A lithium battery aluminum plating resistivity detection method and system

Abstract

The invention provides a method and system for detecting the resistivity of aluminum plating of a lithium battery, and relates to the technical field of measuring electrical variables. The method comprises: obtaining an aluminum-plated structural part of a lithium battery, wherein the aluminum-plated structural part comprises a substrate and an aluminum-plated film; using a non-contact millimeter wave reflection technology to respectively measure the resistivity at the edge of the substrate and the aluminum-plated film; using the resistivity at the edge as a boundary condition, performing finite element analysis to respectively determine the resistivity distribution of the substrate and the aluminum-plated film; constructing a resistivity distribution map of the substrate and a resistivity distribution map of the aluminum-plated film according to the resistivity distribution of the substrate and the aluminum-plated film; fusing the resistivity distribution map of the substrate and the resistivity distribution map of the aluminum-plated film to construct a fused resistivity distribution map; and detecting whether the resistivity of the aluminum-plated structural part is abnormal by using a convolutional neural network according to the fused resistivity distribution map.

Description

Technical Field

[0001] The present invention relates to the technical field of measuring electrical variables, and in particular to a method and system for detecting the resistivity of aluminum plating of a lithium battery. Background Art

[0002] With the development and application of ecological energy, lithium batteries have been widely used, especially in electric vehicles and renewable energy storage systems. In order to reduce weight, improve safety and energy density, it has become an inevitable trend for lithium battery composite current collectors to replace traditional metal current collectors. The positive electrode of lithium battery composite current collectors generally uses polymer materials such as PET, PP, PI, etc., and vacuum deposits aluminum films on the positive and reverse surfaces to form aluminum-plated structures. Aluminum-plated structures can effectively improve the conduction efficiency of internal battery current, reduce resistance, and thus improve the overall conductivity and performance of the battery.

[0003] Aluminum-plated structural parts act as current collectors in lithium batteries, and their resistivity directly affects the electrical performance of the battery. Resistivity measurement can evaluate whether the conductivity of aluminum-plated structural parts meets the design requirements and ensure that the battery can work efficiently.

[0004] Traditional resistivity measurement technology often uses the four-probe method. When using probe measurement, the probe needs to contact the object to be tested, which may leave scratches or other mechanical damage on the surface of the object to be tested. Especially for thinner or more fragile materials such as aluminum-plated structural parts, it may cause irreparable damage.

[0005] With the development of resistivity measurement technology, more and more non-contact resistivity measurement technologies are widely used, the typical one is pulsed eddy current technology. However, pulsed eddy current technology is easily affected by material thickness, especially when measuring thinner materials, it is easy to make inaccurate measurements. In addition, pulsed eddy current technology is sensitive to environmental magnetic fields and electromagnetic interference, and needs to be operated in a clean electromagnetic environment, which will undoubtedly increase the cost of resistivity detection. Summary of the invention

[0006] In order to solve the technical problem that the prior art is easily affected by the thickness of the material, especially when measuring thinner materials, it is easy to make inaccurate measurements. In addition, the pulsed eddy current technology is sensitive to the environmental magnetic field and electromagnetic interference, and needs to be operated in a clean electromagnetic environment, which will undoubtedly increase the cost of resistivity detection, the present invention provides a lithium battery aluminum plating resistivity detection method and system.

[0007] The technical solution provided by the embodiment of the present invention is as follows:

[0008] A first aspect of an embodiment of the present invention provides a method for detecting resistivity of aluminum plating of a lithium battery, comprising:

[0009] S1: obtaining an aluminum-plated structural component of a lithium battery, wherein the aluminum-plated structural component comprises a substrate and an aluminum-plated film;

[0010] S2: Using non-contact millimeter wave reflection technology, respectively measuring the resistivity of the substrate and the edge of the aluminum-plated film;

[0011] S3: using the resistivity at the edge as a boundary condition, performing finite element analysis to determine the resistivity distribution of the substrate and the aluminum-plated film respectively;

[0012] S4: constructing a resistivity distribution map of the substrate and a resistivity distribution map of the aluminum-plated film respectively according to the resistivity distribution of the substrate and the aluminum-plated film;

[0013] S5: fusing the resistivity distribution map of the substrate and the resistivity distribution map of the aluminum-plated film to construct a fused resistivity distribution map;

[0014] S6: According to the fused resistivity distribution map, a convolutional neural network is used to detect whether the resistivity of the aluminum-plated structural component is abnormal.

[0015] Furthermore, the S2 specifically includes:

[0016] S201: Construct a resistivity measurement model based on millimeter wave reflection technology;

[0017] S202: Acquire millimeter wave reflection signals of each edge point;

[0018] S203: Calculate the resistivity of each edge point according to the millimeter wave reflection signal.

