Ultrasonic metal welding quality detection method and device based on multi-signal fusion
Through the neural network model detection method of multi-signal fusion, the problem of unstable solder joint quality in ultrasonic welding technology is solved, and a non-destructive and efficient welding quality evaluation is achieved to ensure high-quality production of IGBT modules.
Patent Information
- Application Number
- CN202510449933.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-10
- Publication Date
- 2025-07-04
Smart Images
Figure CN120257210A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the technical field of ultrasonic metal welding, and particularly to an ultrasonic metal welding quality detection method and device based on multi-signal fusion. Background Art
[0002] With the rapid development of the new energy vehicle industry, the insulated gate bipolar transistor (IGBT) module, as the core component of the electric vehicle power electronics system, its performance and reliability are crucial for the safety and efficiency of the whole vehicle. In the packaging process of the IGBT module, welding technology plays an indispensable role. Due to its advantages such as high reliability, low-resistance connection, and excellent thermal conductivity, ultrasonic welding technology is widely used in the packaging process of IGBT modules to ensure good electrical and mechanical connections.
[0003] However, in practical applications, even when maintaining the same welding parameters, the quality of the solder joints produced by ultrasonic welding still shows significant differences, resulting in unstable product quality. To ensure the quality of the welded joints and reduce the product safety risks caused by welding defects, methods for quality inspection of the joints after welding are commonly used at home and abroad. Traditional detection means mainly include two methods: destructive testing and non-destructive testing. Although destructive testing can provide detailed internal information of the solder joints, this method will damage the solder joints, making it impossible to conduct a comprehensive evaluation of the product and precluding the possibility of repeated testing. Although non-destructive testing can protect the integrity of the solder joints, it is usually complex to operate, costly, and may lead to a decrease in production efficiency. Summary of the Invention
[0004] Based on this, it is necessary to provide an ultrasonic metal welding quality detection method and device based on multi-signal fusion for the above technical problems.
[0005] In a first aspect, an embodiment of this application provides an ultrasonic metal welding quality detection method based on multi-signal fusion, and the method includes:
[0006] Obtain various welding signals during the ultrasonic metal welding process;
[0007] Input the various welding signals into a neural network model for training to obtain a welding quality prediction model. Among them, the neural network model includes a multi-scale feature extraction module, a parallel fusion attention mechanism module, and a classifier. The multi-scale feature extraction module is used to perform multi-scale extraction on each welding signal to obtain corresponding first feature maps, and after fusing the first feature maps, a second feature map is obtained. The parallel fusion attention mechanism module is used to perform feature extraction on the second feature map to obtain a third feature map, and the classifier is used to perform classification based on the third feature map to obtain the welding quality of the metal.
[0008] Input various target welding signals collected during the actual metal welding process into the welding quality prediction model for prediction to obtain the welding quality prediction result.
[0009] In one embodiment, the multi-scale feature extraction module includes multi-path convolutional layers. The multi-scale feature extraction module is used to perform multi-scale extraction on each of the welding signals to obtain corresponding first feature maps, including:
[0010] Input each of the welding signals into the convolutional layers of each path for convolutional operations, and output fourth feature maps of different scales;
[0011] Fuse the fourth feature maps of different scales and perform weighted summation with the welding signals to obtain corresponding first feature maps.
[0012] In one embodiment, the parallel fusion attention mechanism module includes a spatial attention module and a channel attention module. The parallel fusion attention mechanism module is used to perform feature extraction on the second feature map to obtain a third feature map, including:
[0013] Input the second feature map into the spatial attention module to obtain spatial attention feature weights; and input the second feature map into the channel attention module to obtain channel attention feature weights;
[0014] Perform weighted fusion on the second feature map based on the spatial attention feature weights and the channel attention feature weights to obtain a third feature map.
[0015] In one embodiment, the third feature map is used to predict the welding quality, including:
[0016] The third feature map sequentially passes through two convolutional layers, two pooling layers, and two fully connected layers, and finally outputs a prediction result.
[0017] In one embodiment, the step of inputting the various welding signals into the neural network model for training further includes:
[0018] Use the sliding window algorithm to convert each of the welding signals into corresponding two-dimensional images, and input each of the two-dimensional images into the neural network model for training.
