Battery production whole process data monitoring system and method

Abstract

The invention relates to the technical field of data monitoring, in particular to a battery production whole-process data monitoring system and method. The method comprises the following steps: acquiring real-time multi-source data of a battery production line and an assembly station image; analyzing the current abrupt change amplitude and the regional temperature difference standard deviation of the real-time multi-source data of the battery production line; identifying a production stage boundary point of the assembly station image, and performing production process division on the battery by using the current abrupt change amplitude and the regional temperature difference standard deviation to generate battery production process data; and an electrode slurry preparation stage of extracting battery production process data, and performing quantum dot sensor array deployment on the slurry stirring kettle based on the electrode slurry preparation stage to obtain sensor array deployment data. According to the invention, multi-source real-time data fusion, quantum dot sensor array spectrum decoding, three-dimensional particle state reconstruction and digital twinning technologies are utilized, so that the monitoring precision, the abnormity identification capability and the intelligent collaborative response level of the whole battery production process are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data monitoring, and particularly relates to a battery production whole-process data monitoring system and method. BACKGROUND

[0002] Early battery production mainly relies on manual recording and offline analysis, and the data acquisition means is single and time-limited, which is difficult to reflect the production state in time. Subsequently, sensors monitoring and PLC control are introduced into the automatic production line to realize real-time data acquisition at some key nodes, but there is still a lack of overall perception and dynamic response capability for the whole production process. In recent years, with the rise of technologies such as Internet of Things, edge computing and big data analysis, enterprises have gradually built a digital monitoring system covering multiple links such as electrode preparation, slurry stirring, coating, rolling, cutting, assembly, liquid injection and packaging. Through multi-source data fusion, image recognition, AI anomaly detection and other means, the transformation from single-point monitoring to whole-process coordination has been gradually realized. However, at present, as the core link, the existing methods are difficult to capture the micro-particle evolution and agglomeration behavior of the electrode slurry preparation, and most of them rely on offline sampling, which is seriously lagging behind. Moreover, although the traditional system can monitor local data, it lacks spatial mapping of the overall production process and abnormal coordination mechanism, and cannot quickly respond to the influence of slurry abnormalities on subsequent processes. SUMMARY

[0003] Therefore, it is necessary to provide a battery production whole-process data monitoring system and method to solve at least one of the above technical problems.

[0004] To achieve the above-mentioned purpose, a battery production whole-process data monitoring method is provided, and the method comprises the following steps:

[0005] Step S1: acquiring real-time multi-source data of a battery production line and an assembly station image; analyzing the current mutation amplitude and the regional temperature difference standard deviation of the real-time multi-source data of the battery production line; identifying the production stage boundary point of the assembly station image, and dividing the production process of the battery by using the current mutation amplitude and the regional temperature difference standard deviation to generate battery production process data;

[0006] Step S2: extracting the electrode slurry preparation stage of the battery production process data, and deploying a quantum dot sensor array on a slurry stirring kettle based on the electrode slurry preparation stage to obtain sensor array deployment data; decoding the spectral response spectrum by using the sensor array deployment data to generate slurry state spectral feature data;

[0007] Step S3: generating slurry agglomeration behavior data by using the slurry state spectral feature data to perform micro-particle state evolution on the electrode slurry preparation stage; reconstructing the three-dimensional distribution of the slurry agglomeration behavior data, and generating slurry abnormality monitoring data based on the reconstruction result to monitor the micro-unbalance of the slurry stirring;

[0008] Step S4: based on the battery production process data, the digital twin scene is constructed, and battery production process three-dimensional monitoring data is generated; according to the slurry abnormal monitoring data, the battery production process three-dimensional monitoring data is cooperated with the battery production data to perform the battery production whole-process data monitoring operation.

[0009] The application can accurately identify the process boundary of each stage of battery production by collecting real-time multi-source data (such as current and temperature) of the battery production line and combining with the assembly station image, using current mutation amplitude and regional temperature difference standard deviation for production stage division, avoiding the problem of relying on manual judgment or single parameter for production node division in traditional methods, and improving the accuracy and automation level of process identification. In the electrode slurry preparation stage, a quantum dot sensor array is deployed, and high-dimensional slurry state spectral feature data is obtained through spectral response spectrum decoding, effectively enhancing the monitoring capability of the microphysical and chemical state of the slurry, and breaking through the limitation of traditional sensors in sensitivity and spectral resolution. Through the micro-particle state evolution model driven by spectral feature data, the change trend of the slurry agglomeration behavior can be dynamically tracked, and the spatial modeling of the slurry stirring state is realized through three-dimensional distribution reconstruction, effectively identifying the micro-unbalanced problems caused by uneven stirring efficiency or local retention in the stirring process, and realizing early warning of slurry abnormalities. Combined with the battery production process data, a digital twin scene is constructed, and is cooperated with the slurry abnormal monitoring data, which can realize real-time visualization of the production state of each stage in three-dimensional space, realize real-time monitoring, intelligent decision and closed-loop control of the whole process of battery production, and significantly improve the controllability and flexibility of production. Through the real-time monitoring and feedback control mechanism of micro-unbalance, abnormalities can be found early in the electrode slurry preparation stage, timely intervention can be made, electrode defects caused by slurry agglomeration or uneven stirring can be effectively avoided, product quality can be guaranteed from the source, yield can be improved, and rework rate and material waste can be reduced. The method adopts a modular architecture design, which can be flexibly adapted to different types of battery production lines, and is convenient for integration with existing MES systems, SCADA systems or quality traceability systems, and has good engineering landing performance and industrial application prospect. Therefore, by using multi-source real-time data fusion, quantum dot sensor array spectral decoding, three-dimensional particle state reconstruction and digital twin technology, the monitoring accuracy, abnormal identification capability and intelligent collaborative response level of the whole process of battery production are improved.

[0010] Preferably, step S1 comprises the following steps:

[0011] Step S11: acquiring real-time multi-source data of the battery production line and assembly station images;

[0012] Step S12: calculating the current mutation amplitude and the regional temperature difference standard deviation according to the real-time multi-source data of the battery production line, and generating current-temperature difference joint feature data;

[0013] Step S13: edge contour enhancement and structure region extraction are performed on the assembly station image to generate station image structure feature data; a typical component assembly action sequence is recognized according to the station image structure feature data, and production stage boundary point data is generated;

[0014] Step S14: process behavior time window segmentation is performed on the production stage boundary point data by using current-temperature difference joint feature data to generate process feature slice data; key indicator clustering and dynamic stage alignment processing are performed on the process feature slice data to generate a battery production process data.

[0015] The present application realizes deep coupling analysis of process and image data by fusing real-time multi-source process data (current, temperature) and assembly station images, avoids the uncertainty of traditional process identification relying on a single data source, and improves the comprehensiveness and robustness of identification. In step S12, the "current-temperature difference joint feature data" constructed fuses two important process change dimensions: current mutation reflects energy consumption fluctuation, and temperature difference standard deviation reveals thermal field abnormality, which provides stronger process stage resolution capability, helps to accurately capture production behavior change points, and improves the sensitivity to process nodes. Through image structure enhancement and typical assembly action recognition (such as welding, crimping, detection, etc.), production stage boundary point data is generated, which effectively improves the time accuracy and spatial matching degree of production stage division, and avoids the stage deviation or misjudgment caused by relying only on sensing data. In step S14, process behavior time window segmentation and key indicator clustering + dynamic stage alignment processing are proposed to form stable and generalizable process feature slices, effectively solving the problems of inconsistent process length, beat offset, etc., and improving the time alignment and stage consistency of battery production process modeling.

[0016] Preferably, the quantum dot sensor array deployment based on the electrode slurry preparation stage in step S2 includes:

[0017] The quantum dot sensor array deployment based on the electrode slurry preparation stage is performed on the slurry stirring tank to obtain sensor array deployment data, wherein the slurry temperature is controlled at 15 to 80℃, the viscosity range is 500 to 5000 mPa·s, the particle size distribution is 50 to 800 nm, the fluorescence response wavelength is concentrated in 500 to 650 nm, and the fluorescence intensity sensitivity is 10 2 to 10 5 Relative units, the pH value of the slurry is set to be between 6.0 and 9.5, the conductivity range is 0.1 to 5.0 S / m, the vibration acceleration is set to be 0.5 to 3.0 g, the sensor array sampling frequency is set to be 10 to 100 Hz, and the array distribution density is 4 to 16 sensing points per square meter.

[0018] The application realizes all-round perception of the physical and chemical state of the slurry in the stirring process by setting the monitoring range including temperature, viscosity, particle size distribution, pH value, conductivity, vibration acceleration and other key process parameters. The real-time perception ability of this multi-dimensional parameter far exceeds the traditional single physical quantity monitoring method, providing a data basis for slurry quality control. The deployed quantum dot sensor array has a fluorescence response wavelength concentrated in 500 to 650 nm, a fluorescence intensity sensitivity of 10 2 to 10 5 Relative units can accurately respond to changes in the state of micro-particles in the slurry, such as agglomeration, dispersion or interfacial reaction, greatly improving the dynamic monitoring capability of microstructure evolution and effectively breaking through the limitations of traditional sensors in nanoscale perception. The sampling frequency of the sensor array is set to 10 to 100 Hz, and the spatial distribution density is 4 to 16 sensing points per square meter, which can realize spatial uniform coverage and high-frequency dynamic data acquisition, ensuring continuous tracking of physical disturbance, particle flow and thermal changes in the whole cycle of slurry stirring, and improving the integrity and real-time of data acquisition. Set a reasonable slurry working parameter range (such as temperature 1580℃, viscosity 500-5000mPa·s, pH value 6.0-9.5, conductivity 0.1-5.0S / m, particle size 50~800nm), ensure that the sensor deployment environment is in the stable response range of quantum dot fluorescence, avoid signal drift, fluorescence quenching or data distortion due to extreme working conditions, and ensure long-term stable operation of the sensor performance. The real-time output of high-frequency high-density sensor array data and fluorescence spectrum response provides accurate input for subsequent slurry state visualization modeling, stirring uniformity analysis, agglomeration behavior identification, etc., and supports the slurry preparation regulation algorithm driven by digital twin, realizing the closed-loop control capability from "perception→analysis→regulation".

[0019] Preferably, the spectral response spectrum decoding of the sensor array deployment data in step S2 includes:

[0020] Extracting the multi-channel response signal of the sensor array deployment data;

[0021] Performing time domain synchronization and frequency spectrum normalization on the multi-channel original response data to generate standard response spectrum data;

[0022] Performing spectrum difference analysis and high-dimensional noise reduction on the standard response spectrum data to generate effective spectrum response interval data;

[0023] Decoding the composition sensitive band according to the effective spectrum response interval data to generate a slurry composition reflectance matrix;

[0024] Extracting the principal component features of the slurry composition reflectance matrix, and calibrating the chemical absorption peaks of the multi-channel response signal according to the principal component features to generate slurry state spectrum feature data.

