aspect of the present application, a device for performing welding quality diagnosis under multiple welding tasks is provided, the device comprising:

[0011] A module for acquiring first real-time welding data during a welding process, wherein the first real-time welding data matches a first welding task, and the first welding task belongs to one of a plurality of welding tasks;

[0012] A second module is used to determine first welding quality diagnostic information corresponding to the first real-time welding data based on a target welding quality diagnostic model, wherein the target welding quality diagnostic model is used to perform welding quality diagnosis for the multiple welding tasks.

[0013] Compared with the prior art, the present application obtains first real-time welding data during the welding process, wherein the first real-time welding data matches a first welding task, and the first welding task belongs to one of multiple welding tasks; based on a target welding quality diagnosis model, the first welding quality diagnosis information corresponding to the first real-time welding data is determined, wherein the target welding quality diagnosis model is used to perform welding quality diagnosis for the multiple welding tasks. This solution uses incremental learning to train a welding quality diagnosis model for multiple welding tasks, and can perform quality diagnosis on different welding production activities. When executing different welding quality diagnosis tasks, there is no need to redeploy the model, which saves resources and improves diagnosis efficiency. As a result, welding parameters can be adjusted in time during welding to ensure welding production efficiency and quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Other features, objects and advantages of the present application will become more apparent by reading the detailed description of non-limiting embodiments made with reference to the following drawings:

[0015] Figure 1 A flow chart of a method for performing welding quality diagnosis under multiple welding tasks according to an embodiment of the present application is shown;

[0016] Figure 2 A flow chart of a method for welding quality diagnosis according to an embodiment of the present application is shown;

[0017] Figure 3 A structural diagram of a device for performing welding quality diagnosis under multiple welding tasks according to an embodiment of the present application is shown;

[0018] Figure 4 An exemplary system is shown that can be used to implement the various embodiments described in this application.

[0019] The same or similar reference numerals in the drawings represent the same or similar components. DETAILED DESCRIPTION

[0020] The present application is described in further detail below in conjunction with the accompanying drawings.

[0021] In a typical configuration of the present application, the terminal, the device of the service network and the trusted party all include one or more processors (eg, a central processing unit (CPU)), an input / output interface, a network interface and a memory.

[0022] Memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash memory. Memory is an example of a computer-readable medium.

[0023] Computer readable media include permanent and non-permanent, removable and non-removable media that can be used to store information by any method or technology. Information can be computer readable instructions, data structures, modules of programs or other data. Examples of computer storage media include, but are not limited to, Phase-Change Memory (PCM), Programmable Random Access Memory (PRAM), Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), other types of random access memory (RAM), Read-Only Memory (ROM), Electrically-Erasable Programmable Read-Only Memory (EEPROM), Flash memory or other memory technology, Compact Disc Read-Only Memory (CD-ROM), Digital Versatile Disc (DVD) or other optical storage, magnetic cassettes, tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device.

[0024] The devices referred to in this application include but are not limited to user devices, network devices, or devices formed by integrating user devices and network devices through a network. The user devices include but are not limited to any mobile electronic products that can interact with users (for example, interact with users through a touchpad), such as smart phones, tablet computers, etc. The mobile electronic products can use any operating system, such as Android operating system, iOS operating system, etc. Among them, the network device includes an electronic device that can automatically perform numerical calculations and information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application specific integrated circuits (Application Specific Integrated Circuit, ASIC), programmable logic devices (Programmable Logic Device, PLD), field programmable gate arrays (Field Programmable Gate Array, FPGA), digital signal processors (Digital Signal Processor, DSP), embedded devices, etc. The network devices include but are not limited to computers, network hosts, single network servers, multiple network server sets or multiple servers. The cloud is composed of a large number of computers or network servers based on cloud computing (Cloud Computing), wherein cloud computing is a type of distributed computing, a virtual supercomputer composed of a group of loosely coupled computer sets. The network includes but is not limited to the Internet, a wide area network, a metropolitan area network, a local area network, a VPN network, a wireless self-organizing network (Ad Hoc network), etc. Preferably, the device may also be a program running on the user device, the network device, or a device formed by integrating the user device and the network device, the network device, the touch terminal, or the network device and the touch terminal through a network.

[0025] Of course, those skilled in the art should understand that the above-mentioned devices are only examples, and other existing or future devices that are applicable to the present application should also be included in the scope of protection of the present application and are included here by reference.

[0026] In the description of the present application, “plurality” means two or more, unless otherwise clearly and specifically defined.

[0027] Figure 1A flow chart of a method for performing welding quality diagnosis on multiple welding tasks according to an embodiment of the present application is shown, and the method includes: step S11 and step S12. In step S11, the device 1 obtains first real-time welding data during the welding process, wherein the first real-time welding data matches a first welding task, and the first welding task belongs to one of multiple welding tasks; in step S12, the device 1 determines first welding quality diagnosis information corresponding to the first real-time welding data based on a target welding quality diagnosis model, wherein the target welding quality diagnosis model is used to perform welding quality diagnosis for the multiple welding tasks.

