Method and device for detecting pseudo soldering of cylindrical lithium ion battery

Abstract

The invention relates to the technical field of cylindrical lithium ion battery production, and discloses a method for detecting insufficient welding of a cylindrical lithium ion battery, and the method comprises the following steps: S1, appearance inspection; s2, initial internal resistance benchmark test; s3, applying a short-time large-current pulse to the battery by using a pulse generator, simulating an actual load condition, and recording an instantaneous voltage drop and current response curve of the voltage of the battery end during the pulse period; s4, performing infrared thermal imaging identification and data processing; s5, learning the picture and establishing a model; s6, comparing the test picture and the data with the model; s7, dynamic internal resistance calculation; s8, cold solder joint query and sorting processing; and S9, retesting the electrical performance. According to the invention, the technologies of dynamic internal resistance measurement, nondestructive X-ray or ultrasonic detection, rate discharge test and the like are adopted, so that efficient, accurate and nondestructive detection of the cold solder joint of the cylindrical lithium ion battery is realized, and the production efficiency and the detection precision are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of cylindrical lithium-ion battery production, in particular to a method and device for detecting weld defects in cylindrical lithium-ion batteries. Background Art

[0002] In modern society, cylindrical lithium-ion batteries are widely used in various portable electronic devices, electric vehicles, and energy storage systems. As a key process in battery production, welding quality directly impacts battery performance and safety. Cold solder joints are a common defect in the welding process. If not detected and addressed promptly, they can increase the battery's internal resistance, reduce charge and discharge performance, and even cause safety accidents. Therefore, effectively detecting and identifying cold solder joints during production to ensure the high quality of each battery has become a critical task in battery production technology.

[0003] Currently, common technical solutions for detecting weld defects in cylindrical lithium-ion batteries include manual visual inspection, X-ray inspection, OCV testing, and random destructive testing. Manual visual inspection is simple to perform. X-ray inspection can visually detect internal defects in the battery with a certain degree of accuracy. OCV testing indirectly evaluates battery performance by measuring the battery's open-circuit voltage.

[0004] However, the existing technology still has some shortcomings. First, manual visual inspection relies on the experience and visual observation of the inspector, which is highly subjective, has a high rate of missed detection, and has difficulty detecting subtle internal cold solder joint defects. X-ray inspection equipment is expensive and slow to detect, making it unsuitable for rapid inspection on large-scale production lines. Although sampling destructive testing can accurately determine weld quality, it is a destructive test and cannot fully inspect every battery, which cannot meet the requirement of 100% inspection of product quality during the production process. Therefore, the present invention proposes a method and device for detecting cold solder joints in cylindrical lithium-ion batteries, which can quickly and accurately identify cold solder joint problems, greatly improving battery production quality and production efficiency. Summary of the Invention

[0005] In response to the shortcomings of the existing technology, the present invention provides a method and device for detecting weld defects in cylindrical lithium-ion batteries, which solves the problems of low detection efficiency, poor accuracy and inability to achieve non-destructive and rapid detection of the entire battery in the existing technology.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: A method for detecting weld defects in cylindrical lithium-ion batteries, comprising the following steps:

[0007] S1. Appearance inspection: Check the appearance of the battery and confirm that there is no oxidation, cracks or obvious deformation at the welding parts;

[0008] S2. Initial internal resistance benchmark test: Apply a low-amplitude constant test current to the cylindrical lithium-ion battery that has passed the appearance inspection, obtain the voltage change value before and after the load, and calculate its initial internal resistance R0R_0R0 to determine whether there is any abnormality in the electrical connection performance of the welding area.

[0009] The initial internal resistance R0 is calculated by the following formula:

[0010]

[0011] Where: R0 represents the initial internal resistance of the battery, in milliohms (mΩ), which is the equivalent ohmic impedance of the conduction path at the solder joint; ν t It is the voltage value collected instantaneously after applying the test current I, in volts (V), reflecting the terminal voltage after the dynamic response of the battery cell; ν0 is the open circuit voltage in the static state before the current is applied, in volts (V), indicating the natural voltage of the battery cell to be tested; I is the applied test current, in amperes (A), which is a constant low-amplitude pulse current, generally 50mA to 100mA, determined by the battery cell capacity specifications. By controlling the current amplitude to operate within the range where the battery cell is not polarized and does not heat up, the test disturbance is effectively avoided; further comprising temperature correction of the initial internal resistance value to eliminate the influence of the ambient temperature on the internal resistance measurement result, the temperature correction is performed according to the following formula:

[0012]

[0013] Where: R corr is the internal resistance value after temperature correction, in mΩ, indicating the theoretical equivalent value at standard temperature; α is the temperature coefficient, in 1 / °C1, with a value range of 0.003–0.006, which is determined by the conductor material; T is the actual test environment temperature, in degrees Celsius (°C), measured by the environmental sensor; T ref The reference standard temperature is set to 25°C. When the initial internal resistance exceeds the preset qualified range, or the measured value fluctuation exceeds the set threshold, the battery cell is judged to have a risk of poor welding contact and will be rejected or transferred to the subsequent process;

[0014] S3. Pulse current application: Use a pulse generator to apply short, high-current pulses to the battery to simulate actual load conditions, and record the instantaneous voltage drop and current response curve of the battery terminal during the pulse.

[0015] S4. Infrared thermal imaging recognition and data processing: Use an infrared thermal imager to collect infrared thermal images of the welding area in real time to ensure that the dynamic process of temperature changes can be captured. The received infrared thermal images are pre-processed, including image noise reduction, grayscale correction, edge detection, and region segmentation algorithms to extract the temperature distribution area of the welding area;

[0016] S5. Learn images and build models: Using a large amount of experimental data, a support vector machine algorithm is used to build a cold solder joint detection model, which is trained and optimized based on the temperature characteristic data of the welding parts.

[0017] S6. Comparison of test images and data with the model: Input the real-time acquired infrared thermal images and test data into the established cold solder joint detection model for comparison and analysis to determine whether there is a cold solder joint at the welding position;

[0018] S7. Dynamic internal resistance calculation: Dynamically calculate the dynamic internal resistance value under the pulse current application state, and judge whether there is a risk of cold soldering in the battery based on the calculation results;

[0019] S8. Detection and sorting of cold solder joints: If the dynamic internal resistance exceeds the baseline value by more than 20%, the battery is determined to have a risk of cold solder joints. X-ray or ultrasonic testing is then used to confirm the location of the cold solder joints. Automatic sorting is then performed based on the cold solder joint detection results. Batteries with a risk of cold solder joints are separated and sent for destructive testing.

[0020] S9. Retest of electrical performance: Retest the electrical performance of batteries that are determined to have poor solder joints, including charge and discharge cycle tests under different loads to check the actual power output and life performance of the battery.

