Welding gun current anomaly detection and early warning system based on machine learning

Abstract

The invention provides a welding gun current anomaly detection and early warning system based on machine learning. The system comprises the steps that historical current information of a welding gun is restored according to historical welding data of the welding gun, current anomaly characteristics contained in the historical current information are recognized, a current anomaly model of the welding gun is established according to a welding anomaly result corresponding to each current anomaly characteristic, and the current anomaly model of the welding gun is obtained; and real-time welding data of a welding gun is input into the current anomaly model for current detection, real-time current information and current current characteristics of current welding are obtained, when the current characteristics are abnormal, a speculated welding result is deduced according to the real-time current information, an abnormal welding position corresponding to the current current characteristics is positioned in the speculated welding result, and the current welding position is determined. The abnormal attribute corresponding to the abnormal welding position is recognized, the abnormal compensation scheme is generated, and abnormal early warning is carried out, so that welding defects can be effectively prevented, traditional manual detection is replaced with machine learning, the welding defects are effectively prevented, the efficiency and quality of defect detection are improved, and the downtime of a production line is effectively shortened.

Description

Technical Field

[0001] The present invention relates to the technical field of welding quality detection, and particularly to a welding torch current abnormal detection and early warning system based on machine learning. Background Art

[0002] At present, welding processes are involved in many fields, especially in automotive parts production lines. The main tool during welding is a welding torch. A welding torch is a tool used to supply welding current to the welding area during the welding process. During spot welding, a columnar electrode is used to form a weld point between the contact surfaces of two overlapping workpieces. During spot welding, the workpieces are first pressed tightly together, and then the current is switched on. Under the action of the resistance heat, the contact area of the workpieces melts and forms a weld point after cooling. Spot welding is mainly used for welding thin plate components and stamping parts with a thickness of less than 4 mm, and is particularly suitable for welding automotive bodies and carriages. The basic function of a welding torch is to provide a stable arc and pressure to ensure welding quality. In automotive parts production lines, welding work is particularly important. In order to improve the production efficiency of the production line, many manufacturers will open multiple production lines simultaneously. Although the work efficiency is improved, it requires multiple times the amount of manual labor for quality verification during the production process. This not only increases the workload of the manual labor but also easily causes detection omissions. Seriously, it may lead to the suspension of the production line operation, affecting the production quality and reputation of the factory. Therefore, there is an urgent need for a stable, online, and high-precision welding detection system.

[0003] Therefore, the present invention provides a welding torch current abnormal detection and early warning system based on machine learning. Summary of the Invention

[0004] The welding torch current abnormal detection and early warning system based on machine learning of the present invention can effectively prevent products with welding defects from flowing out, replaces traditional manual detection with machine learning, improves the efficiency and quality of defect detection, and effectively reduces the downtime of the production line.

[0005] The present invention provides a welding torch current abnormal detection and early warning system based on machine learning, including: A data analysis module, configured to restore the historical current information of the welding torch according to the historical welding data of the welding torch, and identify the current abnormal features included in the historical current information; A model construction module, configured to establish a current abnormal model of the welding torch according to the welding abnormal results corresponding to each current abnormal feature; A data acquisition module, configured to input the real-time welding data of the welding torch into the current abnormal model for current detection, and obtain the real-time current information and current features of the current welding; A depth detection module, configured to, when the current feature is abnormal, infer a speculative welding result based on the real-time current information, and locate an abnormal welding position corresponding to the current current feature in the speculative welding result; An early warning compensation module, configured to identify an abnormal attribute corresponding to the abnormal welding position, generate an abnormal compensation plan, and issue an abnormal warning.

