A blade battery multi-modal defect detection and random confusion encrypted transmission system

Abstract

The application discloses a kind of blade battery multimode defect detection and random confusion encryption transmission system, belong to power battery intelligent detection and industrial data security technical field.The application fuses five big core technologies of multimode visual perception, robot flexible tactile acquisition, deep learning intelligent identification, dynamic random confusion encryption, block chain credible evidence, constructs the industrial intelligent detection system of integrated integrated accurate detection, low delay data encryption, whole link credible traceability.The system adopts RGB visible light imaging, 3D laser depth imaging, robot tactile perception, four modal data fusion architecture of contour scanning, is combined with attention mechanism optimized YOLOv8 algorithm and cross-modal knowledge distillation technology, realizes blade battery six sides micron level defect detection and high-precision size fitting;While self-developed lightweight dynamic random confusion encryption algorithm, through equipment unique characteristic seed generation, data block tensor confusion, position check code binding mechanism, realize detection data low delay encryption transmission, and encryption delay is controlled within 20ms;Collaborate multi-party alliance chain evidence module, complete detection data hash fingerprint chain evidence, realize data tamper-proof, whole life cycle traceable.

Description

Technical Field

[0001] This invention relates to the fields of intelligent testing of power batteries, machine vision, flexible robot control, industrial data security, and blockchain trusted traceability. Specifically, it relates to a multimodal defect detection and random obfuscation encrypted transmission system for blade batteries, which is applicable to the detection of appearance defects, dimensional calibration, and secure management of testing data throughout the entire life cycle of blade batteries for new energy vehicles. It belongs to the cross-innovation technology of industrial intelligent testing and data security. Background Technology

[0002] As a core power component of new energy vehicles, the blade battery has become the mainstream power battery technology solution in the industry due to its advantages of high safety, long cycle life, and high integration. The appearance quality of the blade battery directly determines its insulation performance, heat dissipation effect, and service life. Defects such as scratches, dents, damaged coatings, misaligned tabs, and deformation on the surface of the cell can easily lead to safety hazards such as thermal runaway and performance degradation, and are the main causes of mass quality accidents in power batteries. Current testing technologies in the blade battery industry suffer from several inherent flaws: traditional manual inspection relies on human experience, resulting in low efficiency, high rates of missed and false detections, and a lack of standardized control over inspection consistency, making it difficult to adapt to the demands of large-scale mass production; traditional single machine vision inspection technology, relying solely on two-dimensional image recognition, cannot be adapted to the highly reflective casing material of blade batteries, has weak capabilities in identifying micron-level defects, and cannot simultaneously perform precise dimensional measurements, leading to cumbersome inspection procedures and poor production line adaptability; furthermore, existing inspection systems focus only on defect identification, neglecting industrial data security issues. Core production data such as inspection images, dimensional data, and quality inspection reports are stored and transmitted in centralized plaintext, posing risks of data leakage, tampering, and forgery, resulting in insufficient credibility of battery quality traceability data, difficulties in cross-entity quality traceability collaboration between upstream and downstream enterprises, and failure to meet the requirements of the Data Security Law and the compliance management of the power battery industry. In existing technologies, some solutions attempt to improve the accuracy of battery defect identification using AI visual inspection, but they generally suffer from problems such as single modality, weak anti-interference ability, and lack of data security protection, failing to meet the integrated requirements of high-precision inspection, high-speed production line adaptation, and secure and reliable data traceability. Therefore, there is an urgent need to develop a blade battery intelligent inspection system that combines high-precision inspection, low-latency secure transmission, and end-to-end reliable traceability to address the pain points of existing technologies in the industry. Summary of the Invention

[0003] Purpose of the invention The purpose of this invention is to address the shortcomings of existing technologies by providing a multimodal defect detection and random obfuscation encrypted transmission system for blade batteries. This system solves the industry pain points of traditional detection technologies, such as low accuracy, poor adaptability to operating conditions, cumbersome procedures, weak data security, and unreliable traceability. It achieves integrated implementation of high-precision intelligent detection of blade batteries, low-latency encrypted transmission of detection data, and trusted blockchain traceability throughout the entire life cycle, while taking into account both production line efficiency and industrial data compliance requirements.