[0025] The present application effectively eliminates the signal distortion problem caused by sampling delay, sensor sensitivity drift, etc. between channels by time domain synchronization and spectral normalization of the multi-channel original response signal collected by the quantum dot sensor array, guarantees the time-frequency consistency and feature comparability of the data used for decoding analysis, and improves the accuracy and repeatability of the decoding result. The introduction of spectral band difference analysis and high-dimensional noise reduction processing can extract effective wavebands with actual component response significance from redundant and complex multi-channel spectral graphs, reduce noise interference and information redundancy, improve the operation efficiency and stability of the decoding model in the subsequent steps, and effectively guarantee the complete preservation of key information in spectral graph processing. By analyzing the effective spectral response interval, identifying the component sensitive spectral band significantly related to the key material components of the slurry (such as binders, conductive agents, active substances, etc.), and decoding to generate a reflection coefficient matrix, the differentiation of different physical components in the multi-channel spectral space can be realized, providing physical support and spectral basis for dynamic identification of slurry components. By extracting the principal component features of the reflection coefficient matrix, capturing the main change patterns in the evolution of the slurry state, and combining the chemical absorption peak position to calibrate the multi-channel response signal, the real-time changes of the micro-particle state, distribution uniformity and chemical reaction of the slurry in the stirring or dispersion process can be finely described, significantly improving the spectral perception ability of the physical-chemical state evolution of the slurry. The generated slurry state spectral feature data as a high-dimensional multi-scale state representation result not only has good representativeness and discrimination, but also can be used as input data for key links such as digital twin modeling, anomaly identification, and agglomeration behavior prediction, forming an efficient bridge from "physical monitoring" to "data cognition".

[0026] Preferably, the micro-particle state evolution of the electrode slurry preparation stage by the slurry state spectral feature data in step S3 includes:

[0027] Performing waveband reflectivity fitting on the slurry state spectral feature data to generate multi-scale waveband reflectivity curve data;

[0028] Performing particle size response mapping on the multi-scale waveband reflectivity curve data to generate particle size response distribution data;

[0029] Performing time window slicing on the particle size response distribution data to generate time-series particle evolution segment data;

[0030] Identifying the particle cluster aggregation trend of the time-series particle evolution segment data;

[0031] Analyzing the particle spacing of the particle cluster aggregation trend data to generate particle spacing distribution data;

[0032] Performing threshold comparison on the particle spacing distribution data based on a preset particle spacing threshold, and when the particle spacing distribution data is greater than or equal to the preset particle spacing threshold, the particle spacing distribution data is marked as agglomeration data.

[0033] spatial density inversion is performed on the agglomeration data to generate micro-agglomeration concentration layer data;

[0034] The micro-agglomeration concentration layer data after spatio-temporal fusion is used to perform stability dispersion analysis on the electrode slurry preparation stage, and finally slurry agglomeration behavior data is generated.

[0035] The present application can indirectly deduce the particle size distribution change by performing band reflectivity fitting and particle size response mapping on the slurry state spectrum characteristic data without directly performing complex physical particle imaging or particle size instrument intervention, significantly improving the real-time analysis capability and perception efficiency of slurry micro-behavior, and providing a new idea for non-invasive quality monitoring. By performing time window slicing on the particle size response mapping result, time sequence particle evolution fragment data is constructed, which not only captures the dynamic trend of particle aggregation and dispersion in the slurry stirring process, but also provides evolution path and periodic analysis basis for subsequent particle agglomeration mode recognition and aggregation strength evaluation. Based on the particle spacing distribution data, and combined with the preset agglomeration threshold for comparison, a scientific and objective agglomeration discrimination logic is constructed, which effectively solves the problem that the traditional particle size statistics cannot accurately describe the degree of "agglomeration" phenomenon, so that the slurry agglomeration recognition is more quantitative, regularized and engineered. Through spatial density inversion on the agglomeration data, the micro-agglomeration concentration layer data reflects the distribution density and strength of the agglomeration region in space; further spatio-temporal fusion and stability dispersion analysis can construct a high-resolution three-dimensional agglomeration evolution map, significantly enhancing the visualization and traceability of the slurry preparation process. Based on the spatio-temporal analysis of the agglomeration behavior, the slurry agglomeration behavior data formed finally can be used to reflect the overall stability, dispersion and micro-mixing uniformity of the slurry system, solving the bottleneck of the previous problem of being difficult to identify micro-agglomeration only relying on macro-viscosity or naked eye judgment, and greatly enhancing the scientificity and sensitivity of micro-state quality control.

[0036] Preferably, the three-dimensional distribution reconstruction of the slurry agglomeration behavior data in step S3 comprises:

[0037] The spatial coordinate index of the slurry agglomeration behavior data is extracted to obtain agglomeration spatial coordinate point set data;

[0038] Point cloud densification interpolation is performed on the agglomeration spatial coordinate point set data to generate particle agglomeration point cloud data;

[0039] The particle agglomeration point cloud data is locally voxelized and coded, and the coded particle agglomeration point cloud data is globally grid spliced and reconstructed to generate slurry agglomeration three-dimensional grid skeleton data;

[0040] Particle size channel attribute mapping is performed on the slurry agglomeration three-dimensional grid skeleton data to generate attribute annotated grid body data;

[0041] Perform multi-view projection rendering on the attribute annotation mesh data to generate particle agglomeration 3D view data;

[0042] Based on the particle agglomeration 3D view data, dynamic sequence frame encoding is performed to generate speed-adjustable particle agglomeration evolution sequence data;

[0043] The particle agglomeration evolution sequence data is processed by stability hotspot annotation to obtain the reconstruction result of three-dimensional distribution reconstruction.

[0044] The present invention reconstructs the dense distribution characteristics of agglomerated particles in the stirring space by extracting the spatial coordinate index of the slurry agglomeration behavior data and performing point cloud densification interpolation, which significantly improves the spatial reconstruction accuracy and local aggregation recognition ability of microscopic agglomeration behavior, and lays the foundation for high-fidelity visualization in subsequent digital twin scenarios. Based on the three-dimensional grid skeleton data generated by local voxel encoding and global grid splicing reconstruction, the slurry agglomeration morphology can be represented in a continuous and connected structure; further introduction of particle size channel attribute mapping and attribute annotation grid body construction can give the three-dimensional grid semantic attributes at the particle level, realizing the leap from "geometric reconstruction" to "physical semantic annotation", and enhancing the depth of understanding of particle agglomeration behavior. Multi-perspective projection rendering of the attribute annotated grid body data can generate realistic particle agglomeration three-dimensional view data, so that the complex agglomeration evolution process can be intuitively displayed at the visual level, which is convenient for operators and system engineers to conduct morphological observation, local tracking and comparative analysis, and effectively improves the interpretability of data-driven judgment. By encoding 3D view data into a speed-adjustable dynamic sequence of frames, particle agglomeration evolution sequence data is generated. This dynamically displays the onset, development, and decay of particle agglomeration behavior at different stirring stages, demonstrating the evolutionary trajectory and mechanism trends of agglomeration from non-existence to formation, from weak to strong, and then to gradual dispersion. Stability hotspot annotation based on evolutionary sequence data automatically identifies unstable hotspots, persistent aggregation hotspots, or areas of high cluster concentration during the agglomeration process, and outputs visual annotation information with risk warning functions, providing a key technical basis for real-time control, parameter adjustment, and stirring strategy intervention.

[0045] Preferably, the slurry stirring micro-imbalance monitoring based on the reconstruction result in step S3 includes:

[0046] Extract the particle structure orientation information in the stirring area based on the reconstruction results of the three-dimensional distribution reconstruction to generate local structure orientation data;

[0047] Perform spatial anisotropy tensor analysis on local structural orientation data to generate orientation tensor deviation data;

[0048] The orientation tensor deviation data is used to perform vector trajectory deduction on the particle flow trend during the electrode slurry preparation stage to generate microscopic flow disturbance data.

[0049] According to the micro flow disturbance data, the local shear stress gradient in the stirring process is calculated to generate a shear imbalance intensity atlas;

[0050] The stability domain boundary of the shear imbalance intensity atlas is identified, and the particle aggregation time window statistics of the stability domain boundary are performed to generate particle group fluctuation interval data;

[0051] The micro disturbance frequency analysis is performed on the particle group fluctuation interval data to finally generate slurry abnormality monitoring data.

[0052] The present application can accurately depict the spatial organization situation of particle arrangement and aggregation in the stirring area by extracting particle structure orientation information from the three-dimensional reconstruction result and generating local structure orientation data, and solves the problem of insufficient perception of orientation change and spatial imbalance in traditional slurry state monitoring, and provides accurate basis for subsequent anisotropy modeling. The structure orientation data is analyzed by spatial anisotropy tensor, and the orientation tensor deviation degree data is extracted, which can quantitatively evaluate the inconsistency and local orientation disorder phenomenon of particle arrangement in different directions, form a high-sensitivity index system for the non-uniformity of microstructure in the stirring area, and significantly improve the abnormality monitoring resolution. Based on the tensor deviation degree, the vector trajectory of the particle flow trend is deduced, the disturbance direction and bending trend of the particle motion in the slurry can be reconstructed, and the micro flow disturbance data can be obtained, so that the particle behavior deviation caused by asymmetric stirring, particle aggregation or shear difference in the local area can be revealed. The local shear stress gradient calculated according to the micro disturbance path forms a shear imbalance intensity atlas, which can clearly reveal the shear difference in different stirring areas, accurately identify typical problem areas such as "shear dead angle" and "stirring flow bias" in the stirring process, and provide reference for stirring parameter calibration and equipment structure optimization. By identifying the stability domain boundary in the shear atlas and performing particle aggregation time window statistics, the particle group fluctuation interval data generated can further reveal the behavior law and concentrated occurrence time of the agglomerated particle group in the stirring cycle, support agglomeration warning, disturbance prediction and process optimization scheduling. Finally, the micro disturbance frequency analysis is performed on the particle group fluctuation interval data, which can effectively identify the high-frequency oscillation area and periodic abnormal aggregation phenomenon, accurately generate slurry abnormality monitoring data, realize dynamic real-time feedback of the stability of the electrode slurry preparation, and has good process control foresight and reaction sensitivity.

[0053] Preferably, the micro disturbance frequency analysis on the particle group fluctuation interval data comprises:

[0054] When any of the following conditions occurs, the particle size concentration abnormality data is obtained: the characteristic particle size distribution range narrows by more than 20%, the D90 / D10 particle size ratio fluctuates by more than ±0.6, and the particle size distribution skewness index deviates from the average value by more than ±15% in three consecutive sampling periods.