[0006] In an implementable manner, The data analysis module includes: A data preprocessing unit, configured to cluster the historical welding data to obtain a number of clustering clusters, respectively identify the attributes of each clustering cluster to obtain several working states of the welding torch, and determine the working frequency corresponding to each working state according to the cluster size corresponding to each clustering cluster; A data sampling unit, configured to establish corresponding sampling times for the corresponding clustering clusters based on the working frequency, perform data sampling on the corresponding clustering clusters based on the sampling times to obtain a number of sampling data, and calculate the isolation degree between each sampling data and the clustering center of the corresponding clustering cluster; A current reduction unit, configured to infer the welding information of the welding torch in the corresponding working state by using the sampling data and its isolation degree, and construct the historical current information corresponding to the welding torch in each working state in combination with the machine parameters of the welding torch; An abnormal identification unit, configured to respectively obtain unstable current sub-information corresponding to each current information, respectively identify the current trend corresponding to each unstable current sub-information, and construct several current abnormal features of the welding torch in combination with the central current of the corresponding clustering center.

[0007] In an implementable manner, The model construction module includes: An abnormal analysis unit, configured to infer the welding result of the welding torch at the corresponding welding moment based on the current abnormal features, and construct an abnormal-result corresponding list; An abnormal training unit, configured to perform supervised learning on the abnormal-result corresponding list by using big data reasoning to construct the current abnormal logic of the welding torch, and convert the current abnormal logic into a model program; A framework training unit, configured to determine several working parameters of the welding torch based on the machine parameters of the welding torch, and at the same time determine the parameter range corresponding to each working parameter to generate a model framework; A model construction unit, configured to input the model program into the model framework to adjust the function of the model framework, and generate the current abnormal model of the welding torch.

[0008] In an implementable manner, The data acquisition module includes: A real-time acquisition unit, configured to acquire real-time welding data generated by the welding torch when the welding torch is in a working state; A current detection unit, configured to input the real-time welding data into the current anomaly model for synchronous detection, obtain anomaly identification information included in the real-time welding data, and determine the anomaly attribute corresponding to each piece of the anomaly identification information; An information recombination unit, configured to obtain the current output information of the current anomaly model, integrate the anomaly identification information into the current output information for anomaly marking, and obtain the real-time current information of the current welding; A feature generation unit, configured to construct the current trend of the current welding process and the current current of the welding torch based on the real-time current information corresponding to different moments during the current welding process, and construct the current current feature.

[0009] In an implementable manner, It further includes: A data arrangement unit, configured to obtain the real-time welding data corresponding to different moments during the current welding process, construct the data stream of the current welding and transmit it to the data synchronization unit for storage; A data synchronization unit, configured to obtain the data stream generated by the welding torch during each welding process, and generate and store the historical welding data of the welding torch after the data stream is updated.

[0010] In an implementable manner, The depth detection module includes: An anomaly capture unit, configured to construct the current transformation information of the current welding based on several current current features of the current welding, construct the current current law of the welding torch, and when the current current feature corresponding to the current moment does not conform to the current current law, regard the current current feature as a suspected anomaly feature; A depth detection unit, configured to input the suspected anomaly feature into the current anomaly model for repeated detection, obtain the current runaway trend and current smooth trend corresponding to the welding torch at the current moment, and when the first slope corresponding to the current runaway trend is greater than the second slope corresponding to the current smooth trend, determine that the suspected anomaly feature belongs to a determined anomaly feature; An anomaly positioning unit, configured to mark the determined anomaly feature in the real-time current information, deduce the speculated welding result of the current welding in a three-dimensional space and draw a conceptual diagram of the speculated welding result in the three-dimensional space, locate the abnormal welding position corresponding to the determined anomaly feature in the speculated welding result, and perform marking in the conceptual diagram of the speculated welding result.

[0011] In an implementable manner, The depth detection unit is further configured to: When the first slope corresponding to the current out-of-control trend is less than or equal to the second slope corresponding to the current smooth trend, it is determined that the suspected abnormal feature belongs to the normal current feature.