[0021] The present invention also provides a device for detecting cold welds in cylindrical lithium-ion batteries, comprising:

[0022] Pulse current generation module: used to apply pulse current and record the instantaneous voltage drop and current response at the battery terminal. This module includes a pulse signal generator, a power amplifier, and a current regulation circuit. The pulse signal generator is responsible for generating a pulse control signal of set frequency and width as the system's driving source; the power amplifier receives and amplifies the pulse signal to generate a large current pulse sufficient to drive the battery; the current regulation circuit monitors the output current in real time and adjusts the power amplifier's drive signal through closed-loop feedback to ensure the output current amplitude;

[0023] Infrared thermal imaging acquisition module: used to collect infrared thermal images of the welding part in real time. This module includes a high-resolution infrared thermal imager that can capture the temperature changes of the welding part and transmit the image data to the subsequent processing module for analysis;

[0024] Image processing and analysis module: used to pre-process infrared thermal images and extract the temperature characteristics of welding parts. This module includes image noise reduction, grayscale correction, edge detection, and region segmentation algorithms. It can extract temperature change information of welding areas from images for subsequent false solder joint determination.

[0025] Battery positioning and transmission module: used to automatically transmit and position batteries to ensure that the welding part is within the detection area. The module consists of a mechanical bracket, a positioning fixture, a transmission guide rail and a drive motor. The positioning fixture is designed according to the outer diameter of the cylindrical lithium-ion battery and adopts an elastic clamp structure. It can not only ensure the accurate positioning of the battery during the detection process, but also adapt to changes in battery specifications within a certain size range. The transmission guide rail adopts a high-precision linear guide rail to ensure the stability and straightness of the battery during the transmission process. The drive motor is connected to the screw of the transmission guide rail through a coupling to realize the control of the battery transmission speed and position.

[0026] The present invention provides a method and device for detecting cold welds in cylindrical lithium-ion batteries. The method has the following beneficial effects:

[0027] 1. This invention utilizes a highly efficient, accurate, and nondestructive method for detecting cold welds in cylindrical lithium-ion batteries. By combining dynamic internal resistance testing with nondestructive testing techniques, it achieves precise identification of battery cold weld defects. Compared to existing manual visual inspection and X-ray inspection methods, this invention significantly improves detection efficiency, avoids the problems of human subjectivity and high missed detection rates, and enables rapid and efficient automated inspection in large-scale production, significantly improving production line efficiency.

[0028] 2. This invention utilizes a combination of constant current charge and discharge testing and rate discharge testing to effectively assess battery power output and internal resistance changes. Compared to traditional OCV testing, the electrical performance retesting method of this invention enables comprehensive electrical performance evaluation of each battery, resolving the existing OCV testing issues of misjudgment and poor contact, ensuring the accuracy and stability of test results, avoiding missed detection and misjudgment of battery performance, and improving product quality traceability.

[0029] 3. After confirming a cold solder joint, the present invention performs non-destructive X-ray imaging or ultrasonic testing on the battery, accurately locating the cold solder joint defect. Compared to existing destructive sampling testing, this present invention achieves non-destructive testing of entire battery batches, avoiding the limitations of destructive testing. It can perform 100% quality inspection on every battery during the production process, effectively preventing defective batteries from entering the market and improving product safety and reliability.

[0030] 4. This invention provides an integrated detection system that combines multiple technical approaches, including dynamic internal resistance measurement, nondestructive imaging, rate discharge, and cycle life testing. Compared to traditional single-method detection methods, this invention achieves multi-angle and all-round detection of cold solder joint defects, with higher accuracy and applicability. The system enables real-time monitoring and rapid response on the production line, enhancing the automation level of battery production, reducing manual intervention, improving production efficiency, and lowering overall detection costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 is a flow chart of the method of the present invention;

[0032] Figure 2 Schematic diagram of the present invention. DETAILED DESCRIPTION

[0033] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the specification of the present invention. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0034] Please see the attached Figure 1 The embodiment of the present invention provides a method for detecting a cold weld in a cylindrical lithium-ion battery, comprising the following steps:

[0035] S1. Appearance inspection: Check the appearance of the battery and confirm that there is no oxidation, cracks or obvious deformation at the welding parts;

[0036] In step S1, a high-precision magnifying glass is used to locally magnify and observe each battery cell under test, focusing on the weld area between the positive and negative terminals and the busbar. The observation process uses manual visual inspection or an image processing system to perform a standardized assessment of the battery weld morphology.

[0037] Specifically, the judgment criteria include whether the solder joints are complete and full, whether the edges are smooth and closed, whether there are obvious signs of oxidation, whether the surface is burned and blackened, whether there is insufficient solder or flash, etc. Cells with any of the following defects will be marked as "appearance abnormal" and exit the main process channel:

[0038] In one possible implementation, the solder joint area is further subjected to contour recognition and geometric quantification to determine whether it meets the morphological stability standard. The present invention uses the following formula to calculate the solder joint roundness factor:

[0039]

[0040] Where: C is the roundness of the solder joint, which is used to indicate the similarity between the solder joint and the ideal circle; A is the projected area of the solder joint area in the image (unit: mm 2 ), which is converted from the pixel area surrounded by the closed boundary contour in the binary image; P is the outer boundary perimeter of the solder joint (unit: mm), which is calculated by the image contour detection algorithm; π is the pi.

[0041] The threshold is usually set to C ≥ 0.80. If it is lower than this threshold, it is determined that the solder joint boundary distortion is obvious and does not meet the process forming standard.

[0042] In addition to the morphological evaluation, in order to identify whether there is surface oxidation on the solder joint, this embodiment further introduces the grayscale mean index as an evaluation basis. The specific calculation formula is as follows:

[0043]

[0044] Among them: G avg is the average grayscale value of the welding area, in grayscale levels (0–255), which is used to measure the brightness of the area; N is the number of valid pixels in the welding area, in dimensionless integers; g i is the grayscale value of the i-th pixel (unit: grayscale level), ranging from 0 (black) to 255 (white).

[0045] Oxidized or ablated areas usually show low grayscale features. avg If the grayscale standard deviation is less than 85G and the regional grayscale standard deviation is greater than the set threshold (such as 20), it can be preliminarily determined that the solder joint has ablation, contamination or oxidation reaction.

[0046] To ensure consistent assessment and system compatibility, some embodiments utilize a standard-sized position-limiting fixture. This ensures that cells of varying diameters maintain consistent solder joint orientation and consistent image acquisition angles during inspection. The fixture utilizes a replaceable bushing and flexible positioning blocks, ensuring both versatility and structural stability.

[0047] To avoid the interference of experience differences in manual inspection on system accuracy, this embodiment also allows the appearance image to be uploaded to the subsequent module for comprehensive judgment. If the edge of the solder joint is abnormal but the initial internal resistance is normal, the sample is allowed to be temporarily retained and enter the dynamic fluctuation recognition module for further comparison and analysis.

[0048] In addition, in an actual engineering deployment environment, the system can also compare the weld spot identification features with the historical welding machine process parameters to build a welding process drift model and achieve reverse tracking of batch process consistency.

[0049] In summary, step S1, under the premise of ensuring the basic stability of the battery welding connection, establishes a structural geometry judgment mechanism, a surface state recognition mechanism and dynamic traceability through multi-dimensional feature recognition means, providing a stable physical input basis for subsequent modules.

[0050] S2. Initial internal resistance benchmark test: Use the constant current pulse method to measure the initial internal resistance of the battery, record it as the benchmark internal resistance value, and compare and analyze the internal resistance distribution of batteries in the same batch to screen out abnormal batteries that deviate from the mean;

[0051] In the proposed method for detecting cold solder joints in cylindrical lithium-ion batteries, step S2 follows step S1. S2 is performed based on further electrical performance screening, assuming the battery cell's external physical condition is satisfactory. Its purpose is to quantitatively assess the resistance level of the weld area's conductivity, eliminating samples with obvious poor contact, abnormally high internal resistance, micro-breaks, or atypical cold solder joints, ensuring stable input conditions for the subsequent image recognition module.