[0055] When the following conditions occur simultaneously, the micro-agglomeration disturbance enhancement phenomenon is determined, and the agglomeration disturbance enhancement data is obtained: the number ratio of agglomerates in the particle group increases by more than 25% within 10 minutes, the average value of the agglomerate particle size exceeds the upper limit of the history by more than 10%, the internal density distribution of the agglomerates presents a double-peak deviation and the duration is more than 20 minutes;

[0056] When the following conditions are met simultaneously, the shear wave fluctuation abnormal state is determined, and the shear wave fluctuation abnormal data is obtained: the stirring system speed fluctuation frequency is more than 2Hz, the particle group micro-disturbance response frequency deviates from the reference spectrum line range by more than ±10% during stirring, the micro-bubble content in the slurry increases by more than 8%, and the duration of this state is more than 30 minutes;

[0057] The particle size concentration abnormality data, the agglomeration disturbance enhancement data and the shear wave fluctuation abnormal data are integrated, and the correlation weight evaluation and disturbance type fusion recognition are performed to finally generate the slurry abnormality monitoring data.

[0058] The present application realizes automatic abnormal identification of multi-scale particle group behavior on the basis of data driving by setting multiple specific and quantifiable judgment conditions (such as particle size distribution narrowing range, D90 / D10 ratio, particle size skewness, agglomerate quantity and density characteristics, stirring frequency and micro-bubble increase, etc.), which effectively overcomes the defects of relying on a single parameter and high misjudgment rate in traditional abnormal identification. By dividing the abnormal behavior in the slurry into three categories: particle size concentration abnormality (reflecting abnormal concentration of particle size distribution), agglomeration disturbance enhancement phenomenon (reflecting sharp increase of agglomeration tendency), and shear fluctuation abnormal state (reflecting imbalance between shear system and micro-response), this mechanism classification system helps to clarify the source and behavior pattern of different abnormal causes in the slurry preparation process, and provides type-based strategy support for precise intervention and system response. Through correlation weight evaluation and disturbance type fusion identification of the three types of abnormal data, not only the collaborative analysis and dynamic superposition judgment of different types of disturbance are realized, but also the global disturbance intensity judgment result is formed by comprehensively considering multiple index weights, which enhances the comprehensive judgment ability of the monitoring system for complex disturbance events. The coupling analysis between agglomeration behavior (such as particle size, density, quantity) and stirring behavior (such as frequency fluctuation, bubble content) in the method can provide early warning when agglomeration is still in the "enhancement" stage rather than "serious out-of-control" stage, realize the process control mechanism of early identification, early adjustment and early intervention, and effectively avoid subsequent quality problems such as agglomerate deposition and uneven coating. By embedding time conditions such as "continuous 3 sampling periods", "increase within 10 minutes" and "last more than 30 minutes" into abnormal judgment, it is ensured that the method not only applies to sudden abnormalities, but also identifies slow-acting and gradual process imbalance states, forming a more complete time sequence diagnosis ability for the dynamic behavior of the slurry system.

[0059] Preferably, step S4 comprises the following steps:

[0060] Step S41: extract each assembly station parameter, key material flow direction and process timeline in the battery production process data for spatial mapping, and generate structured digital scene data of the battery production process;

[0061] Step S42: three-dimensional visualization modeling is performed on the structured digital scene data to generate three-dimensional monitoring data of the battery production process; according to the slurry abnormal monitoring data, the corresponding electrode slurry preparation stage in the three-dimensional monitoring data is precisely positioned in time period to generate abnormal mapping section data;

[0062] Step S43: parameter cross comparison is performed on the abnormal mapping section data and the three-dimensional monitoring data to generate visual abnormal marker layer data; the visual abnormal marker layer data is superimposed on the three-dimensional monitoring data of the battery production process to generate fusion monitoring data with abnormal feedback information;

[0063] Step S44: track the data flow of the fusion monitoring data at the workstation level, identify the potential conduction path of the slurry abnormality to the downstream process nodes, and generate a process-level abnormality conduction map; use the process-level abnormality conduction map to perform data collaboration update of the three-dimensional monitoring data of the battery production process, and finally realize full-process data monitoring operation of the battery production.

[0064] The present application can generate structured digital scene data by extracting assembly station parameters, material flow direction and timeline information, and performing spatial mapping and structure modeling, forming a virtual-real mapping digital base of the real battery production process, providing a basic platform for multi-source data fusion, time sequence process presentation and state synchronization analysis. By using slurry abnormality monitoring data to map and accurately position the corresponding period in the three-dimensional monitoring scene, the local imbalance phenomenon hidden in the microscopic data can be accurately projected into the visual space model, realizing real-time dynamic feedback of the electrode slurry preparation stage. Through parameter cross comparison and layer marking, the abnormal events are visualized and superimposed in the three-dimensional monitoring data in the form of color coding, regional highlighting or icon prompts, etc., to build an abnormal visual feedback interface, improve the intuitive perception ability and quick response efficiency of the operator to abnormal trends. By tracking the data flow of the fusion monitoring data at the workstation level, the path of the slurry abnormality from the preparation source to the downstream key nodes such as coating, compaction and lamination can be identified, and a process-level abnormality conduction map is formed, so that the abnormality monitoring is no longer limited to single-point identification, but is extended to full-process risk prediction and chain response. Using the abnormality conduction map to perform local update and full-process associated adjustment of the three-dimensional monitoring data, a system with closed-loop logic of "abnormality identification→path positioning→impact feedback→data update" is built, which has good self-evolution and dynamic adjustment ability, and is suitable for complex production environment under variable working conditions.

[0065] In the present specification, a battery production full-process data monitoring system is provided for executing the above-mentioned battery production full-process data monitoring method, which comprises:

[0066] A process division module is used to acquire real-time multi-source data of a battery production line and assembly station images; analyze the current mutation amplitude and regional temperature difference standard deviation of the real-time multi-source data of the battery production line; identify the production stage boundary points of the assembly station images, and divide the battery production process using the current mutation amplitude and regional temperature difference standard deviation, to generate battery production process data;

[0067] A slurry analysis module is used to extract the electrode slurry preparation stage of the battery production process data, and deploy a quantum dot sensor array on the slurry agitator based on the electrode slurry preparation stage, to obtain sensor array deployment data; decode the spectral response spectrum graph through the sensor array deployment data, to generate slurry state spectral feature data;

[0068] An imbalance monitoring module is configured to monitor the micro-particle state evolution of the electrode slurry preparation stage through the slurry state spectral feature data, generate slurry agglomeration behavior data, reconstruct the three-dimensional distribution of the slurry agglomeration behavior data, and perform slurry stirring micro-unbalance monitoring based on the reconstruction result to generate slurry abnormality monitoring data.

[0069] A data coordination module is configured to construct a digital twin scene based on the battery production process data to generate three-dimensional monitoring data of the battery production process, and perform battery production data coordination on the three-dimensional monitoring data of the battery production process according to the slurry abnormality monitoring data to perform battery production full-process data monitoring work.

[0070] The beneficial effects of the present application are that through the process division module, the current mutation amplitude and the regional temperature difference standard deviation and other multi-source real-time data are linked and analyzed with the stage boundary points extracted from the assembly station image, the automatic process time window division and accurate process label labeling of each stage of battery production are realized, compared with the traditional stage division method relying on human experience, the division accuracy and dynamic adaptability are significantly improved. The slurry analysis module deploys a quantum dot sensor array, sets the physical and chemical properties of the slurry (temperature, pH, conductivity, particle size distribution, fluorescence response, etc.), and performs spectral decoding and principal component feature extraction on the multi-channel response signal, which can monitor the slurry state with high sensitivity and high timeliness, and solve the pain points of traditional slurry monitoring such as "slow response, little information, and low accuracy". The imbalance monitoring module maps the slurry state spectral data to the particle size evolution process, and combines the three-dimensional distribution reconstruction of the agglomeration behavior, and finally through the orientation tensor analysis, the shear imbalance calculation and the disturbance frequency identification, a complete chain from microscale to macro performance to imbalance identification is constructed, which can effectively early warning key risks such as agglomeration concentration and abnormal stirring, and improve the consistency control ability of the slurry. Through the data coordination module, the slurry abnormality monitoring result is dynamically linked with the three-dimensional visual digital twin scene, a visual abnormality layer is generated in the three-dimensional monitoring data, and the abnormal section is tracked and updated at the process level, forming a closed-loop production monitoring system with abnormal feedback, autonomous update and dynamic coordination ability, which significantly enhances the whole process transparency and risk intervention ability. The four modules in the present scheme cooperate from four aspects of "process identification-state analysis-risk monitoring-data coordination", which breaks through the whole process from front-end identification, process analysis, risk identification to back-end monitoring feedback in the battery production, and has high engineering practical value, effectively promoting the evolution of the battery manufacturing system to intelligent, adaptive and traceable. Therefore, the present application improves the monitoring accuracy, abnormality identification ability and intelligent coordination response level of the whole process of battery production by using multi-source real-time data fusion, quantum dot sensor array spectral decoding, three-dimensional particle state reconstruction and digital twin technology. BRIEF DESCRIPTION OF DRAWINGS

[0071] Figure 1A flowchart showing the steps of a method for monitoring data throughout the entire battery production process;

[0072] Figure 2 for Figure 1 Detailed implementation steps of step S1 in FIG.

[0073] Figure 3 for Figure 1 Detailed implementation steps of step S4 in FIG.

[0074] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0075] The following is a clear and complete description of the technical method of the present invention in conjunction with the accompanying drawings. It is obvious that the embodiments described are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making any creative efforts are within the scope of protection of the present invention.

[0076] In addition, the accompanying drawings are merely schematic illustrations of the present invention and are not necessarily drawn to scale. Identical reference numerals in the figures denote identical or similar parts, and thus repetitive descriptions thereof will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities that do not necessarily correspond to physically or logically separate entities. These functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor and / or microcontroller approaches.

[0077] It should be understood that although the terms "first," "second," and the like may be used herein to describe various elements, these elements should not be limited by these terms. These terms are used solely to distinguish one element from another. For example, a first element may be referred to as a second element, and similarly, a second element may be referred to as a first element, without departing from the scope of the exemplary embodiments. The term "and / or" as used herein includes any and all combinations of one or more of the listed associated items.

[0078] To achieve this, please refer to Figures 1 to 3 A method for monitoring data of the entire battery production process, comprising the following steps:

[0079] Step S1: Obtain real-time multi-source data of the battery production line and assembly station image; analyze the current mutation amplitude and regional temperature difference standard deviation of the real-time multi-source data of the battery production line; identify the production stage boundary point of the assembly station image, and divide the battery production process using the current mutation amplitude and regional temperature difference standard deviation, to generate battery production process data;

[0080] Step S2: Extract the electrode slurry preparation stage of the battery production process data, and deploy a quantum dot sensor array on the slurry stirring tank based on the electrode slurry preparation stage to obtain sensor array deployment data; decode the spectral response spectrum through the sensor array deployment data to generate slurry state spectral feature data;

[0081] Step S3: Evolve the micro-particle state of the electrode slurry preparation stage through the slurry state spectral feature data to generate slurry agglomeration behavior data; reconstruct the three-dimensional distribution of the slurry agglomeration behavior data, and monitor the slurry stirring micro-unbalance based on the reconstruction result to generate slurry anomaly monitoring data;

[0082] Step S4: Construct a digital twin scene based on the battery production process data to generate battery production process three-dimensional monitoring data; coordinate the battery production data according to the slurry anomaly monitoring data to perform battery production full-process data monitoring work.

