Lithium battery defect detection system and method based on PLC control system

Abstract

The invention discloses a lithium battery defect detection system and method based on a PLC control system, relates to the technical field of lithium battery automation equipment, and solves the problems that pure image recognition cannot meet actual production requirements, lithium batteries are different in type and size and different in detection precision, and when the production requirements are frequently changed, the detection precision is poor. Development of an automatic lithium battery defect detection system capable of flexibly adjusting parameters becomes an important demand problem of the industry. The method comprises the following steps: processing an image, analyzing surface data and structural data of a lithium battery, dividing models according to the surface area and the structural volume of image tiling, subsequently generating an exclusive defect detection training data set in a targeted manner for each model, and performing feature extraction on the defect detection training data set by utilizing a detection model integrated with a deep learning algorithm to obtain a defect detection result. According to the method, defects are classified, and precision analysis is carried out, so that targeted defect detection is carried out on lithium batteries of different specifications, and adaptive adjustment of detection standard parameters is realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of lithium battery automation equipment, and specifically relates to a lithium battery defect detection system and method based on a PLC control system. Background Art

[0002] With the widespread application of lithium batteries in various portable electronic devices, electric vehicles and other fields, their safety issues have become increasingly prominent. During the production process of lithium batteries, due to imperfect processes or raw material problems, various defects such as electrode dislocation, electrode damage, electrolyte leakage, etc. may occur. These defects will seriously affect the performance and safety of the battery. Traditional defect detection methods often rely on image processing technology to drag and train image recognition models to perform target detection on lithium batteries;

[0003] However, simple image recognition cannot meet the needs of actual production. Lithium batteries of different types and sizes have different detection accuracy. When production needs change frequently, developing an automated and flexible lithium battery defect detection system becomes an important need of the industry. Summary of the invention

[0004] The present invention aims to solve at least one of the technical problems existing in the prior art; to this end, the present invention proposes a lithium battery defect detection method based on a PLC control system, which is used to solve the technical problems that it is difficult to ensure stable tracking of a target object, the relative position of the target object changes greatly in a short period of time, and it is difficult to ensure accurate monitoring of the target object's posture data.

[0005] To solve the above problems, the first aspect of the present invention provides a lithium battery defect detection method based on a PLC control system, comprising the following steps:

[0006] S1. Obtain real-time images of lithium batteries on the production line, generate surface and structural information of lithium batteries, classify lithium batteries by model on this basis, and establish an initial data set, including plane data and structural data; retrieve the temperature rise curve of lithium battery overheating through the Internet, analyze the acceleration of temperature rise, and establish a temperature anomaly data set;

[0007] S2. For each folder Lj, analyze the initial data set of the lithium battery model corresponding to the folder Lj, and construct a defect detection training data set, including a size accuracy detection data set, a standard data set, and a defect data set; based on the temperature anomaly data set, construct a temperature training data set;

[0008] S3. Establish a recognition model and a comparison model based on a deep algorithm, use the defect detection training set to train the recognition model, and execute the image processing and defect recognition algorithm; use the temperature training data set to train the comparison model and execute the dynamic time warping algorithm;

[0009] S4. Comprehensively analyze the monitoring results of the identification model and the comparison model, connect to the PLC control system, generate control instructions, and eliminate unqualified lithium batteries.

[0010] As a further solution of the present invention: the step S1, establishing an initial data set and a temperature anomaly data set, comprises the following steps:

[0011] The steps to build the initial data set are as follows:

[0012] Several groups of high-speed cameras are set up, and each group of cameras uses multiple different angles to obtain image information of the lithium battery in real time, and the image information is spliced ​​to generate a plan view of the lithium battery;

[0013] Coordinate axis processing is performed on the basis of the plane graph, and a number of regional intervals are generated, which are recorded as Qi [x, y], where x is the horizontal coordinate of the coordinate axis, y is the vertical coordinate of the coordinate axis, and i = 1, 2, 3, ...; the plane data of the lithium battery is formed;

[0014] Based on the plane data, a three-dimensional structure diagram of the lithium battery, i.e., the structural data of the lithium battery, is generated through multi-view image three-dimensional reconstruction technology; wherein each region interval Qi[x, y] corresponds to a plane view of the three-dimensional structure diagram of the lithium battery, and the plane view coordinate axis on the three-dimensional structure diagram is made to coincide with the coordinate axis on the plane diagram;

[0015] Based on the plane data and structural data in the initial data set, the lithium batteries are classified into different models for different plane area sizes and structural size differences, which are recorded as Lj, where j = 1, 2, 3, etc., i.e., the lithium battery model number, and the plane data and structural data of lithium batteries of the same model are placed in a folder, and the folder is named after the lithium battery model;

[0016] The steps to establish the temperature anomaly dataset are as follows:

[0017] Retrieve the temperature rise curve of lithium battery overheating online, analyze the acceleration of temperature rise, and draw the temperature rise acceleration curve;

[0018] In the temperature rise acceleration curve, the starting coordinates where the acceleration changes greater than or equal to r within the time R are defined as the mutation origin. The mutation origin is located and marked, and the temperature rise acceleration curve after the mutation origin is analyzed to establish a temperature anomaly data set.