[0019] Furthermore, the S201 specifically includes:

[0020] Obtaining a correlation between a millimeter wave reflection coefficient and an inherent impedance of a device under test, wherein the device under test includes the substrate and the aluminum-plated film:

[0021] (1)

[0022] Where, Γ represents the reflection coefficient, η represents the inherent impedance of the device under test, and η0 represents the inherent impedance of free space;

[0023] The inherent impedance of the device under test is specifically:

[0024] (2)

[0025] Wherein, σ represents the conductivity of the device under test, μ represents the magnetic permeability of the device under test, ε represents the dielectric constant of the device under test, ω represents the angular frequency, and j represents the imaginary unit;

[0026] The free space inherent impedance is specifically:

[0027] (3)

[0028] Among them, μ0 represents the magnetic permeability of free space, ε0 represents the dielectric constant of free space;

[0029] For non-magnetic materials:

[0030] (4)

[0031] Combining formula (1), formula (2), formula (3) and formula (4), we have:

[0032] (5)

[0033] Wherein, X represents the real part of the reflection coefficient, and Y represents the imaginary part of the reflection coefficient;

[0034] By using the real part X and the imaginary part Y in formula (5) and eliminating the dielectric constant ε of the device under test, the conductivity measurement equation is obtained:

[0035] s = 4 oh e 0 Y ( 1 − X 2 − Y 2 ) [ ( 1 + X ) 2 + Y 2 ] 2 (6);

[0036] The resistivity measurement model is constructed by formula (7):

[0037] r = 1 s = [ ( 1 + X ) 2 + Y 2 ] 2 4 oh e 0 Y ( 1 − X 2 − Y 2 ) (7)

[0038] Here, ρ represents the resistivity.

[0039] Furthermore, the S203 specifically includes:

[0040] According to the millimeter wave reflection signal at each edge point, the real part and the imaginary part of the reflection coefficient are calculated:

[0041] (8)

[0042] Among them, Γ m represents the reflection coefficient at the mth measurement, X m represents the real part of the reflection coefficient at the mth measurement, Y m represents the imaginary part of the reflection coefficient at the mth measurement, A m represents the amplitude at the mth measurement, θ m represents the phase at the mth measurement, and e represents a natural constant;

[0043] The real part and the imaginary part of the reflection coefficient measured multiple times are averaged to obtain the real part average value and the imaginary part average value;

[0044] Substitute the real part average value and the imaginary part average value into formula (7) to calculate the resistivity of each edge point.

[0045] Furthermore, the S2 specifically includes:

[0046] In order to eliminate the error influence of the measurement system, the calculated resistivity is calibrated by the following equation:

[0047] (9)

[0048] Where, σ represents the resistivity of the test piece, C1 and C2 represent the calibration coefficients, and σ * Indicates the resistivity before calibration.

[0049] Furthermore, the S5 is specifically:

[0050] The fused resistivity distribution map is constructed by the following fusion methods:

[0051] (10)

[0052] Among them, f represents the fused resistivity distribution map, f1 represents the resistivity distribution map of the substrate, λ1 represents the fusion coefficient of the resistivity distribution map of the substrate, f2 represents the resistivity distribution map of the aluminum coating film, and λ2 represents the fusion coefficient of the resistivity distribution map of the aluminum coating film.

[0053] Furthermore, the convolutional neural network specifically includes: a convolutional layer, a pooling layer, a fully connected layer and an output layer.

[0054] Furthermore, the S6 specifically includes:

[0055] S601: Using the convolution layer in the convolutional neural network, extracting the features of the fused feature map:

[0056] (11)

[0057] in, Indicates The features of the convolutional layer output using the i-th input feature map, represents the features of the output of the l-1th convolutional layer using the i-th input feature map, Indicates The convolution layer uses the convolution kernel weights when the i-th input feature map is used. Indicates The convolutional layer uses the bias term when the i-th input feature map is used, M j represents the selected input feature map set, J( ) represents the activation function;

[0058] S602: Use the pooling layer in the CNN network to reduce the dimension of the extracted features:

[0059] (12)

[0060] Among them, down( ) represents the downsampling function, represents the features of the output of the lth pooling layer using the i-th input feature map, Represents the features of the output of the l-1th pooling layer using the i-th input feature map;

[0061] S603: Use the fully connected layer in the CNN network to summarize the features and obtain the hidden state;

[0062] S604: Calculate the abnormal value belonging to the aluminum-plated structural part by using the output layer in the CNN network:

[0063] (13)

[0064] Among them, P represents an outlier, Sigmoid () represents the Sigmoid activation function, W represents the weight matrix, h represents the hidden state, and b represents the bias term;

[0065] S605: Determine whether the abnormal value is greater than a preset value; if so, determine that the resistivity of the aluminum-plated structural component is abnormal; otherwise, determine that the resistivity of the aluminum-plated structural component is not abnormal.

[0066] Furthermore, the S6 specifically includes:

[0067] S606: When the resistivity of the aluminum-plated structural component is abnormal, output the abnormal position of the aluminum-plated structural component.