[0028] In step S11, the device 1 obtains first real-time welding data, wherein the first real-time welding data matches a first welding task, and the first welding task belongs to one of multiple welding tasks. In some embodiments, the device 1 includes but is not limited to user devices and network devices with information processing or computing capabilities, such as tablet computers, computers, servers, etc. A model that can be used for welding quality diagnosis is deployed in the device 1, and the model can be used for quality diagnosis of multiple welding tasks. Different welding tasks correspond to different welding conditions. For any of the multiple welding tasks, the device 1 can obtain corresponding real-time welding data to perform diagnosis using the same target welding quality diagnosis model. In some embodiments, the first real-time welding data includes but is not limited to welding parameter information, sensor data (e.g., real-time voltage data, real-time current data, etc.), and / or weld image information corresponding to the first welding task.

[0029] In some embodiments, the step S11 includes: the device 1 receives the real-time acquisition data sent by the data acquisition device in the welding production corresponding to the first welding task, wherein the first welding task belongs to one of multiple welding tasks; based on the real-time acquisition data, the corresponding first real-time welding data is determined. In some embodiments, the data acquisition device includes but is not limited to various sensors and data acquisition gateways deployed in the welding production corresponding to the first welding task. The real-time acquisition data includes but is not limited to welding parameter information, sensor data (for example, real-time voltage data, real-time current data, etc.), and / or weld image information corresponding to the first welding task. The real-time acquisition data is directly obtained from the welding production, and further processing is required to obtain effective first real-time welding data that can be used for welding quality diagnosis. The device 1 can determine the corresponding first real-time welding data from the real-time acquisition data based on corresponding screening conditions (for example, corresponding voltage thresholds, current thresholds, etc.) and corresponding preprocessing methods (for example, missing value processing, outlier processing, normalization, etc.).

[0030] In step S12, the device 1 determines the first welding quality diagnostic information corresponding to the first real-time welding data based on the target welding quality diagnostic model, wherein the target welding quality diagnostic model is used to perform welding quality diagnosis for the multiple welding tasks. In some embodiments, the device 1 inputs the first real-time welding data into the target welding quality diagnostic model to obtain the corresponding first welding quality diagnostic information. In some embodiments, the device 1 can also obtain the real-time welding data corresponding to other welding tasks in the multiple welding tasks, and also use the target welding quality diagnostic model to perform welding quality diagnosis. In some embodiments, the method further includes step S13 (not shown), and the device 1 adjusts the welding parameter information in the welding process in real time according to the first welding quality diagnostic information to improve the welding quality. The welding parameter information includes but is not limited to information such as welding current, welding voltage, and welding shielding gas. For example, the current welding parameter information can be adjusted based on the first welding quality diagnostic information, and the welding quality diagnostic result can be obtained in real time through the target welding quality diagnostic model during the adjustment process, so as to determine the adjustment information of the welding parameter information (for example, the welding parameter information that should be adjusted or the adjustment direction and degree of a certain welding parameter information on the current basis is determined by the change in the number of welding abnormal events in the welding quality diagnostic information). Then, fine-tuning is performed based on the adjustment information, so that the acquired welding quality diagnosis information no longer contains abnormal welding events or the proportion of abnormal welding events is lower than the corresponding threshold. Here, this solution can use the corresponding incremental learning method to train and obtain the target welding quality diagnosis model based on multiple welding tasks, so that the model takes into account different welding tasks and can perform quality diagnosis on different welding production activities.

[0031] In some embodiments, the method further includes: step S14 (not shown), the device 1 determines the target welding quality diagnosis model based on a plurality of welding sample data sets corresponding to the plurality of welding tasks.

[0032] In some embodiments, the data sources corresponding to the multiple welding sample data sets are different. Different welding tasks among the multiple welding tasks correspond to different welding conditions. Generally, welding conditions refer to various specific conditions and environmental factors encountered during the welding process. For example, different welding conditions correspond to different welding processes (for example, different welding methods such as resistance spot welding, arc welding, and submerged arc welding). For another example, the welding process is affected by factors such as the thickness of the parent material, the form of the groove, and the diameter of the welding wire. Even for the same welding process, different process parameters set during welding correspond to different welding conditions.

[0033] In some embodiments, each welding sample data set corresponds to one welding task among multiple welding tasks, and the multiple welding sample data sets are independently collected from the welding production process of the corresponding welding task. The device 1 collects corresponding welding acquisition data from the welding production corresponding to the corresponding welding task, and determines the welding quality label information corresponding to the welding acquisition data; and determines the corresponding welding sample data set in combination with the welding acquisition data and the welding quality label information. The welding acquisition data includes but is not limited to welding parameter information, sensor data (e.g., real-time voltage data, real-time current data, etc.), and / or weld image information corresponding to the welding task. The device 1 can also process the welding acquisition data based on corresponding screening conditions (e.g., corresponding voltage thresholds, current thresholds, etc.) and corresponding preprocessing methods (e.g., missing value processing, outlier processing, normalization, etc.), and obtain effective welding acquisition data to determine the corresponding welding sample data set. In some embodiments, the device 1 is trained based on these welding sample data sets to obtain a target welding quality diagnosis model that can be used for the quality diagnosis of the aforementioned multiple welding tasks.