[0019] In one embodiment, before performing multi-scale extraction on each of the welding signals, the method further includes:
[0020] Input each of the welding signals into a convolutional layer for convolutional operations.
[0021] In one embodiment, the multiple welding signals include at least two of a current signal, a voltage signal, a phase difference signal, and an ultrasonic frequency signal.
[0022] Second, an embodiment of the present application further provides an ultrasonic metal welding quality detection device based on multi-signal fusion. The device includes:
[0023] An acquisition module, configured to acquire multiple welding signals during the ultrasonic metal welding process;
[0024] A training module, configured to input the multiple welding signals into a neural network model for training to obtain a welding quality prediction model. Among them, the neural network model includes a multi-scale feature extraction module and a parallel fusion attention mechanism module. The multi-scale feature extraction module is configured to perform multi-scale extraction on each of the welding signals to obtain corresponding first feature maps, and after the first feature maps are fused, a second feature map is obtained. The parallel fusion attention mechanism module is configured to perform feature extraction on the second feature map to obtain a third feature map, and the classifier is configured to perform classification based on the third feature map to obtain the welding quality of the metal;
[0025] A prediction module, configured to input multiple target welding signals collected during the actual metal welding process into the welding quality prediction model for prediction to obtain a welding quality prediction result.
[0026] Third, an embodiment of the present application further provides a computer device, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the method described in the first aspect above is implemented.
[0027] Fourth, an embodiment of the present application further provides an ultrasonic metal welding process signal acquisition system, including the computer device, a DSP control board, and an ultrasonic metal welding machine described in the third aspect above. The computer device is connected to the DSP control board, and the DSP control board is connected to the ultrasonic metal welding machine. Among them:
[0028] The computer device is configured to set welding parameters and send a start instruction to the DSP control board;
[0029] The DSP control board is configured to receive the start instruction to generate a high-frequency ultrasonic signal and send it to the ultrasonic metal welding machine;
[0030] The ultrasonic metal welding machine is configured to receive the high-frequency ultrasonic signal to start the welding process;
[0031] The DSP control board is further configured to collect various welding signals during the welding process and transmit them to the computer device for storage.
[0032] The above ultrasonic metal welding quality detection method and device based on multi-signal fusion obtain various welding signals during the ultrasonic metal welding process; input the various welding signals into a neural network model for training to obtain a welding quality prediction model; input various target welding signals collected during the actual metal welding process into the welding quality prediction model for prediction to obtain a welding quality prediction result. This can not only accurately evaluate the solder joint quality without damaging the solder joint, but also effectively improve the detection efficiency, reduce costs, and thus ensure the high-quality production and long-term stable operation of the IGBT module.
[0033] Details of one or more embodiments of the present application are set forth in the following drawings and description, so that other features, objects, and advantages of the present application will become more concise and understandable. BRIEF DESCRIPTION OF THE DRAWINGS
[0034] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments and descriptions thereof of the present application are used to explain the present application and do not constitute an improper limitation of the present application. In the drawings:
[0035] Figure 1 is a hardware structure block diagram of a terminal device for an ultrasonic metal welding quality detection method based on multi-signal fusion in an embodiment;
[0036] Figure 2 is a schematic flowchart of an ultrasonic metal welding quality detection method based on multi-signal fusion in an embodiment;
[0037] Figure 3 is a schematic diagram of an R-Inception multi-scale feature extraction module in an embodiment;
[0038] Figure 4 is a schematic diagram of a parallel fusion attention mechanism module (PFAM) in an embodiment;
[0039] Figure 5 is a schematic diagram of a multi-signal feature-level fusion quality prediction model in an embodiment;
[0040] Figure 6 is a structure block diagram of an ultrasonic metal welding quality detection device based on multi-signal fusion in an embodiment;
[0041] Figure 7 is a schematic diagram of a computer device structure in an embodiment. DETAILED DESCRIPTION OF THE EMBODIMENTS
[0042] In order to make the objectives, technical solutions and advantages of the present application more clear and understandable, the present application will be described and explained below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application. Based on the embodiments provided in the present application, all other embodiments obtained by those of ordinary skill in the art without making creative efforts fall within the scope of protection of the present application.