[0068] A second aspect of an embodiment of the present invention provides a lithium battery aluminum plating resistivity detection system, comprising:

[0069] processor;

[0070] A memory having computer-readable instructions stored thereon, wherein when the computer-readable instructions are executed by the processor, the method for detecting the resistivity of aluminum plating of a lithium battery as described in the first aspect is implemented.

[0071] The third aspect

[0072] A third aspect of an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for detecting the resistivity of aluminum plating on a lithium battery as described in the first aspect.

[0073] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0074] (1) In the present invention, the resistivity of the test piece is accurately measured by using non-contact millimeter wave reflection technology, which is not affected by the thickness of the material. For multi-layer materials, the resistivity of both sides of the material can be measured by simply turning the material over, thereby improving the accuracy and efficiency of resistivity measurement, and having low requirements on the measurement environment, which can reduce the cost of resistivity detection.

[0075] (2) Through finite element analysis, the resistivity distribution of the substrate and the aluminum coating can be analyzed in detail, providing global resistivity information rather than just local measurements, ensuring a comprehensive evaluation of the entire structure. At the same time, by fusing the resistivity distribution maps of the substrate and the aluminum coating, a comprehensive resistivity distribution map is obtained, which helps to evaluate the resistivity of the aluminum-coated structural parts as a whole and improve the comprehensiveness and accuracy of the detection. Using convolutional neural networks for automated detection can efficiently identify abnormal areas in the resistivity distribution, improve detection speed and accuracy, and reduce errors caused by human intervention. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0077] Figure 1 A schematic diagram of a process for detecting resistivity of aluminum plating of a lithium battery provided by an embodiment of the present invention;

[0078] Figure 2 A structural schematic diagram of a lithium battery aluminum plating resistivity detection system provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0079] The technical solution of the present invention is described below in conjunction with the accompanying drawings.

[0080] In the embodiments of the present invention, sometimes a subscript such as W1 may be mistakenly written as a non-subscript such as W1. When the difference is not emphasized, the meanings to be expressed are consistent.

[0081] In order to make the technical problems, technical solutions and advantages to be solved by the present invention more clear, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.

[0082] Reference Manual Attached Figure 1 , showing a schematic flow chart of a lithium battery aluminum plating resistivity detection method provided in an embodiment of the present invention.

[0083] The embodiment of the present invention provides a method for detecting the resistivity of aluminum plating on a lithium battery. The method can be implemented by a device for detecting the resistivity of aluminum plating on a lithium battery. The device for detecting the resistivity of aluminum plating on a lithium battery can be a terminal or a server. The processing flow of the method for detecting the resistivity of aluminum plating on a lithium battery can include the following steps:

[0084] S1: Obtain aluminum-plated structural parts of lithium batteries.

[0085] The aluminum-plated structural component includes a substrate and an aluminum-plated film.

[0086] Optionally, the substrate can be a polymer material such as PET, PP, PI, etc.

[0087] It should be noted that aluminum-plated structural parts are mainly used as positive electrode current collectors for lithium batteries. Aluminum-plated structural parts can effectively improve the conduction efficiency of internal current in the battery, reduce resistance, and thus improve the overall conductivity and performance of the battery.

[0088] Furthermore, the aluminum-plated structural parts are mainly manufactured by vacuum evaporation winding aluminum-plating equipment.

[0089] S2: Use non-contact millimeter wave reflection technology to measure the resistivity of the substrate and the edge of the aluminum coating.

[0090] Among them, millimeter wave reflection technology is a method of measuring and detecting material properties using electromagnetic waves in the millimeter wave frequency band (usually between 30 GHz and 300 GHz). This technology has attracted widespread attention in a variety of applications due to its high frequency, short wavelength and high resolution.

[0091] It should be noted that millimeter wave reflection technology is hardly affected by material thickness, which improves the accuracy of resistivity measurement.

[0092] In actual application, the resistivity of the substrate on one side is measured first, and then the resistivity of the aluminum film on the other side can be measured by simply turning it over, which greatly improves the resistivity measurement efficiency.

[0093] At the same time, non-contact technology can also ensure the integrity and cleanliness of aluminum-plated structural parts.

[0094] In a possible implementation, S2 specifically includes:

[0095] S201: Construct a resistivity measurement model based on millimeter wave reflection technology.

[0096] Optionally, S201 specifically includes:

[0097] Obtain the correlation between the millimeter wave reflection coefficient and the inherent impedance of the device under test, which includes the substrate and the aluminum coating:

[0098] (1)

[0099] Where Γ represents the reflection coefficient, η represents the intrinsic impedance of the device under test, and η0 represents the intrinsic impedance of free space.

[0100] The inherent impedance of the device under test is specifically:

[0101] (2)

[0102] Wherein, σ represents the conductivity of the device under test, μ represents the magnetic permeability of the device under test, ε represents the dielectric constant of the device under test, ω represents the angular frequency, and j represents the imaginary unit.

[0103] The inherent impedance of free space is:

[0104] (3)

[0105] Where μ0 represents the magnetic permeability of free space and ε0 represents the dielectric constant of free space.