[0034] In some embodiments, the multiple welding sample data sets are used to train the target welding quality diagnosis model. For example, the training order of each welding sample data set can be determined based on factors such as the complexity of the welding tasks corresponding to the multiple welding sample data sets (for example, the welding tasks with characteristics such as a small number of samples and obvious sample features have low complexity) and similarity. For example, the model is first trained using the welding sample data set corresponding to the welding task with low complexity or a higher similarity to other welding tasks, so as to help the model adapt to new data faster, help the model learn better, and improve the efficiency of model training.

[0035] In some embodiments, if there is a new welding task later (for example, a new welding task is added or the process parameters set for a previous welding task are changed), the welding sample data set corresponding to the new welding task can be used to train the target welding quality diagnosis model previously deployed using incremental learning, and the key parameters of the target welding quality diagnosis model can be fine-tuned to obtain a model that is compatible with the new welding task. There is no need to face the situation where the welding data distribution changes due to changes in the welding task, which in turn makes the model diagnosis effect worse, and the only way is to retrain and deploy a new model that adapts to the new task.

[0036] In some embodiments, the multiple welding sample data sets include a first welding sample data set, and the welding task corresponding to the first welding sample data set belongs to one of the multiple welding tasks; the step S14 includes: step S141 (not shown), the device 1 performs model pruning on the first welding quality diagnosis model to obtain the pruned first welding quality diagnosis model, wherein the first welding quality diagnosis model is obtained based on other welding sample data sets in the multiple welding sample data sets except the first welding sample data set; step S142 (not shown), the device 1 determines the target welding quality diagnosis model based on the pruned first welding quality diagnosis model and the first welding sample data set. In some embodiments, the device 1 can perform model pruning based on the currently trained first welding quality diagnosis model for other welding tasks in the multiple welding tasks except the first welding sample data set, and only retain the most important key model parameters in the first welding quality diagnosis model for the other welding tasks without affecting the recognition accuracy of the first welding quality diagnosis model, so as to reduce the model complexity and facilitate subsequent model training. In some embodiments, the device 1 can perform incremental learning based on the first welding sample data set and the pruned first welding quality diagnosis model, so that the trained target welding quality diagnosis model can adapt to the aforementioned multiple welding tasks at the same time and can perform welding quality diagnosis on the multiple welding tasks. In some embodiments, the first welding quality diagnosis model is obtained based on other welding sample data sets in the multiple welding sample data sets except the first welding sample data set. The acquisition method of the first welding quality diagnosis model is similar to the acquisition method of the target welding quality diagnosis model described in the aforementioned steps S141 and S142: the previously trained welding quality diagnosis model is model pruned, and then the first welding quality diagnosis model is determined in combination with the welding sample data set corresponding to a welding task in the other welding sample data sets. The previously trained welding quality diagnosis model is obtained by training based on other data sets in the other welding sample data sets except the welding task. Therefore, it will not be repeated here, and it is included here by reference. By repeating the aforementioned steps, the welding quality diagnosis model can be continuously iterated and updated, so that the welding quality diagnosis model can adapt to a variety of welding tasks without spending a lot of time training a new model. In the iteration of the aforementioned steps, the initial welding quality diagnosis model that is pruned for the first time can be obtained by training with algorithms such as XGBoost, LightGBM (LightGradient Boosting Machine), AdaBoost (Adaptive Boosting) and random forest based on a welding sample data set corresponding to a welding task among the aforementioned multiple welding tasks.This initial welding quality diagnosis model can only be used for the quality diagnosis of a certain welding task. With the continuous steps of model pruning and training in combination with data sets corresponding to other welding tasks, the obtained model will continue to adapt to more welding tasks.

[0037] For example, for quality diagnosis of different welding tasks, the traditional solution is to train and optimize a model for quality diagnosis for different welding conditions. In the actual welding process, the position of the weld, the glue coating method, the number of base material layers, etc. may change, so the set welding time, welding current, power-on time and other process parameters may change. Once changed, it becomes a new welding task for the model, and the original diagnostic model cannot be effectively generalized and quality diagnosis can no longer be performed. Or due to business needs, quality diagnosis of welding production of other welding methods is required. Usually, users can only reuse the corresponding sample data, retrain, iterate, and optimize, obtain a new model and then deploy it to the production line for quality diagnosis. It often takes a lot of manpower and material resources, but the new model obtained can neither diagnose the welding data of the previous welding task nor adapt to subsequent changes in working conditions. In this solution, based on the sample data of the new welding task collected in a short period of time, according to the method described in the aforementioned steps S141 and S142, the sample data of the new welding task can be automatically learned and the parameters of the model applicable to the old welding task can be fine-tuned on the basis of the model applicable to the old welding task, so that the model can not only perform accurate quality diagnosis on the old welding task, but also support accurate quality diagnosis on the new welding task. There is no need to build a new model from scratch, nor to deploy the new model again, which reduces manual intervention, saves costs, and improves work efficiency.