[0052] Generally, even if a battery cell's solder joints appear intact due to microcracks, oxide films, or solder voids, they may exhibit abnormally high resistance or unstable measurements under static current flow. Accurately obtaining the initial value of internal resistance in this step allows for early identification of these "invisible cold solder joints," establishing a standardized electrical parameter benchmark for subsequent dynamic feature modeling and infrared response analysis.

[0053] In this embodiment, step S2 uses a four-wire Kelvin connection method to measure resistance, applies a small current through constant current excitation, collects the voltage response in real time, and then calculates the initial internal resistance R0 of the battery cell based on the voltage-current ratio.

[0054] Specifically, the internal resistance calculation is based on the following formula:

[0055]

[0056] Where: R0 represents the initial internal resistance of the battery, in milliohms (mΩ), which is the equivalent ohmic impedance of the conduction path at the solder joint; ν t It is the voltage value collected instantaneously after the test current I is applied, in volts (V), reflecting the terminal voltage after the dynamic response of the battery cell; ν0 is the open circuit voltage (OCV) in the static state before the current is loaded, in volts (V), indicating the natural voltage of the battery cell to be tested; I is the applied test current, in amperes (A), which is a constant low-amplitude pulse current, generally 50mA to 100mA, depending on the battery cell capacity specifications.

[0057] This method effectively avoids test disturbances and restores the actual conductivity of the solder joint to the maximum extent by controlling the current amplitude to operate within the range where the battery cell is not polarized and does not generate heat.

[0058] As an option, in order to reduce the interference of contact resistance on the results, the test fixture in this embodiment adopts a four-terminal probe structure, and the current and voltage measurement circuits are independent, which can eliminate the interference impedance of the fixture wires and the external connection ends of the welding piece, so that the measured R0 result is closer to the actual intrinsic conduction state of the solder joint.

[0059] In some embodiments, the system repeatedly samples the test process and sets a difference tolerance to determine whether it is stable. If the maximum difference between three consecutive internal resistance samples is greater than 1.5mΩ, or the variance exceeds a set threshold, the system marks the sample as "abnormal electrical parameter fluctuation" and enters the subsequent feature evolution channel for further observation.

[0060] To further enhance the comparability of different batches of cells under different environments, a temperature correction mechanism is introduced in some systems, and normalization is performed using the following model:

[0061]

[0062] Where: R corr is the internal resistance value after temperature correction, in mΩ, indicating the theoretical equivalent value at standard temperature; α is the temperature coefficient, in 1 / °C1, with a value range of 0.003–0.006, which is determined by the conductor material; T is the actual test environment temperature, in degrees Celsius (°C), measured by the environmental sensor; T ref The reference standard temperature is set to 25°C.

[0063] This correction method ensures a unified judgment standard when conducting cross-batch comparative analysis in factories in the north and south, in different seasons, or under abnormal air-conditioning environments.

[0064] In actual deployments, the system automatically records initial internal resistance data and associates it with the cell's unique identification code, facilitating subsequent traceability, modeling, and machine learning algorithms. Data fields include timestamp, test channel number, voltage response, test current, cell open-circuit voltage, and temperature.

[0065] In some automated production lines, this step uses interconnected communication logic with the upstream appearance screening module. When S1 outputs a "qualified" signal, the robotic arm transfers the cell to S2. The fixture then automatically closes, initiating the measurement process. The data is then written to the control system, and the industrial control platform controls the sample flow.

[0066] Through the above method, step S2 not only provides a quantitative welding conductivity evaluation standard, but also forms a composite prediction system of structure + electrical properties in two dimensions in terms of technical logic with S1, providing a highly reliable initial screening basis for the construction of multi-modal cold solder joint identification in the present invention.

[0067] S3. Pulse current application: Use a pulse generator to apply short, high-current pulses to the battery to simulate actual load conditions, and record the instantaneous voltage drop and current response curve of the battery terminal during the pulse.

[0068] Step S3 simulates the battery's operating conditions under actual high-power discharge conditions to induce transient conduction anomalies that may occur in solder joints under high current flow. This step follows S2 (initial internal resistance test), and its output serves as an important basis for identifying cold solder joint defects in subsequent identification steps.

[0069] Generally, battery solder joints with issues such as insufficient contact area, solder voids, metal oxidation, interface contamination, or cracks can be difficult to detect during low-current static testing. However, once a pulse current is applied, the instantaneous voltage drop will be significantly distorted due to the nonlinear change in local resistance in the current path. Therefore, utilizing a simulated load pulse stress triggering mechanism can effectively improve the sensitivity of cold solder joint detection.

[0070] The system uses a high-frequency programmable pulse source to apply one or more current pulses to the cylindrical lithium-ion battery under test. Each pulse cycle consists of a "power-on phase" and a "rest phase." Throughout this cycle, the system uses a high-speed ADC sampling module to synchronously collect the battery terminal voltage and current curves.

[0071] To ensure comparability of measurement data across different samples, all pulse signal parameters should be uniformly set, primarily including the following three items:

[0072] Pulse peak current I p : The unit is ampere (A), which is the maximum constant current applied during the pulse power-on phase. It is generally set to a current value of 2 to 5 times the nominal capacity of the battery cell. For example, for a battery with a capacity of 2000mAh, I p The value range is 4A to 10A.

[0073] Pulse duration t p : Measured in milliseconds (ms), this value indicates the duration of the pulse current. Common values are 100ms, 200ms, or 300ms to ensure sufficient response time for capture.

[0074] Interval time t off : The unit is milliseconds (ms), which represents the rest time between consecutive pulses to avoid heat accumulation, generally 2–5 times the pulse length.

[0075] The dynamic internal resistance calculation model is as follows:

[0076] During the pulse application process, the system calculates the dynamic internal resistance R of the solder joint conduction area based on the collected voltage-current curve. d , which is defined as follows:

[0077]

[0078] Where: R dThe dynamic internal resistance is expressed in milliohms (mΩ), which reflects the equivalent impedance of the solder joint area under high current pulse load. This value is more likely to reveal nonlinear contact problems than the static internal resistance; ΔV p The instantaneous voltage drop triggered by the pulse is expressed in volts (V). It is defined as the voltage difference between the first stable time point after the pulse is applied and the pulse starting point, that is, ΔV p =V start -V stable ; Where: V start The terminal voltage at the moment the pulse is applied; V stable is the average voltage when the pulse is stable and the sampling time interval between the two does not exceed 5ms; I p The amplitude of the applied pulse current, in amperes (A), is the system setting value and is kept constant to stimulate short-term load response.

[0079] Derivative pressure drop rate of change indicator:

[0080] In order to further enhance the ability to identify abnormal voltage drops, this step introduces the voltage change rate as an auxiliary quantity:

[0081]

[0082] in: is the voltage change rate, in V / ms, which is used to measure the changing trend of the terminal voltage during the pulse application process; V(t) is the instantaneous voltage per unit time t, in volts (V); Δt is the sampling time interval, in milliseconds (ms), which is set to between 0.1ms and 1ms in this embodiment.

