Lithium battery lithium precipitation detection method and system based on laser ultrasonic multi-feature fusion

Abstract

The application provides a lithium battery lithium precipitation detection method and system based on laser ultrasonic multi-feature fusion, and relates to the field of battery detection. The detection method comprises the following steps: controlling a laser generating device to irradiate a lithium battery sample with laser and obtaining time domain data of ultrasonic signals from the lithium battery sample; converting the time domain data into frequency domain data, fusing the time domain data and the frequency domain data through a cross attention mechanism to obtain ultrasonic fusion features; and inputting the ultrasonic fusion features into a trained lithium precipitation prediction model to obtain the lithium precipitation probability of the lithium battery sample. This detection method can realize non-contact nondestructive detection while improving the detection accuracy.

Description

Technical Field

[0001] This invention relates to the field of battery testing, and in particular to a method and system for detecting lithium plating in lithium batteries based on laser-ultrasound multi-feature fusion. Background Technology

[0002] Lithium-ion batteries are widely used in power batteries and energy storage systems due to their high energy density and long cycle life. However, under conditions such as low temperature, high-rate charging, or overcharging, lithium metal deposition easily occurs on the negative electrode surface. Lithium deposition leads to increased internal resistance and capacity decay, and may induce lithium dendrite growth, potentially causing internal short circuits and other safety hazards. Therefore, accurate detection and evaluation of lithium deposition status is of great significance for improving the safety and reliability of lithium-ion batteries.

[0003] The relevant technical solutions make it difficult to achieve both non-destructive testing and testing accuracy. Summary of the Invention

[0004] This invention provides a lithium battery lithium plating detection method and system based on laser and ultrasonic multi-feature fusion, which solves the technical problem of how to improve detection accuracy while performing non-destructive testing.

[0005] The first aspect of this invention provides a method for detecting lithium plating in lithium batteries by laser-ultrasound multi-feature fusion. The method includes: controlling a laser generator to irradiate a lithium battery sample with laser light and acquiring time-domain data of ultrasonic signals from the lithium battery sample; converting the time-domain data into frequency-domain data; fusing the time-domain data and the frequency-domain data through a cross-attention mechanism to obtain ultrasonic fusion features; and inputting the ultrasonic fusion features into a trained lithium plating prediction model to obtain the lithium plating probability of the lithium battery sample.

[0006] In some implementations, fusing the time-domain data and the frequency-domain data through a cross-attention mechanism to obtain ultrasound fusion features includes: performing one-dimensional convolution processing on the time-domain data to obtain a time-domain minimum semantic sequence; obtaining a time-frequency two-dimensional graph based on the time-domain data and the frequency-domain data, and dividing the time-frequency two-dimensional graph into multiple time-frequency data subsets; converting each of the time-frequency data subsets into a two-dimensional time-frequency minimum semantic sequence, expanding the two-dimensional time-frequency minimum semantic sequence into a one-dimensional time-frequency minimum semantic sequence; and combining the time-domain minimum semantic sequence and the one-dimensional time-frequency minimum semantic sequence to obtain the ultrasound fusion feature.

[0007] In some implementations, the step of inputting the ultrasonic fusion features into the trained lithium plating prediction model to obtain the lithium plating probability of the lithium battery sample includes: acquiring laser data and associating the laser data with the ultrasonic fusion features to obtain a condition-guided ultrasonic lithium plating representation; and inputting the condition-guided ultrasonic lithium plating representation into the lithium plating prediction model to obtain the lithium plating probability of the lithium battery sample.

[0008] In some embodiments, the laser data includes the laser pulse width, single-pulse energy, and spot size; acquiring the laser data and associating the laser data with the ultrasound fusion features to obtain a condition-guided ultrasound lithium plating representation includes: constructing the laser pulse width, single-pulse energy, and spot size as an excitation condition vector; inputting the excitation condition vector into an excitation condition encoding module to obtain an excitation condition embedding representation, wherein the excitation condition encoding module includes a multilayer sensing mechanism; and associating the excitation condition embedding representation with the ultrasound fusion features to obtain the condition-guided ultrasound lithium plating representation.