[0043] Obviously, the accompanying drawings in the following description are only some examples or embodiments of the present application. For those of ordinary skill in the art, without making creative efforts, the present application can also be applied to other similar scenarios based on these drawings. In addition, it can also be understood that although the efforts made in this development process may be complex and lengthy, however, for those of ordinary skill in the art related to the content disclosed in the present application, some design, manufacturing or production changes made on the basis of the technical content disclosed in the present application are only conventional technical means and should not be understood as the content disclosed in the present application being insufficient.
[0044] The mention of "embodiment" in the present application means that the specific features, structures or characteristics described in connection with the embodiment may be included in at least one embodiment of the present application. The appearance of this phrase in various positions in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments. Those of ordinary skill in the art explicitly and implicitly understand that the embodiments described in the present application can be combined with other embodiments without conflict.
[0045] Unless otherwise defined, the technical terms or scientific terms involved in this application shall have the ordinary meanings understood by those with ordinary skills in the technical field to which this application belongs. The words such as "a", "one", "kind", "the" and the like involved in this application do not indicate a limitation in quantity and may represent a singular or plural number. The terms "including", "comprising", "having" and any variations thereof involved in this application are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or modules (units) is not limited to the listed steps or units, but may further include unlisted steps or units, or may further include other steps or units inherent to these processes, methods, products or devices. The similar words such as "connected", "coupled" and "joined" involved in this application are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. The "plurality" involved in this application means two or more. "And / or" describes the association relationship of associated objects and indicates that three relationships may exist. For example, "A and / or B" may represent: A exists alone, A and B exist simultaneously, and B exists alone. The character " / " generally represents an "or" relationship between the front and rear associated objects. The terms "first", "second", "third" and the like involved in this application are only used to distinguish similar objects and do not represent a specific order of the objects.
[0046] The method embodiment provided in this embodiment can be executed on a terminal, a computer or a similar computing device. For example, when running on a terminal, Figure 1 is the hardware structure block diagram of the terminal of the ultrasonic metal welding quality detection method based on multi-signal fusion in this embodiment. As Figure 1 shown, the terminal may include one or more ( Figure 1 only one is shown in the figure) processors 102 and a memory 104 for storing data. Among them, the processor 102 may include, but is not limited to, a processing device such as a microprocessor MCU or a programmable logic device FPGA. The above terminal may further include a transmission device 106 for communication functions and an input / output device 108. Those of ordinary skill in the art can understand that Figure 1 the structure shown is only schematic and does not limit the structure of the above terminal. For example, the terminal may further include more or fewer components than those shown in Figure 1 the figure, or have a different configuration from that shown in Figure 1 the figure.
[0047] The memory 104 can be used to store computer programs, such as software programs and modules of application software, such as the computer program corresponding to the ultrasonic metal welding quality detection method based on multi-signal fusion in this embodiment. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, that is, the above method is implemented. The memory 104 may include high-speed random access memory and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memories. In some instances, the memory 104 may further include a memory remotely located relative to the processor 102, and these remote memories can be connected to the terminal through a network. Examples of the above networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.
[0048] The transmission device 106 is used to receive or send data via a network. The above network includes a wireless network provided by a communication provider of the terminal. In one instance, the transmission device 106 includes a network adapter (Network Interface Controller, NIC), which can be connected to other network devices through a base station and thus can communicate with the Internet. In one instance, the transmission device 106 can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.
[0049] An embodiment of the present application provides an ultrasonic metal welding quality detection method based on multi-signal fusion. Taking the case where this method is applied to Figure 1 the terminal in Figure 2 as an example, as
[0050] shown, the method includes the following steps:
[0051] Specifically, a variety of welding signals during the ultrasonic metal welding process are acquired through a welding process signal acquisition system. This system realizes reliable communication between the DSP (Digital Signal Processor) and the terminal through the USB communication protocol, thereby capturing various key signals during the welding process in real time. These signals include voltage, current, phase difference, and ultrasonic frequency, etc.