[0083] This derivative indicator can be approximated using a moving window difference algorithm. If a significant slope jump is observed, the system automatically marks it as an "abnormal trigger sample."

[0084] In some embodiments, to prevent current pulses from causing overheating or damage to the sample, the following protection mechanisms are implemented within the system:

[0085] Current limiting control module: dynamic monitoring I p If the current value deviates from the set range by more than 5%, the current will be automatically cut off.

[0086] Thermal protection module: The temperature sensor monitors the temperature rise in the test area in real time. If the temperature rises by more than 10°C or exceeds the upper limit of 45°C, the system suspends the test.

[0087] As an option, a graded pulse test strategy can be set according to the cell specifications, that is, first use a low amplitude I p1 Perform pre-check and then use high amplitude I p2 Repeat the test and calculate R d1 With Rd2 , establish a nonlinear conduction change curve. If the ratio of the two R d2 / R d1 If the value exceeds the set threshold (such as 1.3), it further indicates the risk of solder joint reliability.

[0088] In summary, step S3 effectively induces dynamic nonlinear behavior of the welding area on a macroscopic scale through a high-bandwidth, short-duration excitation current pulse method.

[0089] S4. Infrared thermal imaging recognition and data processing: Use an infrared thermal imager to collect infrared thermal images of the welding area in real time to ensure that the dynamic process of temperature changes can be captured. The received infrared thermal images are pre-processed, including image noise reduction, grayscale correction, edge detection, and region segmentation algorithms to extract the temperature distribution area of the welding area;

[0090] In the present method for detecting cold welds in cylindrical lithium-ion batteries, step S4, "Infrared Thermal Imaging Recognition and Data Processing," undertakes the critical task of extracting solder joint thermal response characteristics from thermal imaging information and converting them into numerical features useful for defect identification. This step, implemented immediately after S3, uses infrared imaging to capture high-resolution temperature dynamics in the weld area and identifies abnormal solder joint thermal response behavior through image processing and feature extraction.

[0091] Generally speaking, a well-bonded solder joint exhibits good thermal conductivity under heating stimulation, resulting in a rapid and uniform temperature rise. However, areas with solder defects experience a slow or abnormal temperature rise due to a restricted heat conduction path, resulting in a spotty and distorted temperature distribution. Therefore, identifying temperature response patterns in thermal imaging images can effectively assist in identifying cold solder joints.

[0092] In one possible implementation, the infrared thermal imager acquisition frequency is set to 30–60 frames per second, the resolution is set to 320×240 pixels, the thermal sensitivity is no higher than 0.05°C, and the thermal response range covers 0°C to 120°C, ensuring that the true temperature response of the welding part is captured.

[0093] After image acquisition is completed, the preprocessing process begins. Preprocessing includes:

[0094] Image noise reduction: Use median filtering, bilateral filtering, or Gaussian filtering algorithms to remove artifacts caused by sensor noise and background temperature differences;

[0095] Grayscale correction: The pixel values of all frame images are linearly stretched or normalized according to their grayscale range so that they fall between 0 and 255;

[0096] Edge detection: Use the Canny edge detection operator to extract the boundary of the welding target and identify the target area in the image that needs further analysis;

[0097] Region segmentation: A segmentation algorithm based on region growing or K-means clustering is used to extract the pixel set of the solder joint area, which is recorded as

[0098] The solder joint thermal response feature extraction uses the following formula:

[0099]

[0100] Where: S T T is the temperature rise rate of the welding area, in degrees Celsius per second (℃ / s), which indicates the average temperature rise rate from the initial value to the peak value and is a direct reflection of the thermal diffusion capacity; max is the highest temperature of the pixels in the welding area in the entire infrared image sequence, in degrees Celsius (℃); T0 is the initial temperature value of the area, in degrees Celsius (℃), which is the average temperature of the solder joint area in the image before infrared excitation; t0 is the starting time of heating excitation, in seconds (s), which is the timestamp corresponding to the starting frame of infrared recording; t max The time when the temperature reaches the peak value, in seconds (s), is the time when the temperature reaches the peak value in the recording sequence. max The timestamp corresponding to the frame is expressed by the following formula:

[0101]

[0102] Where: T max is the highest temperature of the pixel in the welding area in the entire infrared image sequence; T(x,y,t) represents the temperature value corresponding to the pixel position (x,y) at time t in the image sequence, in degrees Celsius (℃); (x,y) is the image plane coordinate, in pixels; is a pixel set in the solder joint area, which is dynamically identified by the segmentation algorithm; [t0, t1] is the infrared acquisition timing interval, t0 is the start frame time, and t1 is the heating or acquisition end frame time.

[0103] As an option, in some embodiments, a heat diffusion symmetry index γ is also introduced to reflect the spatial stability of the temperature field, and its formula is:

[0104]

[0105] Where: γ is the hot spot concentration index, unitless, ranging from [0,1]; The area of high-temperature pixels whose temperature exceeds the set threshold within the 5×5 pixel range of the geometric center of the welding area; It is the total high temperature area in the entire solder joint area where the temperature exceeds the same threshold.

[0106] Generally speaking, the heat diffusion of a good solder joint is concentrated, which is expressed as γ→1, while the hot spots in the area with a poor solder joint are discrete, γ→0.

[0107] In some embodiments, all image frames, feature values, region outline information, and timestamp data are recorded in a structured format and uploaded to the control platform after being bound to the unique ID of the battery cell. This data can be used to:

[0108] Subsequent classification model discrimination;

[0109] Process retrospective analysis;

[0110] Batch statistics and quality tracking.

[0111] In summary, step S4 extracts the dynamic thermal response characteristics of the welding part by combining thermal image acquisition with image processing, providing a stable and reliable physical indicator system for cold weld detection.

[0112] S5. Learn images and build models: Using a large amount of experimental data, a support vector machine algorithm is used to build a cold solder joint detection model, which is trained and optimized based on the temperature characteristic data of the welding parts.

[0113] Step S5 is the model learning and construction stage, which inherits the temperature characteristic data output by step S4 and is the core link for realizing intelligent defect identification in the entire detection process.

[0114] Generally speaking, the differences in the appearance of cold solder defects in infrared thermal imaging images may be very subtle, making it difficult to distinguish them directly through manual rule definitions. Therefore, using machine learning methods such as support vector machines (SVM) to establish a discriminant model can more effectively identify subtle differences.

[0115] First, construct input samples from the multidimensional temperature feature vector extracted in step S4. Each sample can be defined as:

[0116]

[0117] in: is the temperature rise rate of sample i, in °C / s; is the peak temperature of sample i, in °C; is the initial temperature of sample i, in °C; γ (i) is the hot spot concentration index of sample i, unitless; is the standard deviation of the temperature distribution of sample i, in °C; is the temperature gradient intensity of sample i, in °C / pixel, indicating the thermal diffusion amplitude at the weld zone boundary; Δt (i) is the temperature rise time span of sample i, in seconds (s).

[0118] Each sample x i Corresponding to a label y i∈{-1,+1}, where +1 indicates "cold solder joint" and -1 indicates "normal solder joint".

