Welding quality test method, apparatus and system, and medium and program

Abstract

Provided in the present disclosure are a welding quality test method, apparatus and system, and a medium and a program. The test method comprises: receiving a plurality of groups of welding data collected by a plurality of sensors during welding; pre-processing the plurality of groups of welding data, so as to obtain a plurality of key features; inputting the plurality of key features into a test model, such that the test model determines, by means of a knowledge graph, whether there is a welding defect; when it is determined that there is a welding defect, outputting an identification result for indicating the presence of the welding defect; in response to the welding defect in the identification result belonging to a new welding defect type, adding new nodes into the knowledge graph, wherein the new nodes comprise a plurality of new welding process parameter nodes, which correspond to the plurality of key features, and a new welding defect node; and on the basis of the plurality of groups of welding data, establishing an association relationship between the plurality of new welding process parameter nodes and the new welding defect node.

Description

Methods, devices, systems, media and procedures for testing welding quality

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application is based on the application with CN application number 202411793872.8 and application date December 6, 2024, and claims its priority. The disclosed content of the CN application is hereby introduced as a whole into this application. Technical Field

[0003] The present disclosure relates to the field of industrial welding, and in particular to a method, device, system, medium and program for detecting welding quality. Background Art

[0004] In modern industrial production, welding is an essential manufacturing process, widely used in industries such as automotive, aerospace, shipbuilding, construction, and heavy machinery. Welding quality directly impacts the structural safety, reliability, and service life of a product. Therefore, welding quality testing has become a crucial step in ensuring product quality. Summary of the Invention

[0005] According to one aspect of the present disclosure, a method for detecting welding quality is provided, comprising: receiving multiple sets of welding data collected by multiple sensors during a welding process; preprocessing the multiple sets of welding data to obtain multiple key features; inputting the multiple key features into a detection model so that the detection model determines whether there is a welding defect through a knowledge graph, wherein the knowledge graph comprises multiple nodes and association relationships between the multiple nodes, the multiple nodes comprise welding process parameter nodes and welding defect nodes, the welding process parameter nodes corresponding to the welding defect nodes; in the case of determining that a welding defect exists, outputting an identification result of the existence of the welding defect; in response to the welding defect in the identification result belonging to a new welding defect type, adding a new node to the knowledge graph, the new node comprising multiple new welding process parameter nodes and new welding defect nodes corresponding to the multiple key features; and establishing an association relationship between the multiple new welding process parameter nodes and the new welding defect nodes based on the multiple sets of welding data.

[0006] In some embodiments, based on the multiple sets of welding data, an association relationship is established between the multiple new welding process parameter nodes and the new welding defect node, including: based on the multiple sets of welding data, determining the weight value of each new welding process parameter node of the multiple new welding process parameter nodes relative to the new welding defect node.

[0007] In some embodiments, the detection method further includes: comparing each key feature of the multiple key features with the features of each welding process parameter node in the knowledge graph to obtain the welding process parameter node closest to each key feature; calculating the error feature between each key feature and the feature of the closest welding process parameter node; and based on the error feature, adjusting the parameters of the detection model by gradient descent method to minimize the error feature.

[0008] In some embodiments, the multiple nodes in the knowledge graph further include a quality standard node, and the quality standard node includes characteristics of welding process parameters when the welding quality meets the quality requirements.

[0009] In some embodiments, the plurality of sets of welding data include at least two sets of welding current data, welding voltage data, temperature data of the welding zone, weld seam image data, and acoustic wave data during the welding process.

[0010] In some embodiments, the multiple sets of welding data are preprocessed, including: performing noise reduction processing, normalization processing and feature extraction processing on the multiple sets of welding data to obtain the multiple key features.

[0011] In some embodiments, the detection method also includes: constructing the knowledge graph before receiving the multiple sets of welding data; wherein, constructing the knowledge graph includes: defining the multiple nodes in the knowledge graph, wherein the multiple nodes include multiple welding process parameter nodes and welding defect nodes corresponding to the multiple welding process parameter nodes; using historical data and empirical knowledge to establish an association relationship between the multiple nodes, wherein the association relationship includes the weight value of each welding process parameter node in the multiple welding process parameter nodes relative to the welding defect node.

[0012] According to another aspect of the present disclosure, a device for detecting welding quality is provided, comprising: a receiving module for receiving multiple sets of welding data collected by multiple sensors during a welding process; a preprocessing module for preprocessing the multiple sets of welding data to obtain multiple key features; an identification module for inputting the multiple key features into a detection model so that the detection model determines whether a welding defect exists through a knowledge graph, and outputs an identification result indicating the existence of the welding defect when it is determined that the welding defect exists, wherein the knowledge graph includes multiple nodes and association relationships between the multiple nodes, the multiple nodes include welding process parameter nodes and welding defect nodes, and the welding process parameter nodes correspond to the welding defect nodes; a node setting module for adding a new node to the knowledge graph in response to the welding defect in the identification result belonging to a new welding defect type, the new node including multiple new welding process parameter nodes and new welding defect nodes corresponding to the multiple key features; and a relationship determination module for determining the association relationship between the multiple new welding process parameter nodes and the new welding defect nodes based on the multiple sets of welding data.

[0013] In some embodiments, the relationship determination module is used to determine a weight value of each of the multiple new welding process parameter nodes relative to the new welding defect node based on the multiple sets of welding data.

[0014] In some embodiments, the detection device also includes: an adaptive learning module for comparing each key feature of the multiple key features with the features of each welding process parameter node in the knowledge graph to obtain the welding process parameter node closest to each key feature, calculating the error feature between each key feature and the feature of the closest welding process parameter node, and based on the error feature, adjusting the parameters of the detection model by gradient descent method to minimize the error feature.

[0015] In some embodiments, the multiple nodes in the knowledge graph further include a quality standard node, and the quality standard node includes characteristics of welding process parameters when the welding quality meets the quality requirements.

[0016] In some embodiments, the plurality of sets of welding data include at least two sets of welding current data, welding voltage data, temperature data of the welding zone, weld seam image data, and acoustic wave data during the welding process.

[0017] In some embodiments, the preprocessing module is used to perform noise reduction, normalization and feature extraction on the multiple sets of welding data to obtain the multiple key features.