[0083] The application can accurately identify the process boundary of each stage of battery production by collecting real-time multi-source data (such as current and temperature) of the battery production line and combining the assembly station image, using the current mutation amplitude and regional temperature difference standard deviation to divide the production stage, avoiding the problem of relying on manual judgment or single parameter in the traditional method to divide the production node, and improving the accuracy and automation level of process identification. In the electrode slurry preparation stage, a quantum dot sensor array is deployed, and the high-dimensional slurry state spectral feature data is obtained through spectral response spectrum decoding, effectively enhancing the monitoring ability of the microphysical and chemical state of the slurry, and breaking through the limitation of traditional sensors in sensitivity and spectral resolution. Through the micro-particle state evolution model driven by spectral feature data, the change trend of the slurry agglomeration behavior can be dynamically tracked, and the spatial modeling of the slurry stirring state is realized through three-dimensional distribution reconstruction, effectively identifying the micro-unbalanced problems caused by uneven stirring efficiency or local retention in the stirring process, and realizing early warning of slurry abnormalities. Combined with the battery production process data, a digital twin scene is constructed, and the production state of each stage in the three-dimensional space can be visualized in real time, realizing real-time monitoring, intelligent decision-making and closed-loop control of the whole process of battery production, and significantly improving the controllability and flexibility of production. Through real-time monitoring and feedback control mechanism of micro-unbalance, abnormalities can be found in the early stage of electrode slurry preparation, and timely intervention can be made to effectively avoid electrode defects caused by slurry agglomeration or uneven stirring, ensuring product quality from the source, improving yield, and reducing rework rate and material waste. The method adopts a modular architecture design, which can be flexibly adapted to different types of battery production lines, and is also convenient for integration with existing MES systems, SCADA systems or quality traceability systems, and has good engineering landing performance and industrial application prospect. Therefore, by using multi-source real-time data fusion, quantum dot sensor array spectral decoding, three-dimensional particle state reconstruction and digital twin technology, the monitoring accuracy, abnormal identification ability and intelligent collaborative response level of the whole process of battery production are improved.

[0084] In the embodiment of the application, as shown in the reference Figure 1 The battery production process data monitoring method includes the following steps:

[0085] Step S1: Obtain real-time multi-source data of the battery production line and assembly station images; analyze the current mutation amplitude and regional temperature difference standard deviation of the real-time multi-source data of the battery production line; identify the production stage boundary points of the assembly station images, and divide the battery production process by using the current mutation amplitude and the regional temperature difference standard deviation to generate battery production process data;

[0086] Step S2: extract the electrode slurry preparation stage of the battery production process data, and deploy a quantum dot sensor array to the slurry stirring tank based on the electrode slurry preparation stage to obtain sensor array deployment data; decode the spectral response spectrum through the sensor array deployment data to generate slurry state spectral feature data;

[0087] Step S3: evolve the micro-particle state of the electrode slurry preparation stage through the slurry state spectral feature data to generate slurry agglomeration behavior data; reconstruct the three-dimensional distribution of the slurry agglomeration behavior data, and monitor the micro-unbalance of the slurry stirring based on the reconstruction result to generate slurry anomaly monitoring data;

[0088] Step S4: construct a digital twin scene based on the battery production process data to generate battery production process three-dimensional monitoring data; perform battery production data collaboration on the battery production process three-dimensional monitoring data according to the slurry anomaly monitoring data to execute battery production full-process data monitoring work.

[0089] In the embodiment of the application, a plurality of sensors installed at key nodes of the battery production line collect multi-source data in real time, specifically including current sensors, voltage sensors, temperature sensors and humidity sensors, and the sampling frequency is uniformly set to 1000 Hz to ensure capturing small changes. The data is transmitted to the edge computing node for preprocessing using an industrial protocol (such as OPC UA). At the same time, a high-definition industrial camera is fixed at the battery assembly station, with a resolution of 1920x1080 pixels and a frame rate of 30 frames per second, continuously collecting image data of the assembly process. The image adopts both grayscale and RGB modes for subsequent multi-feature fusion recognition. For real-time multi-source data, a sliding window mechanism (window length of 5 seconds, sliding step of 1 second) is used to calculate the current mutation amplitude, which is the difference between the maximum and minimum currents in the window. At the same time, the standard deviation of the temperature sensor area data is calculated, and the dispersion of the temperature of each point in the 3x3 sensor grid is calculated to reflect the temperature difference fluctuation. Image data is processed by a pre-trained deep convolutional neural network (CNN) model to perform boundary point recognition, with the model input being the preprocessed edge-enhanced image and the output being the coordinates of the key boundary points in the assembly stage. The boundary point accuracy is controlled within ±1 pixel. Using the time series features of the current mutation amplitude and the regional temperature difference standard deviation, combined with image boundary point information, data synchronization is performed according to the timestamp, and a process division algorithm is executed to divide the battery production line data into several continuous production stage intervals. The process division result is stored in the format of timestamp and process identifier, generating complete battery production process data. Based on the electrode slurry preparation stage time period determined in step S1, the quantum dot sensor array of the slurry agitator is deployed. The sensor array adopts a grid layout with a spacing of 5 cm, covering all key mixing areas in the slurry tank, ensuring that the total number of sensors is not less than 64, and ensuring spatial resolution. The spectral response band of the quantum dot sensor covers 400 to 900 nanometers, with a spectral resolution of 1 nanometer. The sensor collects the optical properties of the suspended particles in the slurry in real time, and the output spectral signal is transmitted to the data processing unit through an optical fiber at a sampling frequency of 500 Hz. After time series synchronous sampling of the spectral response signal, denoising processing (wavelet transform denoising, threshold set to 0.02) is performed, and then the spectral data is converted into a multi-dimensional feature vector through a spectral decoding algorithm, including peak position, half-width, integral area, etc. The output forms the slurry state spectral feature data. Using the slurry state spectral feature data of step S2, combined with particle dynamics analysis method, the evolution of the agglomeration state of the micro-particles in the slurry is calculated. Through time series comparison and analysis of particle size distribution and particle size mean change, particle agglomeration behavior data is generated. According to the agglomeration behavior, a three-dimensional reconstruction algorithm (based on voxel mapping technology, voxel size set to 1 cubic millimeter) is used to reconstruct the spatial three-dimensional structure of the particle distribution in the slurry. The reconstruction accuracy is controlled within 1 millimeter.Further utilizing the three-dimensional reconstruction data, uniformity and stability of particles in the slurry stirring tank are analyzed, a micro-particle imbalance index (such as local particle density fluctuation amplitude, and a threshold value of 10%) in the stirring process is calculated, and combined with time series monitoring data, abnormal conditions of the slurry are judged. The output slurry abnormal monitoring data includes abnormal type, abnormal time point and position coordinates. Based on the battery production process data generated in step S1, a three-dimensional digital twin modeling platform is used to build a digital twin scene of the battery production line. By importing the CAD model of the production line, combined with the production process time node, dynamic three-dimensional visualization of the production line is realized, and the model space resolution is set to millimeter level. The slurry abnormal monitoring data of step S3 is mapped to the corresponding stirring tank position of the digital twin scene, realizing real-time marking and three-dimensional display of the abnormal state. The system synchronously updates the slurry physical state and production stage information, and completes the collaborative fusion of data. The three-dimensional monitoring data is refreshed at a frequency of 1 second, supporting production management personnel to view the whole process data and slurry state through the interface, and supporting automatic alarm and decision assistance.

[0090] As an example of the present application, reference is made to Figure 2 In this example, the step S1 includes:

[0091] Step S11: acquiring real-time multi-source data of the battery production line and assembly station images;

[0092] Step S12: calculating current mutation amplitude and regional temperature difference standard deviation according to the real-time multi-source data of the battery production line, and generating current-temperature difference joint feature data;

[0093] Step S13: performing edge contour enhancement and structure region extraction on the assembly station images to generate station image structure feature data; identifying a typical component assembly action sequence according to the station image structure feature data to generate production stage boundary point data;

[0094] Step S14: using the current-temperature difference joint feature data to divide the production stage boundary point data into process behavior time windows to generate process feature slice data; performing key index clustering and dynamic stage alignment processing on the process feature slice data to generate battery production process data.

[0095] In the embodiment of the present application, by adopting a multi-modal sensor system installed at each key node of the battery production line, physical quantity data such as current, temperature and humidity are collected in real time. The sampling frequency of the current sensor is set to 2000 Hz, and the resolution reaches 0.1 mA. The temperature sensor is arranged in a 3x3 matrix grid, covering the surrounding area of the assembly station, and the temperature sampling frequency is 1 Hz with an accuracy of ±0.1°C. A high-resolution industrial camera is installed at the assembly station, with a resolution of 3840x2160 pixels and a frame rate of 30 frames per second, continuously collecting high-definition assembly images. The camera uses a global shutter mode to avoid motion blur, and the images are transmitted to the local processing unit in real time. All collected data have a unified timestamp, and the time synchronization module ensures that the multi-source data is time-sequential. The time synchronization accuracy is controlled within 1 ms. The current signal collected by the battery production line is calculated for mutation amplitude using a sliding time window method, with a time window length of 2 seconds and a step size of 0.5 seconds. The difference between the maximum and minimum values of the current signal in the calculation window is obtained to get the current mutation amplitude data sequence. The temperature sensor data of the assembly station area is calculated for standard deviation using the temperature values in the 3x3 sensor matrix to calculate the dispersion of the temperature data in each time window. The time window length is consistent with the current mutation amplitude calculation. The current mutation amplitude and temperature difference standard deviation data are paired according to the timestamp to generate a current-temperature difference joint feature data sequence, with a data format of timestamp and corresponding feature pair. The assembly station image is preprocessed first, including grayscale conversion, denoising (using Gaussian filter with a standard deviation of 1.5), and enhancing edge features. Then the Canny edge detection algorithm is applied with a low threshold of 50 and a high threshold of 150 to extract the edge contours in the image. Through region growing and morphological processing, structural region features are extracted to form the station image structure feature data. The extracted features include component edge curves, contact surface areas and connection points. Based on the image frame sequence, the dynamic time warping algorithm (DTW) is used to identify the assembly action sequence and determine the start and end time points of the typical component assembly actions. According to the time nodes of the assembly actions, the corresponding production stage boundary point data is generated, and the boundary point data is stored in the form of timestamp and action identification pair. The current-temperature difference joint feature data generated in step S12 is used to divide the production stage boundary points in step S13 into time windows. The time window length is set to 3 seconds with a sliding step size of 1 second to divide the time interval corresponding to the production stage boundary points to form the process feature slice data. The K-means clustering algorithm is applied to the key indicators (such as current mutation amplitude peak value and temperature difference standard deviation trend) in the process feature slice data to divide the different dynamic stages of the process with a preset cluster number of 5. Then the clustering results are dynamically time-aligned to adjust the process stage time axis to ensure smooth transition between stages across time periods and eliminate time misalignment at the slice boundaries. Finally, the battery production process data is output, with a data format including process stage number, corresponding time interval, cluster category and key indicator value, and a storage structure that supports real-time updating and querying.

