Automobile battery welding monitoring method and device, computer equipment and storage medium

Abstract

The embodiment of the invention belongs to the field of welding monitoring, and relates to an automobile battery welding monitoring method which comprises the following steps: when an automobile battery is welded, acquiring optical monitoring data, temperature monitoring data and plasma monitoring data of a welding area, and respectively carrying out time sequence feature extraction and spatial feature extraction; first optical data, first temperature data and first plasma data are generated according to the extracted features; performing interactive fusion on the first optical data, the first temperature data and the first plasma data to obtain second optical data, second temperature data and second plasma data; acquiring a defect type identifier, and adding weights to the second optical data, the second temperature data and the second plasma data according to the defect type identifier; and inputting the second optical data, the second temperature data and the second plasma data with the weights into a defect detection network corresponding to the defect type identifier to obtain a welding defect detection result. According to the invention, the accuracy of automobile battery welding monitoring is improved.

Description

Technical Field

[0001] The present application relates to the technical field of welding monitoring, and particularly to a method, device, computer device and storage medium for monitoring the welding of automotive batteries. Background Art

[0002] With the rapid development of new energy vehicles, the production of automotive power batteries has become a key link. The welding quality of the battery directly affects the safety, performance and service life of the battery. Therefore, the monitoring of the welding process is crucial. At present, the welding monitoring technology of automotive power batteries processes the data of various sensors independently, ignoring the correlation between different sensors, and cannot effectively fuse the data of each sensor, which affects the accurate identification of complex welding defects (such as pores, cracks, false soldering, etc.), and the false detection rate is relatively high. Summary of the Invention

[0003] The purpose of the embodiments of the present application is to propose a method, device, computer device and storage medium for monitoring the welding of automotive batteries to solve the problem of low accuracy in monitoring the welding of automotive batteries.

[0004] To solve the above technical problems, the embodiments of the present application provide a method for monitoring the welding of automotive batteries, and adopt the following technical solutions:

[0005] When welding an automotive battery, obtain optical monitoring data, temperature monitoring data and plasma monitoring data of the welding area;

[0006] Respectively perform time series feature extraction and spatial feature extraction on the optical monitoring data, the temperature monitoring data and the plasma monitoring data, and generate first optical data, first temperature data and first plasma data according to the extracted features;

[0007] Perform interactive fusion on the first optical data, the first temperature data and the first plasma data to obtain second optical data, second temperature data and second plasma data;

[0008] Obtain a defect type identifier, and add weights to the second optical data, the second temperature data and the second plasma data according to the defect type identifier;

[0009] Input the second optical data, the second temperature data and the second plasma data with weights into a defect detection network corresponding to the defect type identifier to obtain a welding defect detection result.

[0010] Further, the step of obtaining optical monitoring data, temperature monitoring data and plasma monitoring data of the welding area includes:

[0011] Collect the molten pool morphology and weld seam morphology of the welding area through an optical sensor to obtain optical monitoring data;

[0012] Collect the molten pool temperature field distribution of the welding area through a temperature sensor to obtain temperature monitoring data;

[0013] Collect the radiation characteristics of the plasma in the welding area through a plasma sensor to obtain plasma monitoring data.

[0014] Furthermore, the steps of respectively performing temporal feature extraction and spatial feature extraction on the optical monitoring data, the temperature monitoring data, and the plasma monitoring data, and generating first optical data, first temperature data, and first plasma data according to the extracted features include:

[0015] Extract the change information of the image sequence in the optical monitoring data as the optical temporal feature, extract the spatial image feature of the image sequence in the optical monitoring data, and obtain the first optical data according to the optical temporal feature and the spatial image feature;

[0016] Extract the temperature change information and temperature gradient of the temperature monitoring data as the temperature temporal feature, extract the temperature field distribution and heat affected zone from the temperature monitoring data as the temperature spatial feature, and obtain the first temperature data according to the temperature temporal feature and the temperature spatial feature;

[0017] Extract the radiation intensity change information and plasma fluctuation information of the plasma monitoring data as the plasma temporal feature, extract the plasma radiation distribution and plasma spectral feature of the plasma monitoring data as the plasma spatial feature, and obtain the first plasma data according to the plasma temporal feature and the plasma spatial feature.

[0018] Furthermore, the steps of performing interactive fusion on the first optical data, the first temperature data, and the first plasma data to obtain second optical data, second temperature data, and second plasma data include:

[0019] Fuse the first temperature data and the first plasma data into the first optical data through a cross-modal attention mechanism to obtain second optical data;

[0020] Fuse the first optical data and the first plasma data into the first temperature data through the cross-modal attention mechanism to obtain second temperature data;

[0021] Fuse the first optical data and the first temperature data into the first plasma data through the cross-modal attention mechanism to obtain second plasma data.

[0022] Further, the step of fusing the first temperature data and the first plasma data into the first optical data through a cross-modal attention mechanism to obtain the second optical data includes:

[0023] Fusing the first optical data and the first temperature data through a cross-modal attention mechanism to obtain a first optical fusion data;

[0024] Fusing the first optical data and the first plasma data through the cross-modal attention mechanism to obtain a second optical fusion data;

[0025] Connecting the first optical fusion data and the second optical fusion data to obtain the second optical data.

[0026] Further, the step of fusing the first optical data and the first temperature data through a cross-modal attention mechanism to obtain the first optical fusion data includes:

[0027] Performing a linear transformation on the first optical data to obtain an optical query vector;

[0028] Calculating the temperature feature key and temperature feature value of the first temperature data;

[0029] Calculating the attention score between the optical query vector and the temperature feature key;

[0030] Weighting the temperature feature value according to the attention score to obtain weighted temperature data;

[0031] Adding the weighted temperature data and the first optical data to obtain the first optical fusion data.

[0032] Further, the step of adding weights to the second optical data, the second temperature data, and the second plasma data according to the defect type identifier includes:

[0033] Obtaining a weight combination corresponding to the defect type identifier, which is obtained by learning in advance;

[0034] Adding weights to the second optical data, the second temperature data, and the second plasma data according to the weight combination; or,

[0035] Calculating the weights of the second optical data, the second temperature data, and the second plasma data through a self-attention mechanism based on the defect type identifier.

