Battery tab laser welding quality online monitoring method and system

Abstract

A method and system for online monitoring of laser welding quality of battery connectors, relating to the field of data processing technology, is disclosed. In this method, thermal signal data during the laser welding process is acquired to construct thermal stress spatiotemporal field data, stress anomaly features are extracted, and potential risk areas are identified. Acoustic emission signals from the potential risk areas are acquired and subjected to dual-modal cross-validation with the thermal stress spatiotemporal field data to determine crack initiation areas. Stress directions are extracted, and local weld images are subjected to directional enhancement processing to extract crack boundary contours. The frequency domain features of the acoustic emission signals are analyzed to obtain crack propagation parameters, and the crack propagation prediction area is calculated based on the crack boundary contours. A coupled evaluation is performed based on the multidimensional risk parameters of the crack propagation prediction area and the crack initiation area to finally determine a list of weld hazard locations. Implementing the technical solution provided in this application enables accurate identification of minute defects during the welding process, improving the online monitoring capability of laser welding quality for battery connectors.

Description

Technical Field

[0001] This invention belongs to the field of data processing technology, specifically a method and system for online monitoring of the quality of laser welding of battery connectors. Background Technology

[0002] With the rapid development of new energy vehicles and the energy storage industry, laser welding technology for battery connectors has become a core process in battery module manufacturing due to its high efficiency and precision. The quality of laser welding directly affects the conductivity, mechanical strength, and lifespan of the battery system; therefore, effective quality monitoring of the welding process is of great significance. Among these, the early identification and precise location of welding defects are crucial for ensuring product reliability, and their monitoring capabilities have a decisive impact on the overall quality of the battery module.

[0003] Currently, online monitoring of the quality of laser welding of battery connectors mainly relies on acquiring characteristic signals during the welding process using a single type of sensor. For example, some methods use temperature sensors to monitor the temperature distribution in the welding area and analyze abnormal changes in the temperature field to identify potential welding defects; other methods utilize vision inspection systems to acquire images of the weld surface and identify surface morphology features through image processing algorithms. These methods directly acquire specific physical quantity information during the welding process to evaluate welding quality, are relatively simple to operate, and are widely used in industrial production.

[0004] However, with increasingly stringent quality control requirements for battery connector welding, the limitations of existing monitoring methods are becoming increasingly apparent. In practical applications, the rapid heating and cooling cycles of laser welding generate complex thermal stress distributions within the material. This accumulation of thermal stress easily induces the initiation of microcracks in and around the weld. Due to a lack of in-depth analysis of the correlation mechanism between thermal stress evolution and crack initiation during welding, signals acquired by a single sensor often fail to accurately capture this dynamic evolution process from stress concentration to crack formation. Especially for early-stage microcracks, their small size and indistinct features make reliable identification difficult using only a single signal source. This insufficient monitoring capability leads to potential welding defects being overlooked during the production stage, making it difficult to accurately locate potentially defective weld areas and their subsequent propagation trends, thus failing to meet the quality control requirements for high-reliability battery module manufacturing. Summary of the Invention

[0005] To address the above problems, this invention provides an online monitoring method and system for the quality of laser welding of battery connectors. This system solves the problem of accurately identifying and locating early micro-cracks and their potential propagation areas in the weld in the prior art. It enables precise identification of micro-defects during the welding process and improves the online monitoring capability of laser welding quality of battery connectors.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: Thermal signal data during the laser welding process of battery connectors is acquired, and thermal stress spatiotemporal field data is constructed based on the thermal signal data; Stress anomaly features are extracted from the thermal stress spatiotemporal field data, and potential risk areas are identified based on the stress anomaly features. Acoustic emission signals corresponding to the potential risk areas are obtained, and dual-modal cross-validation is performed based on the acoustic emission signals and the thermal stress spatiotemporal field data to determine the crack initiation area; Extract the stress direction of the crack initiation region, and perform directional enhancement processing on the local weld image based on the stress direction to extract the crack boundary contour; The crack propagation parameters are obtained by analyzing the frequency domain characteristics of the acoustic emission signal, and the crack propagation prediction region is calculated based on the crack propagation parameters and the crack boundary profile. A coupled assessment is performed based on the multidimensional risk parameters of the crack propagation prediction region and the crack initiation region to determine the list of potential weld hazards.

[0007] By employing the above technical solution, thermal signal data during the laser welding of battery connectors is acquired, and thermal stress spatiotemporal field data is constructed. Furthermore, stress anomaly features are extracted from the thermal stress spatiotemporal field data, and potential risk areas are identified based on these features. This enables the keen detection of early stress concentration phenomena, narrowing the subsequent monitoring range and improving the targeting of defect localization. On this basis, acoustic emission signals corresponding to the potential risk areas are acquired, and dual-modal cross-validation is performed based on the acoustic emission signals and the thermal stress spatiotemporal field data to determine the crack initiation region. Subsequently, the stress direction of the crack initiation region is extracted, and the local weld is analyzed based on the stress direction. The image undergoes targeted enhancement processing, effectively highlighting early cracks that were originally small in size and lacked obvious features, thereby accurately extracting the crack boundary contour. Simultaneously, the frequency domain features of the acoustic emission signal are analyzed to obtain crack propagation parameters. Based on these parameters and the crack boundary contour, a crack propagation prediction region is calculated, overcoming the limitations of traditional methods that can only perform post-event static defect detection. This enables proactive prediction of the subsequent propagation trend of potential welding hazards. Finally, a coupled evaluation is performed based on the multi-dimensional risk parameters of the crack propagation prediction region and the crack initiation region to determine a list of weld hazard locations. This achieves comprehensive quantitative evaluation and precise location of welding defects from initiation, formation, to propagation. In summary, this invention effectively solves the problem of accurately identifying and locating early micro-cracks and their potential propagation regions in welds, improving the online monitoring capability of laser welding quality for battery connectors and meeting the quality control requirements of battery module manufacturing.

[0008] Optionally, the thermal signal data is subjected to spatial domain filtering to obtain temperature spatial distribution data; the thermal expansion coefficient and elastic modulus of the battery connector material are obtained; based on the temperature spatial distribution data, the thermal expansion coefficient and the elastic modulus, the thermal stress tensor components at the corresponding positions of the pixels are calculated to obtain thermal stress spatiotemporal field data containing time series.

[0009] Optionally, a time evolution rate field and a stress spatial gradient field are calculated for the stress values ​​of pixels in the thermal stress spatiotemporal field data; temporal anomalies with evolution rates exceeding a preset rate threshold and spatial anomalies with gradient magnitudes exceeding a preset gradient threshold are identified in the temporal evolution rate field and the stress spatial gradient field, respectively, and the temporal anomalies and the spatial anomalies are used as stress anomaly features; pixels that simultaneously belong to the temporal anomalies and the spatial anomalies are extracted as composite anomalies; the spatial region where the composite anomalies are located is traversed with a sliding window of a preset size, and the proportion of the number of composite anomalies in each sliding window is calculated to obtain the distribution density; the regions where the sliding windows with distribution densities greater than a preset density threshold are filtered, and the filtered regions are merged into connected components, and the merged regions are marked as potential risk regions.

[0010] Optionally, the acoustic emission signal is filtered to extract the effective acoustic emission signal, and the spatial distribution data of the acoustic emission energy of the effective acoustic emission signal is calculated; a spatial coordinate registration relationship between the potential risk area and the spatial distribution data of the acoustic emission energy is established, and the thermal signal characteristic value and acoustic signal characteristic value within the potential risk area are calculated respectively; a thermo-acoustic coupling verification coefficient is calculated based on the thermal signal characteristic value and the acoustic signal characteristic value, and potential risk areas whose thermo-acoustic coupling verification coefficient exceeds a preset coupling threshold are selected as crack initiation areas.

[0011] Optionally, the spatial coordinate range of the potential risk area is obtained, and the spatial coordinate range is mapped to the corresponding spatial location of the acoustic emission energy spatial distribution data to establish a spatial coordinate registration relationship; based on the spatial coordinate registration relationship, stress data within the potential risk area is extracted from the thermal stress spatiotemporal field data, and acoustic emission energy data at the corresponding spatial location is extracted from the acoustic emission energy spatial distribution data; thermal signal characteristic values ​​are calculated based on the stress data, and acoustic signal characteristic values ​​are calculated based on the acoustic emission energy data.