[0004] Technical solution To achieve the above objectives, the present invention adopts the following technical solution: This invention provides a multimodal defect detection and random obfuscation encryption transmission system for blade batteries, comprising a multimodal sensing and acquisition module, an AI intelligent analysis module, a robot flexible control module, a dynamic random obfuscation encryption module, a blockchain trusted evidence storage module, and an application interaction output module. The multimodal sensing and acquisition module performs multi-dimensional synchronous data acquisition and reflection suppression; the robot flexible control module enables non-destructive cell grasping and tactile data acquisition; the AI ​​intelligent analysis module performs defect identification and size calculation based on a multimodal fusion algorithm; the dynamic random obfuscation encryption module achieves low-latency encryption of all types of inspection data; the blockchain trusted evidence storage module ensures tamper-proof data storage and traceability; and the application interaction output module enables data visualization, report generation, and system integration. The modules work together to overcome the accuracy bottleneck of single visual inspection through four-modal data fusion technology, solve industrial data security issues through lightweight dynamic encryption technology, and build a trusted traceability system through consortium blockchain evidence storage technology. The entire process meets industrial indicators such as 10-second ultra-fast full inspection, 5-second size recognition, and more than 95% defect recognition accuracy, while achieving 100% data encryption compliance rate and traceability coverage.

[0005] Beneficial effects Compared with the prior art, the present invention has the following significant advantages: 1. Significantly improved detection accuracy and adaptability: Innovatively integrates RGB, 3D depth, tactile, and contour data, combined with attention-enhanced YOLOv8 algorithm and cross-modal knowledge distillation technology, effectively solving the problems of high reflectivity imaging interference and missed detection of tiny defects in blade batteries. The minimum defect recognition accuracy reaches 0.01mm, and the defect recognition accuracy rate is ≥95%, far exceeding traditional visual inspection solutions. At the same time, it realizes the integrated operation of defect detection and size measurement, streamlines production line processes, and adapts to the 7×24-hour high-speed mass production cycle. 2. Comprehensive upgrade of industrial data security: The self-developed lightweight dynamic random obfuscation encryption mechanism abandons the traditional fixed key encryption method. It generates a dynamic encryption seed based on the unique characteristics of the device, and achieves full data encryption through block obfuscation and location verification. The encryption latency is less than 20ms, which does not affect the production line operation efficiency. It can effectively resist data theft, tampering and replay attacks, and meet the industrial data security compliance requirements. 3. Quality traceability with credibility: A multi-party consortium blockchain evidence storage architecture is built, and the hash fingerprint of the test data is uploaded to the blockchain in real time. The data is permanently immutable and verifiable throughout the entire process, which solves the problems of insufficient credibility and difficulty in cross-entity collaboration in traditional centralized traceability systems, and realizes the full life cycle of blade batteries from production, assembly, after-sales service to recycling with reliable traceability. 4. Highly applicable to industrialization: The overall design adopts a modular and lightweight approach, with mature and easily iterative software and hardware architecture. It is compatible with multiple blade battery specifications and adaptable to both large-scale mass production lines and small-batch customized testing scenarios. Replacing manual testing can significantly reduce enterprise labor and quality control costs, with a short investment return cycle and extremely high market promotion value. Detailed Implementation