[0012] In an implementable manner, The warning compensation module includes: An identification and positioning unit, configured to respectively identify the actual welding position corresponding to each abnormal welding position in the current welded object, and physically mark the actual welding position by using a specified marking method; An attribute identification unit, configured to collect the actual welding image corresponding to the actual welding position, identify the welding presentation information included in the actual welding image, and search for the abnormal attribute corresponding to the welding presentation information in the big data; A solution generation unit, configured to search for the processing method corresponding to the abnormal attribute in the big data and combine it with the specification data of the actual welding position to generate an abnormal compensation solution for the current welded object; A warning execution unit, configured to perform a warning operation after physically marking the actual welding position.

[0013] In an implementable manner, It further includes: A defect compensation module, configured to perform welding compensation on the actual welding position within a specified time period based on the abnormal compensation solution.

[0014] In an implementable manner, It further includes: When the number of abnormal welding positions corresponding to the current welded object is higher than a first preset number or the number of abnormal attributes corresponding to the current welded object is higher than a second preset number, it is determined that the current working state of the welding torch is unqualified, and corresponding warning work is performed.

[0015] The achievable beneficial effects of the above technical solution are as follows: In order to monitor the working quality of multiple welding torches in real time, before quality inspection, the historical current information of the welding torch is restored based on the historical welding data of the welding torch, so as to determine the historical current anomaly characteristics of the welding torch. Then, combined with the welding anomaly results of each current anomaly characteristic, a current anomaly model of the welding torch is established. During quality inspection, the collected real-time welding data is input into the current anomaly model for current detection, obtaining the real-time current information and current characteristics during this welding. When the current characteristics are abnormal, the welding result is inferred, the abnormal welding position is located in the inferred welding result, and the abnormal attribute corresponding to the abnormal welding position is further identified. An abnormal compensation scheme is constructed to repair the defects on the welded object, and corresponding abnormal warnings are given at the same time. In this way, defects can be found and processed in time, effectively preventing welding defects, while improving the quality and efficiency of defect detection, being applicable to various welding scenarios, and improving welding productivity.

[0016] Other features and advantages of the present invention will be described in the following specification, and, in part, will be obvious from the specification, or can be understood by implementing the present invention. The objectives and other advantages of the present invention can be achieved and obtained by the structures specifically pointed out in the written specification and the accompanying drawings.

[0017] The technical solution of the present invention will be further described in detail below through the accompanying drawings and embodiments. Description of the Drawings

[0018] The drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation to the present invention. In the drawings: Figure 1 It is a schematic diagram of the composition of the welding torch current anomaly detection and warning system based on machine learning in the embodiment of the present invention; Figure 2 It is a schematic diagram of the composition of the data analysis module of the welding torch current anomaly detection and warning system based on machine learning in the embodiment of the present invention. Detailed Embodiments

[0019] The following describes the preferred embodiments of the present invention with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention and are not used to limit the present invention.

[0020] Embodiment 1 This embodiment provides a welding torch current anomaly detection and warning system based on machine learning, as Figure 1 shown, including: A data analysis module, which is used to restore the historical current information of the welding torch according to the historical welding data of the welding torch and identify the current abnormal features included in the historical current information; A model construction module, which is used to establish a current abnormal model of the welding torch according to the welding abnormal results corresponding to each current abnormal feature; A data acquisition module, which is used to input the real-time welding data of the welding torch into the current abnormal model for current detection, and obtain the real-time current information and the current current features of the current welding; A depth detection module, which is used to deduce and speculate the welding result according to the real-time current information when the current feature is abnormal, and locate the abnormal welding position corresponding to the current current feature in the speculated welding result; An early warning compensation module, which is used to identify the abnormal attributes corresponding to the abnormal welding position, generate an abnormal compensation plan and issue an abnormal warning.

[0021] In this example, the historical welding data represents the data generated by the welding torch during its previous work. The number of welding torches in this example can be one or more. When the number of welding torches is multiple, each welding torch corresponds to a historical welding data; In this example, the historical current information represents the current change information presented by the welding torch during its previous work; In this example, the current abnormal feature represents the feature presented when the welding torch works abnormally; In this example, the welding abnormal result represents the welding result presented by the item welded when the welding torch works abnormally; In this example, the real-time current information represents the current generated during the current welding work; In this example, the current current feature represents the current feature presented by the welding torch at the current moment; In this example, one speculated welding result corresponds to one abnormal welding position; In this example, the abnormal attributes include: porosity feature, crack feature, slag inclusion feature, lack of fusion feature, deformation feature, overlap feature and transition zone defect feature.