[0019] As a further solution of the present invention: the step S1, classifying the lithium batteries by type, comprises the following steps:

[0020] The classification steps are as follows: compare the surface area and volume of the lithium battery, and set several groups of maximum tolerances based on the comparison results, where each group includes a maximum tolerance for the surface area and a maximum tolerance for the volume; the data samples covered within each group of tolerances are classified as the same model.

[0021] As a further solution of the present invention: the step S2, constructing a defect detection training data set and a temperature training data set, comprises the following steps:

[0022] For the plane data in each folder, use OPEN CV software to generate a grayscale image of the lithium battery surface, analyze the grayscale image, compare the pixel values ​​of each lithium battery in each area interval Qi [x, y], mark all areas with different pixel values ​​and fewer samples as abnormal areas;

[0023] Through the corresponding relationship, the abnormal area is found in the corresponding part of the structural data. In the structural data, the average depression depth h3, the maximum average depression depth h2, and the thickness h1 of the lithium battery in the vertical direction of the abnormal area are calculated through three-dimensional coordinate calculation. Among them, the protruding part is recorded as a negative value; the relationship between h1, h2, and h3 is analyzed to obtain the complex depression coefficient C complex , plot the complex concavity coefficient C complex -Pixel value correspondence table, and generate size accuracy detection data packet;

[0024] In each folder, according to the plane data, the abnormal area is cut out and recorded as a defective data packet, and the other areas in the plane data are recorded as normal areas and standard data packets;

[0025] Based on the temperature anomaly data set in each folder, set up real-time current temperature monitoring. The specific operations are as follows:

[0026] By setting up a charging and discharging circuit, including a protection circuit, a charging circuit and an output circuit, the lithium battery is connected in series, and a current sensor and a temperature sensor are set to monitor and record the lithium battery current and internal temperature changes in real time, draw a real-time temperature acceleration curve, and record it as a temperature training data set.

[0027] As a further solution of the present invention: Analyze the relationship between h1, h2, and h3 to obtain the complex concave coefficient C complex , including the following steps:

[0028] Maximum sag scale factor:

[0029]

[0030] Average Sag Scale Factor:

[0031]

[0032] Define the complex concavity coefficient C complex , which indicates the influence of the depression in the abnormal area on the safe production of lithium batteries. The analysis formula is:

[0033]

[0034] Among them, C complex The symbol and R avg and R max Keep it consistent, k is the amplification factor, γ is the weight factor, and reflects the concave shape in the complex concave coefficient C complex The importance of, S is the factor describing the concave shape, here the value is 0.8;

[0035] The selection of the amplification factor k depends on the sensitivity of the production line to the impact of sag, and its value is 10; the value of γ is 1.38.

[0036] As a further solution of the present invention, the abnormal area is cut out according to the plane data, including the following steps:

[0037] Fill the abnormal area, define a small enough square grid, and smooth the edge of the abnormal area so that the area where the edge of the abnormal area is smaller than the area of ​​the square grid is completely covered by the minimum number of square grids, so that the length of the projection of the abnormal area in any direction is an integer multiple of the side length of the square grid.

[0038] As a further solution of the present invention: the step S3, the training method of the recognition model, comprises the following steps:

[0039] By inputting a standard data package, which includes each model of lithium battery, the feature extraction is performed on each model of lithium battery to determine the standard range of any feature; the specific features include lithium battery surface area, volume, surface scratches, accompanying foreign matter, expansion degree, electrode misalignment and coating loss;

[0040] By inputting a defect data packet including each model of lithium battery, the defects of each model of lithium battery are classified. For each model of lithium battery, the defect types of each model are divided according to the extracted features. Among them, the defects without corresponding features are recorded as other defects;

[0041] By inputting the dimensional accuracy detection data package, which includes each model of lithium battery, the dimensional accuracy of each model of lithium battery is divided; according to the surface scratches and expansion degree of each model of lithium battery, the complex concave coefficient C is set respectively. complex The preset boundary value is set by taking the complex depression coefficient C of this type of lithium battery complex- In the pixel value correspondence table, the complex concave coefficient value corresponding to the lowest pixel value in the abnormal area is the preset boundary value;

[0042] If the complex concave coefficient C of the surface scratch complex If it exceeds the corresponding preset boundary value, it is defined as unqualified in dimensional accuracy; otherwise, it is defined as qualified in dimensional accuracy;

[0043] If the complex concavity coefficient C of the expansion degree complex If it exceeds the corresponding preset boundary value, the dimensional accuracy is defined as unqualified; otherwise, the dimensional accuracy is defined as qualified.