[0036] To solve the above technical problems, an embodiment of the present application further provides an automotive battery welding monitoring device, which adopts the following technical solutions:

[0037] A monitoring and acquisition module, configured to acquire optical monitoring data, temperature monitoring data, and plasma monitoring data of a welding area when welding an automotive battery;

[0038] A feature extraction module, configured to perform temporal feature extraction and spatial feature extraction on the optical monitoring data, the temperature monitoring data, and the plasma monitoring data respectively, and generate first optical data, first temperature data, and first plasma data according to the extracted features;

[0039] An interaction and fusion module, configured to perform interaction and fusion on the first optical data, the first temperature data, and the first plasma data to obtain second optical data, second temperature data, and second plasma data;

[0040] A weight addition module, configured to obtain a defect type identifier, and add weights to the second optical data, the second temperature data, and the second plasma data according to the defect type identifier;

[0041] A defect detection module, configured to input the weighted second optical data, the second temperature data, and the second plasma data into a defect detection network corresponding to the defect type identifier to obtain a welding defect detection result.

[0042] To solve the above technical problems, an embodiment of the present application further provides a computer device, which includes a memory and a processor. Computer-readable instructions are stored in the memory, and when the processor executes the computer-readable instructions, the steps of the above-mentioned automotive battery welding monitoring method are implemented.

[0043] To solve the above technical problems, an embodiment of the present application further provides a computer-readable storage medium, on which computer-readable instructions are stored, and when the computer-readable instructions are executed by a processor, the steps of the above-mentioned automotive battery welding monitoring method are implemented.

[0044] Compared with the prior art, the embodiments of the present application mainly have the following beneficial effects: By extracting the temporal and spatial features of optical monitoring data, temperature monitoring data, and plasma monitoring data, the key changes during the welding process are effectively captured, avoiding the problem that the data of a single sensor cannot comprehensively reflect welding defects; The cross-modal attention mechanism is used to interact and fuse the first optical data, the first temperature data, and the first plasma data, enhancing the correlation between different sensor data and improving the accuracy of defect recognition; Finally, according to the defect type identifier, the weights of the second optical data, the second temperature data, and the second plasma data are dynamically adjusted, making the contribution of each sensor data to the defect detection result more accurate, further reducing the false detection rate, and selecting the corresponding defect detection network for welding defect detection according to the defect type identifier, so that the complex defects during the welding process can be accurately identified, improving the accuracy of automotive battery welding monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the solutions in the present application, the following will briefly introduce the drawings required for the description of the embodiments of the present application. Obviously, the drawings in the following description are some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0046] Figure 1 is an exemplary system architecture diagram to which the present application can be applied;

[0047] Figure 2 is a flowchart of an embodiment of the automotive battery welding monitoring method according to the present application;

[0048] Figure 3 is a schematic structural diagram of an embodiment of the automotive battery welding monitoring device according to the present application;

[0049] Figure 4 is a schematic structural diagram of an embodiment of the computer device according to the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0050] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which this application belongs; The terms used in the description of the present application in this specification are only for the purpose of describing specific embodiments, and are not intended to limit the present application; The terms "including" and "having" and any variations thereof in the specification and claims of this application and the above drawings are intended to cover non-exclusive inclusion. The terms "first", "second", etc. in the specification and claims of this application or the above drawings are used to distinguish different objects, rather than to describe a specific order.

[0051] Reference to "embodiments" in this text means that the specific features, structures, or characteristics described in connection with the embodiments can be included in at least one embodiment of the present application. The phrase appears in various places in the specification and does not necessarily refer to the same embodiment each time, nor is it an independent or alternative embodiment mutually exclusive with other embodiments. Those skilled in the art will explicitly and implicitly understand that the embodiments described herein can be combined with other embodiments.

[0052] To enable those skilled in the art to better understand the solution of this application, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings.

[0053] As Figure 1 shown, the system architecture 100 may include terminal devices 101, 102, 103, a network 104, and a server 105. The network 104 is used to provide a medium for communication links between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired, wireless communication links, or fiber optic cables, etc.

[0054] When welding an automotive power battery, the terminal devices 101, 102, 103 can collect various data in the welding area, such as optical monitoring data, temperature monitoring data, and plasma monitoring data, and interact with the server 105 through the network 104 to receive or send messages, etc. The terminal devices 101, 102, 103 can be, but are not limited to, various industrial computers, personal computers, laptop computers, and sensors, etc. The server 105 may be provided with an automotive battery welding monitoring system.

[0055] It should be noted that the automotive battery welding monitoring method provided by the embodiments of this application is generally executed by the server. Correspondingly, the automotive battery welding monitoring device is generally disposed in the server.

[0056] It should be understood that Figure 1 the numbers of the terminal devices, the network, and the server in

[0057] are merely illustrative. According to the implementation requirements, there can be any number of terminal devices, networks, and servers.

[0057] Continuing to refer to Figure 2 , a flowchart of an embodiment of the automotive battery welding monitoring method according to the present application is shown. The automotive battery welding monitoring method includes the following steps:

[0058] Step S201, when welding an automotive battery, obtain optical monitoring data, temperature monitoring data, and plasma monitoring data in the welding area.

[0059] In this embodiment, the electronic device on which the automotive battery welding monitoring method runs (such as Figure 1The server shown can communicate with the terminal device through a wired connection or a wireless connection. It should be noted that the above wireless connection methods can include, but are not limited to, 3G / 4G / 5G connections, WiFi connections, Bluetooth connections, WiMAX connections, Zigbee connections, UWB (ultra wideband) connections, and other currently known or future-developed wireless connection methods.

[0060] Specifically, when welding an automotive battery (which can be an automotive power battery), the automotive battery welding monitoring system collects monitoring data for the welding area, obtaining optical monitoring data, temperature monitoring data, and plasma monitoring data. Among them, the optical monitoring data contains visual information of the molten pool and the weld seam; the temperature monitoring data contains temperature distribution information; and the plasma monitoring data contains the radiation characteristics of the plasma.

[0061] Furthermore, the steps of obtaining the optical monitoring data, temperature monitoring data, and plasma monitoring data for the welding area can include: collecting the molten pool shape and weld seam shape of the welding area through an optical sensor to obtain optical monitoring data; collecting the temperature field distribution of the molten pool in the welding area through a temperature sensor to obtain temperature monitoring data; and collecting the radiation characteristics of the plasma in the welding area through a plasma sensor to obtain plasma monitoring data.

[0062] Specifically, the automotive battery welding monitoring system can collect monitoring data in three dimensions: optical, temperature, and plasma. The system can use an optical sensor to continuously monitor the molten pool shape and weld seam shape in the welding area to obtain optical monitoring data, providing morphological information for subsequent defect detection. The size, shape, and changes of the molten pool during the welding process are important evaluation criteria for welding quality. For example, an overly large or irregular molten pool may lead to welding defects such as overburning or false soldering. The width, depth, and shape of the weld seam have a direct impact on welding quality.