[0022] Working principle of the above technical solution: In order to monitor the working quality of multiple welding torches in real time, before quality inspection, the historical current information of the welding torch is restored according to the historical welding data of the welding torch, so as to determine the historical current anomaly characteristics of the welding torch. Then, combined with the welding anomaly results of each current anomaly characteristic, an current anomaly model of the welding torch is established. During quality inspection, the collected real-time welding data is input into the current anomaly model for current detection, obtaining the real-time current information and current characteristics during this welding. When the current characteristics are abnormal, the welding result is inferred, the abnormal welding position is located in the inferred welding result, and the abnormal attribute corresponding to the abnormal welding position is further identified. An anomaly compensation scheme is constructed to repair the defects on the welded object, and corresponding anomaly warnings are given at the same time. In this way, defects can be found and processed in time, effectively preventing welding defects, improving the quality and efficiency of defect detection, being applicable to various welding scenarios, and improving welding productivity.

[0023] Embodiment 2 Based on Embodiment 1, the welding torch current anomaly detection and warning system based on machine learning, as Figure 2 shown, the data analysis module includes: A data preprocessing unit, which is used to cluster the historical welding data to obtain several clustering clusters, respectively identify the attributes of each clustering cluster to obtain several working states of the welding torch, and determine the working frequency corresponding to each working state according to the cluster size corresponding to each clustering cluster; A data sampling unit, which is used to establish corresponding sampling times for the corresponding clustering clusters based on the working frequency, perform data sampling on the corresponding clustering clusters based on the sampling times to obtain several sampling data, and calculate the isolation degree between each sampling data and the clustering center of the corresponding clustering cluster; A current restoration unit, which is used to deduce the welding information of the welding torch in the corresponding working state by using the sampling data and its isolation degree, and construct the historical current information corresponding to each working state of the welding torch in combination with the machine parameters of the welding torch; An anomaly identification unit, which is used to respectively obtain the unstable current sub-information corresponding to each current information, respectively identify the current trend corresponding to each unstable current sub-information, and construct several current anomaly characteristics of the welding torch in combination with the central current of the corresponding clustering center.

[0024] In this example, the clustering cluster represents the result of clustering historical welding data with the same characteristics; In this example, the working state represents the state presented by different operations that the welding torch can perform; In this example, the working frequency represents the frequency at which the welding torch presents different working states; In this example, the number of samplings is directly proportional to the working frequency. Generally, the number of samplings is the smallest integer multiple of the working frequency, and the minimum number of samplings is 10. That is, for example, when the working frequency is 0.20, the number of samplings is 2; when the working frequency is 0.35, the number of samplings is 35; In this example, the degree of isolation represents the distance between the sampled data and the corresponding cluster center; In this example, the cluster size represents the size of a clustering cluster, that is, the number of historical welding data in a clustering cluster; In this example, the welding information represents the information presented by the welding torch in its working state; In this example, the unstable current sub - information represents the information presented when the working current of the welding torch is unstable; In this example, the central current represents the current corresponding to the cluster center. When the cluster center is not a data value in the historical data, the current value of the central current is deduced according to the relationship between the cluster center and different data values.

[0025] The working principle and beneficial effects of the above - mentioned technical solution: By clustering historical welding data to distinguish several working states of the welding torch, the efficiency of state recognition is improved. Then, based on the working frequency of each clustering cluster, the number of samplings for each clustering cluster is determined, and thus sampling is carried out. The degree of isolation between the obtained sampled data and its cluster center is analyzed, so as to deduce the welding information in different working states. Combining the machine parameters of the welding torch to determine the historical current information in different working states. Finally, the unstable current sub - information is screened, and the current anomaly characteristics of the welding torch are constructed based on its corresponding current trend. In this way, each welding torch can be accurately analyzed to determine its current anomaly characteristics, laying a foundation for setting up a corresponding model for current anomaly detection later.