[0012] Optionally, the stress gradient vector corresponding to the crack initiation region is extracted from the thermal stress spatiotemporal field data, and the statistical average value of the gradient direction angle is calculated as the stress direction; the stress direction is used as the filtering direction to construct a directional filter, and the directional filter is used to perform convolution operation on the local weld image to obtain a stress direction enhanced image; the gray-level gradient of the pixels in the stress direction enhanced image is calculated, and the gray-level gradient is binarized based on a preset gradient threshold to extract the crack boundary contour.

[0013] Optionally, the effective acoustic emission signal after filtering the acoustic emission signal is obtained; the effective acoustic emission signal is transformed in the frequency domain to obtain a frequency domain power spectral density function; the frequency corresponding to the first peak with the largest amplitude is extracted from the frequency domain power spectral density function as the principal frequency, and the frequency corresponding to the second peak with the second largest amplitude is extracted as the secondary principal frequency; the principal frequency and the secondary principal frequency are used as frequency domain features; a frequency domain attenuation coefficient is calculated based on the principal frequency and the secondary principal frequency, and the potential velocity and influence radius of crack propagation are inferred based on the frequency domain attenuation coefficient; the potential velocity and the influence radius are used as crack propagation parameters; and the crack propagation prediction region is calculated based on the crack propagation parameters and the crack boundary profile.

[0014] Optionally, the current extension angle of the crack boundary profile is extracted, and the predicted direction angle of crack propagation is calculated in combination with the stress direction; a fan-shaped region is constructed based on the predicted direction angle, the potential velocity, and the radius of influence range, starting from the endpoint of the crack boundary profile, and the fan-shaped region is used as the crack propagation prediction region.

[0015] Optionally, based on the crack propagation prediction region and the crack initiation region, multiple pairs of associated regions with a relative distance less than a distance threshold are constructed; the thermoacoustic coupling verification coefficient, peak thermal stress, and area of ​​the initiation region are extracted from the crack initiation region of each pair of associated regions and used as initiation dimension risk parameters; the potential velocity, radius of influence, and predicted direction angle are extracted from the crack propagation prediction region of each pair of associated regions and used as extension dimension risk parameters; the initiation dimension risk parameters and the extension dimension risk parameters are normalized respectively, and the normalized initiation dimension risk parameters and extension dimension risk parameters are weighted and coupled based on a preset weight coefficient matrix to obtain a risk score for each pair of associated regions; the risk scores are classified according to a preset risk grading standard, the spatial coordinates and risk level of each pair of associated regions are extracted, and they are arranged in descending order of risk level to generate the weld hazard location list.

[0016] In a second aspect, embodiments of this application provide an online monitoring system for the quality of laser welding of battery connectors. The online monitoring system for the quality of laser welding of battery connectors includes: one or more processors and a memory; the memory is coupled to the one or more processors and is used to store computer program code, the computer program code including computer instructions, and the one or more processors call the computer instructions to cause the online monitoring system for the quality of laser welding of battery connectors to perform the method described in the first aspect and any possible implementation thereof.

[0017] In summary, one or more technical solutions provided in this application have at least the following technical effects or advantages: By employing the above technical solution, the touch area is divided into edge, center, and transition regions according to spatial location characteristics, and the static capacitance reference value of each region under static reference conditions is obtained. Based on this, a target frequency band filter is constructed by extracting the frequency domain features of the background noise sample sequence, which can specifically filter out frequency domain interference components in the reference values ​​of each region. Furthermore, based on the filtered and enhanced reference values ​​of each region, capacitance distribution characteristic parameters are calculated, and a distributed reference capacitance matrix that continuously transitions from the edge to the center region is constructed. This distributed reference mechanism fully considers the differences in capacitance response characteristics at different locations within the touch area, overcoming the limitation of traditional globally unified references that cannot accurately reflect the actual touch state of each region. When a touch occurs, by combining the distributed reference capacitance matrix with dynamic capacitance sensing values ​​to calculate the signal matching deviation and response difference ratio of each region, the degree of response deviation in each region can be accurately quantified, and the type and region of interference can be identified, achieving precise localization of interference factors. Based on this, a deviation correction coefficient is calculated for each region according to the identified interference type and interference area. This coefficient is then used to specifically compensate for the dynamic capacitive sensing value. When the compensated touch area reaches a preset global equilibrium state, the deviation correction coefficient is locked as a touch consistency configuration parameter, thus establishing a complete zoned adaptive adjustment mechanism. This mechanism eliminates the impact of spatial response differences in touch areas on recognition accuracy by implementing differentiated benchmark adjustments and deviation compensation for different regions, improving the consistency and anti-interference capability of touch response, and enabling the touch system to meet the precise operation requirements of industrial control applications. Attached Figure Description

[0018] Figure 1 This is a schematic flowchart of an online monitoring method for the quality of laser welding of battery connectors disclosed in an embodiment of this application; Figure 2 This is another schematic flowchart of an online monitoring method for the quality of laser welding of battery connectors disclosed in an embodiment of this application; Figure 3This is a schematic diagram of the structure of a system provided in an embodiment of this application.

[0019] In the diagram: 301, Central Processing Unit; 302, Read-Only Memory; 303, Random Access Memory; 304, Bus; 305, Input / Output Interface; 306, Input Section; 307, Output Section; 308, Storage Section; 309, Communication Section; 310, Driver; 311, Removable Media. Detailed Implementation

[0020] To enable those skilled in the art to better understand the technical solution, the present invention will be described in detail below with reference to embodiments. The description in this part is only exemplary and explanatory, and should not be used to limit the scope of protection of the present invention in any way.

[0021] It should be noted that, in this document, the terms "comprising," "including," and any other variations are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Specific examples have been used in this document to illustrate the principles and implementation methods of the present invention. These examples are merely for the purpose of helping to understand the method and core ideas of the present invention. The above are only preferred embodiments of the present invention. It should be pointed out that, due to the limitations of written expression and the objective existence of infinite specific structures, those skilled in the art can make several improvements, modifications, or variations without departing from the principles of the present invention, and can also combine the above technical features in an appropriate manner. These improvements, modifications, variations, or combinations, or the direct application of the concept and technical solution of the present invention to other situations without modification, should all be considered within the scope of protection of the present invention.

[0022] This application provides a method for online monitoring of the quality of laser welding of battery connectors, referring to... Figure 1 , Figure 1 This is a flowchart illustrating an online monitoring method for the laser welding quality of battery connectors according to an embodiment of this application. The method is applied to a system, which refers to a hardware and software integrated platform capable of executing an online monitoring program for the laser welding quality of battery connectors. The system can execute an online monitoring program for the laser welding quality of battery connectors. The method includes steps 101 to 106, as follows: Step 101: Acquire thermal signal data during the laser welding of battery connectors and construct thermal stress spatiotemporal field data based on the thermal signal data.

[0023] In the embodiments of this application, thermal stress spatiotemporal field data refers to a data set that records the internal stress distribution of the material in the welding area due to temperature gradient changes in both time and space dimensions, such as a three-dimensional data matrix containing the stress tensor values ​​of each coordinate point of the weld at different times.

[0024] Specifically, the continuous temperature distribution image sequence acquired by the infrared thermal imaging sensor is used as thermal signal data; the temperature values ​​of each spatial pixel in the thermal signal data are extracted over a continuous time series; the temperature gradient between adjacent spatial pixels is calculated by combining the pre-acquired thermal expansion coefficient, elastic modulus, and Poisson's ratio parameters of the battery connector material; the temperature gradient is multiplied by the thermal expansion coefficient and elastic modulus to calculate the thermal stress tensor at each spatial location at different sampling times; the thermal stress tensors of all spatial pixels at all sampling times are matrix-stitched according to the spatial coordinate axis and the time axis to construct the thermal stress spatiotemporal field data.

[0025] In one possible implementation, thermal signal data is acquired during the laser welding of the battery connector, and thermal stress spatiotemporal field data is constructed based on the thermal signal data. Specifically, this includes steps 1011-1012, as follows: Step 1011: Perform spatial domain filtering on the thermal signal data to obtain spatial temperature distribution data.

[0026] In the embodiments of this application, temperature spatial distribution data refers to the collection of temperature values ​​of various discrete points recorded in a two-dimensional or three-dimensional spatial coordinate system, such as a matrix image containing the absolute temperature values ​​of each pixel on the weld surface after smoothing and denoising.

