Lithium battery packaging state detection system and method based on artificial intelligence

Abstract

The invention discloses a lithium battery packaging state detection system and method based on artificial intelligence, and relates to the field of detection and analysis, and the system comprises a management module which is used for carrying out the general control of a global function module, and carrying out the editing and issuing of a control instruction; the process reading module is used for reading standard information of each packaging process, supporting self-definition to modify the process standard, and obtaining the network distribution and power supply authority of the functional module; the risk analysis module is used for receiving data submitted by the image recognition module and the machine recognition module, comparing the data with a reference template provided by the standard definition module, and obtaining risk judgment data; through real-time collection of images and operation data and comparison with a reference template, potential risks in the packaging process can be recognized and evaluated in real time, problems can be detected at the initial stage of occurrence through a rapid feedback mechanism, the fault rate is effectively reduced, and the function of data collection setting can be automatically adjusted according to a risk analysis result.

Description

Technical Field

[0001] The present invention relates to the field of detection and analysis technology, and specifically to a lithium battery packaging status detection system and method based on artificial intelligence. Background Art

[0002] The lithium battery market is experiencing rapid growth as electric vehicles, renewable energy equipment, and consumer electronics gain widespread popularity. This demand forces manufacturers to improve production efficiency and product quality to meet market demand. Furthermore, lithium batteries are widely used in multiple industries, including automotive, electronics, and smart homes, increasing the requirements for safety and reliability. Therefore, ensuring the packaging quality of lithium batteries and comprehensively monitoring their production processes are particularly important. An efficient detection system can promptly identify potential risks and reduce safety hazards. Currently, the manufacturing industry is undergoing a global transformation towards intelligent manufacturing. By combining the Internet of Things, big data, and artificial intelligence, companies can extract valuable information from data and achieve efficient production process management. AI-based inspection systems are a concrete manifestation of this industry trend. Through real-time data collection and intelligent analysis, they promote the automation and intelligence of manufacturing processes. Traditional detection methods typically require manual operations and multiple steps, resulting in slow response times. Once a problem is discovered, manual analysis and resolution are required, reducing production efficiency. Many existing detection systems lack intelligent risk assessment and traceability analysis tools, making it difficult to deeply mine data for correlations and achieve real-time monitoring of the production process, leading to delays in problem discovery and resolution. Many existing testing solutions lack the ability to be customized for specific process requirements, making it difficult to optimize the testing standards for specific processes. Many existing testing solutions lack the ability to be customized for specific process requirements, making it difficult to optimize the testing standards for specific processes. Summary of the Invention

[0003] In view of the above-mentioned shortcomings of the prior art, the present invention provides a lithium battery packaging status detection system and method based on artificial intelligence, which can effectively solve the problems of the prior art.

[0004] To achieve the above objectives, the present invention is implemented through the following technical solutions: The present invention discloses a lithium battery packaging state detection system based on artificial intelligence, comprising: Management module, used for global control and instruction issuance; Process reading module, used to read and modify packaging process standard information; An image recognition module, which is used to collect image data of the packaging process and convert it into a machine-readable format according to a number of preset levels of collection status; A machine recognition module is used to collect the operation data of the packaging process according to a preset number of levels of collection status and convert it into a machine-readable format; Standard definition module, used to generate benchmark templates for packaging processes; The risk analysis module is used to compare the collected image data and operation data with the benchmark template to obtain risk judgment data. The misalignment module is set in the risk analysis module. It divides the data domain and configures the comparison order based on the benchmark template. It judges risks based on conventional comparison and random comparison strategies, and generates misalignment error distribution data in the random comparison. The upstream module is used to trace abnormal risks, analyze and integrate the time, machine status, operator, image evidence and dislocation error distribution data of abnormal data, and form event groups with similar characteristics through timestamp alignment and correlation analysis; The correction module is used to adjust the collection level settings of the image recognition module and the machine recognition module based on the traceability results, and conduct secondary collection and analysis after the adjustment. The module uses a reinforcement learning algorithm to analyze the success detection rate, false alarm rate, and missed alarm rate of the adjusted results in the secondary collection. The risk results after each adjustment are compared with the preset targets to evaluate the effectiveness of the adjustment. The module obtains the influence coefficient of the changes in key parameters in the benchmark template on the accuracy of the adjustment results. Based on the influence coefficient of reinforcement learning, the module generates and applies new collection level setting rules. The alarm module is used to trigger an alarm and notify the management module when the risk still exists after adjustment.