[0026] Embodiment 3 Based on Embodiment 1, in the welding torch current anomaly detection and early warning system based on machine learning, the model construction module includes: Anomaly analysis unit, used to deduce the welding result of the welding torch at the corresponding welding moment based on the current anomaly characteristics, and construct an anomaly - result corresponding list; Anomaly training unit, used to perform supervised learning on the anomaly - result corresponding list using big data reasoning to construct the current anomaly logic of the welding torch, and convert the current anomaly logic into a model program; Framework training unit, used to determine several working parameters of the welding torch based on the machine parameters of the welding torch, and at the same time determine the parameter range corresponding to each working parameter, and generate a model framework; A model construction unit for inputting the model program into the model framework to adjust the functions of the model framework and generate the current anomaly model of the welding torch.

[0027] In this example, the anomaly-result correspondence list represents the correspondence list between the current anomaly features and the welding results.

[0028] The working principle and beneficial effects of the above technical solution: By using the current anomaly features to deduce the welding results generated by the welding torch at different welding moments, and then constructing the anomaly-result correspondence list, further using big data reasoning to analyze the current logic of the welding torch, thereby generating the model program. At the same time, the model framework is constructed according to the machine parameters of the welding torch. Finally, the model program is input into the model framework to generate the current anomaly model of the welding torch. In this way, the current anomaly model of the welding torch can be constructed to detect the anomalies in the current presented by the welding torch during operation.

[0029] Embodiment 4 Based on Embodiment 1, for the welding torch current anomaly detection and warning system based on machine learning, the data acquisition module includes: A real-time acquisition unit for acquiring the real-time welding data generated by the welding torch when the welding torch is in the working state; A current detection unit for inputting the real-time welding data into the current anomaly model for synchronous detection, obtaining the anomaly identification information contained in the real-time welding data, and determining the anomaly attributes corresponding to each anomaly identification information; An information recombination unit for obtaining the current output information of the current anomaly model, integrating the anomaly identification information into the current output information for anomaly marking, and obtaining the real-time current information of this welding; A feature generation unit for constructing the current trend of this welding process and the current current of the welding torch based on the real-time current information corresponding to different moments during this welding process, and constructing the current current feature.

[0030] The working principle and beneficial effects of the above technical solution: When the welding torch is in the working state, the real-time welding data generated by it is input into the current anomaly model for synchronous detection, obtaining the anomaly identification information contained in the real-time welding data and determining the anomaly attributes corresponding to each anomaly identification information. Further, the anomaly identification information is integrated into the current output information of the current anomaly model to generate the real-time current information of this welding. Then, the current trend of this welding process and the current current of the welding torch are constructed by counting the real-time current information corresponding to different moments. Finally, the current current feature of the welding torch is constructed based on the current trend and the current current. In this way, the current of the welding torch can be fully analyzed to obtain accurate and effective current current features.

[0031] Example 5 Based on Example 4, the machine learning-based abnormal welding torch current detection and warning system further includes: A data sorting unit, configured to obtain the real-time welding data corresponding to different moments during the current welding process, construct a data stream of the current welding, and transmit it to the data synchronization unit for storage; A data synchronization unit, configured to obtain the data stream generated by the welding torch during each welding process, and generate and store the historical welding data of the welding torch after the data stream is updated.

[0032] The working principle and beneficial effects of the above technical solution: Synchronously store the real-time welding data generated during the current welding process, which is convenient for subsequent training of the current abnormal model and improves the detection quality.