[0119] The soft margin support vector machine is used for training, and the optimization problem is as follows:

[0120]

[0121] subject to y i (w T φ(x i )+b)≥1-ξ i ,ξ i ≥0;

[0122] Where: w is the model weight vector, corresponding to the feature dimension; b is the bias term, used to control the position of the classification hyperplane; φ(·) is the kernel function mapping function, used to map the input features to a high-dimensional space; ξ i is a slack variable, indicating the degree of violation of the interval by sample i; C is a regularization parameter (penalty factor), which controls the balance between model complexity and error tolerance; y i is the true label of the sample; N is the total number of training samples.

[0123] In one possible implementation, the kernel function uses a radial basis kernel function (RBF), which has the form:

[0124]

[0125] Where: x i ,x j represents the feature vectors of any two samples; σ is the kernel width parameter, which controls the decay rate of the similarity between samples; K(·,·) is the output value of the kernel function, which represents the nonlinear similarity between samples.

[0126] Through this kernel function, low-dimensional features are mapped to a higher-dimensional space so that nonlinear boundaries can be captured by the linear classifier.

[0127] After training is completed, the prediction function of the support vector machine classification model is defined as follows:

[0128]

[0129] Where: f(x) is the classification result (+1 or -1) of the input feature x; sign(·) is the sign function used to determine the classification label based on the positive or negative decision value; α i is the Lagrange multiplier of the training sample, and only the sample weight corresponding to the support vector is non-zero; K(x i,x) is the kernel function output, which measures the similarity between the support vector and the current input sample; b is the model bias term; N1 is the total number of support vectors (SupportVectors) determined in the training phase; x i is the feature vector of the i-th support vector, which comes from the samples selected as support vectors in the training set, and its feature form is consistent with the input sample x; x is the feature vector of the input sample that needs to be classified and judged, which is usually extracted through the previous steps.

[0130] When the model output is "+1", the system determines that the solder joint is a cold solder joint sample, otherwise it is judged to be normal.

[0131] In some embodiments, the trained model is saved as a structured file in a persistent manner and embedded in the detection system. During detection, the system automatically calls the model and passes in the feature vector generated by S4 to achieve real-time reasoning and classification.

[0132] As an option, the model evaluation phase uses statistical indicators such as accuracy, recall, and F1-score, combined with a cross-validation strategy to optimize the model's generalization ability and improve stability under different battery cell batches.

[0133] In summary, step S5 constructs a supervised classification model with support vector machine as the core, and uses multi-dimensional temperature feature data to identify and judge welding defects.

[0134] S6. Comparison of test images and data with the model: Input the real-time acquired infrared thermal images and test data into the established cold solder joint detection model for comparison and analysis to determine whether there is a cold solder joint at the welding position;

[0135] In the proposed method for detecting cold solder joints in cylindrical lithium-ion batteries, step S6 (comparing test images and data with the model) constitutes the final step in the overall detection process. This step, following the cold solder joint identification model established in step S5, intelligently determines the sample's solder joint quality by analyzing the input of real-time captured images and processed temperature signatures. Its goal is to map and compare these temperature signatures with the learned feature boundaries in the model, quickly determining whether the current battery cell has a cold solder joint.

[0136] Generally, solder joint thermal response exhibits certain process fluctuations, but these fluctuations are effectively standardized through temperature normalization, feature extraction, and model training. Therefore, when a test sample is input into the trained support vector machine model, its thermal behavior can be compared with the classification boundaries using vector mapping to efficiently determine whether the solder joint is defective.

[0137] First, the temperature characteristic value of the solder joint area is extracted from the infrared thermal image sequence collected in the previous step S4, and the test sample feature vector is constructed, which is defined as:

[0138]

[0139] Where: S T The temperature rise rate is expressed in degrees Celsius per second (℃ / s), which reflects how fast the solder point temperature rises. is the maximum temperature of the welding area, in degrees Celsius (℃); T0 is the initial temperature before the start of excitation, in degrees Celsius (℃); γ1 is the hot spot concentration index, which indicates the concentration of high-temperature pixels, and has no unit; σ T is the standard deviation of temperature distribution in degrees Celsius (℃), which is used to reflect the uniformity of heat diffusion; is the temperature gradient intensity, in degrees Celsius (℃), which indicates the degree of temperature mutation at the edge of the solder joint; Δt is the time it takes for the temperature to rise from the initial temperature to the peak value, in seconds (s).

[0140] In one possible implementation, the test feature vector is normalized before input, for example:

[0141]

[0142] Where: x k is the original kth eigenvalue; μ k is the mean of the kth feature in the training set; σ k is the standard deviation of the kth feature in the training set; x′ k is the normalized eigenvalue, used as model input.

[0143] The above processing helps to improve the consistency of the input sample distribution and the training model distribution, and reduce the classification deviation caused by scale differences.

[0144] Next, call the SVM classification model trained in step S5 to perform discrimination

[0145] In some implementations, a probability output mechanism can be introduced to quantify the prediction credibility. For example, the Platt scaling method is used to construct an S-type function to map the original SVM output value to a probability value in the interval [0, 1]:

[0146]

[0147] Where: A, B are the fitting parameters obtained by fitting the validation set results; P(y=1|x test ) indicates that in the input sample x test Under given conditions, the predicted probability value of the sample being judged to have a cold soldering defect; x testThe feature vector of the test sample that needs to be tested; f(x test ) is the output value of the support vector machine (SVM) decision function, reflecting the input sample x test The value is a real number relative to the classification decision boundary, and the positive and negative signs are used to preliminarily determine the category. exp(·) is an exponential function, and the base e of the natural logarithm is the power operation of the base, which is used to map the linear decision value to the nonlinear probability interval.

[0148] This probability can be used as the basis for subsequent control system screening, graded disposal or manual re-inspection.

[0149] In addition, on some automated inspection production lines, the judgment results in this step will be bound to the unique code of the battery cell and synchronously written into the production database to achieve automatic marking, statistical archiving and quality traceability of abnormal samples.

[0150] S7. Dynamic internal resistance calculation: Dynamically calculate the dynamic internal resistance value under the pulse current application state, and judge whether there is a risk of cold soldering in the battery based on the calculation results;

[0151] After completing the model comparison and determination in step S6, further verification and supplementary determination from the perspective of electrical performance are required to improve overall detection accuracy and system stability. Therefore, step S7 (dynamic internal resistance calculation) is included in this process to further confirm the reliability of the weld connection quality through dynamic characteristics analysis of the battery terminal.

[0152] Generally, a solder joint defect not only shows abnormal temperature rise in infrared thermal imaging, but also leads to an abnormal increase in the equivalent internal resistance of the battery connection. Based on this, real-time measurement of dynamic internal resistance can serve as an effective supplementary criterion to assist in solder joint detection, improving the robustness and reliability of the overall judgment.

[0153] As an option, in some embodiments, dynamic internal resistance detection can be used as a secondary verification mechanism for abnormal screening, especially when the thermal imaging model determines that it is close to the boundary.

[0154] In this embodiment, the specific implementation of step S7 is as follows:

[0155] First, when the battery is in the state of applying pulse current, the battery terminal voltage change data is collected in real time.

[0156] To avoid measurement errors affecting the results, the applied current value and voltage response curve were recorded simultaneously when the pulse was applied. Subsequently, the dynamic internal resistance value R was calculated based on the measured data.

[0157] Specifically, the voltage change ΔU can be determined by the following formula:

[0158] ΔU=U before -U after ;

[0159] Among them: U before U is the battery terminal voltage in the steady state before the pulse is applied, in volts (V); after This is the instantaneous voltage reading after the pulse is applied, in volts (V).

