A method and system for analyzing the coating uniformity of positive and negative electrode sheets of lithium ion batteries

Abstract

The application discloses a kind of lithium ion battery positive and negative pole piece coating uniformity analysis method and system, belong to ternary lithium ion battery technical field, including laser source module, multi-point synchronous excitation and dynamic focusing module, light collection and spectral analysis module, data processing and intelligent analysis module and man-machine interface and feedback interface;Realize laser multi-point synchronous excitation by spatial light modulator;Combined with line laser profiler and piezoelectric platform dynamic compensation defocus, improve signal stability;Adopt double-band self-calibration algorithm to eliminate system noise, rely on physical guidance neural network to realize active substance equivalent area density high-precision inversion;Real-time output full-width composition and area density distribution thermal diagram, automatically calculate CV%, CPK and the like index, and send closed-loop control instruction to coater;The lithium ion battery positive and negative pole piece coating uniformity analysis method and system provided by the application realize high-speed, nondestructive, online detection without radioactive source.

Description

Technical Field

[0001] This invention relates to the field of ternary lithium-ion battery technology, and in particular to a method and system for analyzing the coating uniformity of positive and negative electrode sheets in lithium-ion batteries. Background Technology

[0002] Coating uniformity is a core process factor that determines the electrochemical performance, product consistency, and safety of lithium-ion batteries. Uneven coating can lead to battery capacity decay, shortened cycle life, and even safety hazards such as lithium dendrite formation and thermal runaway. It also reduces production yield and increases manufacturing costs. Therefore, it is crucial to achieve rapid and accurate online monitoring of the surface density and composition consistency of the positive and negative electrodes.

[0003] Existing detection technologies have significant limitations: beta-ray / gamma-ray areal density meters can only detect total dry matter, cannot distinguish components, and involve radiation source management; X-ray fluorescence spectroscopy has low sensitivity to light elements such as Li and C, and the equipment is expensive and difficult to integrate; offline laboratory analysis methods are time-consuming, can only perform sampling inspections, and cannot guide real-time process adjustments. Laser-induced breakdown spectroscopy (LIBS) technology can achieve rapid, near-non-destructive testing of multiple elements, but traditional single-point LIBS suffers from slow detection speed, large signal fluctuations, and significant matrix effects, making it difficult to adapt to the online testing requirements of high-speed lithium battery production lines.

[0004] In summary, existing technologies cannot achieve high-speed, non-destructive, online intelligent detection of the uniformity of coating on positive and negative electrodes of lithium-ion batteries, nor can they simultaneously obtain information on coating composition and thickness distribution. There is an urgent need to develop detection technologies and systems that meet the needs of industrial production. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for analyzing the coating uniformity of positive and negative electrode sheets in lithium-ion batteries, which solves the problems of existing technologies such as slow detection speed that cannot match high-speed production lines, poor signal stability, difficulty in simultaneously obtaining coating composition and areal density information, and some technologies involve radiation source management, low sensitivity to light element detection, and can only perform offline sampling inspections without achieving real-time closed-loop control.

[0006] To achieve the above objectives, the present invention provides a method for analyzing the coating uniformity of positive and negative electrode sheets in lithium-ion batteries, comprising the following steps: S1. The positive and negative electrode sheets pass through the inspection station at the speed set by the production line. S2. The line laser profilometer continuously scans to acquire real-time data on the surface morphology of the electrode. S3. The control system dynamically adjusts the focusing lens group based on the topography data through the piezoelectric drive platform to accurately focus the laser sub-beam. S4. The laser source module emits a pulsed laser beam, which is divided into a matrix of a set number of sub-beams by a spatial light modulator (SLM) to simultaneously form ablation points and generate plasma on the surfaces of the positive and negative electrodes. S5. The optical collection system captures the composite emission light of the plasma and transmits it to the spectral analysis module; S6. The high-speed area array spectrometer performs a single exposure, separating and recording the complete spectrum of all ablation points; S7, the data processing and intelligent analysis module performs spatial decoding, dual-band normalization calibration, PINN model inversion calculation, generates surface density distribution map, coefficient of variation CV%, and process capability index CPK, and performs out-of-tolerance judgment; S8, the human-computer interaction and feedback interface displays the surface density distribution map, CV% and CPK in real time, and responds in stages according to CV% and CPK: yellow light warning when there is a slight deviation, yellow light flashing rapidly when there is a significant deviation; red light alarm when there is a dangerous threshold, and automatic adjustment when the deviation can be attributed and closed loop is enabled.

[0007] Preferably, the dual-band normalization calibration formula in step S7 is: in, For the first Dual-band normalized intensity ratio at each detection point; For the first The integral intensity of the characteristic spectral lines of the target element at each detection point; For the first The integral intensity of the characteristic spectral lines of the reference element at each detection point; The wavelength of the characteristic spectral line of the target element; The characteristic wavelength of the reference element.

[0008] Preferably, the total loss function of the PINN model in step S7 is: in, For data fitting terms; This is a monotonic physical constraint loss term; This represents the spatial continuity physical constraint loss term; m The relative weights of the monotonic physical constraint loss term; n The relative weights are the physical constraint loss terms for spatial continuity.