[0038] In some embodiments, the pruning of the first welding quality diagnostic model to obtain the pruned first welding quality diagnostic model includes: device 1 performs a pruning operation on the first welding quality diagnostic model; based on the other welding sample data sets, the first welding quality diagnostic model that performs the pruning operation is trained to obtain the pruned first welding quality diagnostic model. For example, device 1 prunes the model parameters corresponding to the first welding quality diagnostic model. Since the deletion of the aforementioned model parameters may cause the accuracy of the model to decrease, the corresponding sample data can be selected from the other welding sample data sets to train the model on this basis, so as to restore the recognition accuracy of the model on the welding tasks corresponding to the other welding sample data sets. In some embodiments, the aforementioned pruning operation and retraining process can be repeated multiple times to gradually reduce the complexity of the model, and finally only the key model parameters in the first welding quality diagnostic model are retained to obtain the pruned first welding quality diagnostic model.

[0039] In some embodiments, the pruning operation performed on the first welding quality diagnostic model includes: performing corresponding pruning operations based on the model parameters corresponding to the first welding quality diagnostic model, and updating the first welding quality diagnostic model. For example, device 1 determines the characteristic information corresponding to the model parameters in the first welding quality diagnostic model. The characteristic information includes, but is not limited to, indicators such as the sparsity of each model parameter in the first welding quality diagnostic model, the importance of each model parameter relative to other welding sample data sets, and the sensitivity of each model parameter relative to the loss function of the first welding quality diagnostic model. Based on the characteristic information corresponding to the model parameters in the first welding quality diagnostic model, device 1 performs deletions, deletes unimportant / unnecessary model parameters, obtains the first model parameters, and completes the pruning operation on the first welding quality diagnostic model, that is, the first welding quality diagnostic model updated by the pruning operation only retains the first model parameters, and eliminates other redundant model parameters.

[0040] In some embodiments, the step S142 includes: the device 1 updates the pruned first welding quality diagnostic model based on the first welding sample data set using an incremental learning method and a corresponding optimization algorithm to obtain a target welding quality diagnostic model. In some embodiments, the optimization algorithm includes but is not limited to first-order optimization methods such as gradient descent method, stochastic gradient descent method, momentum method, and / or second-order optimization methods such as Newton method and quasi-Newton method. In some embodiments, the device 1 can train the pruned first welding quality diagnostic model based on the first welding sample data set, and use the optimization algorithm combined with the corresponding model parameter update rules to update the pruned first welding quality diagnostic model to obtain a target welding quality diagnostic model. In some embodiments, the model parameter update rule is as follows:

[0041]

[0042] Among them, θ * ={θ 2,i ,i=1,2,3…},L(θ * ) is the update target; L(θ 2 ) is the loss function value corresponding to the first welding sample data set; θ 2,i is the model parameter to be learned using the first welding sample data set, θ 1,i is the model parameter learned using other welding sample datasets, b i ∈[0,+∞], indicating the importance of model parameters to the welding tasks corresponding to other welding sample data sets, b i = 0 means that when updating θ 2,i There is no constraint when 2,i It is impossible to take into account the welding tasks corresponding to other welding sample data sets, b i=+∞ represents θ 2,i =θ 1,i , then the model parameter θ 2,i The welding task corresponding to the first welding sample data set cannot be taken into account.

[0043] In some embodiments, the method of updating the pruned first welding quality diagnosis model based on the first welding sample data set using an incremental learning method and a corresponding optimization algorithm to obtain a target welding quality diagnosis model includes: the device 1 updates the first model parameter corresponding to the pruned first welding quality diagnosis model based on the first welding sample data set using a corresponding optimization algorithm; based on the first welding sample data set, the pruned first welding quality diagnosis model after the first model parameter is updated uses an incremental learning method and a corresponding optimization algorithm to perform model training to obtain a target welding quality diagnosis model. In some embodiments, the device 1 can first be initialized based on the first model parameter corresponding to the pruned first welding quality diagnosis model, and then the model parameter is updated using the first welding sample data set, and then an incremental operation is performed on this basis to add new model parameters for training to obtain a target welding quality diagnosis model. In the updating and training of the aforementioned model parameters, the device 1 uses a corresponding optimization algorithm to complete the update to minimize the loss function corresponding to the welding quality diagnosis model. Taking the XGBoost algorithm for welding quality diagnosis model training as an example, the device 1 can initialize a decision tree model based on the first model parameter in the pruned first welding quality diagnosis model, and the first model parameter includes the depth of the tree, the weight of the leaf node, etc. Based on the first welding sample data set, the first model parameters such as the depth of the tree and the weight value of the leaf node are updated. Then, based on the decision tree structure corresponding to the pruned first welding quality diagnosis model, new leaf nodes are added for update training, and finally the target welding quality diagnosis model is obtained. Here, those skilled in the art should understand that the aforementioned method of training the welding quality diagnosis model by the XGBoost algorithm is only an example, and other existing or future model training algorithms that are applicable to this application should also be included in the scope of protection of this application and are included here by reference.