[0005] Furthermore, the working logic of the dislocation module is as follows: the dislocation module divides the domains into several parts based on the reference template and configures a comparison order for each domain; Under regular comparison, the corresponding domains are matched according to the order in which the submitted data arrives, and then compared one by one according to the configuration order of the current domains. If the error amount exceeds the preset limit within the preset period under the regular comparison of the risk analysis module, the submitted data is considered to be risky. If the error count within a preset period does not exceed the predetermined limit during the regular comparison in the risk analysis module, the currently submitted data will be divided into several domains of the benchmark template for random comparison. The domain participating in the comparison for the first time will avoid the first trigger domain of the regular comparison. Based on historical risk data, the error frequency of different domains in past comparison results and the performance data of the risk level of similar samples are evaluated and obtained. The priority of each domain in the random comparison is dynamically adjusted based on this performance data. The domains with higher risks will be given higher priority in subsequent comparisons. If the amount of errors within a preset period exceeds a predetermined limit under random comparison, the submitted data is considered to be at risk and misalignment error distribution data is generated.

[0006] Furthermore, the working logic of the association analysis of the tracing module is: by extracting and integrating different source data, selecting key features of job image data and job operation data, identifying the association rules between machine job operation features and job image features by applying the association analysis algorithm, and associating several machine job operation data with the defects of specified job image data.

[0007] Furthermore, the correction module adjusts the collection level settings of the image recognition module and the machine recognition module in accordance with the following working logic: if the image data collected by the image recognition module is risky and the machine data collected by the machine recognition module is risk-free, the collection level setting of the machine recognition module is adjusted according to a preset indicator and submitted to the image recognition module; If the machine data collected by the machine recognition module is at risk, and the image data collected by the image recognition module is not at risk, the collection level setting of the image recognition module is adjusted according to the preset indicators; After the adjustment, data is collected again and the risk analysis results of the risk analysis module are received. If the risk still exists, the alarm module is triggered and it is decided whether to trigger the traceback module again. Otherwise, the adjustment is terminated.

[0008] Furthermore, the management module is interactively connected to a storage module via a wireless network, and the storage module is used to provide local and cloud storage of collected data and analyzed data, and generate system operation reports according to a preset period.

[0009] Furthermore, during the operation phase, the risk analysis module compares the data collected by the image recognition module and the machine recognition module with the benchmark template of the standard definition module, and triggers a jump based on the comparison result. If there is a risk, it jumps to the upstream module for further operation; if there is no risk, it jumps to the image recognition module and the machine recognition module for further operation.

[0010] Furthermore, the tracing module is interactively connected to a target definition module via a wireless network. The target definition module is used to preset associated targets of several key features in the tracing result process of the tracing module, and supports custom editing. The associated targets include: any image acquisition component and operation data acquisition component.

[0011] Furthermore, the management module is interactively connected to the process reading module, the image recognition module and the machine recognition module through a wireless network, the risk analysis module is interactively connected to the image recognition module, the machine recognition module, the standard definition module, the tracing module and the correction module through a wireless network, and the alarm module is interactively connected to the tracing module and the correction module through a wireless network.

[0012] A method for detecting the packaging status of a lithium battery based on artificial intelligence comprises the following steps: Step 1: Issue control instructions through the internal IoT network, collect and store data collected and analyzed during the packaging process, and generate system operation reports regularly; Step 2: Read various standard process information during the lithium battery packaging process and support users to customize and adjust the process standards; Step 3: Deploy image acquisition equipment in the corresponding packaging process, set the acquisition position, angle and frequency, collect the operation images, and convert them into a machine-readable format to generate the operation image data of the packaging process; Step 4: Deploy operation data collection equipment, set the collection frequency and target, collect the operation status of the packaging process, and convert it into a machine-readable format to generate equipment operation data during the packaging process; Step 5: Develop standard operating image data and operation data for each packaging process to form a benchmark template; Step 6: Compare the collected image data and operational data with the benchmark template. Through comparison and analysis, risks are identified. If risks are found, further analysis will be triggered. If no risks are found, regular operation image and operational data collection will continue. Step 7: Once a risk is identified, the time of occurrence and the relevant machine status are recorded, and the activities of relevant operators and the generated image evidence are tracked. By integrating data from different sources and performing correlation analysis, key features associated with the risk are identified and event clusters are formed. Step 8: Based on the results of the traceability analysis, adjust the collection settings for image and operational data. If risks are found in the image data but normal machine data, adjust the collection level of operational data according to the preset indicators, and vice versa. Step 9: After the adjustment is completed, a second operation image and operation data collection is carried out. If the data collected in the second collection still identifies risks, corresponding alarm measures will be taken according to the preset risk level and relevant personnel will be notified in real time.