[0063] The temperature sensor can continuously monitor the temperature field distribution of the molten pool during the welding process to obtain temperature monitoring data. The temperature field distribution reflects the heat transfer and heat distribution during the welding process and is an important factor in judging welding quality. During the welding process, the temperature in the molten pool area is a key quality indicator. Too high a temperature may lead to overburning or material loss, while too low a temperature may result in insufficient penetration. The change in the temperature field determines the size of the heat-affected zone, where the temperature is higher than the annealing temperature of the base material, which may cause changes in the physical properties of the material and affect welding quality.

[0064] A plasma sensor (which can be a spectrometer) can monitor the radiation characteristics of the plasma during the welding process, and provide information about the welding by measuring the radiation intensity and wavelength, thereby obtaining plasma monitoring data. The radiation characteristics of the plasma can reveal the energy distribution, metal evaporation, and the possible formation of pores or cracks during the welding process. The intensity and spectral changes of the plasma are often closely related to welding defects (such as pores and cracks).

[0065] In this embodiment, the optical monitoring data reflects the welding morphology, the temperature monitoring data reflects the thermal state, and the plasma monitoring data provides clues about potential defects such as metal evaporation and pores; obtaining multi-dimensional monitoring data is beneficial to more comprehensively and accurately monitor the welding quality.

[0066] Step S202, perform temporal feature extraction and spatial feature extraction on the optical monitoring data, temperature monitoring data, and plasma monitoring data respectively, and generate first optical data, first temperature data, and first plasma data according to the extracted features.

[0067] Specifically, input the optical monitoring data, temperature monitoring data, and plasma monitoring data into a spatio-temporal feature fusion network to extract the temporal features and spatial features in the monitoring data. The monitoring data has time information and can form a time series. The spatio-temporal feature fusion network can use a bidirectional LSTM (Long Short-Term Memory) network to process the time series data and extract the dynamic change rules during the welding process. The spatio-temporal feature fusion network can use a 3D convolutional neural network (3D-CNN) to extract the spatial features of the monitoring data and process the spatial distribution information of the welding area, such as the change of the molten pool morphology.

[0068] Finally, first optical data, first temperature data, and first plasma data are obtained. Among them, the first optical data contains the temporal features and spatial features extracted from the optical monitoring data, the first temperature data contains the temporal features and spatial features extracted from the temperature monitoring data, and the first plasma data contains the temporal features and spatial features extracted from the plasma monitoring data.

[0069] Further, the above step S202 may include: extracting the change information of the image sequence in the optical monitoring data as the optical timing feature, extracting the spatial image feature of the image sequence in the optical monitoring data, and obtaining the first optical data according to the optical timing feature and the spatial image feature; extracting the temperature change information and temperature gradient of the temperature monitoring data as the temperature timing feature, extracting the temperature field distribution and heat affected zone from the temperature monitoring data as the temperature spatial feature, and obtaining the first temperature data according to the temperature timing feature and the temperature spatial feature; extracting the radiation intensity change information and plasma fluctuation information of the plasma monitoring data as the plasma timing feature, extracting the plasma radiation distribution and plasma spectral feature of the plasma monitoring data as the plasma spatial feature, and obtaining the first plasma data according to the plasma timing feature and the plasma spatial feature.

[0070] Specifically, the optical monitoring data is an image sequence with time information, and the change information of the image sequence is extracted as the optical timing feature. The optical timing feature reflects the time information of the image change in the image sequence, that is, the evolution process of the molten pool morphology and weld morphology. The changes in the image sequence reflect the expansion of the molten pool or the offset of the weld, and these changes can be used to identify potential problems in welding (such as pores or cracks). The optical timing feature may include dynamic changes, color changes, or brightness changes in the image, etc.

[0071] The spatial image feature of the image sequence in the optical monitoring data is extracted as the optical spatial feature. The optical spatial feature focuses on the spatial information of the image, such as the size and shape of the molten pool, the width of the weld, the morphology, texture, and brightness distribution of the welding area, etc. Through the extraction of the optical spatial feature, the geometric information and morphological information of the welding area can be used to analyze the stability and quality of welding. The optical timing feature is combined with the spatial image feature to generate the first optical data.

[0072] The temperature change information and temperature gradient of the temperature monitoring data are extracted as the temperature timing feature. The temperature change information can reflect the stability of the heat source and the temperature fluctuation of the molten pool during the welding process. For example, a sudden increase or decrease in temperature may correspond to an abnormality in the welding process (such as overburning or insufficient penetration). The temperature gradient refers to the temperature difference between different time steps, which reflects the rate of temperature change and can detect the non-uniformity of the heat distribution in the welding area, thereby reflecting potential defects in welding (such as pores, false soldering, etc.).

[0073] Extract the temperature field distribution and heat affected zone from the temperature monitoring data as temperature spatial features. The temperature spatial features focus on the temperature distribution within the welding area, including the temperature field distribution (temperature distribution map of the welding area) and the size and shape of the heat affected zone (HAZ). The temperature of the heat affected zone is relatively high, which may affect the microstructure and properties of the material, thereby affecting the welding quality. Combine the temperature temporal features and temperature spatial features to generate the first temperature data, which are used to analyze the thermal state during the welding process.

[0074] Extract the radiation intensity change information and plasma fluctuation information of the plasma monitoring data as plasma temporal features. The radiation intensity change information reflects the change in the radiation intensity of the plasma, which can reflect the temperature fluctuation and metal flow during the welding process, and can be used to identify defects during the welding process, such as the formation of pores and the propagation of cracks. The plasma fluctuation information reflects the unstable part of the welding process. Especially when defects such as pores and cracks occur, the intensity and wavelength of the plasma will fluctuate.

[0075] Extract the plasma radiation distribution and plasma spectral characteristics of the plasma monitoring data as plasma spatial features. The plasma radiation distribution is the spatial distribution of the plasma, which reflects the thermal state and metal flow in different regions during the welding process, and can help detect irregular welding behaviors, thereby judging whether there are welding defects. The plasma spectral characteristics are the measured plasma spectra of different wavelengths, which can judge whether there is abnormal metal evaporation or gas dissolution during the welding process, and then infer the possible defect types. Combine the plasma temporal features and plasma spatial features to generate the first plasma data.