[0160] In one possible implementation, in order to improve the accuracy of dynamic internal resistance calculation, this embodiment introduces a multiple pulse averaging mechanism, that is, the internal resistance value is sampled in multiple pulse periods and the average is calculated:

[0161]

[0162] Where: R avg is the average dynamic internal resistance after multiple pulse sampling, in ohms (Ω); R k is the single dynamic internal resistance calculated under the kth pulse application, in ohms (Ω); n2 is the total number of pulses, a positive integer, usually between 3 and 10 to suppress the influence of accidental errors.

[0163] Generally speaking, when there is a cold soldering defect in the solder joint, the contact resistance increases, which will lead to the dynamic internal resistance R or R avg Obviously higher than the reference threshold under normal working conditions.

[0164] Specifically, in the present invention, in order to realize automatic determination, a threshold value R is set in the dynamic internal resistance detection link. th , and determine the risk of cold soldering according to the following rules:

[0165] IfR avg >R th Then output "false soldering risk";

[0166] else output "welding is normal";

[0167] Where: R th Dynamic internal resistance determination threshold, in ohms (Ω), set based on historical normal sample statistics or determined through experimental calibration.

[0168] In one possible implementation, the threshold R th It can be dynamically adjusted according to different battery cell models, current amplitudes and environmental conditions to adapt to the consistency testing requirements of multiple batches of products.

[0169] As an option, in order to further refine the detection accuracy, some embodiments introduce a dynamic internal resistance change rate indicator ΔR rate , defined as follows:

[0170]

[0171] Where: ΔR rate The dynamic internal resistance change rate is expressed in percentage (%), reflecting the internal resistance relative to the normal reference value R ref The degree of change of R ref It is the standard dynamic internal resistance average value of historical normal solder joint samples, in ohms (Ω).

[0172] When ΔR rate Exceeds the set change rate threshold ΔR rate It can also be used as an auxiliary basis for determining cold solder joints.

[0173] In summary, step S7 effectively supplements the cold solder joint identification method based on thermal imaging and feature model reasoning by real-time measurement and determination of dynamic internal resistance under pulse current application conditions. And its extended applications have realized a technical path to further verify the welding status from the physical and electrical performance level.

[0174] S8. Detection and sorting of cold solder joints: If the dynamic internal resistance exceeds the baseline value by more than 20%, the battery is determined to have a risk of cold solder joints. X-ray or ultrasonic testing is then used to confirm the location of the cold solder joints. Automatic sorting is then performed based on the cold solder joint detection results. Batteries with a risk of cold solder joints are separated and sent for destructive testing.

[0175] After completing step S7, the abnormal cells need to be accurately identified and subsequently processed. Therefore, step S8 is set up in the process to locate the location of the cold solder defect based on the abnormal dynamic internal resistance results, combined with X-ray imaging or ultrasonic testing methods, and realize automated sorting and subsequent destructive verification.

[0176] Generally speaking, relying solely on electrical properties or thermal imaging for determination has certain limitations, especially for weak solder joints or localized poor connections. Introducing non-destructive testing as further verification can effectively improve the overall accuracy and reliability of the solder joint detection system.

[0177] As an option, in actual applications, a grading and sorting strategy can be set up according to production line requirements to handle abnormal cells of different degrees separately, thereby improving production line efficiency and product quality traceability.

[0178] First, according to the dynamic internal resistance calculation result in step S7, cells with abnormal dynamic internal resistance are screened out.

[0179] Specifically, the dynamic internal resistance change rate judgment standard is set as:

[0180] When ΔR rate When the percentage is >20%, it is considered that there is a risk of cold soldering;

[0181] That is, if the dynamic internal resistance increases by more than 20% compared to the baseline value, it is determined that the battery cell has a potential risk of poor soldering and further confirmation is required.

[0182] Specifically, in some embodiments, the battery cells determined to have a risk of cold soldering are first guided into a non-destructive testing unit.

[0183] In the non-destructive testing process, you can choose to use X-ray imaging or ultrasonic scanning technology to confirm the location of the cold weld.

[0184] Generally speaking, X-ray inspection is based on the density differences between different materials and can directly image the solder joint structure and identify connection breaks, holes or discontinuous areas.

[0185] Ultrasonic testing, on the other hand, detects abnormal voids or crack structures inside welds through changes in the reflection characteristics of sound waves, and is suitable for identifying deep welding defects.

[0186] As an option, in X-ray detection mode, the image acquisition process can adopt a multi-angle rotation imaging strategy to improve the coverage of the detection blind area.

[0187] In ultrasonic testing mode, a multi-band combined scanning method can be used to take into account the detection resolution of both shallow and deep defects.

[0188] After the test is confirmed, the cold-soldered cells are sorted according to the test results.

[0189] In one possible implementation, the sorting system is equipped with a robotic arm or an intelligent sorting mechanism to automatically remove battery cells that are judged to have a risk of poor soldering to an independent channel to prevent defective products from flowing into subsequent production links.

[0190] For the sorted cells, this embodiment further implements destructive testing and verification, including but not limited to:

[0191] Disassembly inspection: Physically disassemble the battery cell structure to observe the solder joint connection condition, pad damage, material residue characteristics, etc.

[0192] Tensile test: Perform standardized tensile breaking tests on welds to measure the mechanical strength of welds and confirm whether they meet process specifications.

[0193] Among them, the tensile test results usually record the maximum breaking force value F max , the unit is Newton (N).

[0194] If the actual breaking force is lower than the set process standard F th , it is further confirmed that the battery cell has a cold soldering defect.

[0195] Specifically, the tensile strength can be evaluated according to S7:

[0196] IfF max <F th Then it is determined to be "confirmed cold soldering";

[0197] Among them: F max F is the maximum breaking force measured during the test, in Newton (N); th This is the standard threshold for tensile testing, set according to different battery cell models and process requirements, and the unit is Newton (N).

[0198] In one possible implementation, the destructive testing results will be synchronously recorded in the database and associated with the previous thermal imaging testing, dynamic internal resistance testing and non-destructive testing data to form a complete testing traceability link.

[0199] In summary, step S8 completes the final confirmation and verification of the cold solder joint through dynamic internal resistance anomaly screening, combined with X-ray or ultrasonic non-destructive testing positioning, and then through sorting and destructive testing means.

[0200] S9. Electrical performance retest: Retest the electrical performance of batteries that are found to have poor solder joints, including charge and discharge cycle tests under different loads to check the actual power output and life performance of the battery;

[0201] Following step S8, cold solder joint identification and sorting are completed. To further verify the actual impact of cold solder joint defects on the battery cell's electrical performance, step S9 is set up. This step systematically re-inspects the cold solder joint identification samples through standardized charge and discharge tests, rate tests, and cycle life tests to ensure the accuracy of the test results and provide data support for subsequent product improvements and failure analysis.

[0202] Generally, cold solder joint defects can cause abnormal power drop, temperature rise, or capacity decay in battery cells during high-rate discharge and long cycle life. Therefore, comprehensive electrical performance retesting is important for verifying the accuracy of cold solder joint detection and gaining a deeper understanding of the impact mechanism of cold solder joint defects.