[0044] In some embodiments, the first welding sample data set includes first welding acquisition data and first welding quality label information corresponding to the first welding acquisition data; the updating of the first model parameters corresponding to the pruned first welding quality diagnosis model using a corresponding optimization algorithm based on the first welding sample data set includes: the device 1 determines the welding quality prediction information corresponding to the first welding acquisition data based on the pruned first welding quality diagnosis model; based on the welding quality prediction information and the first welding quality label information, iteratively optimizes the first model parameters corresponding to the pruned first welding quality diagnosis model using a corresponding optimization algorithm. For example, the device 1 inputs the first welding acquisition data into the pruned first welding quality diagnosis model to calculate the corresponding welding quality prediction information. Combined with the first welding quality label information corresponding to the first welding acquisition data, the corresponding loss function value is calculated. Based on the corresponding optimization algorithm, the first model parameters corresponding to the pruned first welding quality diagnosis model are iteratively optimized to minimize the loss function. The loss function used includes but is not limited to the loss function based on distance metric such as mean square error loss function (MSE) and L2 loss function, or the loss function based on probability distribution metric such as relative entropy and cross entropy.

[0045] In some embodiments, the iterative optimization of the first model parameters corresponding to the pruned first welding quality diagnosis model using a corresponding optimization algorithm based on the welding quality prediction information and the first welding quality label information includes: the device 1 determines the gradient and Hessian matrix of the corresponding loss function based on the welding quality prediction information and the first welding quality label information; iteratively optimizes the first model parameters corresponding to the pruned first welding quality diagnosis model based on the gradient of the loss function and the Hessian matrix. For example, the device 1 calculates the gradient and Hessian matrix of the loss function corresponding to the pruned first welding quality diagnosis model based on the welding quality prediction information and the corresponding first welding quality label information, and then combines the gradient descent algorithm, Newton's method and other optimization algorithms to iteratively optimize the first model parameters corresponding to the pruned first welding quality diagnosis model to minimize the loss function.

[0046] In some embodiments, the method of using an incremental learning method and a corresponding optimization algorithm to perform model training based on the first welding quality diagnosis model after the first model parameter is updated based on the first welding sample data set, and obtaining the target welding quality diagnosis model includes: the device 1 determines the target model parameters corresponding to the target welding quality diagnosis model based on the updated first model parameters, wherein the target model parameters include the updated first model parameters and the second model parameters; based on the first welding sample data set, the target model parameters are updated to obtain the target welding quality diagnosis model. For example, taking the use of the XGBoost algorithm for welding quality diagnosis model training as an example, in this solution, the device 1 does not need to rebuild the entire decision tree, but learns on the basis of the decision tree structure corresponding to the pruned first welding quality diagnosis model after the aforementioned first model parameter is updated, and then adds new leaf nodes. That is, the corresponding second model parameters are added on the basis of the updated first model parameters. The device 1 sets the initial weight of the newly added leaf node to 0, uses the first welding sample data set to predict welding quality, and uses the corresponding optimization algorithm to update the weights of each leaf node in the decision tree, thereby obtaining the corresponding target welding quality diagnosis model. Here, those skilled in the art should understand that the aforementioned method of training a welding quality diagnosis model through the XGBoost algorithm is only an example, and other existing or future model training algorithms that may appear, if applicable to the present application, should also be included in the scope of protection of the present application and are included herein by reference.

[0047] In some embodiments, the first welding quality diagnostic model after pruned is updated by using an incremental learning method based on the first welding sample data set in combination with a corresponding optimization algorithm, and obtaining the target welding quality diagnostic model further includes: device 1 performs model pruning on the target welding quality diagnostic model to obtain the pruned target welding quality diagnostic model. In some embodiments, in order to avoid model overfitting, after the target welding quality diagnostic model is obtained by model training using incremental learning, the target welding quality diagnostic model can be model pruned. Here, the process of model pruning the target welding quality diagnostic model is the same or similar to the model pruning described in the aforementioned step S141: performing a pruning operation on the target welding quality diagnostic model; training the target welding quality diagnostic model that performs the pruning operation based on the multiple welding sample data sets corresponding to the multiple welding tasks to obtain the pruned target welding quality diagnostic model, so it will not be repeated here and is included herein by reference.