[0096] Preferably, the step S2 of deploying the quantum dot sensor array based on the electrode slurry preparation stage includes:

[0097] Deploying the quantum dot sensor array based on the electrode slurry preparation stage to the slurry stirred tank to obtain sensor array deployment data, wherein the slurry temperature is controlled at 15 to 80℃, the viscosity range is 500 to 5000 mPa·s, the particle size distribution is 50 to 800 nm, the fluorescence response wavelength is concentrated in 500 to 650 nm, the fluorescence intensity sensitivity is 10 2 to 10 5 Relative units, the pH value of the slurry is set to 6.0 to 9.5, the conductivity range is 0.1 to 5.0 S / m, the vibration acceleration is set to 0.5 to 3.0 g, the sensor array sampling frequency is set to 10 to 100 Hz, and the array distribution density is 4 to 16 sensing points per square meter.

[0098] In the embodiment of the present application, before arranging the quantum dot sensor array in the slurry stirring tank, the physical and chemical parameters of the slurry are ensured to meet the following ranges: the slurry temperature is controlled between 15 degrees Celsius and 80 degrees Celsius to ensure the stability of the sensor material and signal; the slurry viscosity is controlled in the range of 500 to 5000 millipascal seconds to ensure that the slurry flow state is suitable for sensor signal acquisition; the slurry particle size distribution is maintained in the range of 50 nanometers to 800 nanometers to ensure that the quantum dots are sensitive to particle size changes and respond accurately. The quantum dot sensor with a fluorescence response wavelength covering 500 nanometers to 650 nanometers is selected to adapt to the spectral band changes caused by changes in slurry composition. The fluorescence intensity sensitivity range is set to 100 to 100000 relative units to meet the detection needs of different concentration changes. When designing the sensor array, the pH value of the slurry is controlled between 6.0 and 9.5 to ensure stable light emission of the quantum dots and avoid chemical corrosion. The conductivity range is controlled between 0.1 and 5.0 Siemens per meter to ensure that the sensor electrical performance is not affected by abnormal electrolytes. Due to the mechanical vibration during the slurry stirring process, the vibration acceleration is set to 0.5 to 3.0 times the gravity acceleration (g), and the sensor structure design has anti-vibration capability to ensure that there is no distortion in the signal acquisition process. The sensor array is evenly arranged with a density of 4 to 16 sensing points per square meter, covering all key mixing areas in the slurry tank to achieve high spatial resolution detection. The sampling frequency is set to 10 Hz to 100 Hz, taking into account the signal timeliness and data processing load, and reflecting the changes in the slurry state in real time. The fluorescence intensity signals collected by each sensor are transmitted to the local data processing module through optical fibers or wireless high-bandwidth channels, and the data is time-synchronized to ensure that the multi-point data is uniformly time-labeled, facilitating subsequent multi-dimensional analysis. After deployment, the sensor array is calibrated using standard fluorescent liquid samples to correct the sensitivity and baseline drift of each sensing point and ensure the response consistency of the entire array. After calibration, short-term continuous testing is performed to confirm the data stability and repeatability.

[0099] Preferably, the spectral response spectrum decoding of the sensor array deployment data in step S2 includes:

[0100] Extracting the multi-channel response signal of the sensor array deployment data;

[0101] Performing time domain synchronization and frequency spectrum normalization on the multi-channel original response data to generate standard response spectrum data;

[0102] Performing spectral difference analysis and high-dimensional noise reduction on the standard response spectrum data to generate effective spectral response interval data;

[0103] Decoding the composition sensitive spectral band according to the effective spectral response interval data to generate a slurry composition reflectance matrix;

[0104] The principal component features of the reflection coefficient matrix of the slurry composition are extracted, and the multi-channel response signal is calibrated according to the principal component features, to generate slurry state spectral feature data.

[0105] In the embodiment of the application, the multi-channel spectral response signals collected by the quantum dot sensor array are sequentially read according to the channel number. Each channel contains continuous sampling data of spectral intensity changing with time, the sampling frequency is set to 50 Hz, and the sampling accuracy reaches 12-bit conversion. The original data contains noise and baseline drift, which needs to be effectively removed in subsequent processing. The multi-channel response data is processed for time alignment, and the sensor synchronous clock is used for labeling to realize accurate time domain synchronization of the multi-channel signal, and the time error is controlled within 1 millisecond. The spectral intensity of each channel is normalized, and the method is to subtract the average value of the baseline of the channel from the sampling value, and then divide by the maximum response value of the channel, to ensure that all channel spectral data are unified in the range of 0 to 1, which is convenient for subsequent comparison and fusion. The spectral response data is divided into a plurality of preset frequency band intervals, each frequency band has a width of about 5 nanometers, and covers the response wavelength range of the sensor. The mean and variance of the spectral response in each frequency band are calculated to identify the response fluctuation region as an effective spectral response interval. The non-effective interval is processed by principal component analysis (PCA) method for high-dimensional data noise reduction to remove high-frequency random noise and environmental interference signals and improve the signal-to-noise ratio. The reflection coefficient matrix is constructed by combining the characteristic absorption bands of the known main chemical components of the slurry and matching the effective spectral response interval and using the multi-channel data. The matrix takes the sensor channel as the row, the sensitive spectral band as the column, and the matrix element represents the reflection intensity coefficient of the channel in the spectral band. The spectral response intensity corresponding to the component is extracted by matrix decomposition method, the non-target component interference is removed, and the high-accuracy component reflection feature is obtained. The principal component extraction is performed on the component reflection coefficient matrix to obtain the principal component feature vector representing the slurry state change. According to the principal component feature, the absorption peak position and intensity of the multi-channel response signal are calibrated and corrected by combining the known chemical absorption peak position. Finally, the slurry state spectral feature data is output, including the principal component value, the chemical absorption peak information of each sensitive spectral band, and the data structure supports real-time updating and trend analysis, which provides accurate basis for slurry quality monitoring.

[0106] Preferably, the micro-particle state evolution of the electrode slurry preparation stage by the slurry state spectral feature data in step S3 comprises:

[0107] The slurry state spectral feature data is subjected to wave band reflectivity fitting to generate multi-scale wave band reflection curve data;

[0108] The multi-scale wave band reflection curve data is subjected to particle size response mapping to generate particle size response distribution data;

[0109] The particle size response distribution data is subjected to time window slicing to generate time sequence particle size evolution segment data;

[0110] Identify the particle cluster aggregation trend of the time sequence particle size evolution segment data;

[0111] Analyze the particle spacing of the particle agglomeration trend data to generate particle spacing distribution data;

[0112] Threshold comparison of particle spacing distribution data based on preset particle spacing threshold value, when particle spacing distribution data is greater than or equal to the preset particle spacing threshold value, the particle spacing distribution data is marked as agglomeration data;

[0113] Spatial density inversion of agglomeration data to generate micro-agglomeration concentration layer data;

[0114] Spatiotemporal fusion of micro-agglomeration concentration layer data, and according to the fused micro-agglomeration concentration layer data, stability dispersion analysis of electrode slurry preparation stage is carried out, and finally slurry agglomeration behavior data is generated.

[0115] In the embodiment of the present application, by selecting multiple wave bands covering 400 nanometers to 900 nanometers from the spectral feature data of the slurry state, the wave band width is set to 10 nanometers, the spectral reflectivity data in each wave band is fitted by using a curve fitting method, and a continuous multi-scale wave band reflection curve data is generated by using a smoothing spline curve fitting technology. The fitting error is controlled within 0.5%, ensuring that the reflection curve is smooth and accurately reflects the spectral characteristics. According to the response relationship between the particle size of the slurry particles and the corresponding spectral wave band reflectivity obtained by experimental calibration, a particle size response mapping function is established. The function converts the multi-scale wave band reflection curve into the response intensity of the corresponding particle size interval, and the particle size interval is divided into 50 nanometers to 1000 nanometers, with a particle size step of 25 nanometers. The mapping output generates particle size response distribution data, reflecting the relative distribution density of particles of different particle sizes in the slurry. The particle size response distribution data is divided into fixed-length time window segments according to the time sequence, and the length of each time window is set to 5 minutes without overlapping between time windows. The sliced data forms time sequence particle evolution segment data, reflecting the dynamic changes of the micro-particle size state of the slurry over time. Based on the time sequence particle evolution segment data, a spatial clustering algorithm (such as the DBSCAN algorithm, with a neighborhood radius of 50 nanometers and a minimum cluster point number of 10) is used to identify the particle cluster aggregation trend in the particle size distribution, and determine the particle aggregation block and the isolated particle region. For the identified particle aggregation trend data, the Euclidean distance between particles is calculated, and particle distance distribution data is generated. The statistical range is set to 0 nanometers to 500 nanometers, the distance distribution is divided into 20 equal intervals, and the number and density of particles in different distance intervals are obtained. The particle distance threshold is set to 100 nanometers, and the particle distance distribution data is compared. Any particle data in the interval with a distance greater than or equal to the threshold is marked as aggregation data, which is used to reflect the aggregation degree and structural characteristics of the particles. The marked aggregation data is applied to the inversion algorithm to convert the discrete particle aggregation information into a spatial density layer. The spatial resolution is set to 1 square millimeter per unit, and the kernel density estimation method is used to calculate the local particle density to generate a micro-aggregation concentration layer data. The micro-aggregation concentration layer data in multiple time windows is spatio-temporally fused, the fusion weight decreases according to the time proximity, and the spatial adjacent unit weight is equal. After fusion, the micro-aggregation concentration layer is discretely analyzed, the standard deviation statistical index is used to evaluate the spatial stability and temporal fluctuation of the aggregation behavior, and finally the slurry aggregation behavior data is generated, which is an important representation of the microstructure state in the slurry preparation stage.

[0116] Especially important is that identifying the particle cluster aggregation trend of the time sequence particle evolution segment data also includes:

[0117] Performing multi-time window sliding mean processing on the time sequence particle evolution segment data to extract local particle size increment trend data;

[0118] According to the particle size increment trend data, the particle boundary contour is analyzed by using the vector field, and the particle relative motion direction data is generated;

[0119] The particle size superposition block is identified by using the particle relative motion direction data, and the particle contact tightness feature data is extracted;

[0120] The particle contact tightness feature data is clustered and the boundary is identified, and the initial region data of the potential particle cluster is generated;

[0121] Based on the particle cluster initial region data, the consistency verification in continuous frames is carried out, the stable aggregation region is screened, and the particle cluster aggregation trend data is generated.