[0009] Preferably, the CV% calculation formula in step S8 is: in, Standard deviation This is the sample mean; The formula for calculating CPK is: in, This is the upper limit of the specification. This is the lower limit of the specification.

[0010] This invention also provides a system for analyzing the coating uniformity of positive and negative electrode sheets of lithium-ion batteries, including: a laser source module, a multi-point synchronous excitation and dynamic focusing module, a light collection and spectral analysis module, a data processing and intelligent analysis module, a human-computer interaction and feedback interface, and a timing controller; the timing controller coordinates with each module to complete the timing actions of laser emission, morphology scanning, focus compensation, spectral acquisition, and data processing, and the moving positive and negative electrode sheets of the lithium-ion battery under test pass through the detection stations of each module at a speed set by the production line; The laser source module uses a high repetition rate, small pulse width and high stability Nd:YAG laser. The wavelength needs to be selected to balance ablation efficiency and versatility with most materials. The multi-point synchronous excitation and dynamic focusing module includes a spatial light modulator (SLM), a focusing lens group, a line laser profilometer, and a piezoelectric driven translation stage; the spatial light modulator (SLM) is the core component, which is highly programmable and can generate laser matrices of arbitrary shape and number as needed; The light collection and spectral analysis module includes an optical collection system and a high-speed area array spectral imaging system; the optical collection system is a light collection fiber bundle matrix, and the high-speed area array spectral imaging system includes a grating spectrometer and a detector integrated into the grating spectrometer; The data processing and intelligent analysis module incorporates a dual-band self-calibration engine and a physically guided neural network model. The human-computer interaction and feedback interface builds a visualization software platform that displays the composition, trend curves, statistical reports, defect alarms and logs within the characteristic spectral range in real time, and sends key parameters, such as average areal density and maximum deviation location, to the coating machine PLC system.

[0011] Preferably, the spatial light modulator (SLM) in the multi-point synchronous excitation and dynamic focusing module generates a two-dimensional laser matrix of arbitrary shape and number through programming, splitting the laser beam output by the laser source module into laser sub-beams; The focusing lens group focuses the laser sub-beam onto the surface of the moving positive and negative electrode plates of the lithium-ion battery, forming a micron-scale ablation lattice; A line laser profilometer and a piezoelectric-driven translation stage constitute a dynamic focus compensation unit. The line laser profilometer continuously scans the same area of ​​the moving positive and negative electrode sheets of the lithium-ion battery at a speed higher than the main laser frequency, and acquires the electrode surface height map in real time. The control system calculates the distance between the focus of each laser sub-beam and the electrode surface based on the height map, and finely adjusts the focusing lens group through the piezoelectric-driven translation stage to ensure that all laser points are in the optimal focusing state, thereby eliminating signal attenuation caused by foil jitter.

[0012] Preferably, the optical fiber bundle matrix of the light collection and spectral analysis module is a paraxial acquisition system independent of the laser focusing system, which directly collects the plasma light generated by the laser ablation of the electrode; it has high collection efficiency, no missing spectral bands, and the paraxial acquisition system is more sensitive to the position fluctuation of the plasma and the change of the sample surface height; The grating spectrometer is equipped with a high-performance sCMOS camera. This spectrometer can acquire spectral ranges from 200nm to 1000nm. Through fiber bundle arrangement or spatial coding design, the grating spectrometer can achieve spectral resolution at the sub-pm level, so that the pixels in a specific area on the detector correspond one-to-one with the laser ablation points on the moving positive and negative electrodes of the lithium-ion battery, realizing the coupling of spatial information and spectral information.

[0013] Preferably, the dual-band self-calibration engine of the data processing and intelligent analysis module selects Li 610.4nm as the target band, Al I 396.15nm as the reference band for the aluminum foil positive electrode, and Cu I 324.7nm as the reference band for the copper foil negative electrode. The core of self-calibration is to calculate the normalized intensity ratio, which eliminates the common noise of the system caused by laser energy fluctuations, plasma temperature changes, and differences in collection efficiency. The physical-guided neural network model not only learns from historical data, but also embeds known physical laws as soft constraints into the training process.