[0052] Step 202: Input the multiple welding signals into a neural network model for training to obtain a welding quality prediction model. Among them, the neural network model includes a multi-scale feature extraction module, a parallel fusion attention mechanism module, and a classifier. The multi-scale feature extraction module is used to perform multi-scale extraction on each of the welding signals to obtain corresponding first feature maps. After fusing each of the first feature maps, a second feature map is obtained. The parallel fusion attention mechanism module is used to perform feature extraction on the second feature map to obtain a third feature map. The classifier is used to perform classification based on the third feature map to obtain the welding quality of the metal.
[0053] Specifically, a large number of welding signal samples obtained during the metal welding process are used to train the neural network to obtain a welding quality prediction model. The welding quality prediction model integrates a multi-scale feature extraction module and an attention mechanism module. Among them, the multi-scale feature extraction module has the ability to accurately capture subtle changes in welding signals at different scales, so as to more comprehensively and deeply reveal information related to welding quality. The attention mechanism module plays an intelligent adjustment role. It can automatically assign weights according to the importance of different features in the final quality judgment, ensuring that the model can focus more on those features that have a decisive impact on welding quality. Obtaining the welding quality of the metal includes three types: under-welding, qualified, and over-welding.
[0054] Step 203: Input the multiple target welding signals collected during the actual metal welding process into the welding quality prediction model for prediction to obtain a welding quality prediction result.
[0055] In the above steps 201 to 203, by obtaining multiple welding signals during the ultrasonic metal welding process; inputting the multiple welding signals into a neural network model for training to obtain a welding quality prediction model; inputting the multiple target welding signals collected during the actual metal welding process into the welding quality prediction model for prediction to obtain a welding quality prediction result, not only can the quality of the solder joints be accurately evaluated without damaging the solder joints, but also the detection efficiency can be effectively improved and the cost can be reduced, thus ensuring the high-quality production and long-term stable operation of the IGBT module.
[0056] In one embodiment, the multi-scale feature extraction module includes multi-path convolutional layers. The multi-scale feature extraction module is used to perform multi-scale extraction on each of the welding signals, and obtaining the corresponding first feature maps includes the following content: Input each of the welding signals into the convolutional layers of each path for convolution operations to output fourth feature maps of different scales; fuse the fourth feature maps of different scales and perform weighted summation with the welding signals to obtain the corresponding first feature maps.
[0057] Among them, the structure of the multi-scale feature extraction module is as Figure 3As shown, this module adopts a multi-path parallel structure, integrating various types of convolutional layers such as 1x1 convolution, 3x3 convolution, 5x5 depthwise separable convolution, and 7x7 dilated convolution. These convolutional layers focus on extracting feature information of different scales and complexities. The outputs of each path are then fused through 1x1 convolution and weighted sum with the original input features according to certain weights, thus achieving effective enhancement of features.
[0058] In one embodiment, the parallel fusion attention mechanism module includes a spatial attention module and a channel attention module. The parallel fusion attention mechanism module is used to extract features from the second feature map to obtain a third feature map, including the following steps: inputting the second feature map into the spatial attention module to obtain spatial attention feature weights; and inputting the second feature map into the channel attention module to obtain channel attention feature weights; based on the spatial attention feature weights and the channel attention feature weights, performing weighted fusion on the second feature map to obtain a third feature map.
[0059] Among them, the structure of the multi-scale feature extraction module is as Figure 4 shown. The spatial attention mechanism focuses on the spatial distribution of the feature map, identifies and highlights the positions containing key information, and gives higher weights to these important spatial regions. The channel attention mechanism mainly focuses on the importance of different channels in the feature map. By calculating the contribution of each channel to the final result, higher weights are assigned to the key channels to enhance the model's attention to important features. Through the joint processing of the channel and spatial attention mechanisms, the model can express features more precisely, thereby improving the recognition ability of key features and the overall prediction performance.
[0060] In one embodiment, the classifier is used to classify based on the third feature map to obtain the welding quality of the metal, including the following steps: the third feature map sequentially passes through two convolutional layers, two pooling layers, and two fully connected layers, and finally outputs a prediction result.