[0048] In some embodiments, Figure 2The present scheme is shown to perform the whole process of multi-task welding quality diagnosis. During the welding quality diagnosis process, if the welding task changes (for example, the welding process or welding environment changes), which leads to a significant change in the distribution of welding data, so that the previously used welding quality diagnosis model has a poor diagnostic effect on the new welding data, then the device 1 can perform model iteration training on its own based on the currently used welding quality diagnosis model to obtain a model that can take into account the diagnosis of new and old welding data. Specifically, taking the XGBoost algorithm for welding quality diagnosis model training as an example, the device 1 can obtain welding data for new welding tasks from the data acquisition device, filter valid welding data from it through corresponding screening conditions (for example, corresponding voltage thresholds, current thresholds, etc.), and determine the welding quality labels corresponding to the valid welding data. Subsequently, the device 1 can pre-process the aforementioned valid welding data and the corresponding welding quality labels (for example, missing value processing, outlier processing, normalization, etc.) to obtain a sample data set with labels. Then, the device 1 can prune and retrain the initial model (i.e., the welding quality diagnosis model that is only applicable to the old welding task), re-weight the model parameters corresponding to the initial model, and obtain a model framework that retains the key information of the old welding task. Based on the aforementioned sample data set, the device 1 can use the incremental learning method to continuously update the model parameters on the basis of the model framework that retains the key information of the old welding task, so as to obtain a welding quality diagnosis model that contains both the new and old welding task key information. Specifically, the device 1 can first initialize the model parameters of the aforementioned welding quality diagnosis model after pruning and retraining; input the sample data set to obtain the corresponding welding quality prediction information, and then calculate the gradient and Hessian matrix of the loss function to update the model parameters; perform incremental operations on the basis of the aforementioned model structure, build a new decision tree structure by adding new nodes, and then update the node weights using the sample data set; and control the model complexity by pruning the model; and then complete the update of all model parameters to obtain a welding quality diagnosis model that contains both the new and old welding task key information. Subsequently, the welding quality diagnosis model can be used to perform quality inspection on real-time welding data from different welding tasks, without having to redeploy the model according to different welding tasks, thereby improving the inspection efficiency.

[0049] Figure 3A structural diagram of a device for performing welding quality diagnosis on multiple welding tasks according to an embodiment of the present application is shown, wherein the device 1 includes a first module 11 and a second module 12. The first module 11 acquires first real-time welding data during the welding process, wherein the first real-time welding data matches a first welding task, and the first welding task belongs to one of multiple welding tasks; the first module 12 determines first welding quality diagnosis information corresponding to the first real-time welding data based on a target welding quality diagnosis model, wherein the target welding quality diagnosis model is used to perform welding quality diagnosis for the multiple welding tasks. Here, the Figure 3 The specific implementations corresponding to the illustrated one-one module 11 and one-two module 12 are respectively the same as or similar to the specific embodiments of the aforementioned step S11 and step S12=, and thus will not be described in detail and are included herein by reference.

[0050] In some embodiments, the device 1 further includes a module 13 (not shown). The module 13 adjusts the welding parameter information in the welding process in real time according to the first welding quality diagnostic information. Here, the specific implementation of the module 13 is the same or similar to the specific implementation of the aforementioned step S13, so it will not be repeated and is included here by reference.

[0051] In some embodiments, the device 1 further includes a four-module 14 (not shown). The four-module 14 determines the target welding quality diagnosis model based on the multiple welding sample data sets corresponding to the multiple welding tasks. Here, the specific implementation of the four-module 14 is the same or similar to the specific implementation of the aforementioned step S14, so it will not be repeated and is included here by reference.

[0052] In some embodiments, the one-four module 14 includes a one-four-one unit 141 (not shown) and a one-four-two unit 142 (not shown). The multiple welding sample data sets include a first welding sample data set, and the welding task corresponding to the first welding sample data set belongs to one of the multiple welding tasks; the one-four-one unit 141 performs model pruning on the first welding quality diagnosis model to obtain the pruned first welding quality diagnosis model, wherein the first welding quality diagnosis model is obtained based on other welding sample data sets in the multiple welding sample data sets except the first welding sample data set; the one-four-two unit 142 determines the target welding quality diagnosis model based on the pruned first welding quality diagnosis model and the first welding sample data set. Here, the specific implementation methods of the one-four-one unit 141 and the one-four-two unit 142 are respectively the same or similar to the specific embodiments of the aforementioned step S141 and step S142, so they are not repeated here and are included here by reference.

[0053] Figure 4An exemplary system that can be used to implement various embodiments described in this application is shown;

[0054] like Figure 4 In some embodiments shown, the system 300 can be used as any of the devices in the various described embodiments. In some embodiments, the system 300 may include one or more computer-readable media (e.g., system memory or NVM / storage device 320) with instructions and one or more processors (e.g., (one or more) processors 305) coupled to the one or more computer-readable media and configured to execute instructions to implement modules to perform the actions described in this application.

[0055] For one embodiment, system control module 310 may include any suitable interface controller to provide any suitable interface to at least one of processor(s) 305 and / or any suitable device or component in communication with system control module 310 .