[0014] Preferably, the data processing and intelligent analysis module implements graded control based on thresholds of the coefficient of variation (CV%) and the process capability index (CPK); the human-computer interaction and feedback interface is used to display early warning information and status of the graded control; the thresholds for graded control are defined as follows: Normal state: CV% < 1.3%, CPK ≥ 1.67; the human-computer interaction interface displays a solid green light, displays a real-time surface density distribution map, updates CV% and CPK statistical indicators in real time on the statistics panel, records data in real time, and does not generate alarm information; Slight deviation: 1.3% ≤ coefficient of variation (CV%) < 1.8% or 1.33 ≤ process capability index (CPK) < 1.67; the system automatically generates alarm records and stores the alarm time and cause in the database. In the areal density distribution map, the brightness of slightly out-of-tolerance areas is enhanced or the borders are highlighted. Operators determine the type of deviation based on the heat map and trend curve and conduct on-site checks of slurry viscosity / sedimentation, whether there are foreign objects or material accumulation at the die lip, whether the foil conveyor belt is stable, and whether the oven temperature curve is normal. Based on experience, staff can decide whether to manually adjust the die gap or feeding pressure slightly. If the fluctuation is determined to be occasional, the yellow light can be cleared by clicking the "Confirm Known" button, and the system will continue monitoring. Significant deviation: 1.8% ≤ coefficient of variation CV% < 2.0% or 1.0 < process capability index CPK < 1.33; the yellow light flashes rapidly and the alarm is activated, a detailed alarm window pops up, highlighting performance indicators, thermal distribution location and duration, the system records data at high frequency and generates a detailed alarm log, and displays modification suggestions for manual inspection and confirmation or manual adjustment; Danger threshold: CV%≥2.0%, or CPK≤1.0; Displays control commands generated by the physical guided neural network PINN; When automatic control is turned on, the system automatically writes the control commands directly to the coating machine PLC after a 5-second countdown, without manual confirmation, and records the closed-loop intervention log completely; When automatic control is turned off, the window continuously alarms and highlights the message "Automatic closed loop is disabled, please intervene manually immediately," and the operator must immediately stop the machine manually to troubleshoot. When the system continuously detects a certain attributable deviation pattern, it automatically generates a correction command and sends it directly to the PLC of the coating machine via the industrial communication protocol, enabling the system to achieve automatic closed-loop control.

[0015] Therefore, the method and system for analyzing the coating uniformity of positive and negative electrode sheets in lithium-ion batteries, as described above, have the following beneficial effects: (1) Adopting a multi-point synchronous scanning LIBS architecture: Parallel data acquisition is achieved through spatial beam splitting, which solves the contradiction that traditional single-point LIBS is slow and cannot match high-speed production lines, making LIBS feasible for online detection of the coating uniformity of positive and negative electrode sheets of lithium-ion batteries. (2) Dynamic focusing compensation mechanism: By combining line laser profile measurement with piezoelectric driving focusing, real-time adaptive optical focusing on high-speed moving, non-rigid foil surfaces can be achieved, which greatly improves signal stability and detection reliability. (3) Dual-band self-calibration + PINN joint inversion model: It combines the classic internal standard method with cutting-edge AI technology, which not only uses physical laws to suppress noise, but also uses deep learning to mine complex nonlinear relationships, thereby improving the accuracy and robustness of quantitative inversion of actual process parameters from complex LIBS spectra. (4) System integration for industrial scenarios: The entire system design takes into account the factory environment (vibration, dust, electromagnetic interference), ease of use (automated calibration, fault diagnosis) and integration capabilities with existing manufacturing execution systems and data acquisition and monitoring systems; (5) Intrinsic safety and green manufacturing: Using lasers as the detection source eliminates the dependence on radiation sources, making it safer, more environmentally friendly, and without special management burdens.

[0016] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the system structure according to an embodiment of the present invention; Figure 2 This is a flowchart of a method according to an embodiment of the present invention.

[0018] Figure Labels 1. Laser source module; 2. Spatial light modulator (SLM); 3. Focusing lens group; 4. Line laser profilometer; 5. Piezoelectric driven translation stage; 6. Optical fiber bundle matrix for light collection; 7. Grating spectrometer; 8. Data processing and intelligent analysis module; 9. Human-computer interaction and feedback interface; 10. Moving positive and negative electrodes of lithium-ion battery; 11. Timing controller; 12. Detector. Detailed Implementation

[0019] The following detailed description of embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0020] Combination Figure 1 The components are labeled as follows: 1. Laser source module; 2. Spatial light modulator (SLM); 3. Focusing lens group; 4. Line laser profilometer; 5. Piezoelectric driven translation stage; 6. Light collecting fiber bundle matrix; 7. Grating spectrometer; 8. Data processing and intelligent analysis module; 9. Human-computer interaction and feedback interface; 10. Moving lithium-ion battery positive and negative electrodes; 11. Timing controller; 12. Detector; where thick lines represent data transmission and thin lines represent light propagation.

[0021] This system is built based on the principle of laser-induced breakdown spectroscopy and is adapted to high-speed coating production lines for lithium-ion battery positive and negative electrodes. It enables online non-destructive testing of electrode coating uniformity. (Reference) Figure 1 The specific implementation and deployment of each module are as follows: 1. Laser source module 1 The laser source module 1 uses a high repetition rate, narrow pulse width, and high stability Nd:YAG laser. The output wavelength of the laser takes into account both laser ablation efficiency and universality for detecting various types of lithium battery positive and negative electrode materials. It is installed at the beginning of the system's optical path, and its emitting end is arranged corresponding to the incident end of the spatial light modulator (SLM) 2 to ensure the precise incident of the high-energy pulsed laser beam, providing a stable light source for multi-point synchronous excitation. Moreover, the laser is controlled by the timing controller 11 to achieve precise timing control of laser emission.