[0076] In this embodiment, perform temporal and spatial feature extraction on the optical monitoring data, temperature monitoring data, and plasma monitoring data to comprehensively capture the key changes during the welding process; the temporal features help capture the laws of dynamic changes during the welding process, and the spatial features provide the geometric information of the welding process; the multi-dimensional feature extraction method provides more accurate input data for defect detection, improving the accuracy of automotive battery welding monitoring.

[0077] Step S203, perform interactive fusion on the first optical data, first temperature data, and first plasma data to obtain the second optical data, second temperature data, and second plasma data.

[0078] Specifically, the system performs cross-modal fusion on the first optical data, first temperature data, and first plasma data, that is, performs interactive fusion on the temporal and spatial features extracted by different sensors to enhance the correlation between the data.

[0079] Further, the above step S203 may include: fusing the first temperature data and the first plasma data into the first optical data through a cross-modal attention mechanism to obtain second optical data; fusing the first optical data and the first plasma data into the first temperature data through a cross-modal attention mechanism to obtain second temperature data; fusing the first optical data and the first temperature data into the first plasma data through a cross-modal attention mechanism to obtain second plasma data.

[0080] Specifically, the cross-modal attention mechanism is a mechanism that realizes weighting by calculating the similarity between different data sources. This mechanism enables a certain modality (such as optical data) to actively focus on the important information between other modalities (such as temperature data and plasma data), realizing the fusion of information between modalities. In this application, the cross-modal attention mechanism weights the data from different sensors, thereby effectively fusing the features of optical, temperature, and plasma data.

[0081] First, the first temperature data and the first plasma data are fused into the first optical data through the cross-modal attention mechanism: taking the first temperature data and the first plasma data as query information, calculating the influence of these data on the first optical data through the cross-modal attention mechanism, and obtaining the fused optical data (i.e., the second optical data) after dynamic weighting.

[0082] Similarly, the first optical data and the first plasma data are fused into the first temperature data through the cross-modal attention mechanism. The first optical data and the first plasma data will affect the weighting of the first temperature data, obtaining the fused temperature data (i.e., the second temperature data).

[0083] Finally, the first optical data and the first temperature data are fused into the first plasma data through the cross-modal attention mechanism, incorporating the influence of the first optical data and the first temperature data into the first plasma data to generate enhanced plasma data (i.e., the second plasma data).

[0084] In this embodiment, by using the cross-modal attention mechanism to fuse different sensor data, automatically capturing the internal connection between the data, the generated second optical data, second temperature data, and second plasma data after fusion can provide more comprehensive and higher-quality feature information, can better describe the changes and defects in the welding process, and improve the accuracy of welding defect detection.

[0085] Further, the step of fusing the first temperature data and the first plasma data into the first optical data through the cross-modal attention mechanism to obtain the second optical data may include: fusing the first optical data with the first temperature data through the cross-modal attention mechanism to obtain the first optical fusion data; fusing the first optical data with the first plasma data through the cross-modal attention mechanism to obtain the second optical fusion data; connecting the first optical fusion data and the second optical fusion data to obtain the second optical data.

[0086] Specifically, through the cross-modal attention mechanism, the first optical data is fused with the first temperature data to generate the first optical fusion data. The combination of optical data (related to the molten pool shape and weld shape) and temperature data (related to the molten pool temperature change and heat affected zone) can enhance the expression of the influence of temperature on the welding shape and help detect defects caused by temperature fluctuations (such as insufficient penetration or overburn).

[0087] Similarly, the first optical data is fused with the first plasma data to generate the second optical fusion data. By calculating the similarity between the optical data and the plasma data, the plasma data (related to the plasma radiation intensity and fluctuation information) enhances the defect recognition ability of the optical data, especially when detecting defects related to plasma instability (such as porosity), providing more important information.

[0088] The first optical fusion data and the second optical fusion data are connected (concat()) to form the second optical data. The connection operation integrates the features obtained from the two cross-modal fusions to form a rich feature vector containing optical, temperature, and plasma information, which can provide a more comprehensive input for the subsequent defect detection network.

[0089] In this embodiment, through the weighted fusion of the cross-modal attention mechanism, the generated first optical fusion data and second optical fusion data, while expressing the optical data, fuse the key information of the temperature and plasma data. The multi-dimensional feature fusion can provide richer and more comprehensive welding process information, which helps to accurately identify welding defects.

[0090] Further, the step of fusing the first optical data with the first temperature data through the cross-modal attention mechanism to obtain the first optical fusion data may include: performing a linear transformation on the first optical data to obtain an optical query vector; calculating the temperature feature key and temperature feature value of the first temperature data; calculating the attention score between the optical query vector and the temperature feature key; weighting the temperature feature value according to the attention score to obtain weighted temperature data; adding the weighted temperature data and the first optical data to obtain the first optical fusion data.

[0091] Specifically, the first optical data is linearly transformed to extract the most relevant features for fusion with other modalities (such as temperature data), resulting in an optical query vector. The linear transformation can be obtained by multiplying the first optical data with a trainable weight matrix. The linear transformation can map the first optical data from a high-dimensional space to a space more suitable for interacting with other modality data.

[0092] The first temperature data is projected by multiplying it with a key projection matrix to obtain a temperature feature key (Key), and multiplying it with a value projection matrix to obtain a temperature feature value (Value). The temperature feature key contains the key information of the first temperature data and can be matched with the optical query vector; while the temperature feature value is the specific value of the temperature data and is used for the final fusion after weighting.

[0093] The similarity between the optical query vector and the temperature feature key is calculated to obtain an attention score, which represents the correlation between the first temperature data and the first optical data. The first temperature data with a high attention score will play a more important role in the fusion.

[0094] According to the calculated attention score, the temperature feature value is weighted to obtain weighted temperature data. The weighted temperature data can be adjusted according to the query vector of the optical data, so that the most relevant part of the temperature data contributes the most to the final fusion of the optical data.

[0095] The weighted temperature data is added to the first optical data to obtain the first optical fusion data. It contains the comprehensive information of the optical and temperature data, can more accurately reflect the multi-modal characteristics in the welding process, and provides rich data input for subsequent defect detection.

[0096] In this embodiment, through the cross-modal attention mechanism, the first optical data and the first temperature data are organically fused to ensure that the data of each sensor can better play its role in defect detection; the attention calculation between the optical query vector and the temperature feature key enables the system to dynamically adjust the weights according to the correlation of different data, enhance the key information, and ensure that the most relevant part of the first temperature data has the greatest impact on the fusion result of the first optical data. The dynamic weighting mechanism improves the accuracy of the fusion data and effectively enhances the accuracy of welding defect detection.