[0044] As another aspect of the present invention, the step S3, comparing the training method of the model, comprises the following steps:

[0045] By inputting the temperature training data set, which includes each model of lithium battery, the feature extraction of the curve after the mutation origin of the temperature rise acceleration curve is performed for each model of lithium battery, including the overall rising trend and the segmented rising trend. The overall rising trend refers to the curve shape after the mutation origin, and the segmented rising trend is the curve shape of the overall rising trend within a certain time interval, and the longest time interval does not exceed one second.

[0046] As another aspect of the present invention, the step S4, connecting to the PLC control system, generating control instructions, and removing unqualified lithium batteries, includes the following steps:

[0047] The identification model and the comparison model are linked to the PLC control system to determine the model of the lithium battery by its volume and surface area. If any of the following features, such as accompanying foreign matter, electrode misalignment, and coating loss, are detected, the rejection device is coordinated to intercept the target lithium battery.

[0048] If other defects are detected, coordinate the rejection device to intercept the target lithium battery;

[0049] For any type of lithium battery, if it is classified as unqualified, the rejection device will be coordinated to intercept the target lithium battery; otherwise, no action will be taken;

[0050] For any type of lithium battery with surface scratches, if it is classified as unqualified, the rejection device will be coordinated to intercept the target lithium battery; otherwise, no action will be taken;

[0051] Connect the lithium battery in series to the charge and discharge circuit, and repeat S1, S2 and S3 to monitor the expansion degree and temperature rise acceleration of the lithium battery in the full circuit:

[0052] If the similarity between the temperature acceleration rising trend of the lithium battery and the overall rising trend is greater than or equal to 40%, the lithium battery is considered unqualified, and the rejection device is coordinated to intercept the target lithium battery; otherwise, no action is taken;

[0053] If the temperature acceleration of the lithium battery has at least one rising trend, and there are two or more consecutive segmented rising trends with similarities greater than 50%, the lithium battery is considered unqualified, and the rejection device is coordinated to intercept the target lithium battery; otherwise, no action is taken;

[0054] If the temperature acceleration rising trend of the lithium battery and the segmented rising trend have several discontinuous segmented rising trends, and the similarity of each location is greater than or equal to 30%, the complex depression coefficient C of the lithium battery is comprehensively analyzed. complex ; Set the secondary preset boundary value, which is derived from the complex depression coefficient C of the qualified lithium battery when it expands during the charging and discharging process complex The average value of the complex depression coefficient C of the lithium battery within the time range of the segmented rising trend with a similarity greater than or equal to 30%. complex If it is greater than the mean value, the lithium battery is considered unqualified, and the rejection device is coordinated to intercept the target lithium battery; otherwise, no action is taken;

[0055] As another aspect of the present invention, a lithium battery defect detection system based on a PLC control system is provided. The system specifically adopts a lithium battery defect detection method based on a PLC control system to detect lithium battery defects.

[0056] Compared with the prior art, the present invention has the following beneficial effects:

[0057] The present invention analyzes the surface data and structural data of lithium batteries through image processing, divides lithium batteries into models of different specifications according to the surface area and structural volume of the image, and subsequently generates a dedicated defect detection training data set for each model, and uses a detection model based on a deep learning algorithm to perform feature extraction, defect classification, and precision analysis on the defect detection training data set, thereby achieving targeted defect detection for lithium batteries of different specifications and adaptive adjustment of detection standard parameters;

[0058] Analyze the temperature rise acceleration curve of abnormal lithium batteries under temperature out-of-control conditions. Through the trend of the real-time temperature rise acceleration curve of lithium batteries, timely predict the consequences of abnormal lithium battery temperature and effectively avoid production accidents.

[0059] Simply divide the volume gradient and the corresponding surface area gradient of lithium batteries, divide the power storage capacity of lithium batteries, divide the models according to the actual situation, and conduct subsequent analysis on different models of lithium batteries; for different types of lithium batteries in the production process, for the accuracy requirements, types and severity of defects, make automated and more flexible testing standards;

[0060] Pre-set complex sag coefficients that meet the standards. Aiming at the expansion and slight deformation of lithium batteries caused by various elements in the production process, more accurate standards are used to detect whether the lithium batteries have reached the production quality required for qualified use. It is also better to detect whether the size of lithium batteries meets the standards, providing a certain range of tolerance and restriction indicators for actual production.

[0061] The processing of the edge of the abnormal area facilitates the processing and subsequent analysis of the data. The cropping of the abnormal area makes the coordinate interval of the abnormal area better positioned. The abnormal area is covered with a simplified square grid, which optimizes the difficulty of image edge processing. At the same time, the minimum number of square grids approximates the actual abnormal area to a great extent, reducing the difference between the simplified model and the actual value.