[0013] Furthermore, the calculation expression of the comparison analysis process in step 6 is: ; Where, Represents the judgment result, indicating whether there is abnormal risk. Represents the number of machine data involved in the comparison, Represents the i-th image data The weight of Represents the i-th image data, Represents a set of standard parameters, represents the comparison function between image data and standard parameters, Represents the number of image data involved in the comparison, Represents the jth machine parameter The weight of Represents the comparison function between machine parameters and standard parameters, used to evaluate the abnormality score obtained by comparing machine parameters with standard parameters. Represents the j-th machine data; The working logic of the calculation expression of the comparison analysis process is: Receive the collected image data and operating data, score each image data and each machine parameter, and determine their deviation from the standard parameters; All scores are weighted and summed according to the weights; Calculate the final judgment result through the activation function; If the judgment result is greater than the set threshold, it is judged that there is an abnormal risk; otherwise, it is considered that there is no abnormality.

[0014] Compared with the prior art, the technical solution provided by the present invention has the following beneficial effects: 1. Through the real-time collection of images and operating data, and comparison with benchmark templates, potential risks in the packaging process can be identified and evaluated in real time. The rapid feedback mechanism enables problems to be detected in the early stages of their occurrence, effectively reducing the failure rate. The function of automatically adjusting data collection settings based on risk analysis results enables the system to flexibly optimize data collection methods under different production conditions and process parameters, thereby improving detection accuracy.

[0015] 2. By integrating multi-source information from image acquisition and operation data, a more comprehensive risk tracing analysis can be performed. Through correlation analysis technology, the potential connection between machine operation characteristics and operation image characteristics can be identified, and characteristic patterns and rules related to risk occurrence can be discovered, providing more guiding suggestions for risk management in the production process, enabling managers to improve the production process in a targeted manner and enhance overall safety, thereby significantly improving the accuracy of fault prediction.

[0016] 3. Through intelligent risk detection and feedback mechanisms, the system can reduce human errors, improve the degree of automation in the production process, and reduce production costs. At the same time, real-time monitoring and risk warning mechanisms also significantly improve the safety of the lithium battery packaging process, providing reference data support for the company's standardized production. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. Those skilled in the art can also derive other drawings based on these drawings without inventive effort.

[0018] Figure 1 This is a schematic diagram of the framework of a lithium battery packaging status detection system based on artificial intelligence in the present invention; Figure 2 The figure is a flow chart of a lithium battery packaging status detection method based on artificial intelligence in the present invention.

[0019] The numbers in the figure represent: 1. Management module; 2. Process reading module; 3. Image recognition module; 4. Machine recognition module; 5. Standard definition module; 6. Risk analysis module; 7. Traceability module; 8. Correction module; 9. Alarm module; 10. Target definition module; 11. Storage module. DETAILED DESCRIPTION

[0020] To make the purpose, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0021] The present invention will be further described below with reference to the embodiments.