[0106] For non-magnetic materials:

[0107] (4)

[0108] Combining formula (1), formula (2), formula (3) and formula (4), we have:

[0109] (5)

[0110] Wherein, X represents the real part of the reflection coefficient, and Y represents the imaginary part of the reflection coefficient.

[0111] By using the real part X and the imaginary part Y in formula (5) and eliminating the dielectric constant ε of the device under test, the conductivity measurement equation is obtained:

[0112] s = 4 oh e 0 Y ( 1 − X 2 − Y 2 ) [ ( 1 + X ) 2 + Y 2 ] 2 (6).

[0113] The resistivity measurement model is constructed by formula (7):

[0114] r = 1 s = [ ( 1 + X ) 2 + Y 2 ] 2 4 oh e 0 Y ( 1 − X 2 − Y 2 ) (7)

[0115] Here, ρ represents the resistivity.

[0116] In the present invention, by constructing a resistivity measurement model based on millimeter wave reflection technology, high-precision, non-contact, thickness-independent resistivity measurement can be achieved. This method not only improves measurement efficiency and reliability, but also has wide applicability and flexibility, and is of great significance for resistivity measurement of multilayer structures and various materials.

[0117] S202: Acquire millimeter wave reflection signals of each edge point.

[0118] S203: Calculate the resistivity of each edge point according to the millimeter wave reflection signal.

[0119] Optionally, S203 specifically includes:

[0120] According to the millimeter wave reflection signal at each edge point, the real and imaginary parts of the reflection coefficient are calculated:

[0121] (8)

[0122] Among them, Γ m represents the reflection coefficient at the mth measurement, X m represents the real part of the reflection coefficient at the mth measurement, Y m represents the imaginary part of the reflection coefficient at the mth measurement, A m represents the amplitude at the mth measurement, θ m represents the phase at the mth measurement, and e represents a natural constant.

[0123] The real part and the imaginary part of the reflection coefficient measured multiple times are averaged to obtain the real part average value and the imaginary part average value.

[0124] It should be noted that by taking multiple measurements and taking the average value, the random error and noise in a single measurement can be reduced, and the stability and accuracy of the measurement results can be improved.

[0125] Substitute the real part average value and the imaginary part average value into formula (7) to calculate the resistivity of each edge point.

[0126] In the present invention, by measuring multiple times and taking the average of the real and imaginary parts of the reflection coefficient, the accuracy and reliability of resistivity measurement can be significantly improved, providing more accurate data support. This method is highly adaptable and can cope with complex measurement environments and material diversity, and is of great significance for achieving high-precision non-contact resistivity measurement.

[0127] In a possible implementation manner, S2 specifically further includes:

[0128] In order to eliminate the error influence of the measurement system, the calculated resistivity is calibrated by the following equation:

[0129] (9)

[0130] Where, σ represents the resistivity of the test piece, C1 and C2 represent the calibration coefficients, and σ * Indicates the resistivity before calibration.

[0131] In the present invention, the measurement system may have systematic errors in actual operation, such as inaccurate equipment calibration, environmental influences, etc. These errors can be corrected by calibration coefficients, the accuracy of the measurement results can be improved, and the measurement results can be closer to the true value. The calibration formula can effectively eliminate the errors of the measurement system and improve the accuracy, reliability and consistency of resistivity measurement. The calibration process can not only compensate for systematic deviations, but also ensure the standardization and repeatability of measurement data, simplify subsequent data processing and analysis, and provide guarantees for high-precision applications. This method has wide applicability and important significance, and is an important step in achieving high-precision non-contact resistivity measurement.

[0132] Furthermore, since there are only two calibration coefficients, two sets of actual measurement values ​​and standard values ​​can be used to perform a solution to obtain two calibration coefficients.

[0133] S3: Using the resistivity at the edge as the boundary condition, finite element analysis is performed to determine the resistivity distribution of the substrate and the aluminum coating, respectively.

[0134] In a possible implementation manner, the S3 specifically includes:

[0135] S301: Construct the geometric model of aluminum-plated structural parts.

[0136] S302: Using the measured resistivity of the edge points as boundary conditions, setting the resistivity value at the boundary of the geometric model.

[0137] S303: Setting the initial resistivity of the substrate and the aluminum coating.

[0138] S304: Divide the geometric model into a finite element mesh, ensuring that the mesh is fine enough to capture the details of the resistivity distribution.

[0139] S305: Setting the basic battery equation:

[0140]

[0141] in, represents the vector differential operator, represents the electric field strength, represents the electric flux density, represents the magnetic field strength, represents the magnetic induction intensity, represents the current density, q represents the charge density;

[0142] In a static battery:

[0143]

[0144] Among them, ε represents the dielectric constant, μ represents the magnetic permeability, and σ represents the resistivity. , ρ represents resistivity.

[0145] S306: Perform initial solution using initial boundary conditions and material properties to calculate the potential distribution of the entire model. Then, adjust the resistivity distribution of the internal area based on the calculation results and boundary conditions, and iterate the solution repeatedly until the calculation results converge.