[0022] 2. Multi-point synchronous excitation and dynamic focusing module This module consists of a spatial light modulator (SLM) 2, a focusing lens group 3, a line laser profilometer 4, and a piezoelectric driven translation stage 5, and is the core execution module of the system. The spatial light modulator (SLM) 2 is positioned adjacent to the laser source module 1 and features programmable control. It can generate two-dimensional laser dot arrays of arbitrary shapes and quantities according to detection requirements, achieving precise beam splitting of a single laser beam. The focusing lens group 3 is fixedly installed at the execution end of the piezoelectric driven translation stage 5, located between the light-emitting end of the spatial light modulator (SLM) 2 and the moving positive and negative electrodes 10 of the lithium-ion battery. It is used to precisely focus the split laser sub-beams onto the electrode surface to form a micron-level ablation dot array. The line laser profilometer 4 is positioned at the detection station above the electrode, on the same side and adjacent to the focusing lens group 3. It scans the electrode surface at a frequency higher than the main laser emission frequency and collects morphology data in real time. The piezoelectric driven translation stage 5 is controlled by the timing controller 11 and can drive the focusing lens group 3 to make fine adjustments within a range of ±100μm to compensate for defocusing deviations caused by foil jitter and electrode wrinkles, ensuring that the laser sub-beams are always in the optimal focusing state.

[0023] 3. Light collection and spectral analysis module This module consists of a light-collecting fiber bundle matrix 6, a grating spectrometer 7, and a detector 12. The light-collecting fiber bundle matrix 6 adopts a paraxial independent layout, avoiding interference with the laser focusing optical path. It is directly facing the laser ablation area of ​​the electrode, efficiently collecting the composite light signal emitted by the plasma without any spectral band loss, and exhibiting high sensitivity to plasma position fluctuations and electrode height changes. The grating spectrometer 7 is connected to the light-collecting fiber bundle matrix 6 via optical fiber and integrates the detector 12 internally. The grating spectrometer 7 is equipped with a high-performance sCMOS camera and can acquire spectra in the wavelength range of 200nm-1000nm, with a spectral resolution of sub-pm. Through fiber bundle arrangement and spatial coding design, the pixels of the detector 12 correspond one-to-one with the laser ablation points of the electrode, realizing the coupled acquisition of spatial and spectral information. The data output end of the spectrometer is connected to the data processing and intelligent analysis module 8 via thick lines to transmit the raw spectral data.

[0024] 4. Data Processing and Intelligent Analysis Module 8 The data processing and intelligent analysis module 8 is the core processing unit of the industrial control computer. It has a built-in dual-band self-calibration engine and a physical guided neural network PINN model. The dual-band self-calibration engine selects Li 610.4nm as the target band, the aluminum foil positive electrode is adapted to the Al I 396.15nm reference band, and the copper foil negative electrode is adapted to the Cu I 324.7nm reference band. It can calculate the normalized intensity ratio of the dual bands and remove common noise of the system. The PINN model embeds physical laws such as monotonicity and spatial continuity as soft constraints into the training process to optimize the detection residual and improve the accuracy of surface density inversion. The module input is connected to the timing controller 11 and the grating spectrometer 7, respectively, and the output is connected to the human-machine interaction and feedback interface 9 to realize data processing, analysis and transmission.

[0025] 5. Human-computer interaction and feedback interface 9 The human-machine interaction and feedback interface 9 is equipped with a visualization software platform, which has functions such as data display, early warning prompts, command transmission, and log storage. It can display statistical indicators such as spectral composition, coating quality heat map, CV%, and CPK in real time. It is configured with a standard industrial communication protocol interface, which can communicate bidirectionally with the coating machine PLC system to realize graded early warning, manual intervention or automatic closed-loop control. It is installed at the production line control terminal for operators to view and control in real time.

[0026] 6. Timing Controller 11 The timing controller 11 is the core of the system's timing coordination. It is connected to the laser source module 1, the line laser profilometer 4, the piezoelectric driven translation stage 5, the grating spectrometer 7, and the data processing and intelligent analysis module 8 via thick lines. It accurately synchronizes the entire process of laser emission, topography scanning, focus compensation, spectral acquisition, and data processing, ensuring that all modules of the system operate in coordination without timing deviation.

[0027] This embodiment, based on the aforementioned system, conducts an analysis of the coating uniformity of the positive and negative electrode sheets in lithium-ion batteries, referencing... Figure 2 The specific implementation steps are as follows: 1. System initialization and calibration The laser source module 1, grating spectrometer 7, line laser profilometer 4, equipment cooling system and industrial control computer are turned on, and the timing controller 11 starts synchronously. The system automatically completes the self-check of the status of each hardware module to confirm that the communication is smooth and the operation is normal. When the stationary lithium-ion battery positive and negative electrode plates 10 pass through the detection station, blank background spectral signals are collected for background interference subtraction of subsequent online detection data to eliminate detection errors caused by ambient stray light and equipment baseline drift.