[0022] This embodiment is a lithium battery packaging status detection system based on artificial intelligence, such as Figure 1 As shown, including: Management module 1 is used to control the global functional modules and edit and issue control instructions. Management module 1 is interactively connected to storage module 11 via a wireless network. Storage module 11 is used to provide local and cloud storage for collected and analyzed data and generate system operation reports according to preset periods. Generating reports according to preset periods ensures data security and traceability, while facilitating long-term analysis and performance evaluation. Process reading module 2 is used to read the standard information of each packaging process, support customized modification of process standards, and obtain network configuration and power supply permissions for functional modules. This enables the production process to quickly adapt to changing market demands or technological advances, thereby improving process flexibility. Image recognition module 3, used to deploy several image acquisition components in the corresponding packaging process, set the position, angle and acquisition frequency of the image acquisition components, corresponding to several preset levels of acquisition status, perform operation image acquisition, and convert them into machine-readable format; The machine identification module 4 is used to deploy several operation data collection components in the corresponding packaging process, set the collection frequency and collection target of the operation data collection components, preset several levels of collection status, read the operation data of the operation execution components of each packaging process, and convert it into a machine-readable format; Standard definition module 5, used to formulate standard operation image data and standard operation operation data of each packaging process to form a benchmark template; The risk analysis module 6 is used to receive the data submitted by the image recognition module 3 and the machine recognition module 4, compare it with the benchmark template provided by the standard definition module 5, and obtain risk judgment data. During the operation phase, the risk analysis module 6 compares the data collected by the image recognition module 3 and the machine recognition module 4 with the benchmark template of the standard definition module 5, and triggers a jump based on the comparison result. If a risk exists, the module jumps to the upstream module 7 for further operation; if no risk exists, the module jumps to the image recognition module 3 and the machine recognition module 4 for further operation. By comparing the collected data with the benchmark template in real time, the module can quickly determine potential risks and decide on subsequent steps, thereby improving the ability to deal with abnormal risks through instant feedback. The risk analysis module 6 is internally provided with a staggered module, which divides the domains into several domains based on the benchmark template and configures an order for each domain. Under normal comparison, the corresponding domains are matched according to the order in which the submitted data arrives, and are compared one by one according to the configuration order of the current domains. If the error amount exceeds the predetermined limit within the preset period under the normal comparison of the risk analysis module 6, the submitted data is deemed to be at risk. If the error amount does not exceed the predetermined limit within the preset period under the normal comparison of the risk analysis module 6, the currently submitted data is randomly divided into several domains of the benchmark template for comparison. The domain participating in the comparison for the first time avoids the first triggering domain of the normal comparison. Based on the error rate calculation of each domain in the historical risk data in the historical comparison and the number of comparisons marked as risk in each domain, the similarity with the known risk sample is calculated, and the performance data of the error frequency of different domains in the past comparison results and the risk level of similar samples are evaluated. Different weights are assigned to different performance data. A dynamic priority score is calculated for each domain based on the weighted performance data. The priority of each domain in the random comparison is dynamically adjusted based on the performance data. Domains with higher risks will be given higher priority in subsequent comparisons. Recalculate the priority score of each domain at every preset period and update the random comparison order during comparison; If the error count exceeds a predetermined limit within a preset period during random comparison, the submitted data is deemed risky and misalignment error distribution data is generated. By dividing the domains and configuring the order, the comparison strategy can be dynamically adjusted according to actual needs to effectively respond to different production conditions or product types. Random comparison allows the system to introduce a certain degree of randomness into monitoring, reducing false alarms caused by fixed patterns and enhancing the robustness of the system. The tracing module 7 is used to receive the risk judgment content of the risk analysis module 6, conduct traceability analysis on the abnormal risk content, obtain the specific time, relevant machine status, operator, image evidence and dislocation error distribution data in the abnormal risk content, integrate data from different sources, sort and merge them by time, align timestamps, and perform association analysis to determine whether there are association rules related to abnormal risks, and obtain event groups with similar characteristics; the working logic of the association analysis is: by extracting the integrated different source data, selecting the key features of the operation image data and operation operation data, and identifying the association rules between the machine operation characteristics and the operation image characteristics by applying the association analysis algorithm, a number of machine operation operation data are associated with the defects of the specified operation image data; in-depth analysis of the abnormal risk content is carried out, and by integrating data from different sources, the association rules between abnormal events are identified, which helps to track the root cause and provide data support for improvement; Scenario example: The temperature and pressure of a machine are found to be outside the normal range, and the image data shows the characteristics of the defect; During data integration, it was discovered that there were several machine maintenance records before the timeline of these abnormal events; Correlation analysis showed that temperature and pressure did not return to normal parameters after maintenance, and these parameters were strongly correlated with defects; Final verification confirmed that the temperature anomaly caused a change in material properties, resulting in defects; The tracing module 7 is interactively connected to the target definition module 10 via a wireless network. The target definition module 10 is used to preset the associated targets of several key features in the tracing result process of the tracing module 7 and supports custom editing. The associated targets include: any image acquisition component and operation data acquisition component; Correction module 8 is used to receive the traceability analysis content of tracing module 7, adjust the collection settings of image recognition module 3 and machine recognition module 4 according to the traceability analysis results, and perform secondary collection and analysis after the adjustment. It uses the reinforcement learning algorithm to analyze the success detection rate, false alarm rate, and missed alarm rate of the adjusted results in the secondary collection, compare the risk results after each adjustment with the preset target, evaluate the effectiveness of the adjustment, obtain the impact coefficient of the change of key parameters in the benchmark template on the accuracy of the adjustment result, and generate and apply new collection level setting rules based on the reinforcement learning impact coefficient; Among them, the reinforcement learning algorithm is triggered at specific conditional intervals. In its specific implementation process, specific events are set as trigger conditions. Trigger conditions include: the risk of the previous collection result exceeds a certain threshold, the machine status has a negative impact, or a new process standard is released; After each data collection, the reinforcement learning algorithm is not applied immediately. Instead, the data is cached in a temporary storage area until the trigger condition is met, and then the data is comprehensively analyzed. The collected data includes the image recognition and machine recognition results within multiple trigger intervals, as well as the corresponding success detection rate, false alarm rate, and missed alarm rate; Once the trigger conditions are met, a reinforcement learning algorithm is called to analyze the accumulated data. The algorithm will evaluate the impact of each risk adjustment and compare it with the preset target, while calculating the influence coefficient of the change of key parameters in the benchmark template on the accuracy of the result. Based on the output of reinforcement learning, the collection level setting rules are updated to generate new strategies. The adjusted rules can be applied again after the new trigger conditions are met, thereby reducing the burden of frequent system updates. After each reinforcement learning trigger, monitor the adjustment effect to ensure that the generated rules are effectively fed back into the collection and risk analysis mechanism. Through continuous evaluation, ensure that the system is not overly dependent on a single trigger condition, but adapts to the situation; Adjust the data collection strategy based on the traceability analysis results to improve the efficiency and accuracy of data collection. Dynamic adjustment helps optimize the detection process in real time. Its operating logic is as follows: If the image collected by the image recognition module 3 is risky, but the machine data collected is risk-free, the collection level setting of the machine recognition module 4 is adjusted according to the preset indicators and submitted to the image recognition module 3; If the machine data collected by the machine recognition module 4 is risky, but the collected images are not risky, the collection level setting of the image recognition module 3 is adjusted according to the preset indicators; After the adjustment, the secondary data collection is carried out again and the risk analysis results of the risk analysis module 6 are received. If the risk still exists, the alarm module 9 is triggered to select whether to trigger the traceback module 7 again. Otherwise, the adjustment is terminated. The correction module 8 synchronously feeds the recognition results back to the management module 1, classifies the recognition results, and generates a visual report; The alarm module 9 is used to be triggered when the risk analysis module 6 re-identifies the existence of risks after the correction module 8 adjusts and processes. According to the preset risk level, the corresponding alarm method is selected to issue an alarm prompt to the management module 1. According to the risk level, the appropriate alarm method is selected to issue a timely alarm, thereby minimizing potential losses and promoting the rapid handling of problems.