[0203] As an option, in some embodiments, the data from the electrical performance retest can also be used for subsequent modeling to further optimize the parameter settings and judgment criteria of the detection system.

[0204] In this embodiment, the technical implementation of step S9 is specifically as follows:

[0205] First, a constant current charge and discharge test is performed on the battery cells that are determined to have a risk of cold soldering in step S8.

[0206] Specifically, at a standard rate, a constant current is applied to perform charge and discharge cycles, and the output capacity Q of the battery during the discharge process is recorded.discharge And the change of internal resistance R int .

[0207] The calculation formula for discharge capacity is:

[0208] Q discharge =I discharge ×t discharge ;

[0209] Where: Q discharge is the discharge capacity, in ampere-hours (Ah), which indicates the total amount of electricity in a complete discharge process; I discharge is the discharge current in amperes (A), which is set to a fixed value corresponding to the standard rate; t discharge The discharge duration is in hours (h), which is the time required for the actual discharge to complete.

[0210] Generally speaking, the discharge capacity of a poorly soldered cell will decrease to a certain extent compared to a normal sample, and will be accompanied by an abnormal increase in internal resistance.

[0211] Subsequently, a rate discharge test was conducted, i.e., a constant current was applied under different load conditions (e.g., 0.5C, 1C, 2C, and 3C rates), and the terminal voltage change U(t), output power P(t), and temperature rise curve ΔT(t) were recorded in real time.

[0212] The formula for calculating the output power P(t) is:

[0213] P(t)=U(t)×I discharge (t);

[0214] Where: P(t) is the instantaneous output power, in watts (W); U(t) is the terminal voltage at a certain moment in the discharge process, in volts (V); I discharge (t) is the instantaneous current during the discharge process, in amperes (A).

[0215] In one possible implementation, the maximum temperature rise is recorded at the same time:

[0216] ΔT max =T max -T ambient ;

[0217] Where: ΔT max is the maximum temperature rise value, in degrees Celsius (℃); T max T is the maximum temperature of the cell surface during the test, in degrees Celsius (℃); ambient The reference temperature of the test environment, in degrees Celsius (℃).

[0218] The influence of cold soldering defects on the power characteristics of the battery cell can be comprehensively judged by the voltage drop rate at the lower end at different rates, the power output capacity and the abnormal degree of temperature rise.

[0219] Furthermore, a cycle life test is also performed in this embodiment.

[0220] Specifically, the tested battery cell is set to a standard rate charge and discharge cycle, and no less than 100 complete cycle operations are performed.

[0221] At each fixed cycle node (such as the 1st, 20th, 50th, and 100th time), the discharge capacity Q is recorded. n and internal resistance R n , and calculate the capacity retention rate CR and internal resistance change rate ΔR.

[0222] The calculation formula for capacity retention rate is:

[0223]

[0224] Where: CR is the capacity retention rate, expressed in percentage (%), reflecting the capacity maintenance level of the battery after n cycles; Q n is the discharge capacity measured at the nth cycle, in ampere-hours (Ah); Q1 is the initial discharge capacity measured at the first cycle, in ampere-hours (Ah).

[0225] The calculation formula for the internal resistance change rate is:

[0226]

[0227] Where: ΔR is the internal resistance change rate, expressed in percentage (%), indicating the relative magnitude of internal resistance growth; R n is the internal resistance value measured in the nth cycle, in ohm (Ω); R1 is the initial internal resistance value measured in the 1st cycle, in ohm (Ω).

[0228] In one possible implementation, if the capacity retention rate CR is found to be lower than 80% after 100 cycles, or the internal resistance change rate ΔR exceeds 30%, it can be further confirmed as a sample with a cold soldering defect that seriously affects the electrical performance.

[0229] Generally speaking, during the cycle process, the poorly soldered battery cells will age faster due to concentrated local heat and poor contact at the welding points, which will manifest as accelerated capacity decline and abnormal growth of internal resistance.

[0230] In summary, step S9 uses a standardized electrical performance retest process, combined with the three indicators of charge and discharge performance, rate adaptability and cycle life, to systematically verify the actual impact of cold soldering defects on battery performance.

[0231] The device for detecting weld defects in cylindrical lithium-ion batteries described below and the method for detecting weld defects in cylindrical lithium-ion batteries described above can be used in correspondence with each other.

[0232] The present invention also provides a device for detecting cold welds in cylindrical lithium-ion batteries, comprising:

[0233] Pulse current generation module: used to apply pulse current and record the instantaneous voltage drop and current response at the battery terminal. This module includes a pulse signal generator, a power amplifier, and a current regulation circuit. The pulse signal generator is responsible for generating a pulse control signal of set frequency and width as the system's driving source; the power amplifier receives and amplifies the pulse signal to generate a large current pulse sufficient to drive the battery; the current regulation circuit monitors the output current in real time and adjusts the power amplifier's drive signal through closed-loop feedback to ensure the output current amplitude;

[0234] Infrared thermal imaging acquisition module: used to collect infrared thermal images of the welding part in real time. This module includes a high-resolution infrared thermal imager that can capture the temperature changes of the welding part and transmit the image data to the subsequent processing module for analysis;

[0235] Image processing and analysis module: used to pre-process infrared thermal images and extract the temperature characteristics of welding parts. This module includes image noise reduction, grayscale correction, edge detection, and region segmentation algorithms. It can extract temperature change information of welding areas from images for subsequent false solder joint determination.

[0236] Battery positioning and transmission module: used to automatically transmit and position batteries to ensure that the welding part is within the detection area. The module consists of a mechanical bracket, a positioning fixture, a transmission guide rail and a drive motor. The positioning fixture is designed according to the outer diameter of the cylindrical lithium-ion battery and adopts an elastic clamp structure. It can not only ensure the accurate positioning of the battery during the detection process, but also adapt to changes in battery specifications within a certain size range. The transmission guide rail adopts a high-precision linear guide rail to ensure the stability and straightness of the battery during the transmission process. The drive motor is connected to the screw of the transmission guide rail through a coupling to realize the control of the battery transmission speed and position.

[0237] The device of this embodiment can be used to execute the above method embodiment, and its principles and technical effects are similar, so they will not be repeated here.

