Lithium ion battery lithium precipitation ultrasonic nondestructive testing method and system
By combining unsupervised learning with a fully convolutional network and acoustic wave equations, high-resolution and high-precision non-destructive testing of lithium-ion battery lithium plating defects was achieved, solving the problem of low detection accuracy in existing technologies and improving detection efficiency and accuracy.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- BEIHANG UNIV
- Filing Date
- 2023-06-30
- Publication Date
- 2026-07-14
AI Technical Summary
Existing ultrasonic non-destructive testing methods for lithium-ion batteries can only detect larger lithium plating defects, and the detection accuracy is low, which cannot meet the requirements for high precision.
An unsupervised learning method based on a physical model is adopted. Ultrasonic detection data is acquired by a full matrix capture using an ultrasonic phased array transducer. Unsupervised learning is performed through a fully convolutional network. Combined with the acoustic wave equation, a loss function is constructed for iterative training, and an accurate sound velocity prediction model is output to detect lithium plating defects.
It achieves high-resolution and high-precision detection of lithium-ion battery lithium plating defects, improves detection efficiency, avoids damage to the internal structure of the battery, and maintains the accuracy of detection.
Smart Images
Figure CN116818904B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of lithium-ion battery testing technology, and in particular to an ultrasonic non-destructive testing method and system for lithium-ion battery lithium plating. Background Technology
[0002] The statements herein provide only background information in relation to this invention and do not necessarily constitute prior art.
[0003] With rapid economic development, resource shortages are becoming increasingly serious, necessitating the development of various new energy sources, power batteries, and energy storage systems. Lithium-ion batteries, as rechargeable batteries, possess advantages such as high energy density, high output voltage, good cycle performance, and low self-discharge rate, making them highly promising secondary batteries and chemical energy storage power sources. They are widely used in electric vehicles, aerospace, and energy storage systems. However, lithium-ion batteries are prone to lithium plating during charging, especially at low temperatures, during fast charging (i.e., high-rate charging), and during overcharging. Lithium plating in lithium-ion batteries refers to the abnormal phenomenon where lithium ions do not embed in the negative electrode material during charging, but instead precipitate as metallic lithium on the surface of the negative electrode. Because the precipitated metallic lithium often does not form a smooth coating but exists in the form of dendritic crystals, known as lithium dendrites, excessively grown lithium dendrites may pierce the separator, causing a short circuit between the positive and negative electrodes, and even leading to fires and explosions of the lithium-ion battery. Numerous studies on lithium-ion battery failure mechanisms have shown that lithium plating is a significant cause of capacity decay and even internal short circuits leading to thermal runaway. Therefore, lithium plating detection in lithium-ion batteries is of great importance.
[0004] Currently, conventional methods for analyzing the mechanism and contributing factors of lithium plating in lithium-ion batteries mainly rely on battery disassembly and microscopic characterization. However, since lithium-ion batteries typically have sealed packaging, these methods disrupt the originally enclosed internal system, altering internal information and severely impacting analytical accuracy. Therefore, various non-destructive testing methods (i.e., non-disassembly methods) have been proposed, such as detection methods based on changes in the physical properties of the battery cell caused by lithium.
[0005] Among these methods, ultrasonic non-destructive testing (NDT) offers advantages such as broad applicability and high accuracy. It is sensitive to changes in the internal material properties of the structure; variations in the mechanical properties (e.g., density and modulus) of the electrodes during battery charging and discharging affect the propagation characteristics of ultrasonic waves within the battery. Therefore, utilizing ultrasonic propagation characteristics enables NDT of lithium-ion batteries. However, existing ultrasonic NDT methods for lithium-ion batteries primarily rely on ultrasonic time-of-flight analysis, neglecting information such as ultrasonic diffraction at defects (i.e., lithium plating). This method can only detect defects much larger than the incident ultrasonic wavelength. In other words, existing conventional ultrasonic testing methods are only suitable for detecting larger lithium plating defects and have relatively low detection accuracy. Summary of the Invention
[0006] To address the shortcomings of the existing technologies, this invention provides an ultrasonic non-destructive testing method and system for lithium-ion battery lithium plating. Considering the comprehensive response information such as ultrasonic mode conversion and multiple scattering at the defect location, and through an unsupervised learning method based on a physical model, high-resolution and high-precision non-destructive testing of lithium plating defects in lithium-ion batteries is achieved.