[0023] As an implementation method in this embodiment, Figure 1 As shown, the management module 1 is interactively connected with the process reading module 2, the image recognition module 3 and the machine recognition module 4 through a wireless network, the risk analysis module 6 is interactively connected with the image recognition module 3, the machine recognition module 4, the standard definition module 5, the traceability module 7 and the correction module 8 through a wireless network, and the alarm module 9 is interactively connected with the traceability module 7 and the correction module 8 through a wireless network.

[0024] In the specific implementation of this embodiment, the management module 1 controls the global function module, the process reading module 2 reads or edits a number of packaging processes, the image recognition module 3 collects images of the corresponding processes, the machine recognition module 4 collects operating machine parameters, the standard definition module 5 defines standard collected data, and the risk analysis module 6 judges and analyzes the real-time collected data to determine whether there is any abnormal risk. If the image acquisition shows a risk, but the machine acquisition shows no abnormality, then based on the abnormal risk content, the tracing module 7 performs a traceability analysis on the machine acquisition parameters and submits the traceability results to the correction module 8 as a reference for adjusting the machine acquisition threshold of the machine recognition module 4; If there is a risk in machine acquisition but no risk in image acquisition, then based on the abnormal risk content, the correction module 8 adjusts the acquisition frequency and angle of the image acquisition component in the corresponding image recognition module 3, and re-acquires and analyzes. If there is still a risk, the alarm module 9 sends an alarm. If there is no risk, the tracing module 7 performs a traceability analysis on the machine acquisition parameters involved in the risk prompt, and submits the traceability results to the correction module 8 as a reference for adjusting the machine acquisition threshold of the machine recognition module 4.