[0028] Multiple sets of standard NCM ternary cathode plates with known true values ​​of active material surface density, precisely measured by inductively coupled plasma atomic emission spectrometry, were selected. These plates were then subjected to simulated production line conditions, passing through the testing station at a uniform speed of 80 m / min. Spectral data were collected at each measurement point, and the dual-band normalized intensity ratio was calculated. And establish the initial calibration mapping relationship: in, For the first The predicted equivalent surface density of active material at each measurement point; For the first The dual-band normalized intensity ratio calculated from each measurement point. These are residuals or errors that the model cannot explain. The calculation formula is: in, For the first Dual-band normalized intensity ratio at each detection point; For the first The integral intensity of the characteristic spectral lines of the target element at each detection point; For the first The integral intensity of the characteristic spectral lines of the reference element at each detection point; The wavelength of the characteristic spectral line of the target element; The characteristic wavelength of the reference element.

[0029] Based on the above mapping relationship, the initial weight training of the PINN physical guided neural network model is completed.

[0030] 2. Electrode conveying and station positioning When the production line starts, the positive and negative electrode sheets 10 of the lithium-ion battery to be tested enter the station of this testing system from the coating oven at the speed set by the production line.

[0031] 3. Line laser profilometer 4 continuously scans, providing real-time appearance measurement data. The line laser profilometer 4, located above the electrode, continuously projects a laser line perpendicular to the direction of travel onto the moving positive and negative electrode surfaces at a speed much higher than the main laser frequency. A matching camera (such as an sCMOS camera) is used to capture the deformed image of the light, and the surface height corresponding to each pixel on the current scan line is calculated in real time. The continuous line scan data are stitched together to form a miniature three-dimensional topographic map covering the area to be irradiated by LIBS.

[0032] 4. Dynamic Focused Compensation Calculation and Execution The real-time height map data obtained in the third step is sent to the central control unit. The control system calculates the defocus distance of each sub-laser beam focal point in the region relative to the ideal focal point based on the preset focal plane position. The control system sends precise displacement commands to the piezoelectric drive translation stage 5, requiring it to drive the focusing lens group 3 to make fine adjustments (usually within ±100μm) to compensate for the average defocus amount, or to perform tilt / curvature compensation on the local area through algorithms to ensure that all laser sub-beams achieve optimal focus as simultaneously as possible.

[0033] 5. Multi-point synchronous laser excitation and spectral acquisition An Nd:YAG high-energy laser pulser emits a laser pulse. The spatial light modulator (SLM) 2 is programmed to divide the laser pulse into an N×M two-dimensional lattice (the laser points must be spaced a certain distance apart to avoid cross-contamination). The divided sub-beams are then passed through a pre-adjusted focusing lens group 3, causing the sub-beams to simultaneously act on the surfaces of the positive and negative electrodes. This simultaneous action of the sub-beams on the surfaces of the positive and negative electrodes instantaneously generates high-temperature plasma at each point, simultaneously emitting composite light containing characteristic spectral lines of various elements. These characteristic spectral lines are shown in Table 1. Table 1 Elemental Characteristic Spectra

[0034] The rangefinder acquisition system collects all the radiation light and guides it into the grating spectrometer 7, where a single exposure is performed by the camera at the rear. In the two-dimensional image captured by the camera, one dimension represents the wavelength, and the other dimension encodes information about different spatial locations.

[0035] 6. Data transmission and decoding The spectrometer transmits the captured raw two-dimensional image data to the data processing module of the industrial control computer. The software separates the independent regions on the image according to the pre-calibrated spatial-pixel mapping relationship and restores them into independent one-dimensional spectral files corresponding to specific locations on the electrode.

[0036] 7. Spectral preprocessing and self-calibration For each one-dimensional spectrum, noise filtering and background subtraction are performed sequentially to obtain clean emission lines. The software automatically identifies and extracts the intensity of key spectral lines in each spectrum. The intensity of the target element's spectral lines is compared with that of the reference element's spectral lines, and the normalized ratio at each point is calculated, generating a data vector containing all normalized ratios. The core principle of dual-band self-calibration is that factors such as laser energy fluctuations, plasma temperature changes, and differences in collection efficiency simultaneously affect the spectral line intensities of both the target and reference elements. By calculating the normalized intensity ratio of the target element's spectral lines to that of the reference element's spectral lines, these common system noises can be canceled out, thereby obtaining a calibration signal with a stable linear relationship to the element concentration, effectively eliminating the influence of the above three types of interference factors.

[0037] 8. Surface density inversion and thermogram generation The normalized intensity ratio measured in real time Auxiliary parameters such as the current production line speed and batch information are input into a pre-trained physical-guided neural network model. The model performs forward calculations based on the physicochemical laws it has learned internally and outputs the analysis results. Based on the equivalent surface density values ​​of the active material generated at all measurement points, and combined with their spatial coordinates (X, Y), a local mass distribution heatmap covering the scanned area is interpolated. The total loss function calculated by the PINN model inversion is: in, For data fitting terms; This is a monotonic physical constraint loss term; This represents the spatial continuity physical constraint loss term; m The relative weights of the monotonic physical constraint loss term; n The relative weights are the physical constraint loss terms for spatial continuity.