[0062] The present invention performs internal circuit detection on lithium batteries, analyzes the temperature change trend of lithium batteries, diagnoses the status of lithium batteries, and eliminates abnormal lithium batteries. For hidden dangers that cannot be directly judged from the appearance, the temperature change of lithium batteries during the charging and discharging process is compared with the similarity of abnormal cases, so as to better and faster detect unqualified lithium batteries and eliminate them in time. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0064] Figure 1 It is a schematic diagram of the process of the present invention;

[0065] Figure 2 It is a schematic diagram of the system framework of the present invention. DETAILED DESCRIPTION

[0066] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0067] Embodiment 1

[0068] See also Figure 1 The first aspect of the present invention provides a method for detecting defects in a lithium battery based on a PLC control system, comprising the following steps:

[0069] S1. Obtain real-time images of lithium batteries on the production line, generate surface and structural information of lithium batteries, classify lithium batteries by model on this basis, and establish an initial data set, including plane data and structural data; retrieve the temperature rise curve of lithium battery overheating through the Internet, analyze the acceleration of temperature rise, and establish a temperature anomaly data set;

[0070] S2. For each folder Lj, analyze the initial data set of the lithium battery model corresponding to the folder Lj, and construct a defect detection training data set, including a size accuracy detection data set, a standard data set, and a defect data set; based on the temperature anomaly data set, construct a temperature training data set;

[0071] S3. Establish a recognition model and a comparison model based on a deep algorithm, use the defect detection training set to train the recognition model, and execute the image processing and defect recognition algorithm; use the temperature training data set to train the comparison model and execute the dynamic time warping algorithm;

[0072] S4. Comprehensively analyze the monitoring results of the identification model and the comparison model, connect to the PLC control system, generate control instructions, and eliminate unqualified lithium batteries.

[0073] In one embodiment of the present invention, the step S1, establishing an initial data set and a temperature anomaly data set, comprises the following steps:

[0074] The steps to build the initial data set are as follows:

[0075] Several groups of high-speed cameras are set up, and each group of cameras uses multiple different angles to obtain image information of the lithium battery in real time, and the image information is spliced ​​to generate a plan view of the lithium battery;

[0076] Coordinate axis processing is performed on the basis of the plane graph, and a number of regional intervals are generated, which are recorded as Qi [x, y], where x is the horizontal coordinate of the coordinate axis, y is the vertical coordinate of the coordinate axis, and i = 1, 2, 3, ...; the plane data of the lithium battery is formed;

[0077] Based on the plane data, a three-dimensional structure diagram of the lithium battery, i.e., the structural data of the lithium battery, is generated through multi-view image three-dimensional reconstruction technology; wherein each region interval Qi[x, y] corresponds to a plane view of the three-dimensional structure diagram of the lithium battery, and the plane view coordinate axis on the three-dimensional structure diagram is made to coincide with the coordinate axis on the plane diagram;

[0078] Based on the plane data and structural data in the initial data set, the lithium batteries are classified into different models for different plane area sizes and structural size differences, which are recorded as Lj, where j = 1, 2, 3, etc., i.e., the lithium battery model number, and the plane data and structural data of lithium batteries of the same model are placed in a folder, and the folder is named after the lithium battery model;

[0079] The steps to establish the temperature anomaly dataset are as follows:

[0080] Retrieve the temperature rise curve of lithium battery overheating online, analyze the acceleration of temperature rise, and draw the temperature rise acceleration curve;

[0081] In the temperature rise acceleration curve, the starting coordinates where the acceleration changes greater than or equal to r within the time R are defined as the mutation origin. The mutation origin is located and marked, and the temperature rise acceleration curve after the mutation origin is analyzed to establish a temperature anomaly data set.

[0082] By simply dividing the volume gradient and the corresponding surface area gradient of lithium batteries, the power storage capacity of lithium batteries is divided, and the model division is made according to the actual situation, and subsequent analysis is carried out on different models of lithium batteries; for different types of lithium batteries in the production process, the requirements for accuracy, the types and severity of defects, automatic and more flexible inspection standards are made. For example, large power batteries may pay more attention to the alignment of electrodes and electrolyte leakage, while small portable batteries may focus more on surface damage and dimensional accuracy;

[0083] By analyzing the temperature rise acceleration curve of abnormal lithium batteries under temperature out of control conditions, we can timely predict the consequences of abnormal lithium battery temperature through the trend of the real-time temperature rise acceleration curve of the lithium battery and avoid production accidents.

[0084] In one embodiment of the present invention, the step S1, classifying the lithium batteries by model, comprises the following steps:

[0085] The classification steps are as follows: compare the surface area and volume of the lithium battery, and set several groups of maximum tolerances based on the comparison results, where each group includes a maximum tolerance for the surface area and a maximum tolerance for the volume; the data samples covered within each group of tolerances are classified as the same model.