[0027] Specifically, the process involves extracting the two-dimensional pixel matrix corresponding to each sampling time in the thermal signal data; setting a sliding window of a preset size on the two-dimensional pixel matrix and obtaining the initial temperature values ​​of each adjacent pixel within the sliding window; weighting and summing the initial temperature values ​​of each adjacent pixel within the sliding window according to a preset spatial distance weight to calculate the smoothed temperature value of the center pixel of the sliding window; replacing the initial temperature value of the corresponding center pixel in the two-dimensional pixel matrix with the smoothed temperature value; traversing all pixels in the two-dimensional pixel matrix according to a preset step size to complete the spatial domain filtering of the thermal signal data; and using the processed two-dimensional pixel matrix as the temperature spatial distribution data at that sampling time.

[0028] Step 1012: Obtain the coefficient of thermal expansion and elastic modulus of the battery connector material; based on the temperature spatial distribution data, coefficient of thermal expansion and elastic modulus, calculate the thermal stress tensor components at the corresponding positions of the pixels to obtain the thermal stress spatiotemporal field data containing the time series.

[0029] Specifically, the thermal expansion coefficient and elastic modulus values ​​corresponding to the battery connector material are read from a preset material property database; the current temperature value of each pixel in the temperature spatial distribution data is extracted, and the temperature difference between the current temperature value and the preset reference ambient temperature value is calculated; the temperature difference is multiplied by the thermal expansion coefficient to calculate the thermal strain value at the corresponding position of each pixel; the thermal strain value is multiplied by the elastic modulus to calculate the normal stress value and shear stress value of each pixel in different spatial coordinate axis directions, which are used as the thermal stress tensor components at the corresponding position of the pixel; the above calculation process is repeated by extracting the temperature spatial distribution data at multiple consecutive sampling times, and the thermal stress tensor components of all pixels at all sampling times are multidimensionally matrix-stitched according to the time sequence and spatial coordinate system to obtain the thermal stress spatiotemporal field data containing the time series.

[0030] Step 102: Extract stress anomaly features from the thermal stress spatiotemporal field data, and identify potential risk areas based on the stress anomaly features.

[0031] In the embodiments of this application, stress anomaly characteristics refer to local abrupt properties in thermal stress distribution that deviate from the normal physical evolution law or exceed the material's bearing limit, such as a sudden increase in stress peak value in stress concentration area or a drastic fluctuation in stress gradient.

[0032] Specifically, partial derivatives of the thermal stress spatiotemporal field data in both time and space dimensions are calculated to obtain the stress change rate matrix. The statistical variance of each data point in the stress change rate matrix and its neighboring data points is calculated. Data points with statistical variances greater than a preset variance threshold and whose corresponding absolute thermal stress values ​​exceed the material's yield strength are extracted as stress anomaly features. Neighboring data points with stress anomaly features are spatially clustered, and the geometric center coordinates and coverage radius of the clusters are calculated. The two-dimensional spatial range with the geometric center coordinates as the center and the coverage radius as the radius is marked as the potential risk area.

[0033] In one possible implementation, stress anomaly features are extracted from the spatiotemporal field data of thermal stress, and potential risk areas are identified based on these stress anomaly features. Specifically, this includes steps 1021-1023, as follows: Step 1021: Calculate the time evolution rate field and stress spatial gradient field for the stress values ​​of pixels in the thermal stress spatiotemporal field data.

[0034] In this embodiment, the time evolution rate field refers to the spatial distribution data that records the rate of change of physical quantities over time. It is a rate of change matrix formed by calculating the numerical difference between corresponding spatial locations at adjacent time nodes. For example, it is a two-dimensional or three-dimensional matrix containing the rate of increase or decrease of thermal stress of each pixel in the weld within a continuous millisecond time interval.

[0035] Specifically, the stress values ​​of each pixel in the thermal stress spatiotemporal field data are extracted over a continuous time series; the stress value difference between adjacent sampling times of each pixel is calculated, and the stress value difference is divided by the time interval between adjacent sampling times to obtain the time change rate corresponding to each pixel. The time change rates of all pixels are then concatenated into a matrix according to the spatial coordinate system to obtain the time evolution rate field; the stress values ​​of each pixel in the thermal stress spatiotemporal field data at the same sampling time are extracted, and the stress value difference between each pixel and its neighboring pixels in multiple spatial coordinate axes is calculated. The stress value difference is then divided by the corresponding spatial physical distance to obtain the partial derivatives of each pixel in different directions. The partial derivatives of each pixel are then vector-combined according to the spatial coordinate system to obtain the stress spatial gradient field.

[0036] Step 1022: Identify temporal anomalies where the evolution rate exceeds a preset rate threshold and spatial anomalies where the gradient magnitude exceeds a preset gradient threshold in the temporal evolution rate field and stress spatial gradient field, respectively, and use the temporal anomalies and spatial anomalies as stress anomaly features.

[0037] In this application embodiment, a time-domain outlier refers to a discrete spatial coordinate point where the physical quantity changes drastically in the time dimension and exceeds the normal fluctuation range. That is, a pixel point whose absolute value of time evolution rate is greater than the safety boundary value preset by the system. For example, a specific weld location point where the rate of increase of thermal stress exceeds the set megapascal per second threshold within a certain microsecond during the welding process.

[0038] Specifically, the evolution rate values ​​of each pixel in the temporal evolution rate field are traversed, and the absolute values ​​of the evolution rate values ​​are extracted. The absolute values ​​of the evolution rate values ​​are compared with a preset rate threshold, and pixels with absolute values ​​greater than the preset rate threshold are extracted and marked as temporal anomalies. The partial derivatives of each pixel in the stress spatial gradient field in different spatial coordinate axes are extracted, the sum of squares of the partial derivatives in each direction is calculated, and the square root of the sum of squares is performed to obtain the gradient magnitude of each pixel. The gradient magnitude is compared with a preset gradient threshold, and pixels with gradient magnitudes greater than the preset gradient threshold are extracted and marked as spatial anomalies. All marked temporal anomalies and all spatial anomalies are merged into a dataset, and the merged set of points is used as the stress anomaly feature.

[0039] Step 1023: Extract pixels that belong to both temporal and spatial outliers as composite outliers; traverse the spatial region where the composite outliers are located using a sliding window of a preset size, calculate the proportion of composite outliers in each sliding window, and obtain the distribution density; filter the regions where the sliding windows have a distribution density greater than a preset density threshold, merge the filtered regions into connected components, and mark the merged regions as potential risk regions.

[0040] In the embodiments of this application, a composite anomaly point refers to a point in space that exhibits anomalous attributes on multiple evaluation dimensions. That is, an intersection pixel in the spatial coordinate system that satisfies both the time evolution rate exceeding the standard and the spatial gradient change exceeding the standard. For example, a specific coordinate pixel in a thermal stress field where the stress changes drastically over time and has a large difference from the stress in the surrounding area.

[0041] Specifically, the spatial coordinate set of temporal anomalies is compared with the spatial coordinate set of spatial anomalies, and pixels with completely identical spatial coordinates are extracted as composite anomalies. A rectangular sliding window with fixed length and width pixel dimensions is constructed, and the sliding window is moved row by row and column by column within the two-dimensional spatial region where the composite anomalies are located according to a fixed pixel step size. At each position of the sliding window, the number of composite anomalies within the coverage area of ​​the current sliding window is counted, and the number of composite anomalies is divided by the total number of pixels contained in the sliding window to calculate the distribution density at each position. The distribution density is compared with a preset density threshold, and the spatial region covered by the sliding window with a distribution density greater than the preset density threshold is retained. It is determined whether the boundaries between the retained spatial regions are adjacent or overlapping. The spatial regions with adjacent or overlapping boundaries are spliced ​​together to obtain connected components, and the overall spatial range of the merged connected components is marked as a potential risk area.

[0042] Step 103: Obtain the acoustic emission signal corresponding to the potential risk area, and perform dual-modal cross-validation based on the acoustic emission signal and thermal stress spatiotemporal field data to determine the crack initiation area.

[0043] In the embodiments of this application, dual-modal cross-validation refers to the matching process of mutual verification in time synchronization and spatial mapping using sensor data with two different physical properties. For example, thermal stress data reflecting macroscopic force and acoustic wave data reflecting microscopic fracture are spatiotemporally compared.