[0097] It can be understood that referring to the fusion process of the first optical data and the first temperature data, the first optical data and the first plasma data are fused in the same way to obtain the second optical fusion data. The first optical fusion data and the second optical fusion data are concatenated to obtain the second optical data. This application will not elaborate further.

[0098] Refer to the process of fusing the first temperature data and the first plasma data into the first optical data to obtain the second optical data. In the same way, fuse the first optical data and the first plasma data into the first temperature data to obtain the second temperature data; and in the same way, fuse the first optical data and the first temperature data into the first plasma data to obtain the second plasma data. This application will not elaborate further.

[0099] Step S204, obtain the defect type identifier, and add weights to the second optical data, the second temperature data, and the second plasma data according to the defect type identifier.

[0100] Specifically, each defect type has a corresponding defect type identifier, and the system can detect multiple defect types. According to the defect type identifier (such as pores, cracks, poor soldering, etc.), perform weighted processing on the second optical data, temperature data, and plasma data, and dynamically adjust according to the importance of the sensors to better adapt to the characteristics of different defect types. For example: for pore defects, the temperature data may be more important, so a larger weight is given to the temperature data; for crack defects, the optical data may be more critical, so the weight of the optical data is enhanced.

[0101] Further, the step of adding weights to the second optical data, the second temperature data, and the second plasma data according to the defect type identifier may include: obtaining the weight combination corresponding to the defect type identifier, where the weight combination is obtained in advance through learning; adding weights to the second optical data, the second temperature data, and the second plasma data according to the weight combination; or, based on the defect type identifier, calculating the weights of the second optical data, the second temperature data, and the second plasma data through the self-attention mechanism.

[0102] Specifically, the weight combinations corresponding to different defect types can be obtained in advance through training. These weight combinations are learned through machine learning algorithms (such as deep neural networks) during the training phase, based on the different dependencies of each defect type (such as pores, cracks, etc.) on the optical data, temperature data, and plasma data in the training data. Each defect type will have a specific set of weights, indicating the importance of the three data sources under that defect type.

[0103] Obtain the weight combination according to the defect type identifier, and perform weighting on the second optical data, the second temperature data, and the second plasma data. Different defect types correspond to different weight combinations, so that the relevant data (such as temperature data is more important in the defect of insufficient penetration) gets a higher weight, so that the system pays more attention to the contribution of the relevant data when detecting this defect type.

[0104] Alternatively, the self-attention mechanism calculates the weights of the second optical data, the second temperature data, and the second plasma data based on the defect type identifier. By calculating the similarity and correlation between data sources, the weights of each data source are automatically adjusted to improve the detection accuracy for specific defect types. The self-attention mechanism can flexibly handle the feature differences of different types of defects, adjust the weights in real time, and improve the adaptability and accuracy of detection.

[0105] In this embodiment, a weight combination corresponding to the defect type identifier is obtained, and appropriate weights are automatically assigned to each defect type, enabling the system to focus on the most relevant data sources when detecting specific defects, thereby improving the detection accuracy; or through the self-attention mechanism, according to the characteristics of each defect type, the weights of different data sources are dynamically adjusted. The adaptive weighting method enables the system to flexibly handle various welding defects and improves the detection accuracy.

[0106] Step S205: Input the second optical data, the second temperature data, and the second plasma data with weights into a defect detection network corresponding to the defect type identifier to obtain a welding defect detection result.

[0107] Specifically, for different welding defect types, the system uses different neural networks. Different types of defects (such as pores, cracks, and poor soldering) have different physical characteristics and behavior patterns during formation, and may exhibit different time-series characteristics and spatial distribution patterns. For example: Pores are usually closely related to the non-uniformity of the temperature field and may require a higher weight for temperature sensor data; Cracks usually appear as linear or non-linear changes in the image and are more dependent on the spatial characteristics provided by the optical sensor; Poor soldering may manifest as too low energy or temperature in certain areas, and it is necessary to comprehensively analyze the data of multiple sensors. The signal characteristics of each defect type are different, and their detection methods should be customized according to the characteristic differences to improve the diagnostic accuracy.

[0108] For each defect type, different network structures are adopted to optimize the extraction and recognition of its features. For pore detection, a convolutional neural network (CNN) can be designed to analyze the spatial changes in the temperature field, as the temperature changes usually exhibit a certain distribution pattern in space. For cracks, an image recognition network (such as CNN or ResNet) can be used to process the welding images collected by the optical sensor. Cracks usually appear as linear or crack-like features in the image. For poor soldering, regression analysis (such as a fully connected layer) may be combined to analyze the power fluctuations during the welding process and identify areas with unqualified power. Using different neural network models or structures can better adapt to the characteristics of different types of defects, thereby improving the diagnostic accuracy.

[0109] Input the second optical data with weights, the second temperature data, and the second plasma data into a defect detection network corresponding to the defect type identification for welding defect detection. Through the trained deep learning model, accurately classify the defects during the welding process to obtain the welding defect detection result.

[0110] In this embodiment, the extraction of the temporal and spatial features of the optical monitoring data, the temperature monitoring data, and the plasma monitoring data effectively captures the key changes during the welding process, avoiding the problem that the data of a single sensor cannot comprehensively reflect welding defects; the cross-modal attention mechanism is used to interact and fuse the first optical data, the first temperature data, and the first plasma data, enhancing the correlation between the data of different sensors and improving the accuracy of defect recognition; finally, the weights of the second optical data, the second temperature data, and the second plasma data are dynamically adjusted according to the defect type identification, making the contribution of each sensor data to the defect detection result more accurate, further reducing the false detection rate, and selecting the corresponding defect detection network for welding defect detection according to the defect type identification, enabling the complex defects during the welding process to be accurately identified and improving the accuracy of automotive battery welding monitoring.

[0111] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through computer-readable instructions, and the computer-readable instructions can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above methods. Among them, the aforementioned storage medium can be a non-volatile storage medium such as a magnetic disk, an optical disk, a read-only memory (ROM), etc., or a random access memory (RAM), etc.

[0112] It should be understood that although the steps in the flowchart of the accompanying drawings are shown in sequence according to the indication of the arrows, these steps do not necessarily have to be executed in the order indicated by the arrows. Unless there is a clear indication in this article, the execution of these steps has no strict order limit and can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages. These sub-steps or stages do not necessarily have to be executed at the same time, but can be executed at different times, and their execution order does not necessarily have to be sequential, but can be executed alternately or in turn with at least a part of other steps or sub-steps or stages of other steps.