[0018] In some embodiments, the detection device also includes: a graph construction module, used to define the multiple nodes in the knowledge graph, and use historical data and empirical knowledge to establish an association relationship between the multiple nodes, wherein the multiple nodes include multiple welding process parameter nodes and welding defect nodes corresponding to the multiple welding process parameter nodes, and the association relationship includes the weight value of each welding process parameter node in the multiple welding process parameter nodes relative to the welding defect node.

[0019] According to another aspect of the present disclosure, a device for detecting welding quality is provided, comprising: a memory; and a processor coupled to the memory, wherein the processor is configured to execute the aforementioned detection method based on instructions stored in the memory.

[0020] According to another aspect of the present disclosure, a detection system for welding quality is provided, comprising: a current sensor, a voltage sensor, a temperature sensor, an image sensor, a sound sensor and the detection device as described above, wherein the current sensor, the voltage sensor, the temperature sensor, the image sensor and the sound sensor are all electrically connected to the detection device.

[0021] According to another aspect of the present disclosure, a computer-readable storage medium is provided, on which computer instructions are stored. When the computer instructions are executed by a processor, the detection method as described above is implemented.

[0022] According to another aspect of the present disclosure, a computer program is provided, comprising: instructions, which, when executed by a processor, enable the processor to perform the detection method as described above.

[0023] Other features and advantages of the present disclosure will become apparent from the following detailed description of exemplary embodiments of the present disclosure with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments of the present disclosure and, together with the description, serve to explain the principles of the present disclosure.

[0025] The present disclosure can be more clearly understood from the following detailed description with reference to the accompanying drawings, in which:

[0026] FIG1 is a flow chart illustrating a method for detecting welding quality according to some embodiments of the present disclosure;

[0027] FIG2 is a flow chart illustrating a method for detecting welding quality according to other embodiments of the present disclosure;

[0028] FIG3 is a block diagram schematically illustrating a structure of a device for detecting welding quality according to some embodiments of the present disclosure;

[0029] FIG4 is a block diagram schematically illustrating a structure of a device for detecting welding quality according to other embodiments of the present disclosure;

[0030] FIG5 is a block diagram schematically illustrating a structure of a device for detecting welding quality according to other embodiments of the present disclosure;

[0031] FIG6 is a block diagram schematically illustrating a structure of a device for detecting welding quality according to other embodiments of the present disclosure;

[0032] FIG7 is a block diagram schematically illustrating a structure of a system for detecting welding quality according to some embodiments of the present disclosure. DETAILED DESCRIPTION

[0033] Various exemplary embodiments of the present disclosure will now be described in detail with reference to the accompanying drawings. It should be noted that unless otherwise specifically stated, the relative arrangement of components and steps, numerical expressions and numerical values ​​set forth in these embodiments do not limit the scope of the present disclosure.

[0034] At the same time, it should be understood that for the convenience of description, the sizes of the various parts shown in the drawings are not drawn according to the actual proportional relationship.

[0035] The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the present disclosure, its application, or uses.

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

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

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

[0039] The inventors of the present disclosure have found that in the relevant technology, welding quality inspection mainly relies on two methods: manual inspection and automatic inspection. Manual inspection uses visual, ultrasonic, X-ray and other means, and experienced technicians check the welds. However, manual inspection is inefficient and easily affected by the experience and fatigue of the inspectors. The error rate is high, especially when dealing with complex tasks. It is difficult to maintain consistency, and the accuracy rate is about 70%-80%. The automatic inspection system uses technologies such as image processing, ultrasonic detection and infrared imaging. Although it improves efficiency, it has some shortcomings. For example, automatic inspection relies on static models and is difficult to adapt to changes in different welding conditions. Fixed models cannot cover all defect types, affecting the accuracy of detection.

[0040] In view of this, an embodiment of the present disclosure provides a method for detecting welding quality to improve the accuracy of welding defect detection.

[0041] FIG1 is a flow chart illustrating a method for detecting welding quality according to some embodiments of the present disclosure. The detection method can be implemented by a detection device, for example, the detection device is provided in an electronic device. As shown in FIG1 , the detection method includes steps S102 to S112.

[0042] In step S102, multiple sets of welding data collected by multiple sensors during the welding process are received.

[0043] In some embodiments, the plurality of sets of welding data include at least two sets of welding current data, welding voltage data, temperature data of the welding zone, weld seam image data, and acoustic wave data during the welding process.

[0044] For example, the plurality of sensors include at least two of a current sensor, a voltage sensor, a temperature sensor, an image sensor, and a sound sensor.

[0045] The current sensor collects welding current data, I(t), and transmits this data to the detection device. The current sensor collects welding current data at different times, generating a set of welding current data that is then transmitted to the detection device. This allows the detection device to monitor the stability of the welding current.

[0046] The voltage sensor collects welding voltage data, V(t), and transmits this data to the detection device. The voltage sensor collects welding voltage data at different times, generating a set of welding voltage data that is then transmitted to the detection device. This allows the detection device to monitor the stability of the welding voltage and ensure a stable voltage during the welding process.

[0047] The temperature sensor collects temperature data from the weld zone and transmits this data to the detection device. The temperature sensor can collect temperature data at different times, generating a set of temperature data that is transmitted to the detection device as a set of welding data. Temperature changes reflect changes in weld heat input.

[0048] The image sensor is used to capture weld seam image data Img(t) in real time and transmit it to the inspection device. The image sensor can capture weld seam images at different times, generating a set of weld seam images that are then transmitted to the inspection device as a set of weld data. This allows for the inspection of weld seam surface quality, such as weld seam width and surface cracks. For example, the image sensor includes a camera.

[0049] The acoustic sensor is used to collect acoustic wave data S(t) generated during the welding process and transmit this data to a detection device. The acoustic sensor can collect acoustic wave data at different times, thereby forming a set of acoustic wave data, which is transmitted to the detection device as a set of welding data. The detection device can analyze the acoustic characteristics of the welding process, such as the intensity of the arc and the noise generated during the welding process. For example, the acoustic sensor includes a microphone.