[0086] In one embodiment of the present invention, the step S2, constructing a defect detection training data set and a temperature training data set, comprises the following steps:

[0087] For the plane data in each folder, use OPEN CV software to generate a grayscale image of the lithium battery surface, analyze the grayscale image, compare the pixel values ​​of each lithium battery in each area interval Qi [x, y], mark all areas with different pixel values ​​and fewer samples as abnormal areas;

[0088] Through the corresponding relationship, the abnormal area is found in the corresponding part of the structural data. In the structural data, the average depression depth h3, the maximum average depression depth h2, and the thickness h1 of the lithium battery in the vertical direction of the abnormal area are calculated through three-dimensional coordinate calculation. Among them, the protruding part is recorded as a negative value; the relationship between h1, h2, and h3 is analyzed to obtain the complex depression coefficient C complex , plot the complex concavity coefficient C complex -Pixel value correspondence table, and generate size accuracy detection data packet;

[0089] Pre-set complex sag coefficients that meet the standards. Aiming at the expansion and slight deformation of lithium batteries caused by various elements in the production process, more accurate standards are used to detect whether the lithium batteries have reached the production quality required for qualified use. It is also better to detect whether the size of lithium batteries meets the standards, providing a certain range of tolerance and restriction indicators for actual production.

[0090] In each folder, according to the plane data, the abnormal area is cut out and recorded as a defective data packet, and the other areas in the plane data are recorded as normal areas and standard data packets;

[0091] Based on the temperature anomaly data set in each folder, set up real-time current temperature monitoring. The specific operations are as follows:

[0092] By setting up a charging and discharging circuit, including a protection circuit, a charging circuit and an output circuit, the lithium battery is connected in series, and a current sensor and a temperature sensor are set up to monitor and record the lithium battery current and internal temperature changes in real time, draw a real-time temperature acceleration curve, and record it as a temperature training data set;

[0093] The present invention performs internal circuit detection on lithium batteries, analyzes the temperature change trend of lithium batteries, diagnoses the status of lithium batteries, and eliminates abnormal lithium batteries. For hidden dangers that cannot be directly judged from the appearance, the temperature change of lithium batteries during the charging and discharging process is compared with the similarity of abnormal cases, so as to better and faster detect unqualified lithium batteries and eliminate them in time.

[0094] In one embodiment of the present invention, the relationship among h1, h2, and h3 is analyzed to obtain the complex concavity coefficient C complex , including the following steps:

[0095] Maximum sag scale factor:

[0096]

[0097] Average Sag Scale Factor:

[0098]

[0099] Define the complex concavity coefficient C complex , which indicates the influence of the depression in the abnormal area on the safe production of lithium batteries. The analysis formula is:

[0100]

[0101] Among them, C complex The symbol and R avg and R max Keep it consistent, k is the amplification factor, γ is the weight factor, and reflects the concave shape in the complex concave coefficient C complex The importance of, S is the factor describing the concave shape, here the value is 0.8;

[0102] The selection of the amplification factor k depends on the sensitivity of the production line to the impact of sag, and its value is 10; the value of γ is 1.38.

[0103] In one embodiment of the present invention, cutting out the abnormal area according to the plane data includes the following steps:

[0104] Fill the edge of the abnormal area, define a small enough square grid, and smooth the edge of the abnormal area so that the area where the edge of the abnormal area is smaller than the area of ​​the square grid is completely covered by the minimum number of square grids, so that the length of the projection of the abnormal area in any direction is an integer multiple of the side length of the square grid;

[0105] The marginalization of abnormal areas facilitates data processing and subsequent analysis. The cropping of abnormal areas makes it easier to locate the coordinate range of the abnormal areas. Covering the abnormal areas with simplified square grids optimizes the difficulty of image edge processing. At the same time, the minimum number of square grids approximates the actual area of ​​the abnormal areas to the greatest extent, reducing the difference between the simplified model and the actual value.

[0106] In one embodiment of the present invention, the step S3, identifying the training method of the model, comprises the following steps:

[0107] By inputting a standard data package, which includes each model of lithium battery, the feature extraction is performed on each model of lithium battery to determine the standard range of any feature; the specific features include lithium battery surface area, volume, surface scratches, accompanying foreign matter, expansion degree, electrode misalignment and coating loss;

[0108] By inputting a defect data packet including each model of lithium battery, the defects of each model of lithium battery are classified. For each model of lithium battery, the defect types of each model are divided according to the extracted features. Among them, the defects without corresponding features are recorded as other defects;

[0109] By inputting the dimensional accuracy detection data package, which includes each model of lithium battery, the dimensional accuracy of each model of lithium battery is divided; according to the surface scratches and expansion degree of each model of lithium battery, the complex concave coefficient C is set respectively. complex The preset boundary value is set by taking the complex depression coefficient C of this type of lithium battery complex - In the pixel value correspondence table, the complex concave coefficient value corresponding to the lowest pixel value in the abnormal area is the preset boundary value;

[0110] If the complex concave coefficient C of the surface scratch complex If it exceeds the corresponding preset boundary value, it is defined as unqualified in dimensional accuracy; otherwise, it is defined as qualified in dimensional accuracy;

[0111] If the complex concavity coefficient C of the expansion degree complex If it exceeds the corresponding preset boundary value, the dimensional accuracy is defined as unqualified; otherwise, the dimensional accuracy is defined as qualified.