[0009] In some embodiments, obtaining the lithium plating probability of the lithium battery sample by inputting the ultrasonic fusion features into the trained lithium plating prediction model includes: inputting the ultrasonic fusion features of each position of the lithium battery into the trained lithium plating prediction model, the lithium plating prediction model obtaining the lithium plating probability of each position of the lithium battery; mapping the lithium plating probability of each position of the lithium battery to the corresponding position of the lithium battery, and determining the position where the lithium plating probability is higher than the lithium plating threshold as the position where lithium plating exists.

[0010] In some embodiments, before the controlled laser generator irradiates the lithium battery sample with laser and the lithium battery sample acquires time-domain data of the ultrasonic signal, the lithium battery lithium plating detection method further includes: continuously iteratively training a lithium plating prediction model based on a labeled training dataset, obtaining a loss value each time, and iteratively adjusting the lithium plating prediction model according to the loss value until the loss value is continuously less than a loss value threshold. The labeled training dataset includes the actual first-wave time-of-flight offset and the actual temperature of the battery surface. The lithium plating prediction model is used to output the lithium plating probability and stress prediction value. The loss value includes the difference between the actual first-wave time-of-flight offset and the predicted first-wave time-of-flight offset.

[0011] A second aspect of this invention provides a lithium battery series detection system based on laser-ultrasound multi-feature fusion. This lithium battery lithium plating detection system includes: a data acquisition module for controlling a laser generator to irradiate a lithium battery sample with laser light and acquiring time-domain data of ultrasonic signals from the lithium battery sample; a data fusion module for converting the time-domain data into frequency-domain data and fusing the time-domain data and the frequency-domain data through a cross-attention mechanism to obtain ultrasonic fusion features; and a lithium plating prediction module for inputting the ultrasonic fusion features into a trained lithium plating prediction model to obtain the lithium plating probability of the lithium battery sample.

[0012] In some embodiments, the data fusion module is further configured to perform one-dimensional convolution processing on the time-domain data to obtain a time-domain minimum semantic sequence; to obtain a time-frequency two-dimensional map based on the time-domain data and the frequency-domain data, and to divide the time-frequency two-dimensional map into multiple time-frequency data subsets; to convert each of the time-frequency data subsets into a two-dimensional time-frequency minimum semantic sequence, and to expand the two-dimensional time-frequency minimum semantic sequence into a one-dimensional time-frequency minimum semantic sequence; and to combine the time-domain minimum semantic sequence and the one-dimensional time-frequency minimum semantic sequence to obtain the ultrasound fusion feature.

[0013] In some embodiments, the data acquisition module is further configured to acquire laser data; the data fusion module is further configured to correlate the laser data and the ultrasonic fusion features to obtain a condition-guided ultrasonic lithium plating representation; and the lithium plating prediction module is further configured to input the condition-guided ultrasonic lithium plating representation into the lithium plating prediction model to obtain the lithium plating probability of the lithium battery sample.

[0014] In some embodiments, the laser data includes the laser pulse width, single-pulse energy, and spot size; the data fusion module is further configured to construct an excitation condition vector from the laser pulse width, single-pulse energy, and spot size; further configured to input the excitation condition vector into an excitation condition encoding module to obtain an excitation condition embedding representation, wherein the excitation condition encoding module includes a multilayer sensing mechanism; and further configured to associate the excitation condition embedding representation with the ultrasonic fusion feature to obtain the ultrasonic lithium plating representation.