[0006] The present invention will be further described in detail below with reference to specific embodiments. This invention discloses a multimodal defect detection and random obfuscation encrypted transmission system for blade batteries. In actual industrial applications, the entire device is installed at the mass production testing station for power batteries. The hardware uses domestically produced industrial-grade equipment, and the software algorithm is deployed in a modular manner. The specific implementation process is as follows: 1. Equipment initialization and calibration: After the system is powered on, it automatically completes the precision calibration of the 20-megapixel industrial camera and 3D laser sensor, adjusts the brightness and angle of the multi-angle polarized light source, and eliminates interference from the high reflectivity of the battery cells; calibrates the motion trajectory of the 6-axis robot and the gripping force of the dexterous hand to ensure repeatability accuracy of ±0.01mm, and completes the initial deployment of the encryption module and blockchain node. 2. Multimodal data synchronous acquisition: The production line transports the blade battery to the inspection station. The robot's dexterous hand flexibly grasps and precisely fixes the battery cell. Four industrial cameras and 3D sensors are started simultaneously, and the acquisition of RGB images and 3D depth contour data of the battery cell on six sides is completed within 10 seconds. At the same time, the dexterous hand's tactile sensor array completes multi-point pressure sampling to obtain the surface bonding data of the battery cell, forming a four-modal fusion dataset. 3. AI-powered intelligent defect and size analysis: The system performs frequency domain filtering, gamma enhancement, and reflection suppression preprocessing on the collected data. It extracts multi-scale defect features through the attention-enhanced YOLOv8 model and combines visual and tactile data with cross-modal knowledge distillation to accurately identify defects such as scratches, dents, encapsulation damage, and tab abnormalities, completing pixel-level annotation and grade determination. At the same time, it extracts tactile features through 1D-CNN and combines them with a 3D reconstruction algorithm to accurately fit the cell's length, width, thickness, flatness, and pose within 5 seconds, with the size error controlled within ±0.02mm. 4. Dynamic random obfuscation encryption processing: The system automatically distinguishes between structured size data and unstructured image data, extracts the unique hardware features of the device to generate a dynamic encryption seed, performs block cutting and tensor convolution obfuscation on the data, binds a unique position check code, and completes full data encryption. The overall encryption latency is controlled within 20ms, and the encrypted data is transmitted through a dedicated secure channel. 5. Trusted Blockchain Evidence Storage: The encrypted test data is hashed to generate a unique data fingerprint. The data fingerprint, timestamp, device number, production batch, and operator information are packaged and uploaded to multiple consortium blockchain nodes to complete consensus evidence storage. The on-chain confirmation time is less than 1 second, ensuring that the data is permanently immutable. 6. Results Output and Closed-Loop Management: The system automatically sorts OK / NG cells based on the test results. The AI ​​engine automatically calculates defect distribution, defect rate trends, and process deviations, generating standardized PDF / Excel quality inspection reports. All encrypted data, evidence storage information, and test reports are synchronously connected to the factory's MES / ERP system to achieve full closed-loop management of production quality control, data traceability, and compliance archiving. 7. Long-term iterative optimization: During system operation, real defect samples from the production line are continuously collected, and the sample library is expanded through the GAN algorithm to achieve normalized model iteration; at the same time, the encryption and on-chain operation status is monitored in real time to ensure long-term stable operation of the equipment and adapt to the long-term production needs of the production line. Attached Figure Description

[0007] Figure 1 This is a block diagram of the overall system architecture of the present invention; Figure 2 This is a schematic diagram of the layout of the multimodal sensing and acquisition module of the present invention; Figure 3 This is a schematic diagram of the dynamic random obfuscation encryption algorithm of the present invention; Figure 4 This is a schematic diagram of the blockchain data on-chain evidence storage process of the present invention; Figure 5 This is a general flowchart of the overall detection and data security transmission workflow of this invention.

[0008] The technical features not described in detail in this invention are all existing mature technologies in the field. The core innovation of this invention is the combination of a four-modal fusion high-precision defect detection architecture, a vision-tactile integrated size measurement mechanism, a dynamic random obfuscation low-latency encryption algorithm, and a consortium blockchain full-link trusted traceability system. By integrating multiple technologies, it solves the technical pain points of the industry and has outstanding novelty, inventiveness and industrial applicability, which meets the requirements for national invention patent authorization.

Claims

1. A system for multi-modal defect detection and stochastic obfuscated encrypted transmission of a blade battery, comprising: include: Multimodal perception and acquisition module, AI intelligent analysis module, robot flexible control module, dynamic random obfuscation and encryption module, blockchain trusted evidence storage module, application interaction output module; The multimodal sensing and acquisition module is used to simultaneously acquire RGB visible light images, 3D laser depth contour data, robot tactile pressure data, and cell outline scanning data of the blade battery, so as to realize the acquisition of all-round multi-dimensional information of the six sides of the cell, while suppressing the high reflectivity interference of the battery shell through multi-angle polarized light source. The robot's flexible control module is equipped with a 6-axis industrial robot and a multi-degree-of-freedom dexterous hand, and integrates a tactile sensor array to achieve non-destructive flexible grasping, precise positioning, and multi-point tactile sampling of the blade battery, and completes cell size fitting and calibration in conjunction with visual data. The AI ​​intelligent analysis module has a built-in attention-enhanced YOLOv8 detection model, cross-modal knowledge distillation algorithm and GAN data augmentation unit, which is used to perform fusion analysis on multimodal acquired data, automatically identify various defects such as scratches, pits, coating damage, and abnormal tabs on the cell surface, and complete pixel-level defect annotation, defect level judgment and accurate cell size calculation. The dynamic random obfuscation encryption module is used to distinguish between structured size data and unstructured image data. It generates a dynamic encryption seed based on the device's unique feature code and completes real-time encryption of all types of detection data through data block permutation, tensor convolution obfuscation, and position check code matching, thereby achieving low-latency secure transmission. The blockchain trusted evidence storage module is used to perform hash operations on the encrypted test data to generate a unique data fingerprint, and package the data fingerprint, timestamp, equipment number, production batch, and operator information onto the blockchain for evidence storage, so as to realize the immutability of test data and trusted traceability across entities. The application interaction output module is used to realize defect visualization, detection data report generation, and AI defect root cause analysis report output. At the same time, it connects to the factory MES / ERP system to complete standardized data uploading and full-process quality traceability.