[0112] In one embodiment of the present invention, the step S3, comparing the training method of the model, comprises the following steps:

[0113] By inputting the temperature training data set, which includes each model of lithium battery, the feature extraction of the curve after the mutation origin of the temperature rise acceleration curve is performed for each model of lithium battery, including the overall rising trend and the segmented rising trend. The overall rising trend refers to the curve shape after the mutation origin, and the segmented rising trend is the curve shape of the overall rising trend within a certain time interval, and the longest time interval does not exceed one second.

[0114] In one embodiment of the present invention, the step S4, connecting to a PLC control system, generating a control instruction, and removing unqualified lithium batteries, comprises the following steps:

[0115] The identification model and the comparison model are linked to the PLC control system to determine the model of the lithium battery by its volume and surface area. If any of the following features, such as accompanying foreign matter, electrode misalignment, and coating loss, are detected, the rejection device is coordinated to intercept the target lithium battery.

[0116] If other defects are detected, coordinate the rejection device to intercept the target lithium battery;

[0117] For any type of lithium battery, if it is classified as unqualified, the rejection device will be coordinated to intercept the target lithium battery; otherwise, no action will be taken;

[0118] For any type of lithium battery with surface scratches, if it is classified as unqualified, the rejection device will be coordinated to intercept the target lithium battery; otherwise, no action will be taken;

[0119] Connect the lithium battery in series to the charge and discharge circuit, and repeat S1, S2 and S3 to monitor the expansion degree and temperature rise acceleration of the lithium battery in the full circuit:

[0120] If the similarity between the temperature acceleration rising trend of the lithium battery and the overall rising trend is greater than or equal to 40%, the lithium battery is considered unqualified, and the rejection device is coordinated to intercept the target lithium battery; otherwise, no action is taken;

[0121] If the temperature acceleration of the lithium battery has at least one rising trend, and there are two or more consecutive segmented rising trends with similarities greater than 50%, the lithium battery is considered unqualified, and the rejection device is coordinated to intercept the target lithium battery; otherwise, no action is taken;

[0122] If the temperature acceleration rising trend of the lithium battery and the segmented rising trend have several discontinuous segmented rising trends, and the similarity of each location is greater than or equal to 30%, the complex depression coefficient C of the lithium battery is comprehensively analyzed. complex ; Set the secondary preset boundary value, which is derived from the complex depression coefficient C of the qualified lithium battery when it expands during the charging and discharging process complex The average value of the complex depression coefficient C of the lithium battery within the time range of the segmented rising trend with a similarity greater than or equal to 30%. complex If it is greater than the mean, the lithium battery is considered unqualified, and the rejection device is coordinated to intercept the target lithium battery; otherwise, no action is taken.

[0123] Embodiment 2

[0124] In another embodiment of the present invention, please refer to Figure 2 , a lithium battery defect detection system based on a PLC control system, the system specifically adopts a lithium battery defect detection method based on a PLC control system to realize the detection of lithium battery defects:

[0125] Data acquisition unit: obtains real-time images of lithium batteries on the production line, generates surface and structural information of lithium batteries, classifies lithium batteries by model, and establishes an initial data set, including plane data and structural data; retrieves the temperature rise curve of lithium battery overheating through the network, analyzes the acceleration of temperature rise, and establishes a temperature anomaly data set;

[0126] Data integration unit: for each folder Lj, analyze the initial data set of the lithium battery model corresponding to the folder Lj, and construct a defect detection training data set, including a size accuracy detection data set, a standard data set, and a defect data set; based on the temperature anomaly data set, construct a temperature training data set;

[0127] Model training unit: Establish a recognition model based on the deep algorithm, use the defect detection training set to train the recognition model; use the temperature training data set to compare and train the model;

[0128] Inspection unit: Based on the inspection model, it executes image processing and defect recognition algorithms to obtain prediction results; based on the comparison model, it executes dynamic time warping algorithms to generate inspection results;

[0129] PLC control unit: Comprehensively analyze the monitoring results of the identification model and the comparison model, connect to the PLC control system, generate control instructions, and eliminate unqualified lithium batteries.

[0130] Embodiment 3

[0131] As the third embodiment of the present application, the technical solution of this embodiment is to combine the solutions of the above two embodiments for implementation.