[0038] 9. Data integration and visualization The data processing and intelligent analysis module 8 stitches together multiple frames of local quality heatmaps to form a complete, real-time quality monitoring view that extends continuously along the length of the electrode. Simultaneously, it calculates and updates various indicators on the statistics panel, such as the coefficient of variation (CV%) and the process capability index (CPK). The formula for calculating CV% is: in, Standard deviation This is the sample mean; The formula for calculating CPK is: in, This is the upper limit of the specification. This is the lower limit of the specification.

[0039] The thermal map, statistical data, and testing parameters are transmitted to the human-machine interface 9 via high-speed lines, enabling real-time visualization and allowing operators to intuitively grasp the coating quality status. Furthermore, the human-machine interface can also display the trend curves of various testing indicators (such as areal density, CV%, CPK, etc.) over time or electrode length in real time, allowing operators to intuitively grasp the dynamic changes in coating quality.

[0040] 10. Tiered early warning and closed-loop regulation The system monitors whether key parameters exceed preset safety thresholds. Alarms are categorized as minor deviation, significant deviation, dangerous threshold, and attributable deviation. Minor deviations are indicated by a flashing yellow light on the human-machine interface and a log entry, alerting the operator. In this case, the system does not automatically send control commands; the operator must decide whether to adjust. Potential risks are only logged without alarms. Severe defects trigger a red alarm on the human-machine interface, along with a detailed alarm window indicating the cause and logging the error. In this case, the system does not automatically send control commands and requires manual intervention. If a deviation is detected for ≥3 consecutive scanning cycles with a stable trend attributable to a single factor (such as mold head pressure), it is determined to be attributable deviation. If the automatic closed-loop function is enabled, an orange alarm is triggered on the human-machine interface, and corresponding control commands are generated. The criteria for determining a stable trend are: within three consecutive cycles, the direction of the areal density deviation is consistent (all are lower than the target value), the absolute value of the deviation change does not exceed ±0.3%, and the production line travel distance corresponding to a single scanning cycle is ≤6.7cm; The fault diagnosis model based on the process knowledge base clarifies the automatic attribution logic: Establish a database of typical failures in the coating process: die pressure, slurry viscosity, oven temperature, belt tension, etc. The system matches fault types based on the spatial distribution and trend characteristics of the deviation: lateral unilateral deviation is determined to be die head pressure; uniform deviation across the entire width is determined to be slurry viscosity; and periodic fluctuations are determined to be belt tension. Without requiring manual input, it automatically locates specific factors such as "drill head pressure" through feature matching, meeting the requirements of real-time closed-loop control.

[0041] The control commands are sent directly to the PLC (Programmable Logic Controller) of the coating machine through industrial Ethernet communication protocols (such as OPC UA, Modbus TCP / IP, etc.) so that the system can adjust automatically.

[0042] 11. Circular online detection During the continuous movement of the positive and negative electrode sheets, steps three through eleven are repeated to form a continuous, high-speed online detection and feedback closed loop until the entire roll of positive and negative electrode sheets is inspected or the production line stops.

[0043] Example 1: Application of the system in an NCM ternary cathode production line (1) System startup and calibration: Turn on all devices; the system self-test passes.

[0044] Three standard NCM cathode samples, whose Li content had been accurately determined and whose average active material areal density calculated using offline inductively coupled plasma atomic emission spectrometry (ICP-AES), were used for calibration. Their LiBS signals were acquired to establish initial... The calibration curve with areal density is used to initialize the PINN model.

[0045] (2) Online detection process: When the production line starts, the NCM positive electrode sheet passes through the inspection station at a speed of 80 m / min.

[0046] Step A (t=0 ms): The line laser profilometer completes one line scan and obtains the height data of the current section.

[0047] Step B (t=1 ms): The central controller calculates the average defocus amount Δz=+15μm based on the height data and sends a command to the piezoelectric platform to move the lens group upward by 15μm for compensation.

[0048] Step C (t=2 ms): The laser emits a pulse, which is split by the SLM, and 100 sub-beams act synchronously on the surfaces of the positive and negative electrodes, generating 100 plasmas.

[0049] Step D (t=3 ms): The light-collecting fiber bundle captures the composite light and transmits it to the spectrometer. The sCMOS camera completes one exposure and acquires the raw two-dimensional image.

[0050] Step E (t=4-10 ms): Data is transmitted to the industrial computer. Software execution: Spatial decoding → Extraction of 100 spectra → Preprocessing → Calculation of each point .

[0051] Step F (t=11-15 ms): 100 Input the values ​​into the trained PINN model, and the model outputs the predicted "active substance equivalent surface density" at 100 points.

[0052] Step G (t=16-20 ms): The system generates a local heatmap, updates and displays it on the Human-Machine Interface (HMI), and calculates the mean, standard deviation, and CV of the current scan line.

[0053] (3) Example of closed-loop control: 1. During continuous monitoring, the system found that the areal density of the right side (coordinate X of the transverse width of the electrode > 600 mm) was lower than the target value by 1.5% for 5 consecutive sampling points, and the CV% increased from 1.2% to 2.8%.

[0054] 2. After confirming trend stability through 3 scan cycles, where each scan cycle is 50ms (corresponding to a sampling interval of approximately 6.7cm at a linear velocity of 80m / min), the criteria for determining "trend stability" are: within 3 consecutive cycles, the areal density deviation direction is consistent (all below the target value), and the absolute value of the deviation change does not exceed ±0.3%.