[0050] It should be noted that the aforementioned multiple sets of welding data, in addition to welding current data, welding voltage data, weld zone temperature data, weld seam image data, and acoustic wave data during the welding process, may also include other welding data, as long as it can be used for defect detection. Accordingly, the aforementioned sensors, in addition to current sensors, voltage sensors, temperature sensors, image sensors, and acoustic sensors, may also include other sensors, as long as they are used to collect relevant welding data for defect detection.

[0051] In step S104, multiple sets of welding data are pre-processed to obtain multiple key features.

[0052] In some embodiments, step S104 includes: performing noise reduction processing, normalization processing, and feature extraction processing on multiple sets of welding data to obtain multiple key features.

[0053] For example, in the process of denoising raw data, a Kalman filter or wavelet transform can be used to reduce or eliminate noise in the data and improve the purity of the data. For example, a Kalman filter can remove noise collected by a sensor and retain important signal components.

[0054] Data from different sensors has different dimensions and ranges. Therefore, normalization (also known as standardization) can unify the data format and facilitate subsequent analysis. Normalization can be performed using known methods or techniques. For example, the raw data can be normalized based on its mean and variance.

[0055] Next, feature extraction can be performed on the normalized data to obtain key features.

[0056] For example, the peak value, frequency, amplitude, mean and variance of the welding current can be extracted from the welding current data, and these data can be combined into the key feature vector of the welding current.

[0057] For another example, the peak value, frequency, amplitude, mean value, and variance of the welding voltage can be extracted from the welding voltage data, and these data can be combined into a key feature vector of the welding voltage.

[0058] For example, we can extract temperature gradients and peak temperatures from temperature data and combine these data into key feature vectors for the temperature data. The extracted features can then be used for subsequent knowledge graph analysis and detection model optimization.

[0059] For example, the edge features of welds are extracted from image data.

[0060] Another example is extracting spectral features from acoustic wave data. For example, low-frequency components between 20 Hz and 200 Hz can be collected. Low-frequency noise is typically associated with the stability of the welding arc, mechanical fluctuations during the welding process, and the dynamic behavior of the weld pool. Low-frequency components can reveal instabilities or anomalies in the welding process, such as arc fluctuations and weld pool instability. Arc fluctuation frequency can also be extracted. Changes in the welding arc can generate low-frequency noise, especially when the arc is unstable or a short circuit occurs. By analyzing these low-frequency components, arc stability can be determined and potential welding defects can be predicted.

[0061] In the above embodiment, by preprocessing the multiple sets of collected data, the data can be made to have higher quality and consistency, which facilitates the subsequent use of the data for detection and optimization.

[0062] In step S106, the multiple key features are input into the detection model, so that the detection model determines whether a welding defect exists using a knowledge graph. The knowledge graph includes multiple nodes and relationships between the multiple nodes. The multiple nodes include welding process parameter nodes and welding defect nodes, and the welding process parameter nodes correspond to the welding defect nodes.

[0063] The above-mentioned detection model can adopt a detection model known to those skilled in the art. For example, the detection model can be a machine learning model. The detection model can be trained in advance using collected data and knowledge graphs. The detection model performs defect identification, judgment and decision-making in a data-driven manner, which relies on real-time collected data and trained algorithms. The knowledge graph provides a structured knowledge network of welding processes and defects, and helps the detection model understand and reason about possible problems in the welding process by organizing and storing the relationship between different welding parameters and defects. In addition, in the following description, the output results of the detection model (such as defect identification results) can also reversely affect the update of the knowledge graph. The adaptive learning mechanism enables the detection model to update its parameters based on real-time feedback, and the knowledge graph can also be gradually optimized as new information is added in the process. In this way, the two jointly improve the detection capability of the system through continuous feedback and updates.

[0064] The plurality of nodes may include a plurality of welding process parameter nodes. For example, a welding process parameter node may be a node with welding current data as a feature vector, or a node with welding voltage data as a feature vector, or a node with temperature data as a feature vector, etc.

[0065] For example, the above-mentioned welding defect nodes can be defect nodes such as cracks, pores, and lack of fusion.

[0066] The association between each welding process parameter node and the corresponding welding defect node is expressed in the form of a weight value. For example, a welding current parameter node with a weight of 0.3 means that the welding current parameter node has a weight of 0.3 relative to the corresponding defect node.

[0067] In the above steps, by inputting multiple key features into the detection model, the detection model can perform defect recognition through the knowledge graph to determine whether there are welding defects.

[0068] In some embodiments, the multiple nodes in the knowledge graph also include a quality standard node. The quality standard node includes characteristics of welding process parameters when the welding quality meets the quality requirements. This makes it easier for the detection model to determine whether there are welding defects based on the knowledge graph.

[0069] For example, if multiple key features meet the characteristics of the corresponding quality standard node, then there is no welding defect. Taking current as an example, the quality standard node can include the allowable range of the current mean (as a feature), so that it can be detected whether the mean welding current is within the allowable range of the current mean.

[0070] In step S108 , when it is determined that a welding defect exists, a recognition result that a welding defect exists is output.

[0071] In addition, in the case where it is determined that there is no welding defect, a recognition result that there is no welding defect is output.

[0072] In step S110, in response to the welding defect in the recognition result belonging to a new welding defect type, a new node is added to the knowledge graph, and the new node includes a plurality of new welding process parameter nodes and a new welding defect node corresponding to a plurality of key features.

[0073] For example, if an operator determines that a welding defect in the output recognition result belongs to a new welding defect type, that is, such a welding defect does not exist in the knowledge graph, the operator will input information that the welding defect belongs to the new welding defect type into the detection device. For example, the detection device may display information for the operator to determine whether the welding defect belongs to the new welding defect type. After the operator triggers the "Yes" confirmation button, the detection device will determine that the welding defect in the knowledge recognition result belongs to the new welding defect type.

[0074] In response to identifying a welding defect as belonging to a new welding defect type in the identification result, the detection device adds a new node to the knowledge graph. Here, the new nodes include multiple new welding process parameter nodes and new welding defect nodes. The new welding process parameter nodes can correspond to key features. For example, the current, voltage, and temperature features of the multiple key features can be used as new welding process parameter nodes.

[0075] In step S112, association relationships between multiple new welding process parameter nodes and new welding defect nodes are established based on multiple sets of welding data.

[0076] In some embodiments, step S112 includes: determining a weight value of each of a plurality of new welding process parameter nodes relative to a new welding defect node based on multiple sets of welding data.