[0056] The system control module 310 may include a memory controller module 330 to provide an interface to the system memory 315. The memory controller module 330 may be a hardware module, a software module, and / or a firmware module.

[0057] The system memory 315 may be used, for example, to load and store data and / or instructions for the system 300. For one embodiment, the system memory 315 may include any suitable volatile memory, such as a suitable DRAM. In some embodiments, the system memory 315 may include double data rate type four synchronous dynamic random access memory (DDR4 SDRAM).

[0058] For one embodiment, system control module 310 may include one or more input / output (I / O) controllers to provide interfaces to NVM / storage device 320 and communication interface(s) 325 .

[0059] For example, NVM / storage device 320 may be used to store data and / or instructions. NVM / storage device 320 may include any suitable non-volatile memory (e.g., flash memory) and / or may include any suitable non-volatile storage device(s) (e.g., one or more hard disk drives (HDDs), one or more compact disk (CD) drives, and / or one or more digital versatile disk (DVD) drives).

[0060] NVM / storage device 320 may include storage resources that are physically part of the device on which system 300 is installed, or it may be accessible to the device without being part of the device. For example, NVM / storage device 320 may be accessed over a network via communication interface(s) 325.

[0061] Communication interface(s) 325 may provide an interface for system 300 to communicate over one or more networks and / or with any other suitable devices. System 300 may wirelessly communicate with one or more components of a wireless network in accordance with any of one or more wireless network standards and / or protocols.

[0062] For one embodiment, at least one of the processor(s) 305 may be packaged together with the logic of one or more controllers of the system control module 310 (e.g., the memory controller module 330). For one embodiment, at least one of the processor(s) 305 may be packaged together with the logic of one or more controllers of the system control module 310 to form a system-in-package (SiP). For one embodiment, at least one of the processor(s) 305 may be integrated on the same die with the logic of one or more controllers of the system control module 310. For one embodiment, at least one of the processor(s) 305 may be integrated on the same die with the logic of one or more controllers of the system control module 310 to form a system on chip (SoC).

[0063] In various embodiments, the system 300 may be, but is not limited to: a server, a workstation, a desktop computing device, or a mobile computing device (e.g., a laptop computing device, a handheld computing device, a tablet computer, a netbook, etc.). In various embodiments, the system 300 may have more or fewer components and / or a different architecture. For example, in some embodiments, the system 300 includes one or more cameras, a keyboard, a liquid crystal display (LCD) screen (including a touch screen display), a non-volatile memory port, multiple antennas, a graphics chip, an application specific integrated circuit (ASIC), and a speaker.

[0064] In addition to the methods and devices described in the above embodiments, the present application also provides a computer-readable storage medium, which stores computer code. When the computer code is executed, the method described in any of the preceding items is executed.

[0065] The present application also provides a computer program product. When the computer program product is executed by a computer device, the method described in any of the preceding items is executed.

[0066] The present application also provides a computer device, the computer device comprising:

[0067] one or more processors;

[0068] a memory for storing one or more computer programs;

[0069] When the one or more computer programs are executed by the one or more processors, the one or more processors are caused to implement the method as described in any of the preceding items.

[0070] It should be noted that the present application can be implemented in software and / or a combination of software and hardware, for example, can be implemented using an application specific integrated circuit (ASIC), a general purpose computer or any other similar hardware device. In one embodiment, the software program of the present application can be executed by a processor to implement the steps or functions described above. Similarly, the software program of the present application (including relevant data structures) can be stored in a computer-readable recording medium, for example, a RAM memory, a magnetic or optical drive or a floppy disk and similar devices. In addition, some steps or functions of the present application can be implemented using hardware, for example, as a circuit that cooperates with a processor to perform each step or function.

[0071] In addition, a part of the present application may be applied as a computer program product, such as a computer program instruction, which, when executed by a computer, can call or provide the method and / or technical solution according to the present application through the operation of the computer. Those skilled in the art should understand that the existence of computer program instructions in computer-readable media includes but is not limited to source files, executable files, installation package files, etc., and accordingly, the way in which computer program instructions are executed by a computer includes but is not limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to the computer.

[0072] Communication media include media by which communication signals containing, for example, computer readable instructions, data structures, program modules, or other data are transmitted from one system to another. Communication media may include guided transmission media such as cables and wires (e.g., fiber optic, coaxial, etc.) and wireless (unguided transmission) media that can propagate energy waves, such as acoustic, electromagnetic, RF, microwave, and infrared. Computer readable instructions, data structures, program modules, or other data may be embodied as a modulated data signal in, for example, a wireless medium such as a carrier wave or similar mechanism such as embodied as part of spread spectrum technology. The term "modulated data signal" refers to a signal whose one or more characteristics are changed or set in such a manner as to encode information in the signal. Modulation may be analog, digital, or a hybrid modulation technique.