[0044] Specifically, acoustic emission signals within a time window matching the occurrence time of the potential risk area are extracted from the global audio data collected by the acoustic emission sensor; the acoustic emission signals are subjected to time-frequency transformation to extract the timestamp corresponding to the high-frequency burst energy peak; the stress release amount within the potential risk area under the timestamp is extracted from the thermal stress spatiotemporal field data; the product of the high-frequency burst energy peak and the stress release amount is calculated as the cross-validation confidence level; when the cross-validation confidence level is greater than the preset confidence threshold, the cross-validation is deemed to have passed, and the specific spatial coordinate range of the potential risk area is determined as the crack initiation area.

[0045] In one possible implementation, acoustic emission signals corresponding to potential risk areas are acquired, and dual-modal cross-validation is performed based on the acoustic emission signals and thermal stress spatiotemporal field data to determine crack initiation regions. Specifically, this includes steps 1031-1033, as follows: Step 1031: Filter the acoustic emission signal to extract the effective acoustic emission signal, and calculate the spatial distribution data of the acoustic emission energy of the effective acoustic emission signal.

[0046] In the embodiments of this application, acoustic emission energy spatial distribution data refers to a data set that records the distribution of acoustic signal energy at discrete points in a spatial coordinate system. That is, it is a matrix formed by integrating the energy of the acoustic wave signal and mapping it to the corresponding physical location, such as a two-dimensional energy heat map containing the intensity values ​​of acoustic emission signal energy in each region of the battery connector.

[0047] Specifically, the raw acoustic emission signal is acquired, a preset frequency passband range is set, and a bandpass filter is used to filter the raw acoustic emission signal in the frequency domain to remove noise signals that exceed the frequency passband range. The retained signal is then extracted as the effective acoustic emission signal. The amplitude values ​​of the effective acoustic emission signal in the discrete time series are squared, and the squared values ​​are integrated and summed within a preset time window to calculate the acoustic emission energy value corresponding to each acoustic emission source. The spatial position coordinates of the acoustic emission sensor array and the time difference between the reception of the effective acoustic emission signal by each sensor are acquired, and the spatial coordinates of each acoustic emission source are calculated using the time difference of arrival (TDOA) localization algorithm. The acoustic emission energy values ​​are mapped to a preset two-dimensional or three-dimensional spatial grid according to the corresponding spatial coordinates, and interpolation calculations are performed on the grid nodes to obtain the spatial distribution data of the acoustic emission energy of the effective acoustic emission signal.

[0048] Step 1032: Establish the spatial coordinate registration relationship between the potential risk area and the spatial distribution data of acoustic emission energy, and calculate the thermal signal characteristic value and acoustic signal characteristic value within the potential risk area respectively.

[0049] In the embodiments of this application, the spatial coordinate registration relationship is used to represent the position mapping correspondence rules of different data sources under the same physical space reference system. That is, the geometric transformation matrix that aligns heterogeneous data collected by different sensors to a unified coordinate system, such as the mapping relationship of aligning the thermal stress pixel coordinates captured by the infrared thermal imager with the sound source coordinates located by the acoustic emission sensor by translation and rotation.

[0050] Specifically, the first spatial coordinates of the potential risk area in the thermal imager coordinate system and the second spatial coordinates of the acoustic emission energy spatial distribution data in the acoustic emission sensor coordinate system are extracted. A pre-calibrated rotation matrix and translation vector are obtained, and the first spatial coordinates are subjected to an affine transformation using the rotation matrix and translation vector to unify them into the reference coordinate system where the second spatial coordinates are located, establishing a spatial coordinate registration relationship between the potential risk area and the acoustic emission energy spatial distribution data. Based on the spatial coordinate registration relationship, the set of thermal stress values ​​within the coverage area of ​​the potential risk area in the reference coordinate system is extracted, and the average or maximum value of the thermal stress value set is calculated as the thermal signal characteristic value. The set of acoustic emission energy values ​​within the coverage area of ​​the potential risk area in the reference coordinate system is extracted, and the sum or peak value of the acoustic emission energy value set is calculated as the acoustic signal characteristic value.

[0051] In one possible implementation, a spatial coordinate registration relationship is established between the potential risk area and the spatial distribution data of acoustic emission energy. Thermal signal characteristic values ​​and acoustic signal characteristic values ​​within the potential risk area are calculated, specifically including steps 10321-10323, as follows: Step 10321: Obtain the spatial coordinate range of the potential risk area, map the spatial coordinate range to the corresponding spatial location of the acoustic emission energy spatial distribution data, and establish a spatial coordinate registration relationship.

[0052] In the embodiments of this application, spatial coordinate registration relationship refers to the geometric transformation mapping rule between spatial location data in different coordinate systems. That is, the mathematical relationship that transforms the boundary coordinates of a region in one coordinate system to the corresponding position in another coordinate system through operations such as translation, rotation or scaling. For example, the transformation matrix that maps the pixel coordinate system of a potential risk area captured by an infrared camera to the three-dimensional physical coordinate system located by an acoustic emission sensor.

[0053] Specifically, the set of boundary contour pixels of the potential risk area in the first coordinate system is extracted, and the x-coordinate and y-coordinate values ​​of each pixel in the boundary contour pixel set are obtained to determine the spatial coordinate range of the potential risk area. The translation vector parameters and rotation matrix parameters between the pre-calibrated first and second coordinate systems are obtained, where the second coordinate system is the spatial coordinate system where the acoustic emission energy spatial distribution data is located. The x-coordinate and y-coordinate values ​​of each pixel within the spatial coordinate range are multiplied by the rotation matrix parameters, and the results are added by the translation vector parameters to calculate the target x-coordinate and target y-coordinate values ​​in the second coordinate system. The area enclosed by the target x-coordinate and target y-coordinate values ​​is taken as the corresponding spatial location of the acoustic emission energy spatial distribution data, completing the mapping from the spatial coordinate range to the corresponding spatial location and establishing a spatial coordinate registration relationship.

[0054] Step 10322: Based on the spatial coordinate registration relationship, extract stress data in the potential risk area from the thermal stress spatiotemporal field data, and extract acoustic emission energy data at the corresponding spatial location from the acoustic emission energy spatial distribution data.

[0055] In the embodiments of this application, acoustic emission energy data refers to the set of quantified values ​​of the energy carried by the acoustic wave signal within a specific spatial range, that is, the energy integral value of the acoustic emission signal mapped to the corresponding spatial coordinate point after processing, such as the absolute value of acoustic energy at each discrete grid point located in the potential risk area mapping position in the acoustic emission energy spatial distribution map.

[0056] Specifically, based on the established spatial coordinate registration relationship, the first spatial index range corresponding to the potential risk area in the thermal stress spatiotemporal field data and the second spatial index range corresponding to the acoustic emission energy spatial distribution data are determined. The thermal stress spatiotemporal field data is traversed, and according to the first spatial index range, the thermal stress tensor component values ​​of each pixel point located within the potential risk area at the current sampling time are read one by one. All the read thermal stress tensor component values ​​are combined into a data set as the stress data within the potential risk area. The acoustic emission energy spatial distribution data is traversed, and according to the second spatial index range, the energy integral values ​​of each spatial grid node located within the corresponding spatial location are read one by one. All the read energy integral values ​​are collected and spliced ​​as the acoustic emission energy data of the corresponding spatial location.

[0057] Step 10323: Calculate the thermal signal characteristic value based on the stress data, and calculate the acoustic signal characteristic value based on the acoustic emission energy data.

[0058] In this embodiment of the application, the thermal signal characteristic value refers to a statistical representative value used to characterize the degree of concentration or overall intensity of thermal stress distribution in a specific spatial area. That is, it is a single quantitative index obtained by performing mathematical statistical operations on the stress data of multiple discrete points in the area, such as the arithmetic mean or maximum peak value of the thermal stress values ​​of all pixels in the potential risk area.

[0059] Specifically, the thermal stress values ​​of each pixel in the extracted stress data are obtained, and all thermal stress values ​​are summed to obtain the total stress value. The total number of pixels in the stress data is counted, and the total stress value is divided by the total number of pixels to calculate the average stress value. At the same time, the maximum stress value is selected by comparing the values ​​of all thermal stress values. The average stress value and the maximum stress value are weighted and summed according to a preset ratio to obtain the thermal signal feature value. The energy integral values ​​of each spatial grid node in the extracted acoustic emission energy data are obtained, and all energy integral values ​​are summed to obtain the total energy value of the region. The maximum energy value is selected by comparing the values ​​of all energy integral values. The total energy value of the region and the maximum energy value are weighted and summed according to a preset weight to obtain the acoustic signal feature value.