[0077] As previously mentioned, the association relationship is expressed in the form of a weight value. Therefore, here, based on multiple sets of welding data, the weight value of each new welding process parameter node relative to the new welding defect node can be determined, that is, the association relationship between each new welding process parameter node and the new welding defect node can be obtained.

[0078] For example, based on the welding current data, welding voltage data and temperature data in multiple sets of welding data, it is found that the change of welding current data has the greatest impact on defects, the welding voltage data has the second largest impact on defects, and the temperature data has the smallest impact on defects. Therefore, the weight value of the welding current parameter node can be set to the largest, the weight value of the welding voltage parameter node can be set to the second largest, and the weight value of the temperature parameter node can be set to the smallest.

[0079] Thus, a method for detecting welding quality according to some embodiments of the present disclosure is provided. The detection method includes: receiving multiple sets of welding data collected by multiple sensors during a welding process; preprocessing the multiple sets of welding data to obtain multiple key features; inputting the multiple key features into a detection model so that the detection model determines whether a welding defect exists through a knowledge graph, wherein the knowledge graph includes multiple nodes and associations between the multiple nodes, the multiple nodes including welding process parameter nodes and welding defect nodes, and the welding process parameter nodes correspond to the welding defect nodes; if a welding defect is determined to exist, outputting an identification result indicating the presence of a welding defect; in response to the welding defect in the identification result belonging to a new welding defect type, adding a new node to the knowledge graph, the new node including multiple new welding process parameter nodes corresponding to the multiple key features and a new welding defect node; and establishing associations between the multiple new welding process parameter nodes and the new welding defect nodes based on the multiple sets of welding data. In this detection method, the presence of a welding defect can be detected based on the collected data and the knowledge graph, and if the welding defect belongs to a new welding defect type, the knowledge graph can be updated, thereby adapting to changes in different welding conditions, improving the accuracy of welding defect detection, and improving product reliability.

[0080] Traditional systems lack dynamic update capabilities and are unable to adapt to the advancements and sophistication of welding technology, resulting in increased missed detections and misjudgments. When faced with complex or irregular welds, traditional systems have a high misjudgment rate and limited accuracy of approximately 80% in identifying small cracks or pores. Furthermore, their adaptability is poor, making it difficult to meet high-precision inspection requirements.

[0081] In view of this, the detection method of some embodiments of the present disclosure further includes: comparing each key feature of the multiple key features with the features of each welding process parameter node in the knowledge graph to obtain the welding process parameter node closest to each key feature; calculating the error feature between each key feature and the feature of the closest welding process parameter node; and adjusting the parameters of the detection model using a gradient descent method based on the error feature to minimize the error feature. This implements an adaptive learning mechanism.

[0082] Here, the gradient descent method is an algorithm known to those skilled in the art and will not be described in detail here. The parameters of the detection model are also parameters known to those skilled in the art, for example, the parameters may include weight values ​​in the model, etc.

[0083] Through the above-mentioned adaptive learning mechanism, the accuracy, efficiency and system adaptability of welding quality detection can be improved.

[0084] In some embodiments, the above-described detection method further includes constructing a knowledge graph before receiving the multiple sets of welding data. Specifically, before executing the detection method shown in FIG1 , a knowledge graph is constructed. A knowledge graph is a complex network structure that represents various process parameters, quality standards, and defect types in the welding process, as well as the relationships between them. By representing multiple parameters and their relationships in a knowledge graph, the complexity and variability of the welding process can be better reflected, improving the adaptability of detection.

[0085] Optionally, constructing a knowledge graph may include defining multiple nodes in the knowledge graph, wherein the multiple nodes include multiple welding process parameter nodes and welding defect nodes corresponding to the multiple welding process parameter nodes. That is, various nodes are defined in the knowledge graph, including welding process parameter nodes (e.g., current, voltage, temperature) and welding defect nodes (e.g., cracks, pores, lack of fusion, etc.). Furthermore, the aforementioned quality standard node may also be defined in the knowledge graph.

[0086] Optionally, constructing the knowledge graph may also include establishing relationships between multiple nodes using historical data and empirical knowledge. This relationship includes a weight value for each of the multiple welding process parameter nodes relative to the welding defect node. In other words, the relationship between the nodes is established using historical data and the empirical knowledge of welding experts. For example, the causal relationship between current fluctuations and crack formation can be determined using statistical analysis methods (such as association rule mining), and the strength of each relationship in the knowledge graph is represented by a weight value.

[0087] This enables the construction of a knowledge graph. For example, this knowledge graph can be constructed during system initialization. This facilitates subsequent defect detection.

[0088] FIG2 is a flow chart illustrating a method for detecting welding quality according to other embodiments of the present disclosure. The detection method can be implemented by a detection device, for example, the detection device is provided in an electronic device. As shown in FIG2 , the detection method includes steps S202 to S208.

[0089] In step S202, multiple sets of welding data collected by multiple sensors during the welding process are received.

[0090] For example, multiple sets of welding data include: welding current data, welding voltage data, temperature data of the weld zone, weld seam image data, and acoustic wave data during the welding process. During the welding process, multiple sensors are used to synchronously collect various types of data to comprehensively monitor the welding quality.

[0091] In step S204, the nodes in the knowledge graph are updated.

[0092] As mentioned above, multiple key features can be obtained through multiple sets of welding data. After obtaining multiple key features, welding defects are identified through the detection model based on the original knowledge graph. When the system detects that the welding defect belongs to a new welding defect type or the process parameters have changed, it automatically adds a corresponding node to the knowledge graph and establishes a connection with the existing nodes in the knowledge graph (i.e., the original nodes). For example, when the system detects a new defect pattern (such as a crack type caused by a new welding material), it automatically adds a node for that pattern.

[0093] Here, the newly added nodes can represent new welding defect types or changed process parameters, while the existing nodes are the original welding defect types or original process parameters.