[0146] Specifically, an initial solution is performed using initial boundary conditions and material properties to calculate the potential distribution of the entire model.

[0147] Calculate the current density based on the potential distribution and resistivity of the current iteration:

[0148]

[0149] Among them, J t represents the current density at the tth iteration, σ t represents the resistivity distribution at the tth iteration, V t represents the potential distribution at the tth iteration.

[0150] Construct the error function after each iteration:

[0151]

[0152] Among them, e J represents the error function, J a Represents the actual current density distribution measured.

[0153] Update the resistivity distribution by gradient descent:

[0154]

[0155] Among them, σ t+1 represents the resistivity distribution at the t+1th iteration, α t represents the learning rate at the tth iteration, represents the gradient of the error function with respect to the resistivity distribution, β t represents the inertial learning rate at the tth iteration, σ t-1 represents the resistivity distribution at the t-1th iteration.

[0156] It should be noted that the introduction of the inertial learning rate is equivalent to adding a momentum in the update process, which can effectively accelerate the convergence speed of gradient descent, especially when optimizing a large parameter space. The introduction of the inertial term helps to smooth the path of gradient updates and reduce oscillations and instabilities during the update process, especially on complex error surfaces. The combination of adaptive learning rate and inertial term can dynamically adjust the update direction and amplitude according to the current gradient information and historical update information, making the optimization process more flexible and intelligent.

[0157] Optionally, the learning rate is calculated as:

[0158] α t = α 0 / ∑ k = 1 t [ ∇ e J ( s k ) ] 2 + t

[0159] Among them, α t represents the learning rate at the tth iteration, α0 represents the initial learning rate, represents the gradient of the error function with respect to the resistivity distribution at the kth iteration, , τ represents a hyperparameter.

[0160] It should be noted that the learning rate will be dynamically adjusted according to the gradient of the error function. When the gradient is large, the learning rate is small; when the gradient is small, the learning rate is large. This dynamic adjustment can make the optimization process more stable and efficient. The adaptive learning rate can adapt to different optimization stages. In the early stage, a larger step size may be required to quickly decrease. When approaching the optimal value, the step size will automatically decrease to improve the convergence accuracy.

[0161] Determine whether the function value of the error function is less than the preset error value; if so, stop iterating; otherwise, iterate again to update the potential distribution and current density distribution.

[0162] In the present invention, through systematic iterative solution and dynamic adjustment, not only can subtle resistivity changes be captured, but also accurate results can be quickly converged, providing strong support for material property analysis and quality control. This method has wide applicability and important significance in practical applications.

[0163] S4: constructing a resistivity distribution map of the substrate and a resistivity distribution map of the aluminum coating film according to the resistivity distribution of the substrate and the aluminum coating film.

[0164] Among them, the resistivity distribution diagram can represent the resistivity distribution of each point on the aluminum-plated structural part.

[0165] S5: The resistivity distribution map of the substrate and the resistivity distribution map of the aluminum coating film are merged to construct a merged resistivity distribution map.

[0166] In a possible implementation, S5 specifically includes: constructing a fused resistivity distribution map by the following fusion method:

[0167] (10)

[0168] Among them, f represents the fused resistivity distribution map, f1 represents the resistivity distribution map of the substrate, λ1 represents the fusion coefficient of the resistivity distribution map of the substrate, f2 represents the resistivity distribution map of the aluminum coating film, and λ2 represents the fusion coefficient of the resistivity distribution map of the aluminum coating film.

[0169] Among them, those skilled in the art can set the size of the fusion coefficient λ1 of the resistivity distribution diagram of the substrate and the fusion coefficient λ2 of the resistivity distribution diagram of the aluminum-plated film according to actual conditions, and the present invention does not limit this.

[0170] In the present invention, by fusing the resistivity distribution maps of the substrate and the aluminum-plated film, a comprehensive, accurate and intuitive resistivity distribution map can be provided. This method not only improves the measurement accuracy and data processing efficiency, but also optimizes the material design and production process, and enhances the quality control and performance prediction capabilities. This fusion method has important significance and wide applicability in practical applications.

[0171] S6: Based on the fused resistivity distribution map, a convolutional neural network is used to detect whether the resistivity of the aluminum-plated structural parts is abnormal.

[0172] Among them, Convolutional Neural Network (CNN) is a deep learning model that can automatically learn features in images by simulating the visual cortex structure in biological visual systems, and perform classification and prediction through a series of convolution operations, pooling operations, and fully connected layers.

[0173] Optionally, the convolutional neural network specifically includes: a convolution layer, a pooling layer, a fully connected layer and an output layer.