[0060] Step 1033: Calculate the thermo-acoustic coupling verification coefficient based on the thermal signal characteristic value and the acoustic signal characteristic value, and screen potential risk areas where the thermo-acoustic coupling verification coefficient exceeds the preset coupling threshold as crack initiation areas.

[0061] In the embodiments of this application, the thermoacoustic coupling verification coefficient refers to a quantitative index that comprehensively reflects the degree of spatial and temporal correlation between thermal stress anomalies and acoustic emission energy anomalies. It is a dimensionless value obtained by normalizing and fusing the characteristics of thermal and acoustic signals. For example, it is a percentage coefficient used to determine the probability of crack occurrence after multiplying the thermal stress gradient value of a certain region with the peak value of acoustic emission energy and standardizing it.

[0062] Specifically, pre-set thermal signal reference values ​​and acoustic signal reference values ​​are obtained. The thermal signal feature value is divided by the thermal signal reference value to obtain the normalized thermal feature, and the acoustic signal feature value is divided by the acoustic signal reference value to obtain the normalized acoustic feature. The normalized thermal feature is multiplied by a preset thermal weighting coefficient, and the normalized acoustic feature is multiplied by a preset acoustic weighting coefficient. The products are summed to calculate the thermo-acoustic coupling verification coefficient of the potential risk area. The thermo-acoustic coupling verification coefficients of all potential risk areas are iterated, and the values ​​of the thermo-acoustic coupling verification coefficients are compared with the preset coupling threshold. Potential risk areas with thermo-acoustic coupling verification coefficients less than or equal to the preset coupling threshold are eliminated, and potential risk areas with thermo-acoustic coupling verification coefficients greater than the preset coupling threshold are retained. The selected and retained spatial areas are marked as crack initiation areas.

[0063] Step 104: Extract the stress direction in the crack initiation region, and perform directional enhancement processing on the local weld image based on the stress direction to extract the crack boundary contour.

[0064] In the embodiments of this application, targeted enhancement processing refers to an image filtering operation that specifically amplifies the contrast of image pixels along a specific physical force direction, such as an image processing process that enhances image edges along the vertical direction of tensile stress to highlight fine cracks.

[0065] Specifically, the principal stress tensor of the crack initiation region is extracted from the thermal stress spatiotemporal field data, and the eigenvector of the principal stress tensor is calculated to determine the direction of maximum tensile stress. Local weld images containing the crack initiation region are acquired from a visual sensor. An anisotropic Gaussian direction filter is constructed using the angle perpendicular to the direction of maximum tensile stress as the filtering direction. The anisotropic Gaussian direction filter is spatially convolved with the local weld image to enhance the pixel gradient value perpendicular to the direction of tensile stress. Thresholding and edge connection operations are performed on the convolved image to extract continuous pixel connected regions as crack boundary contours.

[0066] In one possible implementation, the stress direction of the crack initiation region is extracted, and the local weld image is subjected to directional enhancement processing based on the stress direction to extract the crack boundary contour. Specifically, this includes steps 1041-1043, as follows: Step 1041: Extract the stress gradient vector corresponding to the crack initiation region from the thermal stress spatiotemporal field data, and calculate the statistical average value of the gradient direction angle as the stress direction.

[0067] In the embodiments of this application, the stress gradient vector refers to the physical quantity of the direction and magnitude of the maximum rate of change of thermal stress in a spatial region. That is, it is a vector data with direction and magnitude formed by taking the partial derivatives of the thermal stress field in different spatial coordinate axes. For example, it is a two-dimensional vector composed of the rate of change of stress along the horizontal and vertical coordinates at a certain point in a two-dimensional thermal stress field.

[0068] Specifically, the thermal stress spatiotemporal field data is traversed to extract the thermal stress values ​​corresponding to each spatial data point within the crack initiation region. The difference in thermal stress values ​​for each spatial data point along the horizontal and vertical coordinate axes is calculated. These differences are combined to obtain the stress gradient vector for each spatial data point. For each stress gradient vector, the arctangent function is used to calculate the ratio of the difference between the vertical and horizontal coordinate axes, yielding the gradient direction angle value for each spatial data point. The gradient direction angle values ​​for all spatial data points within the crack initiation region are summed to obtain the total direction angle value. The total number of spatial data points within the crack initiation region is counted, and the total direction angle value is divided by the total number of spatial data points to calculate the statistical average of the gradient direction angles. This statistical average is then used as the stress direction.

[0069] Step 1042: Construct a directional filter using the stress direction as the filtering direction, and use the directional filter to perform convolution operation on the local weld image to obtain a stress direction enhanced image.

[0070] In this embodiment, a directional filter refers to a convolution kernel matrix that has a high pass rate for image texture or edge features in a specific direction while suppressing features in other directions. In other words, it is a spatial filter template with directional selectivity generated according to a set angle parameter, such as a two-dimensional Gaussian derivative matrix generated according to a specific stress direction to highlight crack texture features in that direction.

[0071] Specifically, the calculated angle values ​​of the stress direction are obtained. Based on the preset filter size and angle values, a two-dimensional weight matrix with corresponding directional response characteristics is generated, and the two-dimensional weight matrix is ​​used as a directional filter. The local weld image to be processed is obtained and converted into a two-dimensional grayscale pixel matrix. The center position of the directional filter is aligned with the starting pixel of the two-dimensional grayscale pixel matrix, and each weight value in the directional filter is multiplied by the corresponding pixel grayscale value in the two-dimensional grayscale pixel matrix. All the multiplication results are summed to obtain the new grayscale value of the pixel at the center position after filtering. According to the preset sliding step size, the directional filter is moved row by row and column by column on the two-dimensional grayscale pixel matrix, and the multiplication and summation operations are repeated to complete the convolution operation on the local weld image. All the new grayscale values ​​obtained by traversing and calculating are combined according to their original spatial positions to obtain the stress direction enhancement image.

[0072] Step 1043: Calculate the gray-level gradient of the pixels in the stress direction enhancement image, and perform binarization on the gray-level gradient based on the preset gradient threshold to extract the crack boundary contour.

[0073] In the embodiments of this application, gray-level gradient refers to the spatial rate of change of gray-level values ​​between adjacent pixels in an image. That is, it is the degree of drastic change in local brightness of the image reflected by calculating the gray-level difference of pixels in the horizontal and vertical directions, such as the absolute value of the gray-level difference at the edge of a crack in the image, where the gray-level value changes from dark to bright or from bright to dark.

[0074] Specifically, the process iterates through each pixel in the stress-direction enhanced image, obtaining the grayscale values ​​of the current pixel and its adjacent horizontal and vertical pixels. The grayscale value of the current pixel is subtracted from the grayscale values ​​of the horizontal and vertical pixels to obtain the horizontal and vertical grayscale differences. The sum of the squares of the horizontal and vertical grayscale differences is calculated, and the square root of the sum is taken to obtain the grayscale gradient value of each pixel in the stress-direction enhanced image. A preset gradient threshold is obtained, and the grayscale gradient value of each pixel is compared with the preset gradient threshold. The grayscale values ​​of pixels with gradient values ​​greater than the preset gradient threshold are set as the maximum grayscale value, and the grayscale values ​​of pixels with gradient values ​​less than or equal to the preset gradient threshold are set as the minimum grayscale value, completing the binarization of the grayscale gradient. The set of connected pixels with the maximum grayscale value after binarization is extracted as the crack boundary contour.

[0075] Step 105: Analyze the frequency domain characteristics of the acoustic emission signal to obtain crack propagation parameters, and calculate the crack propagation prediction region based on the crack propagation parameters and crack boundary profile.

[0076] In the embodiments of this application, crack propagation parameters refer to quantitative indicators that characterize the dynamic trend and physical properties of cracks spreading outward from inside the material, such as a set of values ​​including the expected speed and expected propagation distance of cracks.

[0077] Specifically, a fast Fourier transform is performed on the acoustic emission signal to extract the dominant and secondary frequency components from the spectral data; the energy ratio of the dominant and secondary frequency components is calculated, and the energy ratio is multiplied by a preset acoustic attenuation coefficient to calculate the expected propagation velocity; the duration of the dominant frequency component is calculated, and the expected propagation velocity is multiplied by the duration to calculate the expected extension distance; the expected propagation velocity and the expected extension distance are combined to form the crack propagation parameters; a fan-shaped geometric region is generated along the direction perpendicular to the direction of maximum tensile stress, with the endpoint of the crack boundary profile as the starting point and the expected extension distance as the radius; the fan-shaped geometric region is calculated as the crack propagation prediction region.