[0094] It should be noted that the newly added nodes do not necessarily have to be linked to all existing nodes. For example, they can be linked to existing nodes with significant correlation or similarity. Such links usually depend on factors such as actual data, defect manifestations, and process changes in the welding process. If the newly added node represents a new type of welding defect (such as "porosity crack"), then the newly added node may be associated with some specific process parameter nodes such as welding current and welding temperature. The newly added node may not be linked to existing defect types (such as "cracks" or "incomplete penetration") unless the formation conditions of these defect types are similar to those of the new defect type. If the newly added node is a new welding process parameter (such as the welding speed under a specific process), then the newly added node may be associated with process parameters such as welding current and voltage, and will not be directly linked to the welding defect type, unless this new process parameter plays a decisive role in the occurrence of certain defects.

[0095] In step S206, the relationship between the nodes is updated.

[0096] For example, based on the collected data, the Bayesian update formula is used to dynamically adjust the weights of the nodes in the knowledge graph, thereby adjusting the relationship between the nodes to reflect the latest welding quality status.

[0097] For example, suppose that during the inspection process, it is initially determined that welding current and voltage have a certain influence on crack formation. Therefore, an initial weight (prior probability) is assigned to represent this influence. For example, the initial weight (weight before update) includes: the initial weight (prior probability) of welding current and the initial weight (prior probability) of welding voltage. Assume that the initial weight of welding current is 0.6, indicating that welding current has a greater influence on welding cracks; suppose that the initial weight of welding voltage is 0.4, indicating that voltage has a relatively small influence on welding cracks.

[0098] Suppose new data is obtained and it is detected that the probability of weld cracks increases significantly when the welding current changes significantly, while the voltage change has a smaller impact on cracks. Therefore, the Bayesian update formula can be used to adjust these two weights. This allows the system to adjust the weights in a timely manner when faced with new welding defect types or changing process parameters, thereby improving detection accuracy.

[0099] In this way, through the above steps S202 to S206, based on the dynamic update mechanism, the knowledge graph can be continuously adjusted and optimized according to real-time data and feedback information to cope with changes in welding environment and process conditions, and improve the accuracy and adaptability of detection.

[0100] In step S208 , adaptive learning is performed to optimize the detection model.

[0101] For example, during the detection process, the system compares the features of the collected data and calculates the error between the model output and the expected value in the knowledge graph. Based on the error calculation results, the system uses the gradient descent method to adjust the parameters of the detection model to minimize the error. The system uses a feedback loop mechanism to compare the actual detection results with the expected results in the knowledge graph. If the deviation is large, the system will further optimize the model parameters. For example, to improve learning efficiency, the system introduces reinforcement learning algorithms (for example, Q learning algorithms) to further optimize the detection model.

[0102] The adaptive learning mechanism dynamically adjusts and optimizes the detection model through real-time data feedback, enabling the system to cope with complex and changing welding conditions with greater flexibility and adaptability. This improves detection accuracy and reduces misjudgments and missed detections. In this adaptive learning mechanism, the system optimizes itself without human intervention. This real-time feedback and dynamic adjustment enhance the system's intelligence.

[0103] Thus, a welding quality detection method according to further embodiments of the present disclosure has been provided. This detection method not only dynamically updates the welding quality knowledge graph but also automatically optimizes the detection model under different working conditions through an adaptive learning mechanism, significantly improving the accuracy, efficiency, and adaptability of the system.

[0104] Furthermore, the optimized detection model can also be output.

[0105] In some embodiments, the above-described inspection method may further include generating a welding quality report to record every adjustment and defect detection result during the welding process. The welding quality report may include detection data of internal weld defects and recommended repairs, making it easier for operators to take appropriate measures based on the welding quality report.

[0106] In some embodiments, the above detection method may further include: if a structural defect (e.g., a serious structural defect) is detected, an alarm is issued and an immediate repair or re-welding recommendation is made. This facilitates the operator to be informed of the defect and improve product quality through repair or re-welding.

[0107] The following describes the inspection process by taking the welding quality control of the main arm of an aerial platform vehicle as an example.

[0108] The boom is a critical component in the manufacturing process of aerial work platforms, and its welding quality directly impacts the platform's performance and safety. The boom is typically welded from high-strength steel plates, requiring extremely high weld strength and quality. Traditional weld quality inspection relies primarily on ultrasonic and X-ray testing, but these methods suffer from low efficiency, high costs, and blind spots, making them incapable of promptly detecting defects in production.

[0109] Here, the welding quality is detected using the detection method of the embodiment of the present disclosure, and the detection process includes the following steps.

[0110] In step 1, multi-source data is collected.

[0111] During the main arm welding process, multiple sensors are installed to monitor welding parameters in real time. A current sensor monitors the welding current, mitigating weld quality issues caused by excessive or insufficient current. A voltage sensor monitors the welding voltage, improving the stability of the welding process. A temperature sensor monitors temperature changes in the weld zone, minimizing weld defects caused by overheating or improper cooling. A camera captures images of the weld in real time, detecting surface defects such as cracks and lack of fusion. A microphone collects sound waves generated during welding to identify potential anomalies.

[0112] In step 2, data preprocessing and feature extraction are performed.

[0113] All collected data undergoes preprocessing, including noise reduction, normalization, and feature extraction. For example, edge features are extracted from weld images, and spectral features are extracted from acoustic wave data. This high-quality feature data is then fed into a knowledge graph for analyzing and assessing weld quality.

[0114] In step 3, analysis and real-time detection are performed through the knowledge graph.

[0115] The system uses a constructed knowledge graph to analyze the correlations between various parameters. For example, current fluctuations may be associated with weld crack formation, while temperature variations may be related to lack of weld fusion. By combining historical and real-time data, the knowledge graph automatically identifies the types of defects that may occur during the welding process.

[0116] In step 4, adaptive learning and parameter optimization are performed.

[0117] When the system detects anomalies during welding (such as current fluctuations or abnormal temperatures), it uses an adaptive learning mechanism to adjust welding parameters, such as adjusting the welding current or changing the welding speed, to ensure that the weld quality meets the requirements. Through this feedback mechanism, the system continuously optimizes the knowledge graph, making subsequent welding detection more accurate.

[0118] In step 5, the detection result is obtained and fed back.

[0119] After welding is complete, the system generates an inspection report that includes a quality assessment of the weld, the types and locations of defects found, and the welding parameters adjusted by the system. The operator can use this report to further adjust the welding process. For detected defects, the system provides an automatic alarm, prompting the operator to make repairs.

[0120] The following describes the inspection process using the quality control in the welding of construction crane chassis as an example.