[0238] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A method for detecting cold welds in cylindrical lithium-ion batteries, characterized in that: The following steps are involved: S1. Appearance inspection: Check the appearance of the battery and confirm that there is no oxidation, cracks or obvious deformation at the welding parts; S2. Initial internal resistance benchmark test: Apply a low-amplitude constant test current to the cylindrical lithium-ion battery that has passed the appearance inspection, obtain the voltage change value before and after the load, and calculate its initial internal resistance R0R_0R0 to determine whether there is any abnormality in the electrical connection performance of the welding area. The initial internal resistance R0 is calculated by the following formula: Where: R0 represents the initial internal resistance of the battery, in milliohms (mΩ), which is the equivalent ohmic impedance of the conduction path at the solder joint; ν t It is the voltage value collected instantaneously after applying the test current I, in volts (V), reflecting the terminal voltage after the dynamic response of the battery cell; ν0 is the open circuit voltage in the static state before the current is applied, in volts (V), indicating the natural voltage of the battery cell to be tested; I is the applied test current, in amperes (A), which is a constant low-amplitude pulse current, generally 50mA to 100mA, determined by the battery cell capacity specifications. By controlling the current amplitude to operate within the range where the battery cell is not polarized and does not heat up, the test disturbance is effectively avoided; further comprising temperature correction of the initial internal resistance value to eliminate the influence of the ambient temperature on the internal resistance measurement result, the temperature correction is performed according to the following formula: Where: R corr is the internal resistance value after temperature correction, in mΩ, indicating the theoretical equivalent value at standard temperature; α is the temperature coefficient, in 1 / °C1, with a value range of 0.003–0.006, which is determined by the conductor material; T is the actual test environment temperature, in degrees Celsius (°C), measured by the environmental sensor; T ref The reference standard temperature is set to 25°C. When the initial internal resistance exceeds the preset qualified range, or the measured value fluctuation exceeds the set threshold, the battery cell is judged to have a risk of poor welding contact and will be rejected or transferred to the subsequent process; S3. Pulse current application: Use a pulse generator to apply short, high-current pulses to the battery to simulate actual load conditions, and record the instantaneous voltage drop and current response curve of the battery terminal during the pulse. S4. Infrared thermal imaging recognition and data processing: Use an infrared thermal imager to collect infrared thermal images of the welding area in real time to ensure that the dynamic process of temperature changes can be captured. The received infrared thermal images are pre-processed, including image noise reduction, grayscale correction, edge detection, and region segmentation algorithms to extract the temperature distribution area of the welding area; S5. Learn images and build models: Using a large amount of experimental data, a support vector machine algorithm is used to build a cold solder joint detection model, which is trained and optimized based on the temperature characteristic data of the welding parts. S6. Comparison of test images and data with the model: Input the real-time collected infrared thermal images and test data into the established cold solder joint detection model for comparison and analysis to determine whether there is a cold solder joint at the welding position; S7. Dynamic internal resistance calculation: Dynamically calculate the dynamic internal resistance value under the pulse current application state, and judge whether the battery has the risk of cold soldering based on the calculation results; S8. Detection and sorting of cold solder joints: If the dynamic internal resistance exceeds the baseline value by more than 20%, the battery is determined to have a risk of cold solder joints. X-ray or ultrasonic testing is then used to confirm the location of the cold solder joints. Automatic sorting is then performed based on the cold solder joint detection results. Batteries with a risk of cold solder joints are separated and sent for destructive testing. S9. Retest of electrical performance: Retest the electrical performance of batteries that are determined to have poor solder joints, including charge and discharge cycle tests under different loads to check the actual power output and life performance of the battery.

2. The method for detecting cold welds in cylindrical lithium-ion batteries according to claim 1, wherein: The appearance inspection includes: Use a magnifying glass to observe the morphology of the battery welding points; Eliminate solder joints that are obviously not fused or have insufficient solder; Check the battery's exterior for signs of oxidation, cracks, or obvious deformation.

3. The method for detecting cold welds in cylindrical lithium-ion batteries according to claim 1, wherein: The pulse current application comprises: Use a pulse generator to apply short-term high-current pulses to the battery to simulate the working conditions of the battery under actual load; Record the instantaneous voltage drop and current response curve of the battery terminal to calculate the dynamic internal resistance.

4. The method for detecting cold welds in cylindrical lithium-ion batteries according to claim 1, wherein: The infrared thermal imaging recognition and data processing include: Use an infrared thermal imager to collect infrared thermal images of the welding part in real time with high resolution; Perform image preprocessing, including image noise reduction and grayscale correction, and use edge detection and region segmentation algorithms to extract temperature characteristics of the welding area; Calculate the temperature rise rate and maximum temperature in the welding area.

5. The method for detecting cold welds in cylindrical lithium-ion batteries according to claim 1, wherein: The learning of pictures and building of models include: Collect a large amount of infrared thermal image data, combine it with known types of cold solder joint defects, and use the support vector machine algorithm to establish a detection model; Use data to train the model and optimize detection accuracy to ensure that the model can effectively determine cold solder joints in actual tests.

6. The method for detecting cold welds in cylindrical lithium-ion batteries according to claim 1, wherein: The comparison of the test images and data with the model includes: Input the real-time collected infrared thermal images and test data into the cold solder joint detection model; The model is used for comparative analysis to determine whether there are cold weld defects in the welding area and output the judgment results.

7. The method for detecting cold welds in cylindrical lithium-ion batteries according to claim 1, wherein: The dynamic internal resistance calculation and cold soldering judgment include: According to the formula Calculate the dynamic internal resistance value when the pulse current is applied, where: R is the dynamic internal resistance; ΔU is the instantaneous voltage drop at the battery terminal; and I is the applied pulse current.

8. The method for detecting cold welds in cylindrical lithium-ion batteries according to claim 1, wherein: The virtual weld query and sorting process includes: Use X-ray or ultrasonic testing to confirm the location of the cold solder joint and sort the cold solder joint batteries; Destructive testing is performed on batteries with high internal resistance, including disassembly and tensile testing, to further verify the problem of cold soldering.

9. The method for detecting cold welds in cylindrical lithium-ion batteries according to claim 1, wherein: The electrical performance retest includes: Perform constant current charge and discharge tests on batteries that have been identified as having a cold solder joint, and record the discharge capacity and internal resistance changes of the batteries at standard rates to evaluate the stability of their electrical performance. Perform rate discharge tests under different load conditions, compare the terminal voltage changes, output power, and temperature rise at each rate, and determine the impact of cold soldering on battery power characteristics; A cycle life test was conducted, in which the battery was charged and discharged for 100 cycles to observe the capacity retention rate and internal resistance change trends.

10. A device for detecting cold welds in cylindrical lithium-ion batteries, characterized in that: A method for detecting cold welds in a cylindrical lithium-ion battery according to any one of claims 1 to 9, comprising: Pulse current generation module: used to apply pulse current and record the instantaneous voltage drop and current response at the battery terminal. This module includes a pulse signal generator, a power amplifier, and a current regulation circuit. The pulse signal generator is responsible for generating a pulse control signal of set frequency and width as the system's driving source; the power amplifier receives and amplifies the pulse signal to generate a large current pulse sufficient to drive the battery; the current regulation circuit monitors the output current in real time and adjusts the power amplifier's drive signal through closed-loop feedback to ensure the output current amplitude; Infrared thermal imaging acquisition module: used to collect infrared thermal images of the welding part in real time. This module includes a high-resolution infrared thermal imager that can capture the temperature changes of the welding part and transmit the image data to the subsequent processing module for analysis; Image processing and analysis module: used to pre-process infrared thermal images and extract the temperature characteristics of welding parts. This module includes image noise reduction, grayscale correction, edge detection, and region segmentation algorithms. It can extract temperature change information of welding areas from images for subsequent false solder joint determination. Battery positioning and transmission module: used to automatically transmit and position batteries to ensure that the welding part is within the detection area. The module consists of a mechanical bracket, a positioning fixture, a transmission guide rail and a drive motor. The positioning fixture is designed according to the outer diameter of the cylindrical lithium-ion battery and adopts an elastic clamp structure. It can not only ensure the accurate positioning of the battery during the detection process, but also adapt to changes in battery specifications within a certain size range. The transmission guide rail adopts a high-precision linear guide rail to ensure the stability and straightness of the battery during the transmission process. The drive motor is connected to the screw of the transmission guide rail through a coupling to realize the control of the battery transmission speed and position.

Images