[0007] In one aspect, this disclosure provides an ultrasonic non-destructive testing method for lithium plating in lithium-ion batteries.
[0008] An ultrasonic non-destructive testing method for lithium plating in lithium-ion batteries includes:
[0009] Full-matrix capture ultrasonic testing data of lithium-ion batteries was acquired using an ultrasonic phased array transducer.
[0010] The ultrasonic detection data is input into the initial fully convolutional network, and the output sound velocity prediction model is generated.
[0011] The sound velocity prediction model is used as the initial value input into the acoustic wave equation, and the reconstructed sound velocity data is output.
[0012] The loss function is constructed based on the error between the ultrasonic detection data and the reconstructed sound velocity data. The loss function is continuously iterated until it converges, thus completing the training of the unsupervised fully convolutional network.
[0013] The trained fully convolutional network ultimately outputs an accurate sound velocity prediction model, which is the detection result of lithium plating defects in lithium-ion batteries.
[0014] Secondly, this disclosure provides an ultrasonic non-destructive testing system for lithium-ion battery lithium plating.
[0015] An ultrasonic non-destructive testing system for lithium-ion battery lithium plating includes:
[0016] The data acquisition module is used to acquire full-matrix capture ultrasonic detection data of lithium-ion batteries using an ultrasonic phased array transducer.
[0017] The unsupervised learning fully convolutional network training module is used to input ultrasonic detection data into the initial fully convolutional network and output a sound velocity prediction model; the sound velocity prediction model is used as the initial value to input the acoustic wave equation and output reconstructed sound velocity data; the error between the ultrasonic detection data and the reconstructed sound velocity data is used to construct a loss function, and the process is iterated continuously until the loss function converges, thus completing the training of the unsupervised learning fully convolutional network.
[0018] The lithium plating detection result output module is used to output an accurate sound velocity prediction model from the trained fully convolutional network. The sound velocity prediction model is the detection result of lithium plating defects in lithium-ion batteries.
[0019] Thirdly, this disclosure also provides an electronic device, including a memory and a processor, and computer instructions stored in the memory and running on the processor, wherein the computer instructions, when executed by the processor, perform the steps of the method described in the first aspect.
[0020] Fourthly, this disclosure also provides a computer-readable storage medium for storing computer instructions, which, when executed by a processor, perform the steps of the method described in the first aspect.
[0021] The above one or more technical solutions have the following beneficial effects:
[0022] 1. This invention provides an ultrasonic non-destructive testing method and system for lithium-ion battery lithium plating. Based on the full-matrix capture of full waveform data of ultrasonic testing, an accurate sound velocity prediction model is output through an unsupervised learning method based on a physical model. The sound velocity prediction model can characterize lithium plating defects, thereby achieving high-resolution and high-precision quantitative detection of lithium plating defects in lithium-ion batteries.
[0023] 2. The detection method provided by this invention uses a fully convolutional network with unsupervised learning. Compared with the ultrasonic detection method based on supervised learning, the method of this invention does not require a large training set or a matching relationship between the real sound velocity model and the corresponding ultrasonic detection data in the training set. Therefore, the method proposed by this invention can effectively improve the detection efficiency. Attached Figure Description
[0024] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.
[0025] Figure 1 This is a flowchart of the ultrasonic non-destructive testing method for lithium plating in lithium-ion batteries in an embodiment of the present invention;
[0026] Figure 2 This is a schematic diagram of full-matrix capture of ultrasound detection data in an embodiment of the present invention. Detailed Implementation
[0027] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.
[0028] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.