[0174] In a possible implementation, S6 specifically includes sub-steps S601 to S605:

[0175] S601: Using the convolutional layer in the convolutional neural network, extract the features of the fused feature map:

[0176] (11)

[0177] in, Indicates The features of the convolutional layer output using the i-th input feature map, Indicates The features of the convolutional layer output using the i-th input feature map, Indicates The convolution layer uses the convolution kernel weights when the i-th input feature map is used. Indicates The convolutional layer uses the bias term when the i-th input feature map is used, M j represents the selected input feature map set, and J( ) represents the activation function.

[0178] It should be noted that the convolution layer slides over the input data through the convolution kernel (filter) and performs dot product operations. Each convolution kernel extracts different features (such as edges, textures).

[0179] S602: Use the pooling layer in the CNN network to reduce the dimension of the extracted features:

[0180] (12)

[0181] Among them, down() represents the downsampling function, Indicates The features of the output of the pooling layer using the i-th input feature map, Indicates The features of the output of the pooling layer using the i-th input feature map.

[0182] It should be noted that the pooling layer reduces the dimension of the data by taking the maximum value (maximum pooling) or the average value (average pooling) of the local area, while retaining important features and reducing computational complexity and the risk of overfitting.

[0183] S603: Use the fully connected layer in the CNN network to summarize the features and obtain the hidden state.

[0184] S604: Using the output layer in the CNN network, calculate the outliers belonging to the aluminum-plated structural parts:

[0185] (13)

[0186] Among them, P represents an outlier, Sigmoid () represents the Sigmoid activation function, W represents the weight matrix, h represents the hidden state, and b represents the bias term.

[0187] S605: Determine whether the abnormal value is greater than a preset value. If so, determine that the resistivity of the aluminum-plated structural component is abnormal. Otherwise, determine that the resistivity of the aluminum-plated structural component is not abnormal.

[0188] Among them, those skilled in the art can set the size of the preset value according to actual conditions, and the present invention does not limit it.

[0189] It should be noted that the fully connected layer summarizes the features, the output layer calculates the outliers, and the Sigmoid activation function is combined to convert the output of the model into an intuitive outlier probability value. This method can not only improve the expressiveness of the model, but also simplify the calculation and optimization process, and is suitable for efficient automatic anomaly detection.

[0190] In a possible implementation manner, S6 specifically further includes:

[0191] S606: When the resistivity of the aluminum-plated structural component is abnormal, the abnormal position of the aluminum-plated structural component is output.

[0192] In the present invention, by outputting the abnormal part when the convolutional neural network detects resistivity anomalies, accurate positioning can be achieved and the repair efficiency can be improved. It not only helps to discover and solve current quality problems, but also prevents possible abnormalities in the future through data analysis and process improvements, thereby improving the overall production and management level.

[0193] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0194] (1) In the present invention, the resistivity of the test piece is accurately measured by using non-contact millimeter wave reflection technology, which is not affected by the thickness of the material. For multi-layer materials, the resistivity of both sides of the material can be measured by simply turning the material over, thereby improving the accuracy and efficiency of resistivity measurement, and having low requirements on the measurement environment, which can reduce the cost of resistivity detection.

[0195] (2) Through finite element analysis, the resistivity distribution of the substrate and the aluminum coating can be analyzed in detail, providing global resistivity information rather than just local measurements, ensuring a comprehensive evaluation of the entire structure. At the same time, by fusing the resistivity distribution maps of the substrate and the aluminum coating, a comprehensive resistivity distribution map is obtained, which helps to evaluate the resistivity of the aluminum-coated structural parts as a whole and improve the comprehensiveness and accuracy of the detection. Using convolutional neural networks for automated detection can efficiently identify abnormal areas in the resistivity distribution, improve detection speed and accuracy, and reduce errors caused by human intervention.

[0196] Reference Manual Attached Figure 2 , showing a structural schematic diagram of a lithium battery aluminum plating resistivity detection system provided by the present invention.

[0197] The present invention also provides a lithium battery aluminum plating resistivity detection system 20, which is applied to the above-mentioned lithium battery aluminum plating resistivity detection method, comprising:

[0198] Processor 201.

[0199] The memory 202 stores computer-readable instructions. When the computer-readable instructions are executed by the processor 201, the method for detecting the resistivity of aluminum plating of a lithium battery in the method embodiment is implemented.

[0200] The lithium battery aluminum plating resistivity detection system 20 provided by the present invention can execute the above-mentioned lithium battery aluminum plating resistivity detection method and achieve the same or similar technical effects. To avoid repetition, the present invention will not go into details.

[0201] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0202] (1) In the present invention, the resistivity of the test piece is accurately measured by using non-contact millimeter wave reflection technology, which is not affected by the thickness of the material. For multi-layer materials, the resistivity of both sides of the material can be measured by simply turning the material over, thereby improving the accuracy and efficiency of resistivity measurement, and having low requirements on the measurement environment, which can reduce the cost of resistivity detection.