[0025] In other aspects, this embodiment also provides a lithium battery packaging status detection method based on artificial intelligence, such as Figure 2 As shown, the following steps are included: Step 1: Issue control instructions through the internal IoT network, collect and store data collected and analyzed during the packaging process, and generate system operation reports regularly; Step 2: Read various standard process information during the lithium battery packaging process and support users to customize and adjust the process standards; Step 3: Deploy image acquisition equipment in the corresponding packaging process, set the acquisition position, angle and frequency, collect the operation images, and convert them into a machine-readable format to generate the operation image data of the packaging process; Step 4: Deploy operation data collection equipment, set the collection frequency and target, collect the operation status of the packaging process, and convert it into a machine-readable format to generate equipment operation data during the packaging process; Step 5: Develop standard operating image data and operation data for each packaging process to form a benchmark template; Step 6: Compare the collected image data and operational data with the benchmark template. Through comparison and analysis, risks are identified. If risks are found, further analysis will be triggered. If no risks are found, regular operation image and operational data collection will continue. Step 7: Once a risk is identified, the time of occurrence and the relevant machine status are recorded, and the activities of relevant operators and the generated image evidence are tracked. By integrating data from different sources and performing correlation analysis, key features associated with the risk are identified and event clusters are formed. Step 8: Based on the results of the traceability analysis, adjust the collection settings for image and operational data. If risks are found in the image data but normal machine data, adjust the collection level of operational data according to the preset indicators, and vice versa. Step 9: After the adjustment is completed, a second operation image and operation data collection is carried out. If the data collected in the second collection still identifies risks, corresponding alarm measures will be taken according to the preset risk level and relevant personnel will be notified in real time.

[0026] Compared with existing technologies, this system integrates multiple data sources, enables real-time information sharing and integrated analysis, and provides support for comprehensive and systematic problem solving. Through automated and intelligent data collection and analysis, it significantly reduces errors caused by human factors and improves the accuracy and consistency of detection. Through real-time data collection and automatic risk assessment, it can quickly respond to and promptly identify potential problems, improve the real-time performance and flexibility of the production process, and automatically generate risk analysis reports and decision-making basis, thereby improving the speed and accuracy of data processing. Through in-depth correlation analysis and feature extraction, it is possible to identify potential risk patterns and regularities, providing more scientific decision-making support. Through full-process monitoring of the production process and real-time data feedback, it is possible to adjust the collection strategy in a timely manner, reduce production risks, and improve the overall safety and efficiency of the production line. It allows users to customize process standards and collection strategies according to their own needs, flexibly adapt to different production environments, and improve the adaptability and flexibility of the system.

[0027] In this embodiment, a calculation expression for the comparison analysis process is provided, specifically: ; Where, Represents the judgment result, indicating whether there is abnormal risk, 1 means yes, 0 means no, Represents the number of machine data involved in the comparison, Represents the i-th image data The weight of reflects its importance in the judgment stage. Represents the i-th image data, Represents a set of standard parameters, including standard data within the normal working range, for comparative analysis. Represents a comparison function between image data and standard parameters, used to evaluate the abnormality score obtained when comparing image data with standard parameters. It usually uses similarity to characterize the quality of the image or whether there are defects. Represents the number of image data involved in the comparison, Represents the jth machine parameter The weight of Represents the comparison function between machine parameters and standard parameters, used to evaluate the abnormality score obtained by comparing machine parameters with standard parameters. Represents the j-th machine data; The working logic of the calculation expression of the comparison analysis process is: Receive the collected image data and operating data, score each image data and each machine parameter, and determine their deviation from the standard parameters; All scores are weighted and summed according to the weights; Calculate the final judgment result through the activation function; If the judgment result is greater than the set threshold, it is judged that there is an abnormal risk; otherwise, it is considered that there is no abnormality; This formula takes into account both image data and machine parameters, and can comprehensively evaluate the risks in the packaging process and avoid misjudgments that may be caused by a single data source. This formula takes into account both image data and machine parameters, and can comprehensively evaluate the risks in the packaging process and avoid misjudgments that may be caused by a single data source. It can handle complex relationships more effectively, enhance the system's discrimination ability, help capture small abnormal changes, and freely adjust weights, allowing optimization according to different production environments and needs, enhancing the applicability of the system. Through rapid calculation and evaluation, it can achieve real-time monitoring, timely discover and respond to potential risks, and improve production safety.

[0028] In summary, the present invention provides real-time data collection and analysis, which can quickly identify risks and problems in the production process and ensure timely measures. By setting standard templates for comparative analysis, it can effectively improve the accuracy of detection and reduce misjudgments and missed judgments, thereby ensuring product quality. In addition, by collecting a large amount of image and operation data and using artificial intelligence algorithms for correlation analysis, it can reveal potential risk patterns and improve the scientific nature of production. After identifying risks, the system can automatically adjust the data collection settings according to the analysis results, optimize the monitoring strategy, and improve the flexibility of the system. Once an anomaly is discovered, the system can conduct detailed traceability analysis, integrate multi-source data, help identify the root cause of the problem, and provide a basis for subsequent improvements. Through automated data collection and risk assessment, the need for manual intervention is reduced, and the risks brought by human operational errors are reduced. By quickly identifying and handling problems, downtime and production losses are reduced, overall production efficiency is improved, and possible risks in the packaging process are effectively monitored, thereby increasing the safety of the overall production and use process and reducing the probability of accidents.