[0015] This invention provides a lithium battery lithium plating detection method based on laser-ultrasound multi-feature fusion. This method generates ultrasonic signals by irradiating a lithium battery sample with a laser, and then predicts the lithium plating state based on these ultrasonic signals, achieving non-contact, non-destructive lithium plating prediction. Simultaneously, it fuses the time-domain and frequency-domain data of the ultrasonic signals using a cross-attention mechanism to obtain ultrasonic fusion features. These features allow for the analysis of the correlation between the time-domain and frequency-domain data. Inputting these fusion features into a trained lithium plating prediction model yields a lithium plating probability that more accurately predicts the lithium battery's lithium plating state. In other words, it improves the accuracy of lithium battery lithium plating state detection while performing non-contact, non-destructive measurement. Attached Figure Description

[0016] Figure 1 A flowchart illustrating the first lithium battery lithium plating detection method based on laser-ultrasound multi-feature fusion is provided for embodiments of the present invention. Figure 2 A flowchart illustrating a second lithium battery lithium plating detection method based on laser-ultrasound multi-feature fusion is provided for embodiments of the present invention. Figure 3 A flowchart illustrating a third lithium battery lithium plating detection method based on laser-ultrasound multi-feature fusion is provided for embodiments of the present invention. Figure 4 for Figure 3 A flowchart illustrating step S301 in the process; Figure 5 A flowchart illustrating a fourth lithium battery lithium plating detection method based on laser-ultrasound multi-feature fusion is provided for embodiments of the present invention. Figure 6 A flowchart illustrating the fifth lithium battery lithium plating detection method based on laser-ultrasound multi-feature fusion is provided for embodiments of the present invention. Figure 7 This invention provides an architecture diagram of a lithium battery lithium plating detection system based on laser-ultrasound multi-feature fusion. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0018] The specific technical features described in the various embodiments in the detailed implementation can be combined in various ways without contradiction. For example, different implementation methods can be formed by combining different specific technical features. In order to avoid unnecessary repetition, the various possible combinations of the specific technical features in this invention will not be described separately.

[0019] It should also be noted that, in order to avoid obscuring the present invention with unnecessary details, only the structures and / or processing steps closely related to the present invention are shown in the accompanying drawings, while other details that are not closely related to the present invention are omitted.

[0020] Additionally, it should be noted that the terms "comprising," "including," or any other variations thereof 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. In the following description, the terms "first," "second," etc., are used merely to distinguish different objects and do not indicate any similarity or connection between them. It should be understood that the directional descriptions such as "above," "below," "inside," and "outside" refer to the orientation under normal use conditions.

[0021] In some implementations, such as Figure 1 As shown, the main steps of the lithium battery lithium plating detection method based on laser-ultrasound multi-feature fusion include: Step S101: Control the laser generator to irradiate the lithium battery sample with laser and obtain the time domain data of the ultrasonic signal from the lithium battery sample.

[0022] That is, the laser generator is controlled to irradiate the lithium battery sample, and the lithium battery sample generates an ultrasonic signal under the excitation of the laser. The time domain data of the ultrasonic signal can be obtained through the ultrasonic acquisition device.

[0023] Step S102: Convert the time domain data into frequency domain data, and fuse the time domain data and frequency domain data through a cross-attention mechanism to obtain ultrasound fusion features.

[0024] This can be understood as follows: by converting time-domain data into frequency-domain data through Fourier transform, and then using a cross-attention mechanism to analyze the intrinsic relationship between the frequency-domain and time-domain data, the lithium plating prediction obtained by the fused ultrasonic fusion features can more accurately predict the lithium plating state of lithium batteries. That is, the lithium plating information in laser ultrasound is not hidden in the time domain or frequency domain alone, but in the correspondence between the two. Cross-attention can link these originally scattered information in the two domains to form a more stable and physically meaningful fused characterization. Based on this feature, the lithium plating prediction is more accurate.

[0025] Optionally, time-domain data and frequency-domain data can be cross-fused as a whole through a cross-attention mechanism.

[0026] Step S103: Input the ultrasonic fusion features into the trained lithium plating prediction model to obtain the lithium plating probability of the lithium battery sample.