2. The system of claim 1, wherein, The multimodal sensing and acquisition module includes four 20-megapixel global shutter industrial cameras, a high-precision 3D line laser contour sensor, and a multi-angle ring + strip combined polarization light source. Through a multi-device synchronous triggering mechanism, it can complete the six-sided full-dimensional imaging and contour data acquisition of the blade battery within 10 seconds, with the smallest identifiable defect accuracy reaching 0.01mm.

3. The system of claim 1, wherein, The robot's flexible control module adopts a force-position hybrid control strategy and is equipped with a 12+ degree-of-freedom dexterous hand. It extracts tactile signal features through a 1D-CNN algorithm and completes high-precision calculations of the blade battery's length, width, thickness, flatness, and spatial pose within 5 seconds. The dimensional measurement error is controlled within ±0.02mm, and the repeatability accuracy is ±0.01mm.

4. The system according to claim 1, characterized in that, The algorithm optimization process of the AI ​​intelligent analysis module includes: image preprocessing through frequency domain filtering and gamma image enhancement; strengthening multi-scale feature extraction of micro-defects by relying on the attention mechanism; generating rare defect simulation samples using the GAN algorithm to solve the problem of small sample detection in industry; and improving the robustness of defect recognition under complex working conditions by fusing visual and tactile features through cross-modal knowledge distillation. The overall defect recognition accuracy is ≥95%, the false negative rate is ≤1%, and the false negative rate is ≤2%.

5. The system according to claim 1, characterized in that, The encryption process of the dynamic random obfuscation encryption module is as follows: S1. Classify and identify the data types detected, classifying cell size, defect parameters, etc. as structured data, and imaging images and defect heat maps as unstructured data; S2. Extract the unique feature code of the device hardware as the encryption base, generate a dynamic random confusion tensor seed, and avoid the risk of fixed key cracking. S3. Divide the two types of data into blocks, combine tensor convolution operation to complete data confusion and replacement, and simultaneously bind a unique position check code. S4. Data transmission is completed through a dedicated encrypted channel, with an overall encryption latency of <20ms and a decryption accuracy of 100%, which can resist the risks of brute-force attacks, replay attacks, and side-channel data theft.

6. The system according to claim 1, characterized in that, The blockchain trusted evidence storage module adopts a multi-party consortium blockchain architecture involving production enterprises, quality inspection departments, supply chain manufacturers, and after-sales maintenance. The data on-chain confirmation time is less than 1 second, the data verification response time is less than 100ms, and the stored data is permanently tamper-proof, supporting online traceability, auditing, and compliance verification by multiple entities.

7. The system according to claim 1, characterized in that, The system adopts a six-layer modular collaborative architecture, consisting of a perception and acquisition layer, an AI multimodal analysis layer, a motion control layer, a data security encryption layer, a blockchain evidence storage layer, and an application output layer, from top to bottom. Each module iterates independently and works in tandem, supporting continuous operation 24 / 7 and adapting to various blade battery testing scenarios.

8. A method for multimodal defect detection and random obfuscation encrypted transmission of blade batteries, characterized in that, The system applied to any one of claims 1-7 includes the following steps: Step 1: Equipment initialization and calibration. Complete the precision calibration and light source parameter debugging of industrial cameras, 3D sensors, and robot dexterous hands, and build a multimodal synchronous acquisition link and encrypted transmission channel. Step 2: The robot's dexterous hand flexibly grasps and precisely positions the blade battery, while the multimodal perception module simultaneously collects RGB images of the six sides of the battery cell, 3D depth contours, and tactile pressure data. Step 3: The AI ​​intelligent analysis module performs fusion preprocessing and algorithm reasoning on multimodal data to complete defect identification, annotation, grading and accurate size calculation; Step 4: The dynamic random obfuscation encryption module performs real-time encryption processing on all raw detection data and analysis results data; Step 5: The blockchain evidence storage module performs hash calculations on the encrypted data and packages the core information for blockchain evidence storage. Step 6: The system completes the OK / NG intelligent sorting of battery cells based on the test results, automatically generates a standardized quality inspection report, and uploads encrypted data to the factory's information system simultaneously to support subsequent traceability and compliance verification.

Images