[0029] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A lithium battery packaging status detection system based on artificial intelligence, characterized in that: include: Management module (1), used for global control and instruction issuance; Process reading module (2), used to read and modify packaging process standard information; An image recognition module (3) is used to collect image data of the packaging process according to a predetermined number of levels of collection states and convert the data into a machine-readable format; A machine recognition module (4) is used to collect the operation data of the packaging process according to a predetermined number of levels of collection status and convert it into a machine-readable format; Standard definition module (5), used to generate a benchmark template for the packaging process; a risk analysis module (6) for comparing the collected image data and operation data with the reference template to obtain risk judgment data; a dislocation module, provided in the risk analysis module (6), for dividing the data domain by the reference template and configuring the comparison order, judging the risk based on conventional comparison and random comparison strategies, and generating dislocation error distribution data in the random comparison; The tracing module (7) is used to trace the abnormal risk, analyze and integrate the time, machine status, operator, image evidence and dislocation error distribution data of the abnormal data, and form event groups with similar characteristics through timestamp alignment and correlation analysis; a correction module (8) for adjusting the acquisition level settings of the image recognition module (3) and the machine recognition module (4) according to the traceability results, and performing secondary acquisition and analysis after the adjustment, analyzing the successful detection rate, false alarm rate and missed alarm rate of the adjustment results in the secondary acquisition through a reinforcement learning algorithm, comparing the risk results after each adjustment with the preset target, evaluating the effectiveness of the adjustment, obtaining the influence coefficient of the change of the key parameters in the benchmark template on the accuracy of the adjustment result, and generating and applying a new acquisition level setting rule based on the influence coefficient of the reinforcement learning; The alarm module (9) is used to trigger an alarm and notify the management module (1) when the risk still exists after adjustment.

2. The artificial intelligence-based lithium battery packaging status detection system according to claim 1, characterized in that: The working logic of the dislocation module is as follows: the dislocation module divides the domains into several parts based on the reference template and configures a comparison order for each domain; Under conventional comparison, the corresponding domains are matched according to the order in which the submitted data arrives, and the comparison is performed one by one according to the configuration order of the current domain. If the error amount exceeds the predetermined limit within the preset period under conventional comparison of the risk analysis module (6), the submitted data is considered to be at risk; If the error amount does not exceed the predetermined limit within the preset period in the regular comparison of the risk analysis module (6), the currently submitted data will be divided into several domains of the benchmark template for random comparison. The domain participating in the comparison for the first time will avoid the first triggering domain of the regular comparison. Based on the historical risk data, the error frequency of different domains in the past comparison results and the performance data of the risk level of similar samples will be evaluated and obtained. The priority of each domain in the random comparison will be dynamically adjusted according to the performance data. The domain with higher risk will be given higher priority in the subsequent comparison. If the amount of errors within a preset period exceeds a predetermined limit under random comparison, the submitted data is considered to be at risk and misalignment error distribution data is generated.

3. The artificial intelligence-based lithium battery packaging status detection system according to claim 2, characterized in that: The working logic of the association analysis of the tracing module (7) is as follows: by extracting and integrating different source data, selecting key features of the operation image data and the operation operation data, identifying the association rules between the machine operation characteristics and the operation image characteristics by applying the association analysis algorithm, and associating several machine operation data with the defects of the specified operation image data.

4. The artificial intelligence-based lithium battery packaging status detection system according to claim 3, characterized in that: The working logic of the correction module (8) for adjusting the acquisition level settings of the image recognition module (3) and the machine recognition module (4) is as follows: if the image data collected by the image recognition module (3) has risks and the machine data collected by the machine recognition module (4) has no risks, then the acquisition level setting of the machine recognition module (4) is adjusted according to the preset indicators and submitted to the image recognition module (3); If the machine data collected by the machine recognition module (4) is at risk and the image data collected by the image recognition module (3) is not at risk, adjusting the collection level setting of the image recognition module (3) according to a preset indicator; After the adjustment, the secondary data collection is performed again and the risk analysis results of the risk analysis module (6) are received. If the risk still exists, the alarm module (9) is triggered to select whether to trigger the traceback module (7) again. Otherwise, the adjustment is terminated.