[0027] Optionally, the ultrasonic fusion feature can be the overall feature obtained by fusing the time-domain data and frequency-domain data of the overall ultrasonic signal of the lithium battery sample. Correspondingly, the lithium plating probability output by the lithium plating prediction model is the overall lithium plating probability of the lithium battery sample. If the lithium plating probability is greater than the preset probability threshold, it is considered that the lithium battery sample as a whole has lithium plating. Optionally, the ultrasonic fusion feature can be the fusion feature obtained by fusing the time-domain data and frequency-domain data of different positions of the lithium battery sample. The fusion features of different positions are input into the trained lithium plating prediction model to obtain the lithium plating probability of each position. The lithium plating probability of each position is mapped to the corresponding position of the lithium battery sample, and the lithium plating state of each position of the lithium battery is obtained based on the relationship between the lithium plating probability and the preset probability threshold.

[0028] One method for mapping lithium plating probability to the corresponding position is to record the corresponding coordinate position while obtaining the fused features of each position, and then map the probability to the corresponding coordinate position based on the recorded coordinate position after obtaining the corresponding lithium plating probability. Another method is to arrange the fused features in sequence based on the position corresponding to the fused features to obtain a fused feature matrix, input the fused feature matrix into the trained lithium plating prediction model to obtain a lithium plating probability matrix, and then map the lithium plating probability to the corresponding position of the lithium battery sample based on the position of each lithium plating probability in the matrix.

[0029] This invention provides a lithium battery lithium plating detection method based on laser-ultrasound multi-feature fusion. This method generates ultrasonic signals by irradiating a lithium battery sample with a laser, and then predicts the lithium plating state based on these ultrasonic signals, achieving non-contact, non-destructive lithium plating prediction. Simultaneously, it fuses the time-domain and frequency-domain data of the ultrasonic signals using a cross-attention mechanism to obtain ultrasonic fusion features. These features allow for the analysis of the correlation between the time-domain and frequency-domain data. Inputting these fusion features into a trained lithium plating prediction model yields a lithium plating probability that more accurately predicts the lithium battery's lithium plating state. In other words, it improves the accuracy of lithium battery lithium plating state detection while performing non-contact, non-destructive measurement.

[0030] In some embodiments, such as Figure 2 As shown, with Figure 1 The lithium battery lithium plating detection method shown is different from that of the lithium battery lithium plating detection method shown. Figure 1 Step S102 includes: Step S201: Perform one-dimensional convolution on the time-domain data to obtain the time-domain minimum semantic sequence.

[0031] That is, by processing the temporal data through one-dimensional convolution, the temporal data is transformed into a token sequence (minimum semantic sequence) that satisfies the attention mechanism.

[0032] Step S202: Obtain a two-dimensional time-frequency graph based on time-domain data and frequency-domain data, and divide the two-dimensional time-frequency graph into multiple time-frequency data subsets.

[0033] This can be understood as follows: time-domain data is divided into multiple time-domain data subsets according to predetermined time intervals; each time-domain data subset is subjected to Fourier transform to obtain sub-frequency domain data within the corresponding time period; and the sub-frequency domain data is associated with the time-domain data subsets within the corresponding time period to obtain a two-dimensional time-frequency graph. The time-frequency graph can align the time-domain data and frequency-domain data on the time scale, and can more fully reflect the association between the time-domain data and frequency-domain data within the corresponding time period when associated through the cross-attention mechanism in the subsequent process. At the same time, in order to facilitate subsequent processing, the time-frequency graph is divided into multiple blocks, each block corresponding to a time-frequency data subset.

[0034] Step S203: Convert each time-frequency data subset into a two-dimensional time-frequency minimum semantic sequence, and expand the two-dimensional time-frequency minimum semantic sequence into a one-dimensional time-frequency minimum semantic sequence.

[0035] Specifically, each time-frequency data subset is transformed into a two-dimensional minimal semantic sequence that meets the requirements of the attention mechanism in the form of a two-dimensional matrix. In order to facilitate the association of this sequence with the time-domain minimal semantic sequence, the two-dimensional minimal semantic sequence also needs to be expanded into a one-dimensional time-frequency minimal semantic sequence.