[0113] For further reference Figure 3 to Figure 2 As an implementation of the method shown above, the present application provides an embodiment of an automotive battery welding monitoring device. This device embodiment is related to Figure 2The method embodiments shown correspond to this, and this device can be specifically applied to various electronic devices.

[0114] As Figure 3 shown, the automotive battery welding monitoring device 300 described in this embodiment includes: a monitoring and acquisition module 301, a feature extraction module 302, an interaction and fusion module 303, a weight addition module 304, and a defect detection module 305, where:

[0115] The monitoring and acquisition module 301 is configured to obtain optical monitoring data, temperature monitoring data, and plasma monitoring data of the welding area when welding an automotive battery.

[0116] The feature extraction module 302 is configured to perform temporal feature extraction and spatial feature extraction on the optical monitoring data, temperature monitoring data, and plasma monitoring data respectively, and generate first optical data, first temperature data, and first plasma data according to the extracted features.

[0117] The interaction and fusion module 303 is configured to perform interaction and fusion on the first optical data, first temperature data, and first plasma data to obtain second optical data, second temperature data, and second plasma data.

[0118] The weight addition module 304 is configured to obtain a defect type identifier, and add weights to the second optical data, second temperature data, and second plasma data according to the defect type identifier.

[0119] The defect detection module 305 is configured to input the weighted second optical data, second temperature data, and second plasma data into a defect detection network corresponding to the defect type identifier to obtain a welding defect detection result.

[0120] In this embodiment, the extraction of the temporal features and spatial features of the optical monitoring data, temperature monitoring data, and plasma monitoring data effectively captures the key changes during the welding process, avoiding the problem that the data of a single sensor cannot comprehensively reflect welding defects; the cross-modal attention mechanism is used to perform interaction and fusion on the first optical data, first temperature data, and first plasma data, enhancing the correlation between different sensor data and improving the accuracy of defect recognition; finally, the weights of the second optical data, second temperature data, and second plasma data are dynamically adjusted according to the defect type identifier, making the contribution of each sensor data to the defect detection result more accurate, further reducing the false detection rate, and selecting a corresponding defect detection network for welding defect detection according to the defect type identifier, so that complex defects during the welding process can be accurately identified, improving the accuracy of automotive battery welding monitoring.

[0121] In some alternative implementation manners of this embodiment, the monitoring and acquisition module 301 may include: an optical acquisition sub-module, a temperature acquisition sub-module, and a plasma acquisition sub-module, where:

[0122] The optical acquisition sub-module is configured to collect the molten pool morphology and weld seam morphology of the welding area through an optical sensor, and obtain optical monitoring data.

[0123] The temperature acquisition sub-module is configured to collect the molten pool temperature field distribution of the welding area through a temperature sensor, and obtain temperature monitoring data.

[0124] The plasma acquisition sub-module is configured to collect the radiation characteristics of the plasma in the welding area through a plasma sensor, and obtain plasma monitoring data.

[0125] In this embodiment, the optical monitoring data reflects the welding morphology, the temperature monitoring data reflects the thermal state, and the plasma monitoring data provides clues about potential defects such as metal evaporation and porosity; obtaining multi-dimensional monitoring data is beneficial to more comprehensively and accurately monitor the welding quality.

[0126] In some alternative implementation manners of this embodiment, the feature extraction module 302 may include: an optical extraction sub-module, a temperature extraction sub-module, and a plasma extraction sub-module, where:

[0127] The optical extraction sub-module is configured to extract the change information of the image sequence in the optical monitoring data as optical temporal features, extract the spatial image features of the image sequence in the optical monitoring data, and obtain first optical data based on the optical temporal features and the spatial image features.

[0128] The temperature extraction sub-module is configured to extract the temperature change information and temperature gradient of the temperature monitoring data as temperature temporal features, extract the temperature field distribution and heat affected zone from the temperature monitoring data as temperature spatial features, and obtain first temperature data based on the temperature temporal features and the temperature spatial features.

[0129] The plasma extraction sub-module is configured to extract the radiation intensity change information and plasma fluctuation information of the plasma monitoring data as plasma temporal features, extract the plasma radiation distribution and plasma spectral features of the plasma monitoring data as plasma spatial features, and obtain first plasma data based on the plasma temporal features and the plasma spatial features.

[0130] In this embodiment, temporal and spatial feature extraction is performed on optical monitoring data, temperature monitoring data, and plasma monitoring data to comprehensively capture key changes during the welding process; the temporal features help capture the laws of dynamic changes during the welding process, while the spatial features provide geometric information about the welding process; the multi-dimensional feature extraction method provides more accurate input data for defect detection, improving the accuracy of automotive battery welding monitoring.

[0131] In some alternative implementation manners of this embodiment, the interaction fusion module 303 may include: an optical fusion sub-module, a temperature fusion sub-module, and a plasma fusion sub-module, where:

[0132] The optical fusion sub-module is configured to fuse the first temperature data and the first plasma data into the first optical data through a cross-modal attention mechanism to obtain second optical data.

[0133] The temperature fusion sub-module is configured to fuse the first optical data and the first plasma data into the first temperature data through a cross-modal attention mechanism to obtain second temperature data.

[0134] The plasma fusion sub-module is configured to fuse the first optical data and the first temperature data into the first plasma data through a cross-modal attention mechanism to obtain second plasma data.

[0135] In this embodiment, by using the cross-modal attention mechanism to fuse different sensor data, the internal connections between the data are automatically captured. The generated second optical data, second temperature data, and second plasma data after fusion can provide more comprehensive and higher-quality feature information, can better describe the changes and defects during the welding process, and improve the accuracy of welding defect detection.

[0136] In some alternative implementation manners of this embodiment, the optical fusion sub-module may include: a first fusion unit, a second fusion unit, and a data connection unit, where:

[0137] The first fusion unit is configured to fuse the first optical data and the first temperature data through a cross-modal attention mechanism to obtain first optical fusion data.

[0138] The second fusion unit is configured to fuse the first optical data and the first plasma data through a cross-modal attention mechanism to obtain second optical fusion data.

[0139] The data connection unit is configured to connect the first optical fusion data and the second optical fusion data to obtain second optical data.

[0140] In this embodiment, through the weighted fusion of the cross-modal attention mechanism, the generated first optical fusion data and second optical fusion data fuse the key information of temperature and plasma data on the basis of expressing optical data. The multi-dimensional feature fusion can provide richer and more comprehensive information on the welding process, which helps to accurately identify welding defects.