[0055] Based on the system's built-in fault diagnosis rules (e.g., areal density is consistently low on one side in the horizontal direction and systematically low in the vertical direction, and similar phenomena in historical data are mostly caused by pressure fluctuations on that side of the die head), the system initially infers that the most likely cause is insufficient pressure on the right side of the die head. This embodiment does not model or eliminate other secondary possibilities (such as partial blockage of the exhaust vent).

[0056] 3. The system's automatic closed-loop control function has been enabled by the operator in the human-machine interface (HMI). The HMI interface issues an orange warning and automatically sends instructions to the coating machine's programmable logic controller (PLC) through the Open Platform Communications Unified Architecture (OPC UA) interface.

[0057] 4. The PLC executes the command to adjust the mold head. After about 30 seconds, the system monitors that the density on the right side has recovered to within ±1.0% of the target value (the average deviation has recovered from -1.5% to -0.2%), and the CV% gradually decreases and stabilizes between 1.2% and 1.3%, returning to normal.

[0058] (4) Implementation effect This embodiment successfully achieved full-width, non-destructive, online detection of active material distribution on a 1.2-meter-wide NCM cathode sheet on a high-speed production line at 80 m / min. The measurement resolution reached 5 mm (lateral) × 10 mm (longitudinal), and the areal density prediction accuracy was better than ±1.5 mg / cm³. 2 (Compared to offline weighing method), it fully meets the production requirements of high-quality power batteries.

[0059] Therefore, this invention employs the aforementioned method and system for analyzing the uniformity of coating on positive and negative electrodes of lithium-ion batteries. Through multi-point synchronous LIBS excitation, dynamic focusing compensation, dual-band self-calibration, and physical-guided neural network joint inversion, it achieves high-speed, non-destructive, online, and full-element quantitative detection. It can simultaneously acquire information on the composition distribution and areal density of the electrode coating, effectively solving problems such as insufficient speed, unstable signal, inability to distinguish components, and reliance on radiation sources in traditional detection technologies. This significantly improves the quality consistency, production efficiency, and intrinsic safety of the lithium battery coating process, meeting the high-precision online quality control requirements for the large-scale manufacturing of high-end power batteries.

[0060] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for analyzing the coating uniformity of positive and negative electrode sheets in a lithium-ion battery, characterized in that, Includes the following steps: S1. The positive and negative electrode sheets pass through the inspection station at the speed set by the production line. S2. The line laser profilometer continuously scans to acquire real-time data on the surface morphology of the electrode. S3. The control system dynamically adjusts the focusing lens group through the piezoelectric drive platform based on the topography data to accurately focus the laser sub-beam; S4. The laser source module emits a pulsed laser beam, which is divided into a matrix of a set number of sub-beams by a spatial light modulator (SLM) to simultaneously form ablation points and generate plasma on the surfaces of the positive and negative electrodes. S5. The optical collection system captures the composite emission light of the plasma and transmits it to the spectral analysis module; S6. The high-speed area array spectrometer performs a single exposure, separating and recording the complete spectrum of all ablation points; S7, the data processing and intelligent analysis module performs spatial decoding, dual-band normalization calibration, PINN model inversion calculation, generates surface density distribution map, coefficient of variation CV%, and process capability index CPK, and performs out-of-tolerance judgment; S8, the human-computer interaction and feedback interface displays the surface density distribution map, CV% and CPK in real time, and responds in stages according to CV% and CPK: yellow light warning when there is a slight deviation, yellow light flashing rapidly when there is a significant deviation; red light alarm when there is a dangerous threshold, and automatic adjustment when the deviation can be attributed and closed loop is enabled.

2. The method for analyzing the coating uniformity of positive and negative electrode sheets of a lithium-ion battery according to claim 1, characterized in that, The dual-band normalization calibration formula in step S7 is: in, For the first Dual-band normalized intensity ratio at each detection point; For the first The integral intensity of the characteristic spectral lines of the target element at each detection point; For the first The integral intensity of the characteristic spectral lines of the reference element at each detection point; The wavelength of the characteristic spectral line of the target element; The characteristic wavelength of the reference element.

3. The method for analyzing the coating uniformity of positive and negative electrode sheets of a lithium-ion battery according to claim 1, characterized in that, The total loss function of the PINN model in step S7 is: in, For data fitting terms; This is a monotonic physical constraint loss term; This represents the spatial continuity physical constraint loss term; m The relative weights of the monotonic physical constraint loss term; n The relative weights are the physical constraint loss terms for spatial continuity.

4. The method for analyzing the coating uniformity of positive and negative electrode sheets of a lithium-ion battery according to claim 1, characterized in that, The formula for calculating CV% in step S7 is: in, Standard deviation This is the sample mean; The formula for calculating CPK is: in, This is the upper limit of the specification. This is the lower limit of the specification.