[0078] Step 106: Based on the multidimensional risk parameters of the crack propagation prediction region and the crack initiation region, perform a coupled assessment to determine the list of potential weld hazards.

[0079] In this embodiment of the application, the weld defect location list refers to a structured data table that summarizes and records all weld locations that have been assessed and confirmed to have quality risks and their corresponding severity, such as an electronic spreadsheet containing defect coordinates, defect types and risk levels.

[0080] Specifically, the sum of the area of ​​the crack initiation region and the area of ​​the crack propagation prediction region is calculated as a spatial risk dimension parameter; the expected propagation rate in the crack propagation parameters is extracted as a dynamic risk dimension parameter; the spatial risk dimension parameter is multiplied by a preset spatial weight, and the dynamic risk dimension parameter is multiplied by a preset dynamic weight, and the products are added together to obtain a coupled risk score; the coupled risk score is discretized and mapped according to a preset score interval to determine the corresponding risk level; the center coordinates of the crack initiation region, the boundary coordinates of the crack propagation prediction region, and the risk level are associated and encapsulated according to a preset data structure to determine the weld hazard location list.

[0081] In one possible implementation, a coupled assessment is performed based on multidimensional risk parameters of the crack propagation prediction region and the crack initiation region to determine a list of potential weld hazards. This specifically includes steps 1061-1064, as follows: Step 1061: Based on the crack propagation prediction region and the crack initiation region, construct multiple pairs of related regions with a relative distance less than the distance threshold.

[0082] In the embodiments of this application, an associated region pair refers to a combination of two different attribute regions that meet specific proximity conditions in spatial location. That is, it is a spatial linkage data unit composed of region pairs that are less than a set threshold and whose geometric distance between the crack propagation prediction region and the crack initiation region is calculated. For example, a spatial region consisting of a crack initiation center point within five millimeters and a crack propagation prediction bounding box is paired.

[0083] Specifically, the center coordinates of the crack propagation prediction region and the crack initiation region are obtained separately; the Euclidean distance between the center coordinates of each crack propagation prediction region and the center coordinates of each crack initiation region is calculated, that is, the sum of the squares of the differences in the horizontal and vertical coordinates is calculated and the square root is taken; a pre-set distance threshold value is obtained, and all the calculated Euclidean distance values ​​are compared with the distance threshold value one by one; target center coordinate point combinations with Euclidean distance values ​​less than the distance threshold value are selected; the crack propagation prediction region and crack initiation region corresponding to each target center coordinate point combination are paired and bound to construct multiple associated region pairs with a relative distance less than the distance threshold.

[0084] Step 1062: Extract the thermo-acoustic coupling verification coefficient, peak thermal stress, and area of ​​the initiation region from the crack initiation region of each associated region pair, and use them as initiation dimension risk parameters; extract the potential velocity, radius of influence, and prediction direction angle from the crack propagation prediction region of each associated region pair, and use them as propagation dimension risk parameters.

[0085] In this embodiment of the application, the initiation dimension risk parameter refers to a multi-dimensional feature data set used to quantify the degree of danger in the initial stage of crack initiation. That is, a structured data set extracted from the crack initiation region that includes indicators such as thermoacoustic signal coupling degree, stress extreme value and spatial range. For example, a risk assessment feature vector composed of a thermoacoustic coupling verification coefficient with a value of 0.8, a thermal stress peak value of 300 MPa and an initiation region area of ​​5 square millimeters.

[0086] Specifically, the process iterates through each constructed pair of associated regions, locating the crack initiation region within the current pair. It reads the recorded thermoacoustic signal data, stress distribution data, and boundary coordinate data within the crack initiation region, extracting the corresponding thermoacoustic coupling verification coefficient, peak thermal stress value, and initiation region area value. These values ​​are then combined and concatenated as the initiation dimension risk parameter. Next, the process locates the crack propagation prediction region within the current pair of associated regions, reading the propagation dynamics data and geometric morphology data within the prediction region. The corresponding potential velocity value, influence radius value, and prediction direction angle value are extracted from this data. These values ​​are then combined and concatenated as the propagation dimension risk parameter.

[0087] Step 1063: Normalize the nascent dimension risk parameters and the extended dimension risk parameters respectively. Based on the preset weight coefficient matrix, perform weighted summation and coupling calculation on the normalized nascent dimension risk parameters and the extended dimension risk parameters to obtain the risk score of each associated region pair.

[0088] In this embodiment of the application, the risk score refers to a quantitative evaluation value that comprehensively reflects the possibility of structural failure in a specific area. It is the final danger index obtained by mathematically summing the various risk parameters of the initiation and expansion stages, assigning different weights to them. For example, the regional safety hazard assessment value with a score of 85 points obtained after combining the initiation risk and the expansion risk.

[0089] Specifically, the process involves obtaining the specific values ​​of each element in the nascent and extended dimension risk parameters for each associated region pair; identifying the maximum and minimum values ​​of each element across all associated region pairs; subtracting the minimum value from the current element to obtain the first difference; subtracting the minimum value from the maximum value to obtain the second difference; and dividing the first difference by the second difference to normalize the nascent and extended dimension risk parameters. The process also involves obtaining the weight values ​​of each parameter in the preset weight coefficient matrix; multiplying each value in the normalized nascent dimension risk parameters by its corresponding weight value; and multiplying each value in the normalized extended dimension risk parameters by its corresponding weight value; summing all the multiplied values ​​to complete the weighted summation coupling calculation; and using the final summation value as the risk score for each associated region pair.

[0090] Step 1064: Divide the risk scores into levels according to the preset risk grading standards, extract the spatial coordinates and risk level of each associated area pair, arrange them in order of risk level from high to low, and generate a weld hazard location list.

[0091] In this embodiment of the application, the weld hazard location list refers to a structural defect investigation data table that is sorted according to the degree of danger and contains specific spatial location information. That is, it is a guidance document generated by mapping the risk scores of each associated area pair to discrete risk levels, binding the level with the area coordinate information, and arranging them in descending order. For example, it is a maintenance list that includes items such as "Level 1 risk, coordinate value combination" and "Level 2 risk, coordinate value combination" and is arranged from high to low risk.

[0092] Specifically, the process involves: acquiring multiple score ranges and corresponding level labels from a pre-defined risk grading standard; comparing the risk score value of each associated region pair with each score range to determine the target score range to which the risk score value belongs; assigning the level label corresponding to the target score range to the current associated region pair to complete the risk score grading; reading the horizontal and vertical coordinates of the center point of each associated region pair in the global coordinate system as spatial coordinates, and associating the spatial coordinates with the obtained risk levels; sorting all associated region pairs according to the priority of the extracted risk levels, with the associated region pair with the highest risk level at the top, and so on down to the associated region pair with the lowest risk level; and formatting the combination of all sorted spatial coordinates and risk levels into a text output to generate a weld hazard location list.

[0093] In the above embodiments, the extraction and preliminary localization of crack boundary contours are achieved through the construction of stress direction enhanced images and the binarization processing of grayscale gradients. To further improve the defect evolution prediction capability of the monitoring system in complex welding processes and establish a precise assessment mechanism for crack propagation trends and a dynamic early warning mechanism for potential risk areas, this application also provides an online monitoring method for the quality of laser welding of battery connectors. This method extracts the primary and secondary frequency features and calculates the frequency domain attenuation coefficient by analyzing the frequency domain power spectral density function of the acoustic emission signal. It then uses crack propagation parameters to infer the potential velocity and radius of influence, and constructs a geometric mapping relationship between the crack boundary extension angle and the stress direction to create a fan-shaped prediction region. This enables the monitoring system to more accurately predict the dynamic crack propagation behavior under complex stress environments and pinpoint key evolution areas. The following section combines... Figure 2 Another method for online monitoring of the quality of laser welding of battery connectors in this application embodiment is described below: Please see Figure 2 This is a flowchart illustrating another method for online monitoring of the quality of laser welding of battery connectors in an embodiment of this application.

[0094] Step 201: Obtain the effective acoustic emission signal after filtering the acoustic emission signal, perform frequency domain transformation on the effective acoustic emission signal to obtain the frequency domain power spectral density function, extract the frequency corresponding to the first peak with the largest amplitude as the principal frequency, and the frequency corresponding to the second peak with the second largest amplitude as the secondary principal frequency from the frequency domain power spectral density function, and use the principal frequency and secondary principal frequency as frequency domain features.