[0122] In the embodiment of the application, the particle size distribution data sequence at multiple time points is extracted from the electrode paste preparation stage, and each set of data corresponds to the particle image of the same observation area. The time interval is set to once per second, and a total of 60 frames are collected to form a "time sequence particle size evolution segment data". For this data segment, a sliding time window mean processing method is used, with five frames as a sliding window and a window step of one frame. The local mean of the particle diameter identified in each frame is calculated. This method can smooth the sharp fluctuations in the time sequence while preserving the details of the particle size changes, thereby extracting the local particle size increment trend data. The output data is the particle size mean change sequence of each particle over time, with a unit of microns. Based on the local particle size increment trend data, combined with the particle boundary extraction results in the original image data, the particle boundary profile is analyzed. The relative displacement of the particle boundary in adjacent time frames is calculated by image difference method to determine the offset path of each particle in the time continuous frames. These path data are used to construct the aggregation vector field of the particles, and each vector of the vector field represents the relative motion direction of the boundary between two adjacent particles. The direction points from the center of one particle to the center of the adjacent particle, and the length represents the distance of the two particles approaching in unit time, with a unit of microns per second. This vector field can be regarded as a mapping of the trend of the aggregation or separation of particles, and the output is "particle relative motion direction data". According to the particle relative motion direction data, the boundary overlap or proximity area of the particles in the continuous time frames is identified. If the center distance between two particles is less than 1.2 times the average diameter of the particles, and the number of overlapping pixels between the boundaries exceeds 5 pixels, it is considered that there is a particle size superposition block. The number of adjacent particles in a unit area for each particle is counted, and the minimum distance between the boundaries is calculated, combined with the boundary overlap degree, to construct the contact tightness feature data. This feature is represented by a numerical value, ranging from 0 (no contact) to 1 (complete fit), and each particle corresponds to a contact tightness value. Using the contact tightness feature data, a density-based spatial clustering method (such as the fixed threshold method) is used for clustering boundary recognition. If the contact tightness of a plurality of consecutive particles in a region exceeds 0.6, and the distance between the particles is less than 2 times the average particle diameter, the region is determined as a potential cluster. The output of the clustering is "particle cluster initial region data", and the data format is a plurality of closed contour regions, each region containing its particle number, position coordinates, local density value, etc. attributes. The particle cluster initial region is verified for time consistency. If the same cluster region exists stably in the continuous 10 frames of images (i.e. the particle number and position overlap degree exceeds 80%), the region is determined as a "stable aggregation region". Finally, these time-verified cluster regions are output as "particle cluster aggregation trend data", each trend data containing cluster position, duration, average particle size, aggregation speed (particle distance shortening speed), etc. All information is recorded in a structured table form for further analysis of the dispersion or agglomeration trend of the paste.

[0123] Preferably, the three-dimensional distribution reconstruction of the slurry agglomeration behavior data in step S3 includes:

[0124] Extract the spatial coordinate index of the slurry agglomeration behavior data to obtain the agglomeration spatial coordinate point set data;

[0125] Perform point cloud densification interpolation on the cluster space coordinate point set data to generate particle cluster point cloud data;

[0126] Local voxel encoding of particle agglomeration point cloud data, and global grid splicing and reconstruction of the encoded particle agglomeration point cloud data to generate slurry agglomeration three-dimensional grid skeleton data;

[0127] Map the particle size channel attributes of the slurry agglomerated three-dimensional grid skeleton data to generate attribute annotated grid data;

[0128] Perform multi-view projection rendering on the attribute annotation mesh data to generate particle agglomeration 3D view data;

[0129] Based on the particle agglomeration 3D view data, dynamic sequence frame encoding is performed to generate speed-adjustable particle agglomeration evolution sequence data;

[0130] The particle agglomeration evolution sequence data is processed by stability hotspot annotation to obtain the reconstruction result of three-dimensional distribution reconstruction.

[0131] In an embodiment of the present invention, the three-dimensional spatial coordinates of all particle agglomeration positions are extracted from the slurry agglomeration behavior data. The coordinates are measured with millimeter-level accuracy, and the coordinate range corresponds to the internal space size of the slurry agitator, which is approximately 1000 mm × 1000 mm × 500 mm. The generated spatial coordinate point set is stored in the form of a structured array for easy subsequent processing. The extracted spatial coordinate point set is densified and interpolated using an interpolation algorithm based on the inverse distance weighting method. The interpolation radius is set to 20 mm to ensure that the density of the interpolated point cloud is evenly distributed, fill the original sampling gaps, and generate continuous and high-density particle agglomeration point cloud data. The average distance between points is controlled within 5 mm. The particle agglomeration point cloud data is divided into cubic voxel units with a side length of 10 mm. The number of points contained in each voxel unit is counted and encoded as a voxel density value. All local voxel units are spatially spliced ​​using an adjacency algorithm to form a global three-dimensional grid skeleton data. The grid side length is also 10 mm to ensure spatial continuity and structural integrity. Based on the corresponding particle size information in the raw slurry agglomeration behavior data, the particle size is mapped as an attribute to the corresponding grid unit. The particle size range is divided into five levels, corresponding to the following ranges: 50-150 nanometers, 151-300 nanometers, 301-500 nanometers, 501-800 nanometers, and 801-1000 nanometers. Each particle size level is assigned a different attribute value to generate attribute annotated grid data to reflect the particle size distribution characteristics. The attribute annotated grid data is rendered in two dimensions using six fixed viewing angles (front, back, left, right, top, and bottom), and the resolution is set to 1024×1024 pixels. The rendering uses volume rendering technology, supports translucent effects, realizes color mapping of particle size attributes, and generates three-dimensional view data of particle agglomeration. The three-dimensional view data of particle agglomeration collected at multiple time points is arranged in chronological order, with a frame rate set to 30 frames per second. Inter-frame differential encoding is used to compress the data, preserving information on spatial position and particle size attribute changes. This generates a particle agglomeration evolution sequence data with adjustable playback speed, supporting dynamic analysis operations such as fast forward and slow playback. Based on the particle agglomeration evolution sequence data, the local area stability index is calculated by statistically analyzing the changes in grid density within consecutive time frames. For areas where the stability index is higher than the threshold, the threshold is set to a density fluctuation of less than 5% within 10 consecutive frames, and hot spots are marked to form three-dimensional stability hot spot data, realizing the reconstruction of the spatial distribution of slurry agglomeration.

[0132] Preferably, the slurry stirring micro-imbalance monitoring based on the reconstruction result in step S3 includes:

[0133] Extract the particle structure orientation information in the stirring area based on the reconstruction results of the three-dimensional distribution reconstruction to generate local structure orientation data;

[0134] Perform spatial anisotropy tensor analysis on local structural orientation data to generate orientation tensor deviation data;

[0135] The orientation tensor deviation data is used to deduce the particle flow trend vector trajectory in the electrode slurry preparation stage, and micro flow disturbance data is generated;

[0136] According to the micro flow disturbance data, the local shear stress gradient in the stirring process is calculated, and a shear imbalance intensity atlas is generated;

[0137] The stability domain boundary of the shear imbalance intensity atlas is identified, and the particle aggregation time window statistics of the stability domain boundary is performed, and particle group fluctuation interval data is generated;

[0138] The particle group fluctuation interval data is subjected to micro disturbance frequency analysis, and finally slurry abnormal monitoring data is generated.

[0139] In the embodiment of the application, based on the three-dimensional distribution reconstruction of the slurry aggregate voxel data, the principal component analysis (PCA) method is used to calculate the feature vector of the particle point cloud in each voxel unit, and the principal direction vector of the local particle structure is obtained. The spatial resolution is set to a cube with a side length of 10 mm, and the extraction result forms a local structure orientation data set, and the accuracy is controlled within the direction change range of the nano-scale particle structure. A three-dimensional anisotropy tensor matrix is constructed for the local structure orientation data, and the matrix reflects the uniformity and deviation of the orientation distribution. By calculating the eigenvalue distribution of the tensor, orientation tensor deviation data is generated, and the deviation reflects the difference between the local structure and the ideal uniform distribution, and the numerical range is between 0 and 1, and the larger the value represents the more significant the deviation. The orientation tensor deviation data is used in combination with the fluid dynamics simulation algorithm to simulate the flow trend of the particles in the slurry affected by stirring. The particle flow vector trajectory is generated by the vector field integration method, the trajectory time step is 0.1 seconds, and the trajectory length covers the entire stirring period, and a micro flow disturbance data set is generated. Based on the micro flow disturbance data, the local shear stress gradient distribution is calculated using a finite element shear stress model. The model parameters include slurry viscosity, shear rate, etc., and the shear stress gradient spatial resolution is 10 mm, and a shear imbalance intensity atlas is generated, and the atlas numerical range corresponds to the shear stress gradient size, and reflects the shear uniformity in the stirring area. The edge detection algorithm (such as Canny algorithm) is applied to the shear imbalance intensity atlas to identify the stability boundary of the shear balanced and imbalanced regions. For the regions within the boundary, the time window length is set to 30 seconds, the number of particle aggregation changes and the time distribution are counted, and particle group fluctuation interval data is generated, which reflects the dynamic stability of the microstructure of the slurry. The frequency spectrum analysis method (such as fast Fourier transform) is used to calculate the disturbance frequency distribution of the particle group fluctuation interval data, and the high-frequency disturbance area and the low-frequency stable area are identified. Through frequency threshold screening, slurry abnormal monitoring data is generated, which points out the micro imbalance and abnormal fluctuation points existing in the slurry stirring process.

[0140] Of particular importance is the use of orientation tensor deviation data to perform vector trajectory deduction of particle flow trends during the electrode slurry preparation stage, which also includes:

[0141] The particle flow direction vector field is constructed based on the orientation tensor deviation data to generate a local particle flow vector distribution map;

[0142] Perform multi-scale grid interpolation on the particle flow vector distribution map to generate particle flow continuity trajectory data;

[0143] Use particle flow continuity trajectory data to perform vector trajectory dynamic deduction, identify abnormal disturbance nodes, and generate microscopic trajectory disturbance identification data;

[0144] The disturbance intensity evaluation and direction decomposition of the microscopic trajectory disturbance identification data are performed to finally generate the microscopic flow direction disturbance data.