[0033] Example 6 Based on Example 1, in the machine learning-based abnormal welding torch current detection and warning system, the depth detection module includes: An abnormal capture unit, configured to construct the current transformation information of the current welding according to several current current characteristics of the current welding, construct the current law of the current welding torch, and when the current current characteristic corresponding to the current moment does not conform to the current law, regard the current current characteristic as a suspected abnormal characteristic; A depth detection unit, configured to input the suspected abnormal characteristic into the current abnormal model for repeated detection, obtain the current out-of-control trend and current smooth trend corresponding to the welding torch at the current moment, and when the first slope corresponding to the current out-of-control trend is greater than the second slope corresponding to the current smooth trend, determine that the suspected abnormal characteristic belongs to a determined abnormal characteristic; An abnormal positioning unit, configured to mark the determined abnormal characteristic in the real-time current information, deduce the speculated welding result of the current welding in the three-dimensional space and draw a conceptual diagram of the speculated welding result in the three-dimensional space, locate the abnormal welding position corresponding to the determined abnormal characteristic in the speculated welding result, and mark it in the conceptual diagram of the speculated welding result.

[0034] In this example, the current out-of-control trend represents the trend of the current getting out of control at the current moment; In this example, the current smooth trend represents the trend that the current is smooth with the current of the previous moment at the current moment; In this example, the first slope represents the slope of the current out-of-control trend, and the second slope represents the slope of the current smooth trend; In this example, the conceptual diagram of the speculated welding result represents the speculated welding result presented in the three-dimensional space.

[0035] Working principle and beneficial effects of the above technical solution: By using the current characteristics of the current welding to construct the current change information of the current welding to determine the current law of the welding torch, timely capture the suspected abnormal characteristics that do not match the current law of the current characteristics, and then input the suspected abnormal characteristics into the current abnormal model for repeated detection, obtaining the current out-of-control trend and current smooth trend of the welding torch at the current moment. According to the slopes of the two trends, determine whether the suspected abnormal characteristics are definite abnormal characteristics, and then mark the definite abnormal characteristics in the real-time current information. At the same time, draw a conceptual diagram of the speculated welding result in three-dimensional space, locate its abnormal welding position and make corresponding markings. In this way, the welding result can be analyzed while welding, timely locate the abnormal welding position, facilitate timely welding repair, and reduce the probability of generating defective products.

[0036] Example 7 Based on Example 6, for the welding torch current abnormal detection and early warning system based on machine learning, the depth detection unit is further used for: When the first slope corresponding to the current out-of-control trend is less than or equal to the second slope corresponding to the current smooth trend, it is determined that the suspected abnormal characteristics belong to normal current characteristics.

[0037] Working principle and beneficial effects of the above technical solution: When the first slope corresponding to the current out-of-control trend is less than or equal to the second slope corresponding to the current smooth trend, it indicates that the suspected abnormal characteristics belong to normal current characteristics.

[0038] Example 8 Based on Example 1, for the welding torch current abnormal detection and early warning system based on machine learning, the early warning compensation module includes: An identification and positioning unit, used to respectively identify the actual welding position corresponding to each abnormal welding position in the current welding object, and use a specified marking method to physically mark the actual welding position; An attribute identification unit, used to collect the actual welding image corresponding to the actual welding position, identify the welding presentation information contained in the actual welding image, and find the abnormal attributes corresponding to the welding presentation information in the big data; A solution generation unit, used to find the processing method corresponding to the abnormal attributes in the big data and combine the specification data of the actual welding position to generate an abnormal compensation solution for the current welding object; An early warning execution unit, used to execute the early warning work after physically marking the actual welding position.

[0039] In this example, the physical marking can be: light marking, drawing marking, etc.; In this example, the actual welding image refers to the image obtained by photographing the actual welding position. In this example, the specification data represents the specifications of the welding joint at the actual welding position.

[0040] The working principle and beneficial effects of the above technical solution: By identifying the actual welding position of the welded object and making corresponding physical marks on it, then photographing it, determining the abnormal attributes of the current welded object by identifying the welding presentation information in the image, and then generating a corresponding abnormal compensation plan in combination with big data to guide relevant personnel to make corresponding defect corrections to the current welded object, and at the same time giving a defect warning to remind relevant personnel to handle the defects in time.