[0132] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A lithium battery defect detection method based on a PLC control system, characterized in that: The following steps are involved: S1. Obtain real-time images of lithium batteries on the production line, generate surface and structural information of lithium batteries, classify lithium batteries by model, and establish an initial data set, including plane data and structural data; Retrieve the temperature rise curve of lithium battery overheating online, analyze the acceleration of temperature rise, and establish a temperature anomaly data set; S2. For each folder Lj, analyze the initial data set of the lithium battery model corresponding to the folder Lj, and construct a defect detection training data set, including a size accuracy detection data set, a standard data set, and a defect data set; Based on the temperature anomaly dataset, a temperature training dataset is constructed; S3. Establish a recognition model and a comparison model based on a deep algorithm, use the defect detection training set to train the recognition model, and execute the image processing and defect recognition algorithm; use the temperature training data set to train the comparison model and execute the dynamic time warping algorithm; S4. Comprehensively analyze the monitoring results of the identification model and the comparison model, connect to the PLC control system, generate control instructions, and eliminate unqualified lithium batteries.

2. A lithium battery defect detection method based on a PLC control system according to claim 1, characterized in that: The step S1, establishing an initial data set and a temperature anomaly data set, comprises the following steps: The steps to build the initial data set are as follows: Several groups of high-speed cameras are set up, and each group of cameras uses multiple different angles to obtain image information of the lithium battery in real time, and the image information is spliced ​​to generate a plan view of the lithium battery; Coordinate axis processing is performed on the basis of the plane graph, and a number of regional intervals are generated, which are recorded as Qi [x, y], where x is the horizontal coordinate of the coordinate axis, y is the vertical coordinate of the coordinate axis, and i = 1, 2, 3, ...; the plane data of the lithium battery is formed; Based on the plane data, a three-dimensional structure diagram of the lithium battery, i.e., the structural data of the lithium battery, is generated through multi-view image three-dimensional reconstruction technology; wherein each region interval Qi[x, y] corresponds to a plane view of the three-dimensional structure diagram of the lithium battery, and the plane view coordinate axis on the three-dimensional structure diagram is made to coincide with the coordinate axis on the plane diagram; Based on the plane data and structural data in the initial data set, the lithium batteries are classified into different models for different plane area sizes and structural size differences, which are recorded as Lj, where j = 1, 2, 3, etc., i.e., the lithium battery model number, and the plane data and structural data of lithium batteries of the same model are placed in a folder, and the folder is named after the lithium battery model; The steps to establish the temperature anomaly dataset are as follows: Retrieve the temperature rise curve of lithium battery overheating online, analyze the acceleration of temperature rise, and draw the temperature rise acceleration curve; In the temperature rise acceleration curve, the starting coordinates where the acceleration changes greater than or equal to r within the time R are defined as the mutation origin. The mutation origin is located and marked, and the temperature rise acceleration curve after the mutation origin is analyzed to establish a temperature anomaly data set.

3. A lithium battery defect detection method based on a PLC control system according to claim 2, characterized in that: The step S1, classifying the lithium batteries by type, comprises the following steps: The classification steps are as follows: compare the surface area and volume of the lithium battery, and set several groups of maximum tolerances based on the comparison results, where each group includes a maximum tolerance for the surface area and a maximum tolerance for the volume; the data samples covered within each group of tolerances are classified as the same model.

4. A lithium battery defect detection method based on a PLC control system according to claim 1, characterized in that: The step S2, constructing a defect detection training data set and a temperature training data set, comprises the following steps: For the plane data in each folder, use OPEN CV software to generate a grayscale image of the lithium battery surface, analyze the grayscale image, compare the pixel values ​​of each lithium battery in each area interval Qi [x, y], mark all areas with different pixel values ​​and fewer samples as abnormal areas; Through the corresponding relationship, the abnormal area is found in the corresponding part of the structural data. In the structural data, the average depression depth h3, the maximum average depression depth h2, and the thickness h1 of the lithium battery in the vertical direction of the abnormal area are calculated through three-dimensional coordinate calculation. Among them, the protruding part is recorded as a negative value; the relationship between h1, h2, and h3 is analyzed to obtain the complex depression coefficient C complex , plot the complex concavity coefficient C complex -Pixel value correspondence table, and generate size accuracy detection data packet; In each folder, according to the plane data, the abnormal area is cut out and recorded as a defective data packet, and the other areas in the plane data are recorded as normal areas and standard data packets; Based on the temperature anomaly data set in each folder, set up real-time current temperature monitoring. The specific operations are as follows: By setting up a charging and discharging circuit, including a protection circuit, a charging circuit and an output circuit, the lithium battery is connected in series, and a current sensor and a temperature sensor are set to monitor and record the lithium battery current and internal temperature changes in real time, draw a real-time temperature acceleration curve, and record it as a temperature training data set.

5. A lithium battery defect detection method based on a PLC control system according to claim 4, characterized in that: Analyze the relationship between h1, h2, and h3 to obtain the complex concave coefficient C complex , including the following steps: Maximum sag scale factor: Average Sag Scale Factor: Define the complex concavity coefficient C complex , which indicates the influence of the depression in the abnormal area on the safe production of lithium batteries. The analysis formula is: Among them, C complex The symbol and R avg and R max Keep it consistent, k is the amplification factor, γ is the weight factor, and reflects the concave shape in the complex concave coefficient C complex The importance of is, S is a factor describing the shape of the concave.