[0095] In the embodiments of this application, the frequency domain power spectral density function refers to the mathematical function of the power distribution of the signal at each frequency component in the frequency domain. That is, after performing frequency domain transformation operations such as Fourier transform on the time domain signal, it reflects the data set that reflects the law of signal energy changing with frequency. For example, after a sound emission signal containing multiple frequency components is transformed, the horizontal axis is the frequency value and the vertical axis is the corresponding power magnitude, which is a two-dimensional curve data.

[0096] Specifically, the time-domain amplitude data sequence of the effective acoustic emission signal after processing by a preset filter is read; the time-domain amplitude data sequence is converted to frequency using a fast Fourier transform algorithm to obtain a frequency-domain data sequence containing complex results; the square of the modulus of the complex number corresponding to each frequency point in the frequency-domain data sequence is calculated, and the square value is divided by the total number of sampling points of the effective acoustic emission signal to obtain the power spectral density value corresponding to each frequency point, thus forming a frequency-domain power spectral density function; all power spectral density values ​​in the frequency-domain power spectral density function are traversed, and the first peak value with the largest value and the second peak value with the second largest value are selected by comparing their magnitudes; the horizontal axis frequency value corresponding to the first peak value is extracted as the principal frequency, and the horizontal axis frequency value corresponding to the second peak value is extracted as the secondary principal frequency; the extracted principal and secondary principal frequencies are combined, and the combined data set is used as the frequency domain feature.

[0097] Step 202: Calculate the frequency domain attenuation coefficient based on the main frequency and the secondary main frequency, and inversely deduce the potential velocity and radius of influence of crack propagation based on the frequency domain attenuation coefficient, and use the potential velocity and radius of influence as crack propagation parameters.

[0098] In the embodiments of this application, the frequency domain attenuation coefficient refers to a physical quantity that measures the rate of energy attenuation of acoustic emission signals at different frequency components. That is, it is a characteristic value that reflects the ratio of high-frequency and low-frequency energy dissipation during the crack propagation process inside the material by calculating the difference between the dominant frequency and the secondary dominant frequency of the signal. For example, it is a dimensionless constant used to characterize the energy loss rate when sound waves propagate in a medium, calculated by the difference or ratio between the dominant frequency and the secondary dominant frequency.

[0099] Specifically, the process involves obtaining the extracted values ​​of the primary and secondary primary frequencies; subtracting the secondary primary frequency from the primary frequency to obtain the frequency difference; dividing the frequency difference by the primary frequency to obtain the frequency domain attenuation coefficient; obtaining the pre-calibrated base propagation velocity and base influence radius; dividing the base propagation velocity by the frequency domain attenuation coefficient to calculate the potential crack propagation velocity; multiplying the base influence radius by the reciprocal of the frequency domain attenuation coefficient to calculate the radius of influence of crack propagation; and concatenating the calculated potential velocity and radius of influence, using the combined data as the crack propagation parameter.

[0100] Step 203: Calculate the crack propagation prediction region based on the crack propagation parameters and crack boundary profile.

[0101] In the embodiments of this application, the crack propagation prediction region refers to the spatial range in which the crack may extend and propagate within a certain period of time in the future. That is, it is a two-dimensional or three-dimensional geometric region in the spatial coordinate system that includes potential propagation paths and influence boundaries, which is calculated by combining the current geometric boundary characteristics of the crack and the dynamic properties of crack propagation. For example, it is a closed polygonal spatial range formed by extending outward based on the current crack boundary according to the propagation speed and influence radius.

[0102] Specifically, the process involves obtaining the potential velocity and influence radius values ​​from the crack propagation parameters, as well as the spatial coordinates of each discrete boundary pixel in the crack boundary profile; obtaining the preset prediction time length value, multiplying the potential velocity value by the prediction time length value to obtain the predicted propagation distance value; adding the predicted propagation distance value to the influence radius value to obtain the comprehensive propagation radius value; traversing each discrete boundary pixel in the crack boundary profile, using the spatial coordinates of the current discrete boundary pixel as the center and the comprehensive propagation radius value as the radius, generating a corresponding circular propagation sub-region in a two-dimensional spatial coordinate system; extracting the outer envelope boundary lines of all generated circular propagation sub-regions, and using the closed spatial range enclosed by the outer envelope boundary lines as the crack propagation prediction region.

[0103] In one possible implementation, the crack propagation prediction region is calculated based on crack propagation parameters and crack boundary profile, specifically including steps 2031-2032, as follows: Step 2031: Extract the current extension angle of the crack boundary profile and calculate the predicted direction angle of crack propagation in combination with the stress direction.

[0104] In the embodiments of this application, the predicted direction angle refers to the spatial pointing angle value along which the crack will propagate in the future. That is, it is a value that represents the future cracking direction of the crack tip, calculated in a given coordinate system by comprehensively considering the existing geometric extension trend of the crack and the direction of the external stress field. For example, the crack future propagation direction angle of 37.5 degrees is calculated by combining the current extension angle of 30 degrees and the external stress direction of 45 degrees.

[0105] Specifically, the coordinate values ​​of multiple consecutive boundary pixels near the crack tip are obtained in the crack boundary contour; the slope values ​​of the lines connecting adjacent boundary pixels are calculated, the average value of all the line slope values ​​is calculated, and the average value is converted into the corresponding angle value as the current extension angle; the specific angle value of the pre-measured stress direction is obtained; the preset geometric extension weight value and stress action weight value are obtained, and the sum of the geometric extension weight value and the stress action weight value is one; the current extension angle value is multiplied by the geometric extension weight value to obtain the first product value, and the angle value of the stress direction is multiplied by the stress action weight value to obtain the second product value; the first product value and the second product value are added together, and the angle value obtained by the addition is used as the predicted direction angle of crack propagation.

[0106] Step 2032: Starting from the endpoint of the crack boundary profile, construct a fan-shaped region based on the predicted direction angle, potential velocity, and radius of influence, and use the fan-shaped region as the crack propagation prediction region.

[0107] In the embodiments of this application, the fan-shaped region refers to the geometric space bounded by an arc and two radii passing through the two ends of the arc in a two-dimensional plane. That is, it is a closed shape that represents the potential influence range of the crack by expanding to both sides at a certain angle with the crack tip as the center, the predicted propagation distance as the radius, and the predicted direction as the central axis. For example, it is a fan-shaped space with a 30-degree central angle formed by expanding to the left and right at 15 degrees along the predicted direction with the crack tip as the vertex and a radius of 10 millimeters as the apex.

[0108] Specifically, the coordinate data of the crack boundary profile is traversed, and the coordinate points at the outermost edge are extracted as the endpoints of the crack boundary profile. The coordinates of the endpoints are set as the reference points for the center of the constructed shape. A preset prediction time length value is obtained, and the potential velocity value is multiplied by the prediction time length value to obtain the predicted extension distance value. The predicted extension distance value is added to the radius of the influence range value, and the result is used as the sector radius value. A preset sector unfolding angle value is obtained, and half of the sector unfolding angle value is subtracted from the predicted direction angle value to obtain the starting angle value. Half of the sector unfolding angle value is added to the predicted direction angle value to obtain the ending angle value. In the two-dimensional coordinate system, a closed boundary is drawn from the starting angle value to the ending angle value, with the reference point of the circle as the starting point and the sector radius value as the radius, to construct the sector region. The generated sector region is directly used as the crack propagation prediction region.

[0109] The following describes an online monitoring system for the quality of laser welding of battery connectors according to an embodiment of the present invention from the perspective of hardware processing. Please refer to [link / reference needed]. Figure 3 This is a schematic diagram of an online monitoring system for the quality of laser welding of battery connectors in an embodiment of this application.

[0110] It should be noted that, Figure 3 The structure of the online monitoring system for laser welding quality of battery connectors shown is merely an example and should not impose any limitations on the functionality and scope of application of the embodiments of the present invention.

[0111] like Figure 3 As shown, an online monitoring system for the quality of laser welding of battery connectors includes a central processing unit (CPU) 301, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 302 or a program loaded from a storage section 308 into a random access memory (RAM) 303, such as executing the method described in the above embodiment. The RAM 303 also stores various programs and data required for system operation. The CPU 301, ROM 302, and RAM 303 are interconnected via a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0112] The following components are connected to I / O interface 305: input section 306 including audio input devices, push-button switches, etc.; output section 307 including a liquid crystal display (LCD) and audio output devices, indicator lights, etc.; storage section 308 including a hard disk, etc.; and communication section 309 including a network interface card such as a LAN (Local Area Network) card, modem, etc. Communication section 309 performs communication processing via a network such as the Internet. Drive 310 is also connected to I / O interface 305 as needed. Removable media 311, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., are installed on drive 310 as needed so that computer programs read from them can be installed into storage section 308 as needed.