[0061] In one embodiment, the step of inputting the various welding signals into the neural network model for training further includes: using a sliding window algorithm to convert each welding signal into a corresponding two-dimensional image, and inputting each two-dimensional image into the neural network model for training.
[0062] In one embodiment, before performing multi-scale extraction on each welding signal, the method further includes: inputting each welding signal into a convolutional layer for convolution operation.
[0063] In one embodiment, the various welding signals include at least two of a current signal, a voltage signal, a phase difference signal, and an ultrasonic frequency signal.
[0064] In one embodiment, training to obtain a welding quality prediction model includes the following steps:
[0065] Step 301, obtain various welding signals during the ultrasonic metal welding process, including current signals, voltage signals, phase difference signals, ultrasonic frequency signals, etc.
[0066] Step 302, adopt the sliding window conversion technology to convert the collected one-dimensional signals into two-dimensional grayscale images for more intuitively displaying signal features. First, use the linear interpolation method to interpolate the collected original one-dimensional signals (including voltage, current, power, and phase difference) until each signal is expanded to 4096 sample points. Next, adopt the sliding window technology (Sliding Window Transition, SWT) to convert the one-dimensional signals into a 64×64 (4096) two-dimensional grayscale image and construct a data set that meets the requirements of neural network model training.
[0067] Step 303, construct a neural network model based on multi-signal feature-level fusion. Figure 5 This is the specific structure of the neural network model: Each welding signal first passes through a convolutional layer and then enters an R-Inception multi-scale feature extraction module. After that, these four signals are fused and input into the parallel fusion attention mechanism module (PFAM). Then, the data sequentially passes through two convolutional layers, two pooling layers, and two fully connected layers, and finally outputs the classification result.
[0068] In one embodiment, as Figure 6 shown, the embodiment of the present application further provides an ultrasonic metal welding quality detection device based on multi-signal fusion. The device includes:
[0069] An acquisition module 10 for acquiring various welding signals during the ultrasonic metal welding process;
[0070] A training module 20 for inputting the various welding signals into the neural network model for training to obtain a welding quality prediction model; wherein, the neural network model includes a multi-scale feature extraction module and a parallel fusion attention mechanism module. The multi-scale feature extraction module is used to perform multi-scale extraction on each of the welding signals to obtain corresponding first feature maps, and the first feature maps are fused to obtain a second feature map. The parallel fusion attention mechanism module is used to perform feature extraction on the second feature map to obtain a third feature map, and the classifier is used to classify based on the third feature map to obtain the welding quality of the metal;
[0071] A prediction module 30 for inputting various target welding signals collected during the actual metal welding process into the welding quality prediction model for prediction to obtain a welding quality prediction result.
[0072] In one embodiment, the multi-scale feature extraction module includes convolutional layers with multiple paths, and the training module 20 is further configured to: input each of the welding signals into the convolutional layers of each path for convolution operations to output fourth feature maps of different scales; fuse the fourth feature maps of different scales and perform weighted summation with the welding signals to obtain corresponding first feature maps.
[0073] In one embodiment, the parallel fusion attention mechanism module includes a spatial attention module and a channel attention module, and the training module 20 is further configured to: input the second feature map into the spatial attention module to obtain spatial attention feature weights; and input the second feature map into the channel attention module to obtain channel attention feature weights; perform weighted fusion on the second feature map based on the spatial attention feature weights and the channel attention feature weights to obtain a third feature map.
[0074] In one embodiment, the training module 20 is further configured to: sequentially pass the third feature map through two convolutional layers, two pooling layers, and two fully connected layers, and finally output a prediction result.
[0075] In one embodiment, the training module 20 is further configured to: convert each of the welding signals into a corresponding two-dimensional image by using a sliding window algorithm, and input each of the two-dimensional images into a neural network model for training.
[0076] In one embodiment, before performing multi-scale extraction on each of the welding signals, the training module 20 is further configured to: input each of the welding signals into a convolutional layer for convolution operations.
[0077] In one embodiment, the multiple welding signals include at least two of a current signal, a voltage signal, a phase difference signal, and an ultrasonic frequency signal.