6. A lithium battery defect detection method based on a PLC control system according to claim 4, characterized in that: According to the plane data, the abnormal area is cut out, including the following steps: Fill the abnormal area, define a small enough square grid, and smooth the edge of the abnormal area so that the area where the edge of the abnormal area is smaller than the area of ​​the square grid is completely covered by the minimum number of square grids, so that the length of the projection of the abnormal area in any direction is an integer multiple of the side length of the square grid.

7. A lithium battery defect detection method based on a PLC control system according to claim 1, characterized in that: The step S3, identifying the training method of the model, comprises the following steps: By inputting a standard data package, which includes each model of lithium battery, the feature extraction is performed on each model of lithium battery to determine the standard range of any feature; the specific features include lithium battery surface area, volume, surface scratches, accompanying foreign matter, expansion degree, electrode misalignment and coating loss; By inputting a defect data packet including each model of lithium battery, the defects of each model of lithium battery are classified. For each model of lithium battery, the defect types of each model are divided according to the extracted features. Among them, the defects without corresponding features are recorded as other defects; By inputting the dimensional accuracy detection data package, which includes each model of lithium battery, the dimensional accuracy of each model of lithium battery is divided; according to the surface scratches and expansion degree of each model of lithium battery, the complex concave coefficient C is set respectively. complex The preset boundary value is set by taking the complex depression coefficient C of this type of lithium battery complex - In the pixel value correspondence table, the complex concave coefficient value corresponding to the lowest pixel value in the abnormal area is the preset boundary value; If the complex concave coefficient C of the surface scratch complex If it exceeds the corresponding preset boundary value, it is defined as unqualified in dimensional accuracy; otherwise, it is defined as qualified in dimensional accuracy; If the complex concavity coefficient C of the expansion degree complex If it exceeds the corresponding preset boundary value, the dimensional accuracy is defined as unqualified; otherwise, the dimensional accuracy is defined as qualified.

8. The method for detecting defects in a lithium battery based on a PLC control system according to claim 1, characterized in that: The step S3, comparing the training method of the model, comprises the following steps: By inputting the temperature training data set, which includes each model of lithium battery, the feature extraction of the curve after the mutation origin of the temperature rise acceleration curve is performed for each model of lithium battery, including the overall rising trend and the segmented rising trend. The overall rising trend refers to the curve shape after the mutation origin, and the segmented rising trend is the curve shape of the overall rising trend within a certain time interval, and the longest time interval does not exceed one second.

9. The method for detecting lithium battery defects based on a PLC control system according to claim 1, characterized in that: The step S4, connecting to the PLC control system, generating control instructions, and removing unqualified lithium batteries, includes the following steps: The identification model and the comparison model are linked to the PLC control system to determine the model of the lithium battery by its volume and surface area. If any of the following features, such as accompanying foreign matter, electrode misalignment, and coating loss, are detected, the rejection device is coordinated to intercept the target lithium battery. If other defects are detected, coordinate the rejection device to intercept the target lithium battery; For any type of lithium battery, if it is classified as unqualified, the rejection device will be coordinated to intercept the target lithium battery; otherwise, no action will be taken; For any type of lithium battery with surface scratches, if it is classified as unqualified, the rejection device will be coordinated to intercept the target lithium battery; otherwise, no action will be taken; Connect the lithium battery in series to the charge and discharge circuit, and repeat S1, S2 and S3 to monitor the expansion degree and temperature rise acceleration of the lithium battery in the full circuit: If the similarity between the temperature acceleration rising trend of the lithium battery and the overall rising trend is greater than or equal to 40%, the lithium battery is considered unqualified, and the rejection device is coordinated to intercept the target lithium battery; otherwise, no action is taken; If the temperature acceleration of the lithium battery has at least one rising trend, and there are two or more consecutive segmented rising trends with similarities greater than 50%, the lithium battery is considered unqualified, and the rejection device is coordinated to intercept the target lithium battery; otherwise, no action is taken; If the temperature acceleration rising trend of the lithium battery and the segmented rising trend have several discontinuous segmented rising trends, and the similarity of each location is greater than or equal to 30%, the complex depression coefficient C of the lithium battery is comprehensively analyzed. complex ; Set the secondary preset boundary value, which is derived from the complex depression coefficient C of the qualified lithium battery when it expands during the charging and discharging process complex The average value of the complex depression coefficient C of the lithium battery within the time range of the segmented rising trend with a similarity greater than or equal to 30%. complex If it is greater than the mean, the lithium battery is considered unqualified, and the rejection device is coordinated to intercept the target lithium battery; otherwise, no action is taken.

10. A lithium battery defect detection system based on a PLC control system, the system specifically adopts the method described in any one of claims 1 to 10 to realize the detection of lithium battery defects.

Images