[0121] The chassis of a construction crane, as the crane's foundational structure, bears the entire crane's weight and creates concentrated stress during operation. Therefore, the quality of its welding is crucial to the overall stability of the crane. Chassis welding is a complex process, prone to quality issues such as incomplete weld penetration, undercuts, and porosity. If not promptly detected and corrected, this can lead to insufficient chassis structural strength, impacting the equipment's service life and safety.

[0122] Here, the welding quality is detected using the detection method of the embodiment of the present disclosure, and the detection process includes the following steps.

[0123] In step 1, multi-source data is collected.

[0124] During the chassis welding process, current, voltage, and temperature sensors are deployed, along with a welding monitoring system to collect welding parameters and weld seam images in real time. The system also uses an ultrasonic detector to monitor weld quality, specifically detecting internal defects such as incomplete penetration and porosity.

[0125] In step 2, data preprocessing and feature extraction are performed.

[0126] The collected data is pre-processed to eliminate noise interference and normalize it. The ultrasonic data is then processed to extract the characteristics of internal defects in the weld, such as the intensity and frequency of the echo signal.

[0127] The geometric features of the weld surface (e.g., undercut depth, weld width, etc.) are extracted from the weld image, and the heat input fluctuation information of the welding process is extracted from the current and temperature data.

[0128] In step 3, the knowledge graph is constructed and applied.

[0129] A welding quality knowledge graph for construction crane chassis was constructed. This knowledge graph includes welding process parameters, weld defect types, and quality standards. The knowledge graph records the possible defect types and corresponding countermeasures for various welding parameter combinations.

[0130] The system feeds real-time data into a detection model, which uses the knowledge graph to determine weld quality. If porosity is detected within the weld, the system suggests that the welding temperature is too low, and recommends adjusting the temperature.

[0131] In step 4, adaptive learning and optimization are performed.

[0132] The system uses an adaptive learning mechanism to adjust welding process parameters based on actual weld defects detected. For example, if a lack of penetration defect is detected, the system will adjust the welding speed and current intensity to increase the weld penetration depth.

[0133] Dynamically updating the knowledge graph enables the system to adapt to materials of different thicknesses and different welding processes.

[0134] In step 5, the detection result is obtained and fed back.

[0135] The system outputs a detailed welding quality report that records every adjustment and defect detection result during the welding process. The report includes the detection data of internal defects in the weld and the repair suggestions.

[0136] If a serious structural defect is detected, the system can sound an alarm and recommend immediate repair or re-welding.

[0137] In some cases, welding quality inspection technology is developing towards intelligence, automation, and adaptation to overcome the limitations of traditional methods. First, multi-source data fusion utilizes data from multiple sensors such as current, voltage, temperature, image, and sound waves to comprehensively reflect the dynamic changes in the welding process and improve the ability to identify different welding defects. Second, knowledge graph technology forms a complex knowledge network by combining welding processes, parameters, defect types, and inspection results, enhancing the intelligence level of the inspection system. Furthermore, the adaptive learning mechanism can adjust the parameters and strategies of the inspection model in real time according to changes in welding conditions, improving the adaptability and accuracy of the system under different working conditions. It also gradually optimizes the model through real-time feedback and improves the ability to identify new defect types.

[0138] In the embodiments of this disclosure, a knowledge graph-based method for dynamic welding quality updates and adaptive learning is proposed. By collecting and analyzing welding data in real time, the knowledge graph is dynamically updated, and an adaptive learning mechanism is introduced to improve the accuracy, efficiency, and system adaptability of welding quality inspection. This innovative method not only adapts to complex and changing welding conditions but also allows for the continuous updating and optimization of inspection models, significantly improving the overall level of welding quality inspection.

[0139] FIG3 is a block diagram schematically illustrating a device for detecting welding quality according to some embodiments of the present disclosure. As shown in FIG3 , the device includes a receiving module 302 , a pre-processing module 304 , an identification module 306 , a node setting module 308 , and a relationship determination module 310 .

[0140] The receiving module 302 is used to receive multiple sets of welding data collected by multiple sensors during the welding process.

[0141] For example, the multiple sets of welding data include at least two sets of welding current data, welding voltage data, temperature data of the welding zone, weld seam image data, and sound wave data during the welding process.

[0142] The pre-processing module 304 is used to pre-process multiple sets of welding data to obtain multiple key features.

[0143] For example, the preprocessing module is used to perform noise reduction, normalization and feature extraction on multiple sets of welding data to obtain multiple key features.

[0144] Identification module 306 is used to input multiple key features into the detection model, so that the detection model determines whether a welding defect exists through the knowledge graph. If a welding defect is determined to exist, the detection model outputs an identification result indicating the presence of the welding defect. The knowledge graph includes multiple nodes and associations between the multiple nodes. The multiple nodes include welding process parameter nodes and welding defect nodes, and the welding process parameter nodes correspond to the welding defect nodes.

[0145] In some embodiments, the multiple nodes in the knowledge graph further include a quality standard node, which includes features of welding process parameters when the welding quality meets quality requirements.

[0146] The node setting module 308 is used to add new nodes to the knowledge graph in response to the welding defect in the recognition result belonging to a new welding defect type. The new nodes include multiple new welding process parameter nodes and new welding defect nodes corresponding to multiple key features.

[0147] The relationship determination module 310 is used to determine the association relationship between multiple new welding process parameter nodes and new welding defect nodes based on multiple sets of welding data.

[0148] For example, the relationship determination module 310 is used to determine the weight value of each new welding process parameter node of multiple new welding process parameter nodes relative to the new welding defect node based on multiple sets of welding data.

[0149] Thus, a welding quality detection device according to some embodiments of the present disclosure is provided. The detection device includes a receiving module, a preprocessing module, an identification module, a node setting module, and a relationship determination module. This detection device can detect the presence of welding defects based on collected data and a knowledge graph. Furthermore, if a welding defect belongs to a new welding defect type, the knowledge graph can be updated. This allows the device to adapt to changes in welding conditions, improve the accuracy of welding defect detection, and enhance product reliability.

[0150] FIG4 is a block diagram schematically illustrating a structure of a device for detecting welding quality according to other embodiments of the present disclosure.