[0203] (2) Through finite element analysis, the resistivity distribution of the substrate and the aluminum coating can be analyzed in detail, providing global resistivity information rather than just local measurements, ensuring a comprehensive evaluation of the entire structure. At the same time, by fusing the resistivity distribution maps of the substrate and the aluminum coating, a comprehensive resistivity distribution map is obtained, which helps to evaluate the resistivity of the aluminum-coated structural parts as a whole and improve the comprehensiveness and accuracy of the detection. Using convolutional neural networks for automated detection can efficiently identify abnormal areas in the resistivity distribution, improve detection speed and accuracy, and reduce errors caused by human intervention.

[0204] An embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein when the program is executed by a processor, the method for detecting the resistivity of aluminum plating of a lithium battery as described in the method embodiment is implemented.

[0205] A computer-readable storage medium provided by the present invention can implement the steps and effects of the lithium battery aluminum plating resistivity detection method of the above method embodiment. To avoid repetition, the present invention will not go into details.

[0206] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0207] (1) In the present invention, the resistivity of the test piece is accurately measured by using non-contact millimeter wave reflection technology, which is not affected by the thickness of the material. For multi-layer materials, the resistivity of both sides of the material can be measured by simply turning the material over, thereby improving the accuracy and efficiency of resistivity measurement, and having low requirements on the measurement environment, which can reduce the cost of resistivity detection.

[0208] (2) Through finite element analysis, the resistivity distribution of the substrate and the aluminum coating can be analyzed in detail, providing global resistivity information rather than just local measurements, ensuring a comprehensive evaluation of the entire structure. At the same time, by fusing the resistivity distribution maps of the substrate and the aluminum coating, a comprehensive resistivity distribution map is obtained, which helps to evaluate the resistivity of the aluminum-coated structural parts as a whole and improve the comprehensiveness and accuracy of the detection. Using convolutional neural networks for automated detection can efficiently identify abnormal areas in the resistivity distribution, improve detection speed and accuracy, and reduce errors caused by human intervention.

[0209] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

[0210] The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto.

Claims

1. A method for detecting the resistivity of aluminum plating of a lithium battery, characterized in that: include: S1: obtaining an aluminum-plated structural component of a lithium battery, wherein the aluminum-plated structural component comprises a substrate and an aluminum-plated film; S2: Using non-contact millimeter wave reflection technology, respectively measuring the resistivity of the substrate and the edge of the aluminum coating; S3: Taking the resistivity at the edge as the boundary condition, finite element analysis is performed to determine the resistivity distribution of the substrate and the aluminum-plated film respectively; S3 specifically includes: S301: Construct the geometric model of aluminum-plated structural parts; S302: Using the measured resistivity of the edge points as boundary conditions, setting the resistivity value at the boundary of the geometric model; S303: setting the initial resistivity of the substrate and the aluminum coating; S304: Divide the geometric model into a finite element mesh, ensuring that the mesh is fine enough to capture the details of the resistivity distribution; S305: Setting the basic battery equation: in, represents the vector differential operator, represents the electric field strength, represents the electric flux density, represents the magnetic field strength, represents the magnetic induction intensity, represents the current density, q represents the charge density; In a static battery: in, ε represents the dielectric constant, μ represents the magnetic permeability, σ represents the resistivity, where , ρ represents resistivity; S306: Perform initial solution using initial boundary conditions and material properties to calculate the potential distribution of the entire model, and then adjust the resistivity distribution of the internal area according to the calculation results and boundary conditions, and repeatedly iterate the solution until the calculation results converge; Calculate the current density based on the potential distribution and resistivity of the current iteration: in, J t Indicates t The current density of the iteration is, σ t Indicates t The resistivity distribution at the iteration, V t Indicates t The potential distribution at the iteration; Construct the error function after each iteration: in, e J represents the error function, J a Indicates the actual current density distribution measured; Update the resistivity distribution by gradient descent: in, σ t+1 Indicates t Resistivity distribution at +1 iteration, α t Indicates t The learning rate at the iteration, represents the gradient of the error function with respect to the resistivity distribution, β t Indicates t The inertia learning rate at the iteration, σ t-1 Indicates t -Resistivity distribution at 1 iteration; The learning rate is calculated as: in, α t Indicates t The learning rate at the iteration, α 0 represents the initial learning rate, Indicates k The gradient of the error function with respect to the resistivity distribution at the iteration, , τ represents a hyperparameter; S4: constructing a resistivity distribution map of the substrate and a resistivity distribution map of the aluminum-plated film respectively according to the resistivity distribution of the substrate and the aluminum-plated film; S5: fusing the resistivity distribution map of the substrate and the resistivity distribution map of the aluminum-plated film to construct a fused resistivity distribution map; S6: According to the fused resistivity distribution map, a convolutional neural network is used to detect whether the resistivity of the aluminum-plated structural component is abnormal.

2. The method for detecting resistivity of aluminum plating of lithium battery according to claim 1, characterized in that: The S2 specifically includes: S201: Construct a resistivity measurement model based on millimeter wave reflection technology; S202: Acquire millimeter wave reflection signals of each edge point; S203: Calculate the resistivity of each edge point according to the millimeter wave reflection signal.