5. A system for performing the method for analyzing the coating uniformity of positive and negative electrode sheets of a lithium-ion battery according to any one of claims 1-4, characterized in that, include: The system includes a laser source module, a multi-point synchronous excitation and dynamic focusing module, a light collection and spectral analysis module, a data processing and intelligent analysis module, a human-machine interaction and feedback interface, and a timing controller. The timing controller coordinates with each module to complete the timing actions of laser emission, morphology scanning, focus compensation, spectral acquisition, and data processing. The moving positive and negative electrode sheets of the lithium-ion battery under test pass through the detection stations of each module at the speed set by the production line. The laser source module uses a YAG laser to output a laser beam; The multi-point synchronous excitation and dynamic focusing module includes a spatial light modulator (SLM), a focusing lens group, a line laser profilometer, and a piezoelectric driven translation stage. The light collection and spectral analysis module includes an optical collection system and a high-speed area array spectral imaging system; the optical collection system is a light collection fiber bundle matrix, and the high-speed area array spectral imaging system includes a grating spectrometer and a detector integrated into the grating spectrometer; The data processing and intelligent analysis module incorporates a dual-band self-calibration engine and a physically guided neural network model. The human-computer interaction and feedback interface builds a visualization software platform that displays the composition, trend curves, statistical reports, defect alarms and logs within the characteristic spectral range in real time, and sends key parameters to the coating machine PLC system.

6. The system for analyzing the coating uniformity of positive and negative electrode sheets of a lithium-ion battery according to claim 5, characterized in that, The spatial light modulator (SLM) in the multi-point synchronous excitation and dynamic focusing module generates two-dimensional laser matrices of arbitrary shape and number through programming, splitting the laser beam output from the laser source module into laser sub-beams; The focusing lens group focuses the laser sub-beam onto the surface of the moving positive and negative electrode plates of the lithium-ion battery, forming a micron-scale ablation lattice; A line laser profilometer and a piezoelectric-driven translation stage constitute a dynamic focus compensation unit. The line laser profilometer continuously scans the same area of ​​the moving positive and negative electrode sheets of the lithium-ion battery at a speed higher than the main laser frequency, and acquires the electrode surface height map in real time. The control system calculates the distance between the focus of each laser sub-beam and the electrode surface based on the height map, and finely adjusts the focusing lens group through the piezoelectric-driven translation stage to ensure that all laser points are in the optimal focusing state, thereby eliminating signal attenuation caused by foil jitter.

7. The system for analyzing the coating uniformity of positive and negative electrode sheets of a lithium-ion battery according to claim 6, characterized in that, The optical fiber bundle matrix of the optical collection and spectral analysis module is a paraxial acquisition system independent of the laser focusing system, which directly collects the plasma light generated by the laser ablation of the electrode. The grating spectrometer is equipped with a high-performance sCMOS camera. Through fiber bundle arrangement or spatial coding design, the grating spectrometer makes the pixels in a specific area on the detector correspond one-to-one with the laser ablation points on the moving positive and negative electrodes of the lithium-ion battery, thus realizing the coupling of spatial information and spectral information.

8. The system for analyzing the coating uniformity of positive and negative electrode sheets of a lithium-ion battery according to claim 7, characterized in that, The dual-band self-calibration engine of the data processing and intelligent analysis module eliminates common system noise caused by laser energy fluctuations, plasma temperature changes, and differences in collection efficiency by calculating the normalized intensity ratio; the physics-guided neural network model embeds known physical laws as soft constraints into the training process.

9. The system for analyzing the coating uniformity of positive and negative electrode sheets of a lithium-ion battery according to claim 8, characterized in that, The data processing and intelligent analysis module implements hierarchical control based on the thresholds of the coefficient of variation (CV%) and the process capability index (CPK). The threshold for graded regulation is defined as follows: Under normal conditions, the human-computer interaction interface displays a solid green light, shows a real-time surface density distribution map, updates CV% and CPK statistical indicators in real time on the statistics panel, records data in real time, and does not generate alarm information. When there is a slight deviation, the human-machine interface and feedback interface will flash a yellow light to warn the system. The system will automatically generate an alarm record and store the alarm time and alarm reason in the database. The color block of the slightly out-of-tolerance area in the surface density distribution map will be brightened or the border will be highlighted. The operator can judge the type of deviation based on the heat map and trend curve and conduct on-site inspection. When there is a significant deviation, the yellow light flashes rapidly and the alarm is activated. A detailed alarm window pops up, highlighting the performance indicators, the location and duration of heat distribution. The system records data at high frequency and generates a detailed alarm log, displaying modification suggestions for manual inspection and confirmation or manual adjustment. When the indicator reaches or exceeds the danger threshold, the human-machine interaction and feedback interface issues a red alarm and displays the control command generated by the physical guided neural network PINN; when automatic control is turned on, the system automatically writes the control command directly to the coating machine PLC after a 5-second countdown, without manual confirmation, and records the closed-loop intervention log completely; when automatic control is turned off, the window continuously alarms and highlights the prompt, and the operator immediately stops the machine manually to check. When the system continuously detects a certain attributable deviation pattern, it automatically generates a correction command and sends it directly to the PLC of the coating machine via the industrial communication protocol, enabling the system to achieve automatic closed-loop control.

Images