[0078] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as Figure 7As shown in the figure. The computer device includes a processor, a memory, a communication interface, a display screen, and an input device connected via a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, a mobile cellular network, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements a method for detecting the quality of ultrasonic metal welding based on multi-signal fusion. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a button, a trackball, or a touchpad provided on the housing of the computer device, or an external keyboard, a touchpad, or a mouse, etc.
[0079] Those skilled in the art can understand that Figure 7 the structure shown in the figure is only a block diagram of some structures related to the solution of this application, and does not constitute a limitation on the computer device to which the solution of this application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine some components, or have a different component layout.
[0080] In an exemplary embodiment, a computer device is provided, including a memory and a processor. A computer program is stored in the memory. When the processor executes the computer program, the following steps are implemented:
[0081] Obtain various welding signals during the ultrasonic metal welding process;
[0082] Input the various welding signals into a neural network model for training to obtain a welding quality prediction model; wherein, the neural network model includes a multi-scale feature extraction module, a parallel fusion attention mechanism module, and a classifier. The multi-scale feature extraction module is used to perform multi-scale extraction on each of the welding signals to obtain corresponding first feature maps, and the second feature map is obtained after fusing each of the first feature maps. The parallel fusion attention mechanism module is used to perform feature extraction on the second feature map to obtain a third feature map, and the classifier is used to classify based on the third feature map to obtain the welding quality of the metal;
[0083] Input various target welding signals collected during the actual metal welding process into the welding quality prediction model for prediction to obtain a welding quality prediction result.
[0084] In one embodiment, a signal acquisition system for ultrasonic metal welding process is further provided, including the computer device, DSP control board, and ultrasonic metal welding machine as described in the third aspect above. The computer device is connected to the DSP control board, and the DSP control board is connected to the ultrasonic metal welding machine, where: The computer device is configured to set welding parameters and send a start instruction to the DSP control board; The DSP control board is configured to receive the start instruction to generate a high-frequency ultrasonic signal and send it to the ultrasonic metal welding machine; The ultrasonic metal welding machine is configured to receive the high-frequency ultrasonic signal to start the welding process; The DSP control board is further configured to collect various welding signals during the welding process and transmit them to the computer device for storage.
[0085] In the embodiment of the present application, an ultrasonic metal welding machine, a DSP control board, and a computer device are used to build a complete signal acquisition system for ultrasonic metal welding process. The working process of this system is as follows: First, key parameters of the ultrasonic transducer are set through the computer device, including ultrasonic amplitude, welding time, welding pressure, etc. Subsequently, the computer device sends an instruction to the DSP control board to indicate it to start generating a high-frequency ultrasonic signal. This signal will drive the ultrasonic transducer to start the welding process. During the welding process, the DSP control board is responsible for collecting various signals during the welding process. These signals are transmitted to the computer device in real time through the USB interface. After receiving these signals, the computer device will display and store them for subsequent data analysis and processing.
[0086] In one embodiment, the computer device includes a memory and a processor. A computer program is stored in the memory. When the processor executes the computer program, the following steps are implemented:
[0087] Obtain various welding signals during the ultrasonic metal welding process;
[0088] Input the various welding signals into a neural network model for training to obtain a welding quality prediction model. Wherein, the neural network model includes a multi-scale feature extraction module, a parallel fusion attention mechanism module, and a classifier. The multi-scale feature extraction module is configured to perform multi-scale extraction on each of the welding signals to obtain corresponding first feature maps. After the first feature maps are fused, a second feature map is obtained. The parallel fusion attention mechanism module is configured to perform feature extraction on the second feature map to obtain a third feature map. The classifier is configured to perform classification based on the third feature map to obtain the welding quality of the metal;
[0089] Input various target welding signals collected during the actual metal welding process into the welding quality prediction model for prediction to obtain a welding quality prediction result.
[0090] The technical features of the above-described embodiments can be combined arbitrarily. For the sake of concise description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope recorded in this specification.
[0091] The above-described embodiments only represent several implementation manners of the present application. The description is relatively specific and detailed, but it should not be construed as a limitation on the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the appended claims.