[0029] Example 1
[0030] This embodiment provides an ultrasonic non-destructive testing method for lithium-ion battery lithium plating. It utilizes ultrasonic waves to perform mode switching and multiple scattering responses at the point where deposited metallic lithium reacts with the electrolyte to release gas. Combined with deep learning algorithms, this method achieves high-resolution non-destructive quantitative detection and evaluation of lithium plating defects in lithium-ion batteries. Specifically, during the lithium-ion battery charging process, a full matrix capture (FMC) data, i.e., ultrasonic detection data, is acquired using an ultrasonic phased array transducer. This data is input into an initial unsupervised fully convolutional network (FCN), which outputs an ultrasonic sound velocity prediction model. This model is then used as initial values in the acoustic wave equation to generate reconstructed sound velocity data. The error between the ultrasonic full matrix capture data (i.e., ultrasonic detection data) and the reconstructed sound velocity data is used to construct a loss function. This process is iterated until the loss function converges, completing the training of the fully convolutional network. During the unsupervised training of the fully convolutional network, the network ultimately outputs an accurate sound velocity prediction model, which represents the detected and characterized lithium plating defect. Through the above scheme, this embodiment can achieve high-resolution and high-precision detection of lithium-ion battery lithium plating defects.
[0031] The ultrasonic non-destructive testing method for lithium plating in lithium-ion batteries proposed in this embodiment is as follows: Figure 1 As shown, the specific steps include:
[0032] Full-matrix capture ultrasonic testing data of lithium-ion batteries was acquired using an ultrasonic phased array transducer.
[0033] The ultrasonic detection data is input into the initial fully convolutional network, and the output sound velocity prediction model is generated.
[0034] The sound velocity prediction model is used as the initial value input into the acoustic wave equation, and the reconstructed sound velocity data is output.
[0035] The loss function is constructed based on the error between the ultrasonic detection data and the reconstructed sound velocity data. The loss function is continuously iterated until it converges, thus completing the training of the unsupervised fully convolutional network.
[0036] The trained fully convolutional network ultimately outputs an accurate sound velocity prediction model, which is the detection result of lithium plating defects in lithium-ion batteries.
[0037] The fully convolutional network is trained using unsupervised learning. Unsupervised learning does not rely on a large training set; instead, it combines the physical model (i.e., the acoustic wave equation) with the full-matrix ultrasound data captured in a single acquisition to uncover intrinsic features. Compared to supervised learning, which requires a large training set, this embodiment improves detection efficiency through unsupervised learning. That is, unlike conventional supervised learning-based ultrasound detection methods, the method described in this embodiment does not require a large training set or a matching relationship between the real sound velocity model and the corresponding ultrasound detection data within the training set, thus improving detection efficiency. Figure 1 As shown, the training process of the unsupervised learning fully convolutional network based on the physical model specifically includes the following steps:
[0038] Step S1: With the lithium-ion battery charging, firstly, full matrix capture (FMC) data, i.e., ultrasonic testing data, is acquired using an ultrasonic phased array transducer. In fact, ultrasonic non-destructive testing can be performed on lithium-ion batteries regardless of their state. In this embodiment, ultrasonic non-destructive testing of a lithium-ion battery in a charging state is used as an example. Full matrix capture (FMC) is a specific data acquisition process using an ultrasonic phased array transducer. For an array probe with N crystals, each crystal is sequentially excited, and all crystals simultaneously receive signals. These signal data are organized into a matrix S containing all acquired signals, such as... Figure 2 As shown, S ij This represents the A-scan signal emitted by chip i and received by chip j.
[0039] Preferably, the acquired full-matrix capture ultrasound detection data is preprocessed, such as noise reduction to improve the signal-to-noise ratio.
[0040] Step S2: Construct an unsupervised learning fully convolutional network. Input the ultrasonic detection data into the initial fully convolutional network and output a sound velocity prediction model. Specifically, the fully convolutional network mainly includes convolutional layers, upsampling layers, and skip layers. Through the fully convolutional network, the mapping relationship between the ultrasonic detection data (input) and the sound velocity prediction model (output, which can characterize lithium plating defects) is realized. The mathematical expression of this process is as follows:
[0041]
[0042] In the formula, d represents the ultrasound detection data, and FCN represents a fully convolutional network. This represents the sound speed prediction model after passing through a fully convolutional network (FCN).
[0043] In fact, the sound velocity prediction model takes advantage of the fact that the sound velocity at the defect (i.e., lithium plating) is obviously different from the sound velocity of the lithium-ion battery itself, and uses the sound velocity map to quantitatively characterize the lithium plating defect.