[0041] Embodiment 9 Based on Embodiment 8, the welding torch current abnormal detection and warning system based on machine learning further includes: A defect compensation module, configured to perform welding compensation on the actual welding position within a specified time period based on the abnormal compensation plan.

[0042] In this example, the specified time represents the duration set in advance by relevant personnel, generally 10 minutes.

[0043] The working principle and beneficial effects of the above technical solution: By using the abnormal compensation plan to perform welding compensation on the actual welding position, the generation rate of defective products can be effectively reduced.

[0044] Embodiment 10 Based on Embodiment 8, the welding torch current abnormal detection and warning system based on machine learning further includes: When the number of the abnormal welding positions corresponding to the current welded object is higher than a first preset number or the number of the abnormal attributes corresponding to the current welded object is higher than a second preset number, it is determined that the current working state of the welding torch is unqualified, and corresponding warning work is carried out.

[0045] In this example, the first preset number is 10 and the second preset number is 4.

[0046] The working principle and beneficial effects of the above technical solution: When there are too many abnormal welding positions or too many abnormal attributes on the current welded object, it is determined that the corresponding welding torch is faulty, and corresponding warning information is generated in time to remind relevant personnel to repair or replace the welding torch in time, so as to improve the welding efficiency.

[0047] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention is also intended to include these modifications and variations.

Claims

1. A welding torch current anomaly detection and warning system based on machine learning, characterized in that, Including: A data analysis module, configured to restore the historical current information of the welding torch according to the historical welding data of the welding torch, and identify the current anomaly features included in the historical current information; A model construction module, configured to establish a current anomaly model of the welding torch according to the welding anomaly results corresponding to each current anomaly feature; A data acquisition module, configured to input the real-time welding data of the welding torch into the current anomaly model for current detection, and obtain the real-time current information and the current features of the current welding; A depth detection module, configured to, when the current features are abnormal, deduce the welding result when the welding torch performs the next welding operation, and screen the abnormal welding positions with abnormal welding structures to deduce the welding result; A warning compensation module, configured to identify the abnormal attributes corresponding to the abnormal welding positions, generate an abnormal compensation scheme and perform abnormal warning.

2. The machine learning-based abnormal welding torch current detection and warning system according to claim 1, wherein The data analysis module includes: A data preprocessing unit, configured to cluster the historical welding data to obtain a plurality of clustering clusters, respectively identify the attributes of each clustering cluster to obtain several working states of the welding torch, and determine the working frequency corresponding to each working state according to the cluster scale corresponding to each clustering cluster; A data sampling unit, configured to establish corresponding sampling times for the corresponding clustering clusters based on the working frequency, perform data sampling on the corresponding clustering clusters based on the sampling times to obtain a plurality of sampling data, and respectively calculate the isolation degree between each sampling data and the clustering center of the corresponding clustering cluster; A current restoration unit, configured to deduce the welding information of the welding torch in the corresponding working state by using the sampling data and its isolation degree, and construct the historical current information corresponding to each working state of the welding torch in combination with the machine parameters of the welding torch; An anomaly identification unit, configured to respectively obtain the unstable current sub-information corresponding to each current information, respectively identify the current trend corresponding to each unstable current sub-information, and construct several current anomaly features of the welding torch in combination with the central current of the corresponding clustering center.

3. The machine learning-based abnormal welding torch current detection and warning system according to claim 1, characterized in that The model construction module includes: An anomaly analysis unit, configured to deduce the welding result of the welding torch at the corresponding welding moment based on the current anomaly features, and construct an anomaly-result correspondence list; An anomaly training unit, configured to perform supervised learning on the anomaly-result correspondence list by using big data inference to construct the current anomaly logic of the welding torch, and convert the current anomaly logic into a model program; A framework training unit, configured to determine several working parameters of the welding torch based on the machine parameters of the welding torch, and at the same time determine the parameter range corresponding to each working parameter to generate a model framework; A model construction unit, configured to input the model program into the model framework to adjust the function of the model framework, and generate a current anomaly model of the welding torch.