[0073] By way of example and not limitation, computer-readable storage media may include volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information such as computer-readable instructions, data structures, program modules or other data. For example, computer-readable storage media include, but are not limited to, volatile memory, such as random access memory (RAM, DRAM, SRAM); and non-volatile memory, such as flash memory, various read-only memories (ROM, PROM, EPROM, EEPROM), magnetic and ferromagnetic / ferroelectric memories (MRAM, FeRAM); and magnetic and optical storage devices (hard disks, magnetic tapes, CDs, DVDs); or other media now known or later developed that can store computer-readable information / data for use by a computer system.

[0074] Here, according to an embodiment of the present application, a device is included, which includes a memory for storing computer program instructions and a processor for executing the program instructions, wherein, when the computer program instructions are executed by the processor, the device is triggered to run the methods and / or technical solutions based on the aforementioned multiple embodiments of the present application.

[0075] It is obvious to those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and that the present application can be implemented in other specific forms without departing from the spirit or basic features of the present application. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-restrictive, and the scope of the present application is limited by the attached claims rather than the above description, so it is intended to include all changes that fall within the meaning and scope of the equivalent elements of the claims in the present application. Any figure mark in the claims should not be regarded as limiting the claims involved. In addition, it is obvious that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices stated in the device claim can also be implemented by one unit or device through software or hardware. The words first, second, etc. are used to indicate names, and do not indicate any particular order.

Claims

1. A method for welding quality diagnosis under multiple welding tasks, wherein: The method comprises: Acquire first real-time welding data during the welding process, wherein the first real-time welding data matches a first welding task, and the first welding task belongs to one of a plurality of welding tasks; Determining first welding quality diagnostic information corresponding to the first real-time welding data based on a target welding quality diagnostic model, wherein the target welding quality diagnostic model is used to perform welding quality diagnosis for the multiple welding tasks; The method further comprises: determining the target welding quality diagnosis model based on a plurality of welding sample data sets corresponding to the plurality of welding tasks, wherein the plurality of welding sample data sets include a first welding sample data set, and the welding task corresponding to the first welding sample data set belongs to one of the plurality of welding tasks; The determining the target welding quality diagnosis model based on the multiple welding sample data sets corresponding to the multiple welding tasks comprises: performing model pruning on a first welding quality diagnosis model to obtain a pruned first welding quality diagnosis model, wherein the first welding quality diagnosis model is obtained based on other welding sample data sets among the multiple welding sample data sets except the first welding sample data set; determining the target welding quality diagnosis model based on the pruned first welding quality diagnosis model and the first welding sample data set; Wherein, determining the target welding quality diagnostic model based on the pruned first welding quality diagnostic model and the first welding sample data set includes: based on the first welding sample data set, using an incremental learning method and a corresponding optimization algorithm, updating the pruned first welding quality diagnostic model to obtain a target welding quality diagnostic model; Wherein, based on the first welding sample data set, using an incremental learning method and a corresponding optimization algorithm, updating the pruned first welding quality diagnosis model to obtain a target welding quality diagnosis model comprises: based on the first welding sample data set, using a corresponding optimization algorithm to update a first model parameter corresponding to the pruned first welding quality diagnosis model; based on the first welding sample data set, using an incremental learning method and a corresponding optimization algorithm to perform model training on the basis of the pruned first welding quality diagnosis model after the first model parameter is updated, to obtain a target welding quality diagnosis model; Among them, the first welding sample data set includes first welding acquisition data and first welding quality label information corresponding to the first welding acquisition data; the updating of the first model parameters corresponding to the pruned first welding quality diagnosis model based on the first welding sample data set using a corresponding optimization algorithm includes: determining the welding quality prediction information corresponding to the first welding acquisition data based on the pruned first welding quality diagnosis model; based on the welding quality prediction information and the first welding quality label information, iteratively optimizing the first model parameters corresponding to the pruned first welding quality diagnosis model using a corresponding optimization algorithm.

2. The method according to claim 1, wherein: The step of acquiring first real-time welding data during the welding process, wherein the first real-time welding data matches a first welding task, and the first welding task belongs to one of a plurality of welding tasks, comprises: Receiving real-time collected data sent by a data collection device in welding production corresponding to the first welding task, wherein the first welding task is one of a plurality of welding tasks; Based on the real-time collected data, corresponding first real-time welding data is determined.

3. The method according to claim 1, wherein: The method of performing model training based on the first welding sample data set and the pruned first welding quality diagnosis model after the first model parameter is updated by using an incremental learning method and a corresponding optimization algorithm to obtain a target welding quality diagnosis model comprises: Based on the updated first model parameters, determining target model parameters corresponding to the target welding quality diagnosis model, wherein the target model parameters include the updated first model parameters and the second model parameters; Based on the first welding sample data set, the target model parameters are updated to obtain the target welding quality diagnosis model.

4. A computer device for performing welding quality diagnosis under multiple welding tasks, comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 3.

5. A computer-readable storage medium having a computer program / instruction stored thereon, characterized in that: When the computer program / instructions are executed by a processor, the steps of the method according to any one of claims 1 to 3 are implemented.

Images