[0036] Step S204: Combine the time-domain minimum semantic sequence and the one-dimensional time-frequency minimum semantic sequence to obtain the ultrasound fusion feature.

[0037] Optionally, a linear channel can be used as a channel mapping, and fusion can be performed through a cross-attention mechanism to allow key waveforms in the time-domain branch to be fused and represented, thereby achieving time-frequency information alignment and complementary enhancement.

[0038] In some embodiments, such as Figure 3 As shown, with Figure 1 The lithium battery lithium plating detection method shown is different from that of the lithium battery lithium plating detection method shown. Figure 1 Step S103 includes: Step S301: Acquire laser data and correlate the laser data with ultrasonic fusion features to obtain a condition-guided ultrasonic lithium plating representation.

[0039] This can be understood as follows: ultrasonic data is related to laser data. By using laser data as an input condition and associating it with ultrasonic fusion features, we can further explore the relationship between the input condition and ultrasonic fusion features. Thus, by using this condition-guided ultrasonic series representation as a lithium plating prediction model, we can more accurately obtain the lithium plating state of lithium battery samples.

[0040] Optionally, the laser data may include laser energy and energy distribution. The laser data and ultrasound fusion features can be fused by storing them in the same vector to obtain a conditionally guided ultrasound series representation.

[0041] Step S302: Input the condition-guided ultrasonic lithium plating representation into the lithium plating prediction model to obtain the lithium plating probability of the lithium battery sample.

[0042] In some embodiments, laser parameters include the laser pulse width, single-pulse energy, and spot size, such as... Figure 4 As shown, Figure 3 Step S301 includes: Step S401: Construct the laser pulse width, single pulse energy, and spot size into an excitation condition vector.

[0043] That is, by using the laser pulse width, single pulse energy, and spot size to represent the action time, single pulse energy, and single pulse range of a single laser pulse, respectively, constructing these three parameters into an excitation condition vector can more fully reflect the laser's effect on the lithium battery sample.

[0044] Step S402: Input the excitation condition vector into the excitation condition encoding module to obtain the excitation condition embedded representation.

[0045] The excitation condition encoding module includes a multilayer perceptron, which uses the multilayer perceptron to deeply mine the excitation condition vector to obtain the correlation between various laser parameters, thereby obtaining the excitation condition embedded representation.

[0046] Step S403: Associate the excitation condition embedding representation with the ultrasonic fusion feature to obtain the condition-guided ultrasonic lithium plating representation.

[0047] Specifically, this embedding representation combines feature splicing or feature linear conditional modulation modules with ultrasonic fusion features to form a condition-guided ultrasonic lithium plating representation. This enables the network to learn the correspondence between laser ultrasonic response and lithium plating state under different excitation conditions. Compared to analyzing only ultrasonic data, introducing excitation condition guidance can fully reflect the complex interface and microstructure changes caused by lithium plating, further improving the accuracy of lithium plating detection.

[0048] In some embodiments, such as Figure 5 As shown, with Figure 1 The lithium battery lithium plating detection method shown is different from that of the lithium battery lithium plating detection method shown. Figure 1 Step S103 includes: Step S501: Input the ultrasonic fusion features of each location of the lithium battery into the trained lithium plating prediction model.

[0049] The lithium plating prediction model is used to output the lithium plating probability at each location of the lithium battery. It can be understood that by applying a laser to different locations of the lithium battery sample, ultrasonic data of different detection locations of the lithium battery sample is obtained. Based on the ultrasonic data, ultrasonic fusion data of each detection location is obtained. Inputting the ultrasonic fusion data of each location into the lithium plating prediction model can obtain the lithium plating probability of each detection location.