5. The artificial intelligence-based lithium battery packaging status detection system according to claim 1, characterized in that: The management module (1) is interactively connected to a storage module (11) via a wireless network. The storage module (11) is used to provide local and cloud storage for collected and analyzed data, and to generate a system operation report according to a preset period.

6. The artificial intelligence-based lithium battery packaging status detection system according to claim 1, characterized in that: During the operation phase, the risk analysis module (6) compares the data collected by the image recognition module (3) and the machine recognition module (4) with the reference template of the standard definition module (5), and triggers a jump based on the comparison result. If there is a risk, the module jumps to the upstream module (7) for further operation; if there is no risk, the module jumps to the image recognition module (3) and the machine recognition module (4) for further operation.

7. The artificial intelligence-based lithium battery packaging status detection system according to claim 1, characterized in that: The tracing module (7) is interactively connected to a target definition module (10) via a wireless network. The target definition module (10) is used to preset associated targets of several key features in the tracing result process of the tracing module (7), and supports custom editing. The associated targets include: any image acquisition component and any operation data acquisition component.

8. The artificial intelligence-based lithium battery packaging status detection system according to claim 1, characterized in that: The management module (1) is interactively connected to the process reading module (2), the image recognition module (3) and the machine recognition module (4) via a wireless network; the risk analysis module (6) is interactively connected to the image recognition module (3), the machine recognition module (4), the standard definition module (5), the tracing module (7) and the correction module (8) via a wireless network; and the alarm module (9) is interactively connected to the tracing module (7) and the correction module (8) via a wireless network.

9. A method for detecting the packaging status of a lithium battery based on artificial intelligence, wherein the method is an implementation method of a lithium battery packaging status detection system based on artificial intelligence according to any one of claims 1 to 8, characterized in that: The following steps are involved: Step 1: Issue control instructions through the internal IoT network, collect and store data collected and analyzed during the packaging process, and generate system operation reports regularly; Step 2: Read various standard process information during the lithium battery packaging process and support users to customize and adjust the process standards; Step 3: Deploy image acquisition equipment in the corresponding packaging process, set the acquisition position, angle and frequency, collect the operation images, and convert them into a machine-readable format to generate the operation image data of the packaging process; Step 4: Deploy operation data collection equipment, set the collection frequency and target, collect the operation status of the packaging process, and convert it into a machine-readable format to generate equipment operation data during the packaging process; Step 5: Develop standard operating image data and operation data for each packaging process to form a benchmark template; Step 6: Compare the acquired image data and operational data with the benchmark template. Through comparison and analysis, risks are identified. If risks are found, further analysis will be triggered. If no risks are found, then regular job image and operational data collection continues; Step 7: Once a risk is identified, the time of occurrence and the relevant machine status are recorded, and the activities of relevant operators and the generated image evidence are tracked. By integrating data from different sources and performing correlation analysis, key features associated with the risk are identified and event clusters are formed. Step 8: Based on the results of the traceability analysis, adjust the collection settings for image and operational data. If risks are found in the image data but normal machine data, adjust the collection level of operational data according to the preset indicators, and vice versa. Step 9: After the adjustment is completed, a second operation image and operation data collection is carried out. If the data collected in the second collection still identifies risks, corresponding alarm measures will be taken according to the preset risk level and relevant personnel will be notified in real time.

10. The method for detecting lithium battery packaging status based on artificial intelligence according to claim 9, characterized in that: The calculation expression of the comparison analysis process in step 6 is: ； Where, Represents the judgment result, indicating whether there is abnormal risk. Represents the number of machine data involved in the comparison, Represents the i-th image data The weight of Represents the i-th image data, Represents a set of standard parameters, represents the comparison function between image data and standard parameters, Represents the number of image data involved in the comparison, Represents the jth machine parameter The weight of Represents the comparison function between machine parameters and standard parameters, used to evaluate the abnormality score obtained by comparing machine parameters with standard parameters. Represents the j-th machine data; The working logic of the calculation expression of the comparison analysis process is: Receive the collected image data and operating data, score each image data and each machine parameter, and determine their deviation from the standard parameters; All scores are weighted and summed according to the weights; Calculate the final judgment result through the activation function; If the judgment result is greater than the set threshold, it is judged that there is an abnormal risk; otherwise, it is considered that there is no abnormality.

Images