Claims
1. An ultrasonic metal welding quality detection method based on multi-signal fusion, characterized in that The method includes: Obtaining various welding signals during ultrasonic metal welding; Inputting the various welding signals into a neural network model for training to obtain a welding quality prediction model; wherein, the neural network model includes a multi-scale feature extraction module, a parallel fusion attention mechanism module, and a classifier. The multi-scale feature extraction module is used to perform multi-scale extraction on each of the welding signals to obtain corresponding first feature maps. After fusing each of the first feature maps, a second feature map is obtained. The parallel fusion attention mechanism module is used to perform feature extraction on the second feature map to obtain a third feature map. The classifier is used to perform classification based on the third feature map to obtain the welding quality of the metal; Inputting various target welding signals collected during the actual metal welding process into the welding quality prediction model for prediction to obtain a welding quality prediction result.
2. The method according to claim 1, characterized in that, The multi-scale feature extraction module includes multi-path convolutional layers. The multi-scale feature extraction module is used to perform multi-scale extraction on each of the welding signals to obtain corresponding first feature maps, including: Inputting each of the welding signals into the convolutional layers of each path for convolution operations to output fourth feature maps of different scales; Fusing the fourth feature maps of different scales and performing weighted summation with the welding signals to obtain corresponding first feature maps.
3. The method according to claim 2, characterized in that, The parallel fusion attention mechanism module includes a spatial attention module and a channel attention module. The parallel fusion attention mechanism module is used to perform feature extraction on the second feature map to obtain a third feature map, including: Inputting the second feature map into the spatial attention module to obtain spatial attention feature weights; and inputting the second feature map into the channel attention module to obtain channel attention feature weights; Performing weighted fusion on the second feature map based on the spatial attention feature weights and the channel attention feature weights to obtain a third feature map.
4. The method according to claim 3, characterized in that The third feature map is used to predict the welding quality, including: The third feature map sequentially passes through two convolutional layers, two pooling layers, and two fully connected layers, and finally outputs a prediction result.
5. The method according to claim 1, wherein The step of inputting the various welding signals into the neural network model for training further includes: Using a sliding window algorithm to convert each of the welding signals into a corresponding two-dimensional image, and inputting each of the two-dimensional images into the neural network model for training.
6. The method according to claim 1, characterized in that, Before performing multi-scale extraction on each of the welding signals, the method further includes: Inputting each of the welding signals into a convolutional layer for convolution operations.
7. The method according to claim 1, characterized in that, The various welding signals include at least two of a current signal, a voltage signal, a phase difference signal, and an ultrasonic frequency signal.
8. An ultrasonic metal welding quality detection device based on multi-signal fusion, characterized in that, The device includes: An acquisition module, configured to acquire various welding signals during ultrasonic metal welding; A training module for inputting the multiple welding signals into a neural network model for training to obtain a welding quality prediction model. Wherein, the neural network model includes a multi-scale feature extraction module, a parallel fusion attention mechanism module, and a classifier. The multi-scale feature extraction module is used for multi-scale extraction of each of the welding signals to obtain corresponding first feature maps. After the first feature maps are fused, a second feature map is obtained. The parallel fusion attention mechanism module is used for feature extraction of the second feature map to obtain a third feature map. The classifier is used for classification based on the third feature map to obtain the welding quality of the metal. A prediction module for inputting multiple target welding signals collected during the actual metal welding process into the welding quality prediction model for prediction to obtain a welding quality prediction result.
9. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, it implements the method according to any one of claims 1 to 7.
10. An ultrasonic metal welding process signal acquisition system, characterized in that, It includes a computer device, a DSP control board, and an ultrasonic metal welding machine according to claim 9. The computer device is connected to the DSP control board, and the DSP control board is connected to the ultrasonic metal welding machine, wherein: The computer device is used for setting welding parameters and sending a start instruction to the DSP control board. The DSP control board is used for receiving the start instruction to generate a high-frequency ultrasonic signal and sending it to the ultrasonic metal welding machine. The ultrasonic metal welding machine is used for receiving the high-frequency ultrasonic signal to start the welding process. The DSP control board is further used for collecting various welding signals during the welding process and transmitting them to the computer device for storage.