[0050] Step S502: Map the series probabilities of each position of the lithium battery to the corresponding position of the lithium battery, and determine the position where the lithium plating probability is higher than the lithium plating threshold as the position where lithium plating exists.

[0051] Optionally, the lithium plating prediction model is also used to output the severity of lithium plating and the predicted stress. The lithium plating prediction model also includes an exponential output module, which includes a multilayer perceptron. The exponential output module obtains the overall lithium plating state of the lithium battery sample by combining the lithium plating probability, the severity of lithium plating, the predicted stress, and the relative positional relationship of the detection positions at each detection location.

[0052] In some embodiments, such as Figure 6 As shown, with Figure 1 The lithium battery detection method shown differs from others in that, prior to step S101, the lithium battery lithium plating detection method further includes: Step S601: Continuously train the lithium plating prediction model iteratively based on the labeled training dataset. Each training iteration yields a loss value, and the lithium plating prediction model is iteratively adjusted based on the loss value until the loss value is consistently less than the loss value threshold.

[0053] This can be understood as follows: data from the training dataset is sequentially input into an untrained lithium plating prediction model. The lithium plating prediction model outputs predicted data, and a loss value is obtained based on the difference between the actual data and the predicted data. The thresholds of each node in the lithium plating prediction model are adjusted based on the loss value until the loss value reaches its minimum. This process is called one training iteration. The next labeled training data is input into the model that has completed one training iteration, and the above process is repeated. This is called iterative training of the model. If the loss value of multiple consecutive generations is less than the preset loss value threshold during iterative training, the model is considered to have completed training.

[0054] The labeled training dataset includes time-domain data of actual ultrasonic signals, actual first-wave time-of-flight offsets (FFOs) of lithium battery samples, and actual surface temperatures of the batteries. The lithium plating prediction model outputs the lithium plating probability and stress prediction values. The predicted FFOs can be calculated using the stress prediction values ​​and the actual surface temperature of the batteries. Specifically, the formula for calculating the predicted FFOs is as follows: In the formula, The first-wave flight time offset is calculated based on predicted stress values ​​and actual battery surface temperature. The acoustic elastic coefficient, This is a reference value for the first wave of flight times. This is the predicted stress value. T is the temperature coefficient, T is the actual surface temperature of the battery, and T0 is the reference temperature (the reference temperature is generally selected as room temperature).

[0055] The actual first-wave flight time offset is obtained through the time domain data of the ultrasonic signal. Specifically, the time point corresponding to the first peak in the data threshold of the ultrasonic signal is determined as the first-wave flight time. The actual first-wave flight time offset is obtained by subtracting the first-wave flight time from the preset reference first-wave flight time.

[0056] The loss function is constructed based on the difference between the actual first-wave flight time offset and the predicted first-wave flight time offset. Physical constraints can be added during the training process to ensure that the first-wave flight time offset and the battery surface stress estimation satisfy a physical relationship, thereby further improving the accuracy of the lithium plating function output by the lithium plating prediction model after training.

[0057] Optionally, the loss function can be expressed as: L phys =|ΔToF-g( ,T)| L phys The loss function represents the physical constraints, ΔToF is the actual first-wave flight time offset, and g( ,T) represents the predicted first-wave flight time offset.

[0058] This invention also provides a lithium battery lithium plating detection system, which is used to achieve the following: Figures 1 to 6 The lithium battery lithium plating detection method shown in any one of the images.

[0059] In some embodiments, such as Figure 7 As shown, the lithium battery lithium plating detection system includes: a data acquisition module 100, a data fusion module 200, and a lithium plating prediction module 300. The data acquisition module 100 controls a laser generator to irradiate a lithium battery sample with laser light and acquires time-domain data of ultrasonic signals from the sample. The data fusion module 200 converts the time-domain data into frequency-domain data and fuses the time-domain and frequency-domain data using a cross-attention mechanism to obtain ultrasonic fusion features. The lithium plating prediction module 300 inputs the ultrasonic fusion features into a trained lithium plating prediction model to obtain the lithium plating probability of the lithium battery sample.