[0044] Step S3: Input the sound velocity prediction model obtained in step S2 as the initial value into the acoustic wave equation, and output the reconstructed sound velocity data. Specifically, the data expression for this process is as follows:
[0045]
[0046] In the formula, f -1 This represents the forward modeling operator in the acoustic wave equation. This indicates the reconstruction of sound speed data.
[0047] The aforementioned acoustic wave equations are standard equations in this field, derived from Maxwell's equations, describing the characteristics of electromagnetic field waves. They are used to describe various wave phenomena in nature (such as sound waves, light waves, and water waves), including transverse and longitudinal waves. By reconstructing sound velocity data, a better initial value can be provided for the acoustic wave equations. In fact, this is an inversion problem; if the given initial values are not accurate enough, the inversion result of the acoustic wave equations cannot be obtained.
[0048] Step S4: Construct a loss function based on the error between the ultrasonic detection data and the reconstructed sound velocity data. Iterate continuously until the loss function converges, completing the training of the unsupervised fully convolutional network. Specifically, the loss function L is used as the difference between the ultrasonic detection data d and the reconstructed sound velocity data. Error criteria are used to achieve high-precision lithium plating defect detection. The loss function L is composed of pixel loss L... pw and perceived loss L pl It consists of two parts, namely:
[0049]
[0050] The above pixel loss L pw The definition is as follows:
[0051]
[0052] Here, λ1 and λ2 represent two hyperparameters used to control relative importance; L1 and L2 represent the L1 norm and L2 norm, respectively.
[0053] The above-mentioned perceived loss L p The definition is as follows:
[0054]
[0055] Where λ3 and λ4 represent two hyperparameters; This represents the network architecture parameters of a fully convolutional network (FCN). Compared to pixel loss, perceptual loss captures region structure better.
[0056] Finally, during the training of the unsupervised fully convolutional network FCN, the network outputs an accurate sound velocity prediction model that can characterize lithium plating defects, thus obtaining the detection results of lithium plating defects in lithium-ion batteries.
[0057] In this embodiment, the method described above considers the comprehensive response information such as ultrasonic mode conversion and multiple scattering at the lithium plating site, and combines it with deep learning algorithms. Based on the acoustic wave equation, it performs unsupervised learning training of a fully convolutional network. Through the trained fully convolutional network, an accurate sound velocity prediction model for lithium-ion batteries is obtained, achieving high-resolution and high-precision detection of lithium plating defects in lithium-ion batteries.
[0058] Example 2
[0059] This embodiment provides an ultrasonic non-destructive testing system for lithium-ion battery lithium plating, including:
[0060] The data acquisition module is used to acquire full-matrix capture ultrasonic detection data of lithium-ion batteries using an ultrasonic phased array transducer.
[0061] The unsupervised learning fully convolutional network training module is used to input ultrasonic detection data into the initial fully convolutional network and output a sound velocity prediction model; the sound velocity prediction model is used as the initial value to input the acoustic wave equation and output reconstructed sound velocity data; the error between the ultrasonic detection data and the reconstructed sound velocity data is used to construct a loss function, and the process is iterated continuously until the loss function converges, thus completing the training of the unsupervised learning fully convolutional network.
[0062] The lithium plating detection result output module is used to output an accurate sound velocity prediction model from the trained fully convolutional network. The sound velocity prediction model is the detection result of lithium plating defects in lithium-ion batteries.
[0063] Example 3
[0064] This embodiment provides an electronic device, including a memory and a processor, as well as computer instructions stored in the memory and running on the processor. When the processor executes the computer instructions, it completes the steps in the ultrasonic non-destructive testing method for lithium-ion battery lithium plating as described above.
[0065] Example 4
[0066] This embodiment also provides a computer-readable storage medium for storing computer instructions, which, when executed by a processor, complete the steps in the ultrasonic non-destructive testing method for lithium plating of lithium-ion batteries as described above.
[0067] The steps and methods involved in Embodiments 2 to 4 above correspond to those in Embodiment 1. For specific implementation details, please refer to the relevant description section of Embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood as including any medium capable of storing, encoding, or carrying an instruction set for execution by a processor and enabling the processor to perform any of the methods in this invention.
[0068] Those skilled in the art will understand that the modules or steps of the present invention described above can be implemented using general-purpose computer devices. Optionally, they can be implemented using computer-executable program code, thereby allowing them to be stored in a storage device for execution by a computer device, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. The present invention is not limited to any particular combination of hardware and software.