4. The machine learning-based abnormal welding torch current detection and warning system according to claim 1, characterized in that The data acquisition module includes: A real-time acquisition unit, configured to acquire the real-time welding data generated by the welding torch when the welding torch is in a working state; A current detection unit, configured to input the real-time welding data into the current anomaly model for synchronous detection, obtain the anomaly recognition information included in the real-time welding data, and determine the anomaly attributes corresponding to each piece of the anomaly recognition information; An information recombination unit, configured to obtain the current output information of the current anomaly model, integrate the anomaly recognition information into the current output information for anomaly marking, and obtain the real-time current information of the current welding; A feature generation unit, configured to construct the current trend of the current welding process and the current current of the welding torch based on the real-time current information corresponding to different moments during the current welding process, and construct the current current feature.

5. The machine learning-based abnormal welding torch current detection and warning system according to claim 4, wherein It further includes: A data sorting unit, configured to obtain the real-time welding data corresponding to different moments during the current welding process, construct the data stream of the current welding and transmit it to the data synchronization unit for storage; A data synchronization unit, configured to obtain the data stream generated by the welding torch during each welding process, and generate and store the historical welding data of the welding torch after the data stream is updated.

6. The machine learning-based abnormal welding torch current detection and early warning system according to claim 1, wherein, The depth detection module includes: An anomaly capture unit, configured to construct the current transformation information of the current welding based on several of the current current features of the current welding, construct the current current pattern of the welding torch, and when the current current feature corresponding to the current moment does not conform to the current current pattern, regard the current current feature as a suspected anomaly feature; A depth detection unit, configured to input the suspected anomaly feature into the current anomaly model for repeated detection, obtain the current runaway trend and the current smooth trend corresponding to the welding torch at the current moment, and when the first slope corresponding to the current runaway trend is greater than the second slope corresponding to the current smooth trend, determine that the suspected anomaly feature belongs to a determined anomaly feature; An anomaly positioning unit, configured to mark the determined anomaly feature in the real-time current information, deduce the speculated welding result of the current welding in a three-dimensional space and draw a conceptual diagram of the speculated welding result in the three-dimensional space, locate the abnormal welding position corresponding to the determined anomaly feature in the speculated welding result, and make a mark in the conceptual diagram of the speculated welding result.

7. The machine learning-based abnormal welding torch current detection and early warning system according to claim 6, characterized in that, The depth detection unit is further configured to: When the first slope corresponding to the current runaway trend is less than or equal to the second slope corresponding to the current smooth trend, determine that the suspected anomaly feature belongs to a normal current feature.

8. The machine learning-based abnormal welding torch current detection and warning system according to claim 1, characterized in that, The warning compensation module includes: An identification and positioning unit, configured to respectively identify the actual welding positions corresponding to each abnormal welding position in the current welding object, and physically mark the actual welding positions using a specified marking method; An attribute identification unit, configured to collect the actual welding image corresponding to the actual welding position, identify the welding presentation information included in the actual welding image, and search for the anomaly attributes corresponding to the welding presentation information in big data; A solution generation unit, configured to search for the processing methods corresponding to the anomaly attributes in big data and combine them with the specification data of the actual welding position to generate an anomaly compensation solution for the current welding object; An early warning execution unit, which is used to perform physical marking on the actual welding position and then execute the early warning work.

9. The machine learning-based abnormal welding torch current detection and early warning system according to claim 8, characterized in that, It further includes: A defect compensation module, which is used to perform welding compensation on the actual welding position within a specified time period based on the abnormal compensation scheme.

10. The machine learning-based welding torch current anomaly detection and warning system according to claim 8, wherein, It further includes: When the number of abnormal welding positions corresponding to the current weldment is higher than a first preset number or the number of abnormal attributes corresponding to the current weldment is higher than a second preset number, it is determined that the current working state of the welding torch is unqualified, and corresponding warning work is carried out.

Images