[0141] In some alternative implementation manners of this embodiment, the first fusion unit may include: a linear transformation subunit, a key-value calculation subunit, a score calculation subunit, a value weighting subunit, and an optical fusion subunit, where:

[0142] The linear transformation subunit is configured to perform a linear transformation on the first optical data to obtain an optical query vector.

[0143] The key-value calculation subunit is configured to calculate a temperature feature key and a temperature feature value of the first temperature data.

[0144] The score calculation subunit is configured to calculate an attention score between the optical query vector and the temperature feature key.

[0145] The value weighting subunit is configured to weight the temperature feature value according to the attention score to obtain weighted temperature data.

[0146] The optical fusion subunit is configured to add the weighted temperature data and the first optical data to obtain the first optical fusion data.

[0147] In this embodiment, through the cross-modal attention mechanism, the first optical data and the first temperature data are organically fused to ensure that the data of each sensor can better play a role in defect detection; the attention calculation between the optical query vector and the temperature feature key enables the system to dynamically adjust the weights according to the correlation of different data, enhance the key information, and ensure that the most relevant part of the first temperature data has the greatest impact on the fusion result of the first optical data. The dynamic weighting mechanism improves the accuracy of the fusion data and effectively improves the accuracy of welding defect detection.

[0148] In some alternative implementation manners of this embodiment, the weight addition module 304 may include: a combination acquisition sub-module, a weight addition sub-module, and a weight calculation sub-module, where:

[0149] The combination acquisition sub-module is configured to acquire a weight combination corresponding to the defect type identifier, and the weight combination is obtained through learning in advance.

[0150] The weight addition sub-module is configured to add weights to the second optical data, the second temperature data, and the second plasma data according to the weight combination.

[0151] A weight calculation sub-module, configured to calculate the weights of the second optical data, the second temperature data, and the second plasma data based on the defect type identifier through a self-attention mechanism.

[0152] In this embodiment, a weight combination corresponding to the defect type identifier is obtained, and appropriate weights are automatically assigned to each defect type, enabling the system to focus on the most relevant data sources when detecting specific defects, thereby improving the detection accuracy; or through a self-attention mechanism, according to the characteristics of each defect type, the weights of different data sources are dynamically adjusted, and the adaptive weighting method enables the system to flexibly handle various welding defects and improves the detection accuracy.

[0153] To solve the above technical problems, the embodiments of the present application also provide a computer device. For details, please refer to Figure 4 , Figure 4 which is the basic structural block diagram of the computer device in this embodiment.

[0154] The computer device 4 includes a memory 41, a processor 42, and a network interface 43 that are communicatively connected to each other through a system bus. It should be noted that only the computer device 4 with components 41-43 is shown in the figure, but it should be understood that it is not required to implement all the shown components, and more or fewer components can be alternatively implemented. Among them, those skilled in the art of the present technology can understand that a computer device here is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to a microprocessor, an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), a digital signal processor (DSP), an embedded device, etc.

[0155] The computer device can be a computing device such as a desktop computer, a notebook, a palm computer, and a cloud server. The computer device can perform human-computer interaction with the user through a keyboard, a mouse, a remote control, a touchpad, a voice control device, or the like.

[0156] The memory 41 includes at least one type of readable storage medium, which includes flash memory, hard disk, multimedia card, card-type memory (such as SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 41 may be an internal storage unit of the computer device 4, such as the hard disk or memory of the computer device 4. In other embodiments, the memory 41 may also be an external storage device of the computer device 4, such as a plug-in hard disk, Smart Media Card (SMC), Secure Digital (SD) card, FlashCard, etc. equipped on the computer device 4. Of course, the memory 41 may also include both the internal storage unit and the external storage device of the computer device 4. In this embodiment, the memory 41 is generally used to store the operating system and various application software installed in the computer device 4, such as computer-readable instructions of the automotive battery welding monitoring method. In addition, the memory 41 may also be used to temporarily store various data that have been output or will be output.

[0157] In some embodiments, the processor 42 may be a central processing unit (CPU), controller, microcontroller, microprocessor, or other data processing chip. The processor 42 is generally used to control the overall operation of the computer device 4. In this embodiment, the processor 42 is used to run the computer-readable instructions stored in the memory 41 or process data, such as running the computer-readable instructions of the automotive battery welding monitoring method.

[0158] The network interface 43 may include a wireless network interface or a wired network interface, and this network interface 43 is generally used to establish a communication connection between the computer device 4 and other electronic devices.

[0159] The computer device provided in this embodiment can execute the above-mentioned automotive battery welding monitoring method. Here, the automotive battery welding monitoring method may be the automotive battery welding monitoring method of each of the above embodiments.

[0160] In this embodiment, the extraction of the temporal and spatial features of the optical monitoring data, temperature monitoring data, and plasma monitoring data effectively captures the key changes during the welding process, avoiding the problem that single-sensor data cannot comprehensively reflect welding defects; the cross-modal attention mechanism is used to interactively fuse the first optical data, first temperature data, and first plasma data, enhancing the correlation between different sensor data and improving the accuracy of defect recognition; finally, the weights of the second optical data, second temperature data, and second plasma data are dynamically adjusted according to the defect type identifier, making the contribution of each sensor data to the defect detection result more accurate, further reducing the false detection rate, and selecting the corresponding defect detection network for welding defect detection according to the defect type identifier, enabling the accurate identification of complex defects during the welding process and improving the accuracy of automotive battery welding monitoring.

[0161] The present application also provides another implementation manner, that is, to provide a computer-readable storage medium storing computer-readable instructions that can be executed by at least one processor, so that the at least one processor executes the steps of the automotive battery welding monitoring method as described above.

[0162] In this embodiment, the extraction of the temporal and spatial features of the optical monitoring data, temperature monitoring data, and plasma monitoring data effectively captures the key changes during the welding process, avoiding the problem that single-sensor data cannot comprehensively reflect welding defects; the cross-modal attention mechanism is used to interactively fuse the first optical data, first temperature data, and first plasma data, enhancing the correlation between different sensor data and improving the accuracy of defect recognition; finally, the weights of the second optical data, second temperature data, and second plasma data are dynamically adjusted according to the defect type identifier, making the contribution of each sensor data to the defect detection result more accurate, further reducing the false detection rate, and selecting the corresponding defect detection network for welding defect detection according to the defect type identifier, enabling the accurate identification of complex defects during the welding process and improving the accuracy of automotive battery welding monitoring.