[0060] In some embodiments, such as Figure 7As shown, the data fusion module 200 is further used to perform one-dimensional convolution processing on time-domain data to obtain a time-domain minimum semantic sequence; it is also used to obtain a time-frequency two-dimensional map based on time-domain data and frequency-domain data, and divide the time-frequency two-dimensional map into multiple time-frequency data subsets; it is also used to transform each time-frequency data subset into a two-dimensional time-frequency minimum semantic sequence, and expand the two-dimensional time-frequency minimum semantic sequence into a one-dimensional time-frequency minimum semantic sequence; it is also used to combine the time-domain minimum semantic sequence and the one-dimensional time-frequency minimum semantic sequence to obtain ultrasound fusion features.

[0061] In some embodiments, such as Figure 7 As shown, the data acquisition module 100 is also used to acquire laser data; the data fusion module 200 is also used to associate the laser data and ultrasonic fusion features to obtain a condition-guided ultrasonic lithium plating representation; the lithium plating prediction module 300 is also used to input the condition-guided ultrasonic lithium plating representation into the lithium plating prediction model to obtain the lithium plating probability of the lithium battery sample.

[0062] In some embodiments, such as Figure 7 As shown, the laser data includes the laser pulse width, single pulse energy, and spot size; the data fusion module 200 is also used to construct an excitation condition vector from the laser pulse width, single pulse energy, and spot size; it is also used to input the excitation condition vector into the excitation condition encoding module to obtain the excitation condition embedded representation, wherein the excitation condition encoding module includes a multilayer sensing mechanism; it is also used to associate the excitation condition embedded representation with the ultrasonic fusion features to obtain the ultrasonic lithium plating representation.

[0063] The above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for detecting lithium plating in lithium batteries based on laser-ultrasound multi-feature fusion, characterized in that, The lithium battery lithium plating detection method includes: The laser generator is controlled to irradiate a lithium battery sample with a laser, and the time-domain data of the ultrasonic signal is obtained from the lithium battery sample. The time-domain data is converted into frequency-domain data, and the time-domain data and the frequency-domain data are fused through a cross-attention mechanism to obtain ultrasound fusion features. The lithium plating probability of the lithium battery sample is obtained by inputting the ultrasonic fusion features into the trained lithium plating prediction model.

2. The lithium battery lithium plating detection method according to claim 1, characterized in that, The method of fusing the time-domain data and the frequency-domain data through a cross-attention mechanism to obtain ultrasound fusion features includes: The time-domain data is processed by one-dimensional convolution to obtain the time-domain minimum semantic sequence; A time-frequency two-dimensional graph is obtained based on the time-domain data and the frequency-domain data, and the time-frequency two-dimensional graph is divided into multiple time-frequency data subsets; Each of the time-frequency data subsets is transformed into a two-dimensional time-frequency minimum semantic sequence, and the two-dimensional time-frequency minimum semantic sequence is expanded into a one-dimensional time-frequency minimum semantic sequence. The ultrasonic fusion feature is obtained by combining the time-domain minimum semantic sequence and the one-dimensional time-frequency minimum semantic sequence.

3. The lithium battery lithium plating detection method according to claim 1 or 2, characterized in that, The process of inputting the ultrasonic fusion features into the trained lithium plating prediction model to obtain the lithium plating probability of the lithium battery sample includes: Acquire laser data and correlate the laser data with the ultrasonic fusion features to obtain a condition-guided ultrasonic lithium plating representation; The condition-guided ultrasonic lithium plating representation is input into the lithium plating prediction model to obtain the lithium plating probability of the lithium battery sample.