[0151] As shown in FIG. 4 , the detection device includes a receiving module 302 , a pre-processing module 304 , an identification module 306 , a node setting module 308 and a relationship determination module 310 .

[0152] In some embodiments, as shown in FIG4 , the detection device further includes an adaptive learning module 412. The adaptive learning module 412 is configured to compare each key feature of the multiple key features with the features of each welding process parameter node in the knowledge graph to obtain the welding process parameter node closest to each key feature, calculate the error feature between each key feature and the feature of the closest welding process parameter node, and adjust the parameters of the detection model by a gradient descent method based on the error feature to minimize the error feature.

[0153] In some embodiments, as shown in FIG4 , the detection device further includes a graph construction module 414. The graph construction module 414 is used to define multiple nodes in a knowledge graph and establish association relationships between the multiple nodes using historical data and empirical knowledge, wherein the multiple nodes include multiple welding process parameter nodes and welding defect nodes corresponding to the multiple welding process parameter nodes, and the association relationships include weight values ​​of each of the multiple welding process parameter nodes relative to the welding defect node.

[0154] FIG5 is a block diagram schematically illustrating a device for detecting welding quality according to other embodiments of the present disclosure. The device includes a memory 510 and a processor 520.

[0155] The memory 510 may be a disk, a flash memory, or any other non-volatile storage medium. The memory is used to store the instructions in the embodiments corresponding to FIG. 1 and / or FIG. 2 .

[0156] The processor 520 is coupled to the memory 510 and can be implemented as one or more integrated circuits, such as a microprocessor or a microcontroller. The processor 520 is used to execute instructions stored in the memory to improve the accuracy of welding defect detection and improve product reliability.

[0157] In some embodiments, as shown in FIG6 , the detection device 600 may include a memory 610 and a processor 620. The processor 620 is coupled to the memory 610 via a BUS 630. The detection device 600 may also be connected to an external storage device 650 via a storage interface 640 to access external data, and may also be connected to a network or another computer system (not shown) via a network interface 660, which will not be described in detail here.

[0158] In this embodiment, data instructions are stored in a memory and then processed by a processor, thereby improving the accuracy of welding defect detection and enhancing product reliability.

[0159] FIG7 is a block diagram schematically illustrating a structure of a system for detecting welding quality according to some embodiments of the present disclosure.

[0160] As shown in FIG7 , the detection system includes a current sensor 710, a voltage sensor 720, a temperature sensor 730, an image sensor 740, a sound sensor 750, and a detection device 760. The current sensor 710, the voltage sensor 720, the temperature sensor 730, the image sensor 740, and the sound sensor 750 are all electrically connected to the detection device 760. For example, the detection device 760 may be the detection device shown in FIG3 , FIG4 , FIG5 , or FIG6 .

[0161] The embodiments of the present disclosure significantly improve the accuracy, adaptability and efficiency of welding quality detection by constructing a welding quality detection device or detection system based on a knowledge graph, and combining advanced technologies such as multi-source data fusion, dynamic update and adaptive learning.

[0162] First, multi-source data fusion technology enables the system to comprehensively analyze multi-dimensional data such as current, voltage, temperature, images, and sound waves during the welding process, thereby more comprehensively and accurately identifying welding defects. Compared with traditional detection methods that rely on a single data source, the methods, devices, or systems of the disclosed embodiments have improved the accuracy of welding defect identification to 90%-95% in industrial testing, exceeding the 70%-85% of traditional methods.

[0163] Secondly, the introduction of a dynamic knowledge graph enables the system to update and optimize the inspection model in real time. By continuously learning from new data, the device or system automatically adds or adjusts nodes and relationships in the graph to ensure it always reflects the latest welding quality standards and defect patterns. This dynamic update capability ensures high adaptability to different welding conditions and can reduce inspection errors by 25%-30%.

[0164] In addition, the adaptive learning mechanism further enhances the automation and intelligence level of the device or system. The device or system can automatically adjust the detection model parameters based on real-time feedback, reducing the need for manual intervention. This not only improves detection efficiency (for example, the efficiency is increased by about 30%-40%), but also reduces the false positive and missed detection rates (for example, the false positive rate is reduced by 20%). The device or system can adapt to a variety of complex welding conditions, such as welding operations under different materials, processes and environments, improving the efficiency and reliability of quality inspection.

[0165] Furthermore, the continuous optimization capabilities and high scalability of the disclosed embodiments enable the device or system to continuously improve performance over long-term applications, adapt to changes in new materials and processes, and form a highly intelligent and automated welding quality inspection platform. In summary, the methods, devices, and / or systems of the disclosed embodiments have brought significant technological advancements in the field of welding quality inspection and have broad industrial application prospects.

[0166] In some embodiments, the present disclosure further provides a computer-readable storage medium (e.g., a non-transitory computer-readable storage medium) having computer program instructions stored thereon, which, when executed by a processor, implement the steps of the method in the embodiment corresponding to Figure 1 and / or Figure 2. It should be understood by those skilled in the art that the embodiments of the present disclosure can be provided as methods, devices, or computer program products. Therefore, the present disclosure can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present disclosure can take the form of a computer program product implemented on one or more computer-usable non-transitory storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0167] The present disclosure is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems) and computer program products according to the embodiments of the present disclosure. It should be understood that each process and / or box in the flowchart and / or block diagram and the combination of the processes and / or boxes in the flowchart and / or block diagram can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate a device for implementing the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram.

[0168] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a product including an instruction device that implements the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram.

[0169] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram.

[0170] In some embodiments of the present disclosure, a computer program product is further provided. The computer program product includes a computer program or instructions. When the computer program or instructions are executed by a processor, the detection method as described above is implemented.

[0171] In some embodiments of the present disclosure, a computer program is further provided, comprising: instructions, which, when executed by a processor, cause the processor to perform the detection method as described above.

[0172] The present disclosure has been described in detail so far. To avoid obscuring the concept of the present disclosure, some details known in the art have not been described. Based on the above description, those skilled in the art can fully understand how to implement the technical solutions disclosed herein.

[0173] Although some specific embodiments of the present disclosure have been described in detail by way of examples, those skilled in the art will appreciate that the above examples are for illustrative purposes only and are not intended to limit the scope of the present disclosure. Those skilled in the art will appreciate that modifications may be made to the above embodiments without departing from the scope and spirit of the present disclosure. The scope of the present disclosure is defined by the appended claims.