Claims

1. A method for online monitoring of the quality of laser welding of battery connectors, characterized in that, The method includes: Thermal signal data during the laser welding process of battery connectors is acquired, and thermal stress spatiotemporal field data is constructed based on the thermal signal data; Stress anomaly features are extracted from the thermal stress spatiotemporal field data, and potential risk areas are identified based on the stress anomaly features. Acoustic emission signals corresponding to the potential risk areas are obtained, and dual-modal cross-validation is performed based on the acoustic emission signals and the thermal stress spatiotemporal field data to determine the crack initiation area; Extract the stress direction of the crack initiation region, and perform directional enhancement processing on the local weld image based on the stress direction to extract the crack boundary contour; The crack propagation parameters are obtained by analyzing the frequency domain characteristics of the acoustic emission signal, and the crack propagation prediction region is calculated based on the crack propagation parameters and the crack boundary profile. A coupled assessment is performed based on the multidimensional risk parameters of the crack propagation prediction region and the crack initiation region to determine the list of potential weld hazards.

2. The method for online monitoring of the quality of laser welding of battery connectors according to claim 1, characterized in that, The acquisition of thermal signal data during the laser welding of battery connectors, and the construction of thermal stress spatiotemporal field data based on the thermal signal data, includes: Spatial domain filtering is performed on the thermal signal data to obtain spatial temperature distribution data; Obtain the coefficient of thermal expansion and elastic modulus of the battery connector material; Based on the temperature spatial distribution data, the coefficient of thermal expansion, and the elastic modulus, the thermal stress tensor components at the corresponding positions of the pixels are calculated to obtain thermal stress spatiotemporal field data containing a time series.

3. The method for online monitoring of the quality of laser welding of battery connectors according to claim 1, characterized in that, The step of extracting stress anomaly features based on the thermal stress spatiotemporal field data and identifying potential risk areas based on the stress anomaly features includes: The temporal evolution rate field and stress spatial gradient field are calculated for the stress values ​​of pixels in the thermal stress spatiotemporal field data. In the time evolution rate field and the stress spatial gradient field, time-domain anomalies with evolution rates exceeding a preset rate threshold and spatial domain anomalies with gradient magnitudes exceeding a preset gradient threshold are identified respectively, and the time-domain anomalies and the spatial domain anomalies are used as stress anomaly features. Pixels that simultaneously belong to the temporal domain anomalies and the spatial domain anomalies are extracted as composite anomalies. The spatial region where the composite anomalies are located is traversed by a sliding window of a preset size, and the proportion of the composite anomalies in each sliding window is calculated to obtain the distribution density. The regions containing sliding windows with a distribution density greater than a preset density threshold are selected, and the selected regions are merged into connected components. The merged regions are then marked as potential risk regions.

4. The method for online monitoring of the quality of laser welding of battery connectors according to claim 1, characterized in that, The step of acquiring the acoustic emission signal corresponding to the potential risk area and performing dual-modal cross-validation based on the acoustic emission signal and the thermal stress spatiotemporal field data to determine the crack initiation region includes: The acoustic emission signal is filtered to extract the effective acoustic emission signal, and the spatial distribution data of the acoustic emission energy of the effective acoustic emission signal is calculated. Establish a spatial coordinate registration relationship between the potential risk area and the spatial distribution data of acoustic emission energy, and calculate the thermal signal characteristic value and acoustic signal characteristic value within the potential risk area respectively; The thermo-acoustic coupling verification coefficient is calculated based on the thermal signal characteristic value and the acoustic signal characteristic value, and potential risk areas where the thermo-acoustic coupling verification coefficient exceeds a preset coupling threshold are selected as crack initiation areas.

5. The method for online monitoring of the quality of laser welding of battery connectors according to claim 4, characterized in that, The process of establishing a spatial coordinate registration relationship between the potential risk area and the acoustic emission energy spatial distribution data, and calculating the thermal signal characteristic values ​​and acoustic signal characteristic values ​​within the potential risk area, includes: Obtain the spatial coordinate range of the potential risk area, map the spatial coordinate range to the corresponding spatial location of the acoustic emission energy spatial distribution data, and establish a spatial coordinate registration relationship; Based on the spatial coordinate registration relationship, stress data within the potential risk area is extracted from the thermal stress spatiotemporal field data, and acoustic emission energy data corresponding to the spatial location is extracted from the acoustic emission energy spatial distribution data. The thermal signal characteristic value is calculated based on the stress data, and the acoustic signal characteristic value is calculated based on the acoustic emission energy data.

6. The method for online monitoring of the quality of laser welding of battery connectors according to claim 1, characterized in that, The step of extracting the stress direction of the crack initiation region and performing directional enhancement processing on the local weld image based on the stress direction to extract the crack boundary contour includes: Extract the stress gradient vector corresponding to the crack initiation region from the thermal stress spatiotemporal field data, and calculate the statistical average value of the gradient direction angle as the stress direction; A directional filter is constructed using the stress direction as the filtering direction. The directional filter is then used to perform a convolution operation on the local weld image to obtain a stress-direction enhanced image. Calculate the grayscale gradient of the pixels in the stress direction enhancement image, and binarize the grayscale gradient based on a preset gradient threshold to extract the crack boundary contour.

7. The method for online monitoring of the quality of laser welding of battery connectors according to claim 1, characterized in that, The process of analyzing the frequency domain characteristics of the acoustic emission signal to obtain crack propagation parameters, and calculating the crack propagation prediction region based on the crack propagation parameters and the crack boundary profile, includes: The effective acoustic emission signal after filtering the acoustic emission signal is obtained, and the effective acoustic emission signal is transformed in the frequency domain to obtain the frequency domain power spectral density function. The frequency corresponding to the first peak with the largest amplitude is extracted from the frequency domain power spectral density function as the principal frequency, and the frequency corresponding to the second peak with the second largest amplitude is extracted as the secondary principal frequency. The principal frequency and the secondary principal frequency are used as frequency domain features. The frequency domain attenuation coefficient is calculated based on the main frequency and the secondary main frequency, and the potential velocity and radius of influence of crack propagation are inferred from the frequency domain attenuation coefficient. The potential velocity and the radius of influence are used as crack propagation parameters. The crack propagation prediction region is calculated based on the crack propagation parameters and the crack boundary profile.

8. The method for online monitoring of the quality of laser welding of battery connectors according to claim 7, characterized in that, The step of calculating the crack propagation prediction region based on the crack propagation parameters and the crack boundary profile includes: Extract the current extension angle of the crack boundary profile, and calculate the predicted direction angle of crack propagation in combination with the stress direction; Starting from the endpoint of the crack boundary profile, a fan-shaped region is constructed based on the predicted direction angle, the potential velocity, and the radius of the influence range, and the fan-shaped region is used as the crack propagation prediction region.

9. The method for online monitoring of the quality of laser welding of battery connectors according to claim 1, characterized in that, The multidimensional risk parameters based on the crack propagation prediction region and the crack initiation region are coupled for evaluation to determine the weld defect location list, including: Based on the crack propagation prediction region and the crack initiation region, multiple pairs of associated regions with a relative distance less than a distance threshold are constructed; The thermoacoustic coupling verification coefficient, peak thermal stress, and area of ​​the initiation region are extracted from the crack initiation region of each of the associated region pairs and used as initiation dimension risk parameters; the potential velocity, radius of influence, and prediction direction angle are extracted from the crack propagation prediction region of each of the associated region pairs and used as propagation dimension risk parameters. The nascent dimension risk parameter and the extended dimension risk parameter are normalized respectively. Based on the preset weight coefficient matrix, the normalized nascent dimension risk parameter and the extended dimension risk parameter are weighted and coupled to obtain the risk score of each of the associated region pairs. The risk scores are classified according to the preset risk grading standard, the spatial coordinates and risk level of each associated region are extracted, and the regions are arranged in descending order of risk level to generate the weld hazard location list.

10. An online monitoring system for the quality of laser welding of battery connectors, characterized in that, The online monitoring system for laser welding quality of battery connectors includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to cause the online monitoring system for laser welding quality of battery connectors to perform the method as described in any one of claims 1-9.

Images