[0163] Through the description of the above embodiments, those skilled in the art can clearly understand that the above-described embodiment methods can be implemented by means of software plus a necessary general hardware platform. Of course, they can also be implemented by hardware, but in many cases, the former is a better implementation manner. Based on such an understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product stored in a storage medium (such as ROM / RAM, magnetic disk, optical disc), including several instructions for causing a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of the present application.

[0164] Obviously, the embodiments described above are only a part of the embodiments of this application, rather than all of them. The preferred embodiments of this application are shown in the accompanying drawings, but they do not limit the patent scope of this application. This application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosed content of this application more thorough and comprehensive. Although this application has been described in detail with reference to the foregoing embodiments, for those skilled in the art, they can still modify the technical solutions described in the foregoing specific embodiments, or perform equivalent replacements on some of the technical features. Any equivalent structure that makes use of the content of this application's specification and the accompanying drawings, directly or indirectly applied in other related technical fields, is similarly within the scope of patent protection of this application.

Claims

1. A method for monitoring automobile battery welding, characterized in that: The steps include: When welding automotive batteries, obtain optical monitoring data, temperature monitoring data, and plasma monitoring data of the welding area; Performing time series feature extraction and space feature extraction on the optical monitoring data, the temperature monitoring data and the plasma monitoring data respectively, and generating first optical data, first temperature data and first plasma data according to the extracted features; Interactively fusing the first optical data, the first temperature data and the first plasma data to obtain second optical data, second temperature data and second plasma data; Acquire a defect type identifier, and add weights to the second optical data, the second temperature data, and the second plasma data according to the defect type identifier; The second optical data, the second temperature data and the second plasma data with weights are input into a defect detection network corresponding to the defect type identifier to obtain a welding defect detection result.

2. The automotive battery welding monitoring method according to claim 1, characterized in that: The step of obtaining optical monitoring data, temperature monitoring data and plasma monitoring data of the welding area includes: The molten pool shape and weld shape of the welding area are collected by optical sensors to obtain optical monitoring data; The temperature sensor is used to collect the temperature field distribution of the molten pool in the welding area to obtain temperature monitoring data; The radiation characteristics of the plasma in the welding area are collected by a plasma sensor to obtain plasma monitoring data.

3. The automotive battery welding monitoring method according to claim 1, characterized in that: The steps of respectively performing time series feature extraction and space feature extraction on the optical monitoring data, the temperature monitoring data and the plasma monitoring data, and generating first optical data, first temperature data and first plasma data according to the extracted features include: Extracting change information of the image sequence in the optical monitoring data as an optical time sequence feature, extracting a spatial image feature of the image sequence in the optical monitoring data, and obtaining first optical data according to the optical time sequence feature and the spatial image feature; Extracting temperature change information and temperature gradient of the temperature monitoring data as temperature time series features, extracting temperature field distribution and heat affected zone from the temperature monitoring data as temperature space features, and obtaining first temperature data according to the temperature time series features and the temperature space features; The radiation intensity variation information and plasma fluctuation information of the plasma monitoring data are extracted as plasma timing features, the plasma radiation distribution and plasma spectrum features of the plasma monitoring data are extracted as plasma space features, and the first plasma data are obtained according to the plasma timing features and the plasma space features.

4. The automotive battery welding monitoring method according to claim 1, characterized in that: The step of interactively fusing the first optical data, the first temperature data and the first plasma data to obtain second optical data, second temperature data and second plasma data comprises: fusing the first temperature data and the first plasma data into the first optical data through a cross-modal attention mechanism to obtain second optical data; fusing the first optical data and the first plasma data into the first temperature data through the cross-modal attention mechanism to obtain second temperature data; The first optical data and the first temperature data are fused into the first plasma data through the cross-modal attention mechanism to obtain second plasma data.

5. The method for monitoring automobile battery welding according to claim 4, characterized in that: The step of fusing the first temperature data and the first plasma data into the first optical data through a cross-modal attention mechanism to obtain second optical data comprises: fusing the first optical data with the first temperature data through a cross-modal attention mechanism to obtain first optical fused data; fusing the first optical data with the first plasma data through the cross-modal attention mechanism to obtain second optical fused data; The first optical fusion data and the second optical fusion data are connected to obtain second optical data.

6. The method for monitoring automobile battery welding according to claim 5, characterized in that: The step of fusing the first optical data with the first temperature data through a cross-modal attention mechanism to obtain first optical fusion data includes: Performing a linear transformation on the first optical data to obtain an optical query vector; Calculate a temperature characteristic key and a temperature characteristic value of the first temperature data; calculating an attention score between the optical query vector and the temperature feature key; weighting the temperature feature value according to the attention score to obtain weighted temperature data; The weighted temperature data and the first optical data are added to obtain first optical fusion data.

7. The automotive battery welding monitoring method according to claim 1, characterized in that: The step of adding weights to the second optical data, the second temperature data and the second plasma data according to the defect type identification comprises: Obtaining a weight combination corresponding to the defect type identifier, wherein the weight combination is obtained in advance through learning; adding weights to the second optical data, the second temperature data and the second plasma data according to the weight combination; or, Based on the defect type identification, weights of the second optical data, the second temperature data, and the second plasma data are calculated through a self-attention mechanism.

8. An automobile battery welding monitoring device, characterized in that: include: A monitoring acquisition module, used to acquire optical monitoring data, temperature monitoring data and plasma monitoring data of the welding area when welding the automobile battery; a feature extraction module, configured to perform temporal feature extraction and spatial feature extraction on the optical monitoring data, the temperature monitoring data and the plasma monitoring data, respectively, and generate first optical data, first temperature data and first plasma data according to the extracted features; An interactive fusion module, used for interactively fusing the first optical data, the first temperature data and the first plasma data to obtain second optical data, second temperature data and second plasma data; a weight adding module, used for acquiring a defect type identifier and adding weights to the second optical data, the second temperature data and the second plasma data according to the defect type identifier; The defect detection module is used to input the second optical data, the second temperature data and the second plasma data with weights into a defect detection network corresponding to the defect type identifier to obtain a welding defect detection result.

9. A computer device, comprising a memory and a processor, wherein the memory stores computer-readable instructions, and the processor implements the steps of the automotive battery welding monitoring method according to any one of claims 1 to 7 when executing the computer-readable instructions.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-readable instructions, and when the computer-readable instructions are executed by a processor, the steps of the automotive battery welding monitoring method according to any one of claims 1 to 7 are implemented.

Images