[0145] In the embodiments of the present application, the particle orientation tensor deviation data in different observation regions is calculated based on the multi-dimensional spectral response map collected by the quantum dot sensor array. The deviation calculation is based on the angular deviation between the tensor principal direction and the theoretical uniform flow direction, and the numerical range is set to 0 to 90 degrees. Taking each group of sensors as the basic unit in the spatial unit (the volume of each unit is about 1 cubic centimeter), the main flow direction distribution of the particles in the three-dimensional space is summarized. The tensor principal direction is converted into a standard three-dimensional vector form, and combined with the deviation value, a local particle flow direction vector field is constructed. In the actual construction process, each vector has a specific position coordinate, direction component (X, Y, Z), and deviation label. The output result is a particle flow vector distribution map, in which each vector represents the main motion direction of the particles in the form of an arrow, and the deviation degree is represented by a color gradient (for example: blue to red), with red indicating a larger deviation degree and blue indicating a trend towards uniform flow. Based on the particle flow direction vector field data, a multi-layer grid system is divided according to different spatial scales. The basic grid unit size is set to 2 cubic millimeters, the medium scale is set to 5 cubic millimeters, and the large scale is set to 1 cubic centimeter, and the three levels are used to cover the particle motion characteristics of different scales. A three-dimensional spline interpolation method is used to estimate the continuity of the vector direction in each layer of grid, eliminating the local data gaps caused by uneven distribution of sensors. After interpolation, particle flow continuity trajectory data is generated, which is recorded in the form of a set of path points, each trajectory is composed of consecutive spatial points, and the velocity direction and gradient change value are recorded at each point. This trajectory data is used for subsequent dynamic evolution analysis. Using the particle flow continuity trajectory data, dynamic trajectory deduction is performed. The deduction is based on two key indicators: the velocity field change rate (i.e. the trajectory curvature change) and the trajectory angle deviation rate: if the angle change between three consecutive points of a certain trajectory exceeds 45 degrees, or the curvature radius is less than 2 millimeters, it is marked as a "possible disturbance section"; if the change overlaps in more than three trajectories, it is determined as a disturbance node. These abnormal deformation sections are numbered, positioned, and corresponded to the spatial coordinates one by one, and finally the micro trajectory disturbance identification data is generated. This data includes the location (XYZ coordinates) of the disturbance node, the corresponding time point, the disturbance type (sharp turn, speed mutation, direction reversal, etc.), and the visualization identifier. The disturbance intensity of each micro disturbance node is quantitatively evaluated. The evaluation adopts three dimensions: direction deviation angle: the angle with the average flow direction exceeds 30 degrees; local speed fluctuation: the speed standard deviation exceeds 20% of the overall average speed; curvature change rate: the curvature change rate in unit time is greater than 10 degrees / millisecond. Each disturbance node is scored according to the above three indicators, and the disturbance intensity level is calculated by adding the scores, which is divided into low disturbance (less than 1 point), medium disturbance (1-2 points), and high disturbance (more than 2 points). At the same time, the disturbance direction is decomposed into XYZ direction components to analyze whether there is a directional deviation concentration trend.The final output of the micro-flow disturbance data is a structured data table, including disturbance location, disturbance type, disturbance intensity level, direction component information, fluctuation duration, and associated trajectory number, etc.

[0146] Preferably, the micro-disturbance frequency analysis on the particle swarm fluctuation interval data includes:

[0147] When any of the following conditions occurs, the particle size concentration abnormality data is obtained: the characteristic particle size distribution range narrows by more than 20%, the D90 / D10 particle size ratio fluctuates by more than ±0.6, and the particle size distribution skewness index deviates from the average value by more than ±15% in three consecutive sampling periods.

[0148] When the following conditions occur simultaneously, the micro-agglomeration disturbance enhancement phenomenon is determined, and the agglomeration disturbance enhancement data is obtained: the number ratio of agglomerates in the particle swarm increases by more than 25% within 10 minutes, the average value of the agglomerate particle size exceeds the historical upper limit by more than 10%, the internal density distribution of the agglomerates presents a double-peak deviation and the duration is more than 20 minutes.

[0149] When the following conditions are met simultaneously, the shear fluctuation abnormal state is determined, and the shear fluctuation abnormal data is obtained: the stirring system speed fluctuation frequency is more than 2Hz, the particle swarm micro-disturbance response frequency deviates from the reference spectrum line range by more than ±10% during stirring, the micro-bubble content in the slurry increases by more than 8%, and the duration of this state is more than 30 minutes.

[0150] The particle size concentration abnormality data, agglomeration disturbance enhancement data, and shear fluctuation abnormal data are integrated, and the correlation weight evaluation and disturbance type fusion recognition are performed to finally generate the slurry abnormality monitoring data.

[0151] In the embodiment of the present application, by dynamically analyzing the particle size distribution data in the particle swarm fluctuation interval, a sliding time window method is adopted, and the time window length is set to be 3 continuous sampling periods (usually 5 minutes per period, a total of 15 minutes): the change of the characteristic particle size distribution range is monitored, if the range narrows more than 20%, that is, the difference between the maximum and minimum particle sizes in the current time window is reduced by more than one fifth compared with the historical average, it is determined to be abnormal; the ratio of D90 and D10 particle sizes is calculated, if the fluctuation amplitude of the ratio exceeds plus or minus 0.6, it is considered to be abnormal fluctuation; the skewness index of particle size distribution is calculated, if the deviation from the historical average exceeds 15% in the continuous 3 sampling periods, it is considered to be skewness abnormality. When any of the above conditions is met, the system automatically records the current abnormal period and the corresponding data, and generates particle size concentration abnormality data. In a statistical window with a time scale of 10 minutes, the number and properties of particle agglomerates are analyzed: if the number ratio of agglomerates increases by more than 25% compared with the previous period, it is considered to be an abnormal increase in number; the average value of agglomerate particle size is monitored, if it exceeds the historical maximum value by more than 10%, it is determined to be an agglomerate particle size abnormality; the internal density distribution characteristics of the agglomerates are analyzed, if a double-peak deviation (i.e. the density distribution graph presents two obvious peaks and the position deviates) occurs and the state lasts for more than 20 minutes, it is confirmed that the agglomerate disturbance is enhanced. When the above three conditions are met at the same time, agglomerate disturbance enhancement data is generated. The state of the slurry stirring system and the micro-response of the slurry are monitored, and the following determination conditions are set: the fluctuation frequency of the stirring system speed is more than 2 times per second; the particle swarm micro-disturbance response frequency (measured by vibration sensor or particle size change frequency) deviates from the reference frequency spectrum by more than ±10%; the micro-bubble content in the slurry increases by more than 8% compared with the normal level; and the duration of the above state is not less than 30 minutes. When all the conditions are met, it is determined to be a shear fluctuation abnormality, and the corresponding abnormal data is generated. The particle size concentration abnormality, agglomerate disturbance enhancement and shear fluctuation abnormality data are compared and correlated in time and space, and the weight of the correlation is evaluated: the weight index is calculated based on the frequency, duration and intensity of the abnormal data; the disturbance type fusion identification is performed on different types of abnormal data by using a multi-factor fusion algorithm (such as weighted average fusion, fuzzy logic reasoning); and finally the structured slurry abnormality monitoring data is formed, including abnormal type classification, severity level and corresponding time period identification.

[0152] As an example of the present application, reference is made to Fig. 1, wherein in the present example the step S4 comprises: Figure 3

[0153] Step S41: extracting each assembly station parameter, key material flow direction and process timeline in the battery production process data for spatial mapping, generating structured digital scene data of the battery production process;

[0154] ​Step S42: three-dimensional visual modeling is performed on the structured digital scene data to generate three-dimensional monitoring data of the battery production process; and according to the abnormal slurry monitoring data, the corresponding electrode slurry preparation stage in the three-dimensional monitoring data is precisely positioned in a time period to generate abnormal mapping section data;

[0155] Step S43: parameter cross comparison is performed on the abnormal mapping section data and the three-dimensional monitoring data to generate visual abnormal marking layer data; the visual abnormal marking layer data is superimposed on the three-dimensional monitoring data of the battery production process to generate fusion monitoring data with abnormal feedback information;

[0156] Step S44: station-level data flow tracking is performed on the fusion monitoring data to identify a potential conduction path of the slurry abnormality to downstream process nodes to generate a process-level abnormal conduction map; and the process-level abnormal conduction map is used to perform data collaborative updating on the three-dimensional monitoring data of the battery production process to finally realize full-process data monitoring operation of the battery production.

[0157] In the embodiment of the application, by calling the production line real-time database interface, the process parameter data of each assembly station in the battery production process is collected, including but not limited to current, voltage, temperature, pressure, etc., and the time sampling frequency is set to once per second to ensure data integrity. At the same time, by using the material tracking system of the assembly station, the flow path and assembly sequence information of the key materials (such as slurry, electrode sheet, etc.) are obtained, and the time stamp is accurate to the millisecond level. The process timeline is extracted through the process management system to clearly define the start and end time and duration of each process, with a time error controlled within 100 milliseconds. After the above data is processed synchronously in time and space, it is mapped to a unified spatial coordinate system, and the assembly station layout is represented by XYZ three-dimensional coordinates, completing the spatial mapping of station parameters, material flow and timeline, generating structured digital scene data, and the data format uses a three-dimensional vector model with a time sequence label, with a file size controlled within 500 MB to ensure real-time processing capability. By using professional three-dimensional modeling software (such as Unity 3D or Unreal Engine), the structured digital scene data generated in step S41 is imported to perform high-precision modeling, with a model space accuracy of 1 millimeter to ensure that the spatial position and size of the assembly station accurately reflect the real production environment. Combined with the slurry abnormal monitoring data, the abnormal events are accurately mapped to the corresponding assembly station and time period of the three-dimensional model according to the time stamp, with a positioning time period error controlled within ±0.5 seconds to generate abnormal mapping section data. This step includes time synchronization and spatial comparison of abnormal events, and uses a time window mechanism to expand each 5 seconds before and after the abnormal time point to cover the influence range of the abnormality. For abnormal mapping section data, the change trend of key parameters (current, voltage, temperature, etc.) of the corresponding assembly station is extracted, and abnormal feature points are identified through parameter threshold comparison and change rate calculation. Set the abnormal judgment threshold, such as current mutation amplitude exceeding 20 amperes, temperature difference exceeding 3 degrees Celsius, etc., to automatically mark abnormal nodes. The abnormal mark is generated as a visual abnormal mark layer in the form of a semi-transparent red highlight layer, with a layer spatial resolution consistent with the three-dimensional model. By using layer superimposition technology, the abnormal mark layer is accurately superimposed on the battery production process three-dimensional monitoring data to ensure the real-time and spatial accuracy of the display, generating fusion monitoring data with abnormal feedback information. Based on the fusion monitoring data, the production line material flow and process flow database is called to track the downstream process nodes of the abnormal station along the time axis, with a time tracking accuracy controlled within 1 second to ensure the accuracy of the tracking path. The abnormal conduction path is identified by using topological analysis method, including material flow conduction and process parameter influence transmission. The tracking results are structured into a process-level abnormal conduction map, with nodes including station identification, abnormal level and conduction time, and edges representing conduction relationship and influence strength, with influence strength represented by a decimal between 0 and 1. The map drives the dynamic update of three-dimensional monitoring data to reflect the abnormal influence range and propagation trend in real time, ensuring the integrity and real-time of data collaboration, and finally realizing the battery production full-process data monitoring operation.

[0158] Thus, the embodiments should be considered in all respects as illustrative and not restrictive, the scope of the application being indicated by the appended claims rather than by the description given above, and all changes which come within the meaning and range of equivalency of the claims are intended to be embraced therein.