[0069] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
[0070] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.
Claims
1. An ultrasonic non-destructive testing method for lithium plating in lithium-ion batteries, characterized in that, include: Full-matrix capture ultrasonic testing data of lithium-ion batteries was acquired using an ultrasonic phased array transducer. Ultrasonic detection data is input into an initial fully convolutional network, which outputs a sound velocity prediction model. This fully convolutional network represents the mapping relationship between the ultrasonic detection data and the sound velocity prediction model, as shown in the formula: in, This represents ultrasound detection data; FCN represents a fully convolutional network. This represents the sound speed prediction model output after passing through a fully convolutional network (FCN). The sound speed prediction model is used as the initial value input to the acoustic wave equation, and the reconstructed sound speed data is output. The acoustic wave equation is a set of differential equations derived from Maxwell's equations that describe the characteristics of electromagnetic field waves and are used to describe various wave phenomena in nature. A loss function is constructed based on the error between ultrasonic detection data and reconstructed sound velocity data. The loss function includes pixel loss and perception loss. The loss function is continuously iterated until it converges, thus completing the training of the unsupervised fully convolutional network. The fully convolutional network is trained using unsupervised learning, which combines the acoustic wave equation with the full matrix capture data of a single ultrasonic acquisition to achieve the mining of intrinsic features. The perceived loss for: The pixel loss for: in, , and This represents four hyperparameters. This represents the network architecture parameters of a fully convolutional network. and They represent Norm and Norm, This represents ultrasound test data. This indicates the reconstruction of sound speed data; The trained fully convolutional network ultimately outputs an accurate sound velocity prediction model, which is the detection result of lithium plating defects in lithium-ion batteries.
2. The ultrasonic non-destructive testing method for lithium plating in lithium-ion batteries as described in claim 1, characterized in that, The acquired full-matrix capture ultrasound detection data is preprocessed, including noise reduction.
3. An ultrasonic non-destructive testing system for lithium-ion battery lithium plating, characterized in that, include: The data acquisition module is used to acquire full-matrix capture ultrasonic detection data of lithium-ion batteries using an ultrasonic phased array transducer. The unsupervised learning fully convolutional network training module is used to input ultrasound detection data into the initial fully convolutional network and output a sound velocity prediction model. The fully convolutional network represents the mapping relationship between the ultrasound detection data and the sound velocity prediction model, as shown in the formula: in, This represents ultrasound detection data; FCN represents a fully convolutional network. This represents the sound speed prediction model output after passing through a fully convolutional network (FCN). The sound speed prediction model is used as the initial value input to the acoustic wave equation, and the reconstructed sound speed data is output. The acoustic wave equation is a set of differential equations derived from Maxwell's equations that describe the characteristics of electromagnetic field waves and are used to describe various wave phenomena in nature. A loss function is constructed based on the error between ultrasonic detection data and reconstructed sound velocity data. The loss function includes pixel loss and perception loss. The process is iterated continuously until the loss function converges, thus completing the training of the unsupervised fully convolutional network. The fully convolutional network is trained using unsupervised learning, combining the acoustic wave equation with the full matrix capture data of a single ultrasonic acquisition to achieve the mining of intrinsic features. The perceived loss for: The pixel loss for: in, , and This represents four hyperparameters. This represents the network architecture parameters of a fully convolutional network. and They represent Norm and Norm, This represents ultrasound test data. This indicates the reconstruction of sound speed data; The lithium plating detection result output module is used to output an accurate sound velocity prediction model from the trained fully convolutional network. The sound velocity prediction model is the detection result of lithium plating defects in lithium-ion batteries.
4. An electronic device, characterized in that, It includes a memory and a processor, as well as computer instructions stored in the memory and running on the processor, which, when executed by the processor, complete the steps of an ultrasonic non-destructive testing method for lithium plating of a lithium-ion battery as described in any one of claims 1-2.
5. A computer-readable storage medium, characterized in that, Used to store computer instructions, which, when executed by a processor, complete the steps of an ultrasonic non-destructive testing method for lithium plating in lithium-ion batteries as described in any one of claims 1-2.