3. The method for detecting resistivity of aluminum plating of lithium battery according to claim 2, characterized in that: The S201 specifically includes: Obtaining a correlation between a millimeter wave reflection coefficient and an inherent impedance of a device under test, wherein the device under test includes the substrate and the aluminum-plated film: Where Γ is the reflection coefficient, η represents the inherent impedance of the device under test, η 0 represents the inherent impedance of free space; The inherent impedance of the device under test is specifically: in, σ Indicates the conductivity of the test piece. μ Indicates the magnetic permeability of the test piece. ε Indicates the dielectric constant of the device under test, ω represents the angular frequency, j represents an imaginary unit; The free space inherent impedance is specifically: in, μ 0 represents the permeability of free space, ε 0 represents the dielectric constant of free space; For non-magnetic materials: Combining formula (1), formula (2), formula (3) and formula (4), we have: in, X represents the real part of the reflection coefficient, Y represents the imaginary part of the reflection coefficient; Through the real part of formula (5) X and the imaginary part Y , and eliminate the dielectric constant of the DUT ε , we get the conductivity measurement equation: ； The resistivity measurement model is constructed by formula (7): in, ρ Represents resistivity.

4. The method for detecting resistivity of aluminum plating of lithium battery according to claim 3, characterized in that: The S203 specifically includes: According to the millimeter wave reflection signal at each edge point, the real part and the imaginary part of the reflection coefficient are calculated: Among them, Γ m Indicates m The reflection coefficient at the time of the measurement is X m Indicates m The real part of the reflection coefficient at the time of the measurement, Y m Indicates m The imaginary part of the reflection coefficient at the time of the measurement, A m Indicates m The amplitude at the time of the measurement, θ m Indicates m The phase at the time of the second measurement, e represents a natural constant; The real part and the imaginary part of the reflection coefficient measured multiple times are averaged to obtain the real part average value and the imaginary part average value; Substitute the real part average value and the imaginary part average value into formula (7) to calculate the resistivity of each edge point.

5. The method for detecting resistivity of aluminum plating of lithium battery according to claim 4, characterized in that: The S2 specifically includes: In order to eliminate the error influence of the measurement system, the calculated resistivity is calibrated by the following equation: in, σ Indicates the resistivity of the test piece. C 1. C 2 represents the calibration factor, σ * Indicates the resistivity before calibration.

6. The method for detecting resistivity of aluminum plating of lithium battery according to claim 1, characterized in that: The S5 is specifically: The fused resistivity distribution map is constructed by the following fusion methods: in, f represents the fusion resistivity distribution map, f 1 represents the resistivity distribution diagram of the substrate, λ 1 represents the fusion coefficient of the resistivity distribution map of the substrate, f 2 represents the resistivity distribution diagram of the aluminum coating. λ 2 represents the fusion coefficient of the resistivity distribution diagram of the aluminum coating.

7. The method for detecting resistivity of aluminum plating of lithium battery according to claim 1, characterized in that: The convolutional neural network specifically includes: a convolutional layer, a pooling layer, a fully connected layer and an output layer.

8. The method for detecting resistivity of aluminum plating of lithium battery according to claim 7, characterized in that: The S6 specifically includes: S601: Using the convolution layer in the convolutional neural network, extracting the features of the fused feature map: in, Indicates l The output of the convolutional layer is i The features of the input feature map, Indicates l -1 convolutional layer output adopts the i The features of the input feature map, Indicates l The convolutional layer uses i The convolution kernel weights for each input feature map, Indicates l The convolutional layer uses i The bias term when inputting feature maps, M j represents the selected set of input feature maps, J ( ) represents the activation function; S602: Use the pooling layer in the CNN network to reduce the dimension of the extracted features: in, down ( ) represents the downsampling function, Indicates l The output of the pooling layer is i The features of the input feature map, Indicates l -1 pooling layer output adopts the i Features when input feature mapping; S603: Use the fully connected layer in the CNN network to summarize the features and obtain the hidden state; S604: Calculate the abnormal value belonging to the aluminum-plated structural part by using the output layer in the CNN network: in, P represents an outlier, Sigmoid ( ) represents the Sigmoid activation function, W represents the weight matrix, h represents the hidden state, b represents the bias term; S605: Determine whether the abnormal value is greater than a preset value; if so, determine that the resistivity of the aluminum-plated structural component is abnormal; otherwise, determine that the resistivity of the aluminum-plated structural component is not abnormal.

9. The method for detecting resistivity of aluminum plating of lithium battery according to claim 8, characterized in that: The S6 specifically also includes: S606: When the resistivity of the aluminum-plated structural component is abnormal, output the abnormal position of the aluminum-plated structural component.

10. A lithium battery aluminum plating resistivity detection system, characterized in that: include: processor; A memory having computer-readable instructions stored thereon, wherein when the computer-readable instructions are executed by the processor, the method for detecting the resistivity of aluminum plating of a lithium battery as claimed in any one of claims 1 to 9 is implemented.

Images