4. The lithium battery lithium plating detection method according to claim 3, characterized in that, The laser data includes the laser pulse width, single pulse energy, and spot size; The step of acquiring laser data and correlating the laser data with the ultrasonic fusion features to obtain a condition-guided ultrasonic lithium plating representation includes: The pulse width of the laser, the single pulse energy, and the spot size are constructed as an excitation condition vector; The excitation condition vector is input into the excitation condition encoding module to obtain the excitation condition embedded representation, wherein the excitation condition encoding module includes a multilayer sensing mechanism; The conditional embedding representation is associated with the ultrasonic fusion feature to obtain the condition-guided ultrasonic lithium plating representation.

5. The lithium battery lithium plating detection method according to claim 1, characterized in that, The process of inputting the ultrasonic fusion features into the trained lithium plating prediction model to obtain the lithium plating probability of the lithium battery sample includes: The ultrasonic fusion features at each location of the lithium battery are input into the trained lithium plating prediction model, and the lithium plating prediction model obtains the lithium plating probability at each location of the lithium battery. The lithium plating probability at each location of the lithium battery is mapped to the corresponding location of the lithium battery, and the location where the lithium plating probability is higher than the lithium plating threshold is determined as the location where lithium plating exists.

6. The lithium battery lithium plating detection method according to claim 1, characterized in that, Before the laser generator irradiates the lithium battery sample with laser light and the time-domain data of the ultrasonic signal are acquired from the lithium battery sample, the lithium battery lithium plating detection method further includes: The lithium plating prediction model is continuously trained iteratively based on a labeled training dataset. Each training iteration yields a loss value, and the model is iteratively adjusted based on this loss value until it is consistently less than a loss threshold. The labeled training dataset includes the actual first-wave flight time offset and the actual battery surface temperature. The lithium plating prediction model outputs the lithium plating probability and stress prediction values. The loss value includes the difference between the actual first-wave flight time offset and the predicted first-wave flight time offset.

7. A lithium battery lithium plating detection system, characterized in that, The lithium plating battery detection system includes: The data acquisition module is used to control the laser generator to irradiate the lithium battery sample with laser and acquire the time-domain data of the ultrasonic signal from the lithium battery sample. The data fusion module is used to convert the time-domain data into frequency-domain data, and fuse the time-domain data and the frequency-domain data through a cross-attention mechanism to obtain ultrasound fusion features; The lithium plating prediction module is used to input the ultrasonic fusion features into the trained lithium plating prediction model to obtain the lithium plating probability of the lithium battery sample.

8. The lithium battery lithium plating detection system according to claim 7, characterized in that, The data fusion module is further configured to perform one-dimensional convolution processing on the time-domain data to obtain a time-domain minimum semantic sequence; to obtain a time-frequency two-dimensional map based on the time-domain data and the frequency-domain data, and to divide the time-frequency two-dimensional map into multiple time-frequency data subsets; to convert each of the time-frequency data subsets into a two-dimensional time-frequency minimum semantic sequence, and to expand the two-dimensional time-frequency minimum semantic sequence into a one-dimensional time-frequency minimum semantic sequence; and to combine the time-domain minimum semantic sequence and the one-dimensional time-frequency minimum semantic sequence to obtain the ultrasound fusion feature.

9. The lithium battery lithium plating detection system according to claim 7 or 8, characterized in that, The data acquisition module is also used to acquire laser data; The data fusion module is also used to correlate the laser data and the ultrasonic fusion features to obtain a condition-guided ultrasonic lithium plating representation; The lithium plating prediction module is further used to input the condition-guided ultrasonic lithium plating representation into the lithium plating prediction model to obtain the lithium plating probability of the lithium battery sample.

10. The lithium battery lithium plating detection system according to claim 9, characterized in that, The laser data includes the laser pulse width, single pulse energy, and spot size; The data fusion module is further configured to construct an excitation condition vector from the pulse width of the laser, the single pulse energy, and the spot size; to input the excitation condition vector into the excitation condition encoding module to obtain an excitation condition embedding representation, wherein the excitation condition encoding module includes a multilayer sensing mechanism; and to associate the excitation condition embedding representation with the ultrasonic fusion feature to obtain the ultrasonic lithium plating representation.

Images