[0159] The foregoing is considered as illustrative only of the principles of the application. Numerous modifications and changes will readily occur to those skilled in the art, and the generic principles defined herein can be applied to other embodiments without departing from the spirit or scope of the application. Therefore, the application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for monitoring data of the entire battery production process, characterized in that: The following steps are involved: Step S1: Acquire real-time multi-source data of the battery production line and assembly station images; analyze the current mutation amplitude and regional temperature difference standard deviation of the real-time multi-source data of the battery production line; identify the production stage boundary points of the assembly station images, and use the current mutation amplitude and regional temperature difference standard deviation to divide the battery production process, and generate battery production process data; Step S2: extracting the electrode slurry preparation stage of the battery production process data, and deploying a quantum dot sensor array on the slurry stirring kettle based on the electrode slurry preparation stage to obtain sensor array deployment data; decoding the spectral response spectrum using the sensor array deployment data to generate slurry state spectral characteristic data; Step S3: analyzing the microscopic particle state evolution during the electrode slurry preparation stage through the slurry state spectral characteristic data to generate slurry agglomeration behavior data; Reconstruct the three-dimensional distribution of slurry agglomeration behavior data, and monitor the microscopic imbalance of slurry stirring based on the reconstruction results to generate slurry abnormality monitoring data; Step S4: construct a digital twin scenario based on the battery production process data to generate three-dimensional monitoring data of the battery production process; Based on the slurry abnormality monitoring data, the battery production process three-dimensional monitoring data is coordinated with the battery production data to perform battery production process data monitoring operations.

2. The battery production full process data monitoring method according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: Acquire real-time multi-source data of the battery production line and assembly station images; Step S12: Calculate the current mutation amplitude and regional temperature difference standard deviation based on the real-time multi-source data of the battery production line to generate current-temperature difference joint feature data; Step S13: performing edge contour enhancement and structural region extraction on the assembly station image to generate station image structural feature data; identifying a typical component assembly action sequence based on the station image structural feature data to generate production stage boundary point data; Step S14: Use the current-temperature difference joint feature data to divide the production stage boundary point data into process behavior time windows to generate process feature slice data; perform key indicator clustering and dynamic stage alignment processing on the process feature slice data to generate battery production process data.

3. The battery production full process data monitoring method according to claim 1, characterized in that: In step S2, the quantum dot sensor array is deployed on the slurry stirring tank based on the electrode slurry preparation stage, including: Based on the electrode slurry preparation stage, quantum dot sensor array deployment was performed on the slurry stirring tank to obtain sensor array deployment data. The slurry temperature was set to be controlled between 15 and 80 °C, the viscosity range was 500 to 5000 mPa·s, the particle size distribution was between 50 and 800 nm, the fluorescence response wavelength was concentrated between 500 and 650 nm, and the fluorescence intensity sensitivity was 10 2 to 10 5 Relative units, the pH value of the slurry is set between 6.0 and 9.5, the conductivity range is 0.1 to 5.0 S / m, the vibration acceleration is set between 0.5 and 3.0 g, the sensor array sampling frequency is set between 10 and 100 Hz, and the array distribution density is 4 to 16 sensing points per square meter.

4. The battery production full process data monitoring method according to claim 1, characterized in that: Decoding the spectral response spectrum using the sensor array deployment data in step S2 includes: Extract multi-channel response signals from sensor array deployment data; Perform time domain synchronization and spectrum normalization on multi-channel raw response data to generate standard response spectrum data; Perform spectral segment difference analysis and high-dimensional noise reduction on the standard response spectrum data to generate effective spectral response interval data; Decode the component sensitive spectral bands based on the effective spectral response interval data to generate the slurry component reflection coefficient matrix; The principal component characteristics of the slurry component reflection coefficient matrix are extracted, and the chemical absorption peaks of the multi-channel response signals are calibrated according to the principal component characteristics to generate the slurry state spectral characteristic data.

5. The battery production full process data monitoring method according to claim 1, characterized in that: In step S3, the microscopic particle state evolution of the electrode slurry preparation stage is performed using the slurry state spectral characteristic data, including: Perform band reflectivity fitting on the slurry state spectral characteristic data to generate multi-scale band reflectance curve data; Perform particle size response mapping on multi-scale band reflectance curve data to generate particle size response distribution data; Slice the particle size response distribution data into time windows to generate time series particle size evolution fragment data; Identify the trend of particle clustering in time-series particle size evolution segment data; Analyze the particle distance of particle agglomeration trend data to generate particle distance distribution data; Performing a threshold comparison on the particle spacing distribution data based on a preset particle spacing threshold, and marking the particle spacing distribution data as agglomeration data when the particle spacing distribution data is greater than or equal to the preset particle spacing threshold; Perform spatial density inversion on the agglomeration data to generate microscopic agglomeration concentration layer data; The micro-agglomeration concentration layer data is fused temporally and spatially, and the stability discrete analysis of the electrode slurry preparation stage is performed based on the fused micro-agglomeration concentration layer data, and finally the slurry agglomeration behavior data is generated.

6. The battery production full process data monitoring method according to claim 1, characterized in that: The three-dimensional distribution reconstruction of the slurry agglomeration behavior data in step S3 includes: Extract the spatial coordinate index of the slurry agglomeration behavior data to obtain the agglomeration spatial coordinate point set data; Perform point cloud densification interpolation on the cluster space coordinate point set data to generate particle cluster point cloud data; Local voxel encoding of particle agglomeration point cloud data, and global grid splicing and reconstruction of the encoded particle agglomeration point cloud data to generate slurry agglomeration three-dimensional grid skeleton data; Map the particle size channel attributes of the slurry agglomerated three-dimensional grid skeleton data to generate attribute annotated grid data; Perform multi-view projection rendering on the attribute annotation mesh data to generate particle agglomeration 3D view data; Based on the particle agglomeration 3D view data, dynamic sequence frame encoding is performed to generate speed-adjustable particle agglomeration evolution sequence data; The particle agglomeration evolution sequence data is processed by stability hotspot annotation to obtain the reconstruction result of three-dimensional distribution reconstruction.

7. The battery production full process data monitoring method according to claim 1, characterized in that: The slurry stirring micro-imbalance monitoring based on the reconstruction result in step S3 includes: Extract the particle structure orientation information in the stirring area based on the reconstruction results of the three-dimensional distribution reconstruction to generate local structure orientation data; Perform spatial anisotropy tensor analysis on local structural orientation data to generate orientation tensor deviation data; The orientation tensor deviation data is used to perform vector trajectory deduction on the particle flow trend during the electrode slurry preparation stage to generate microscopic flow disturbance data. The local shear stress gradient during the stirring process is calculated based on the microscopic flow disturbance data to generate a shear imbalance intensity map; Identify the stability domain boundary of the shear imbalance intensity map, and perform particle aggregation time window statistics on the stability domain boundary to generate particle group fluctuation interval data; The micro-disturbance frequency analysis is performed on the particle group fluctuation interval data to finally generate the slurry abnormality monitoring data.

8. The battery production full process data monitoring method according to claim 7, characterized in that: The micro-disturbance frequency analysis of particle swarm fluctuation interval data includes: Perform micro-disturbance frequency analysis on the particle size fluctuation interval data. When any of the following situations occurs, it is determined to be abnormal particle size concentration and the abnormal particle size concentration data is obtained: the characteristic particle size distribution range is narrowed by more than 20%, the D90 / D10 particle size ratio fluctuates by more than ±0.6, and the particle size distribution skewness index deviates from the average value by more than ±15% within three consecutive sampling periods; When the following conditions occur simultaneously, it is determined to be a micro-agglomeration disturbance enhancement phenomenon and agglomeration disturbance enhancement data are obtained: the ratio of the number of aggregates in the particle group increases by more than 25% within 10 minutes, the average agglomerate particle size exceeds the historical upper limit by more than 10%, and the internal density distribution of the agglomerates shows a bimodal shift and the duration exceeds 20 minutes; When the following conditions are met at the same time, it is determined to be a shear fluctuation abnormal state and the shear fluctuation abnormal data is obtained: the fluctuation frequency of the stirring system speed exceeds 2Hz, the perturbation response frequency of the particle group during stirring deviates from the reference spectrum range by more than ±10%, the microbubble content in the slurry increases by more than 8%, and the state lasts for more than 30 minutes; The abnormal data of particle size concentration, agglomeration disturbance enhancement and shear fluctuation are integrated and the associated weight evaluation and disturbance type fusion identification are performed to finally generate the abnormal slurry monitoring data.

9. The battery production full process data monitoring method according to claim 1, characterized in that: Step S4 includes the following steps: Step S41: extracting the parameters of each assembly station, key material flow direction and process timeline from the battery production process data for spatial mapping to generate structured digital scene data of the battery production process; Step S42: Performing three-dimensional visual modeling on the structured digital scene data to generate three-dimensional monitoring data of the battery production process; accurately locating the time period of the corresponding electrode slurry preparation stage in the three-dimensional monitoring data based on the slurry abnormality monitoring data to generate abnormal mapping segment data; Step S43: Cross-check the parameters of the abnormal mapping segment data with the three-dimensional monitoring data to generate visual abnormality marking layer data; superimpose the visual abnormality marking layer data on the three-dimensional monitoring data of the battery production process to generate fused monitoring data with abnormality feedback information; Step S44: Track the workstation-level data flow of the fused monitoring data, identify the potential transmission path of slurry anomalies to downstream process nodes, and generate a process-level anomaly transmission map; use the process-level anomaly transmission map to perform data collaborative updates of the three-dimensional monitoring data of the battery production process, and ultimately realize data monitoring operations for the entire battery production process.

10. A battery production full process data monitoring system, characterized in that: For executing the battery production full process data monitoring method according to claim 1, the battery production full process data monitoring system comprises: The process division module is used to obtain real-time multi-source data on the battery production line and assembly station images; analyze the current mutation amplitude and regional temperature difference standard deviation of the real-time multi-source data of the battery production line; identify the production stage boundary points of the assembly station images, and use the current mutation amplitude and regional temperature difference standard deviation to divide the battery production process, generating battery production process data; The slurry analysis module is used to extract the electrode slurry preparation stage of the battery production process data, and deploy the quantum dot sensor array on the slurry stirring kettle based on the electrode slurry preparation stage to obtain the sensor array deployment data; the spectral response spectrum is decoded through the sensor array deployment data to generate the slurry state spectral characteristic data; The imbalance monitoring module is used to analyze the microscopic particle state evolution during the electrode slurry preparation stage using slurry state spectral characteristic data to generate slurry agglomeration behavior data; reconstruct the slurry agglomeration behavior data into three dimensions, and monitor the microscopic imbalance of slurry stirring based on the reconstruction results to generate slurry abnormality monitoring data; The data collaboration module is used to build digital twin scenarios based on battery production process data and generate three-dimensional monitoring data of the battery production process; it collaborates with the three-dimensional monitoring data of the battery production process based on the slurry abnormality monitoring data to perform data monitoring operations for the entire battery production process.

Images