Claims

1. A method for detecting welding quality, comprising: Receive multiple sets of welding data collected by multiple sensors during the welding process; Preprocessing the multiple sets of welding data to obtain multiple key features; Inputting the multiple key features into a detection model so that the detection model determines whether a welding defect exists through a knowledge graph, wherein the knowledge graph includes multiple nodes and association relationships between the multiple nodes, the multiple nodes include welding process parameter nodes and welding defect nodes, and the welding process parameter nodes correspond to the welding defect nodes; If it is determined that a welding defect exists, outputting an identification result indicating the presence of the welding defect; In response to the welding defect in the recognition result belonging to a new welding defect type, adding new nodes to the knowledge graph, the new nodes including a plurality of new welding process parameter nodes and a new welding defect node corresponding to the plurality of key features; and Based on the multiple sets of welding data, an association relationship is established between the multiple new welding process parameter nodes and the new welding defect nodes.

2. The detection method according to claim 1, wherein Establishing, based on the multiple sets of welding data, association relationships between the multiple new welding process parameter nodes and the new welding defect nodes, including: A weight value of each of the multiple new welding process parameter nodes relative to the new welding defect node is determined based on the multiple sets of welding data.

3. The detection method according to claim 1 or 2, further comprising: Comparing each key feature of the multiple key features with the features of each welding process parameter node in the knowledge graph to obtain a welding process parameter node closest to each key feature; Calculating an error characteristic between each key characteristic and a characteristic of the closest welding process parameter node; and Based on the error characteristics, the parameters of the detection model are adjusted by gradient descent method to minimize the error characteristics.

4. The detection method according to any one of claims 1 to 3, wherein The multiple nodes in the knowledge graph also include a quality standard node, which includes characteristics of welding process parameters when the welding quality meets the quality requirements.

5. The detection method according to any one of claims 1 to 4, wherein The plurality of welding data groups include at least two groups of welding current data, welding voltage data, temperature data of the welding zone, weld seam image data, and sound wave data during the welding process.

6. The detection method according to any one of claims 1 to 5, wherein Preprocessing the multiple sets of welding data includes: The multiple sets of welding data are subjected to noise reduction processing, normalization processing and feature extraction processing to obtain the multiple key features.

7. The detection method according to any one of claims 1 to 6, further comprising: Before receiving the plurality of sets of welding data, constructing the knowledge graph; The construction of the knowledge graph includes: defining the plurality of nodes in the knowledge graph, wherein the plurality of nodes include a plurality of welding process parameter nodes and welding defect nodes corresponding to the plurality of welding process parameter nodes; An association relationship between the multiple nodes is established by using historical data and empirical knowledge, wherein the association relationship includes a weight value of each welding process parameter node in the multiple welding process parameter nodes relative to the welding defect node.

8. A device for detecting welding quality, comprising: A receiving module, used for receiving multiple sets of welding data collected by multiple sensors during the welding process; A preprocessing module, configured to preprocess the plurality of welding data sets to obtain a plurality of key features; an identification module, configured to input the plurality of key features into a detection model so that the detection model determines whether a welding defect exists through a knowledge graph, and output an identification result indicating the presence of the welding defect if the welding defect is determined to exist, wherein the knowledge graph includes a plurality of nodes and associations between the plurality of nodes, the plurality of nodes including welding process parameter nodes and welding defect nodes, the welding process parameter nodes corresponding to the welding defect nodes; a node setting module, configured to add new nodes to the knowledge graph in response to the welding defect in the recognition result belonging to a new welding defect type, the new nodes including a plurality of new welding process parameter nodes and a new welding defect node corresponding to the plurality of key features; and The relationship determination module is used to determine the association relationship between the multiple new welding process parameter nodes and the new welding defect nodes based on the multiple sets of welding data.

9. The detection device according to claim 8, wherein: The relationship determination module is used to determine a weight value of each of the multiple new welding process parameter nodes relative to the new welding defect node based on the multiple sets of welding data.

10. The detection device according to claim 8 or 9, further comprising: An adaptive learning module is used to compare each key feature of the multiple key features with the features of each welding process parameter node in the knowledge graph to obtain a welding process parameter node closest to each key feature, calculate an error feature between each key feature and the features of the closest welding process parameter node, and adjust the parameters of the detection model by a gradient descent method based on the error feature to minimize the error feature.

11. The detection device according to any one of claims 8 to 10, wherein: The multiple nodes in the knowledge graph also include a quality standard node, which includes characteristics of welding process parameters when the welding quality meets the quality requirements.

12. The detection device according to any one of claims 8 to 11, wherein: The plurality of welding data groups include at least two groups of welding current data, welding voltage data, temperature data of the welding zone, weld seam image data, and sound wave data during the welding process.

13. The detection device according to any one of claims 8 to 12, wherein: The preprocessing module is used to perform noise reduction, normalization and feature extraction on the multiple sets of welding data to obtain the multiple key features.

14. The detection device according to any one of claims 8 to 13, further comprising: A graph construction module is used to define the multiple nodes in the knowledge graph and use historical data and empirical knowledge to establish an association relationship between the multiple nodes, wherein the multiple nodes include multiple welding process parameter nodes and welding defect nodes corresponding to the multiple welding process parameter nodes, and the association relationship includes a weight value of each welding process parameter node in the multiple welding process parameter nodes relative to the welding defect node.

15. A device for detecting welding quality, comprising: Memory; as well as A processor coupled to the memory, wherein the processor is configured to execute the detection method according to any one of claims 1 to 7 based on instructions stored in the memory.

16. A welding quality detection system comprising: A current sensor, a voltage sensor, a temperature sensor, an image sensor, a sound sensor, and a detection device according to any one of claims 8 to 15, wherein the current sensor, the voltage sensor, the temperature sensor, the image sensor, and the sound sensor are all electrically connected to the detection device.

17. A computer-readable storage medium having computer instructions stored thereon, wherein the computer instructions, when executed by a processor, implement the detection method according to any one of claims 1 to 7.

18. A computer program comprising: Instructions, when executed by a processor, cause the processor to perform the detection method according to any one of claims 1 to 7.

Images