A method and system for detecting NG product marking paper by using a camera instead of a color mark sensor

Through the intelligent camera system and dynamic threshold adjustment, the problem that the color mark sensor cannot detect the position and type of label paper is solved, the detection accuracy and stability are improved, and the accurate detection and prevention of label paper during the lithium battery electrode manufacturing process is achieved. Sexual maintenance.

CN118654575BActive Publication Date: 2024-11-19GUANGZHOU EHOLLY INTELLIGENT EQUIP CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411140821.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-20
Publication Date
2024-11-19
Estimated Expiration
2044-08-20

AI Technical Summary

Technical Problem

The existing color mark sensors cannot effectively detect the sticking position and skew of the label paper during the lithium battery electrode manufacturing process, and identify the type of label paper, resulting in production line failure and quality problems, and are not adapted to environmental changes and are cumbersome to adjustments. Increased maintenance costs.

Method used

The intelligent camera system is used to replace the color standard sensor, and through environmental data acquisition and dynamic threshold adjustment, label images are captured in real time, label parameters are calculated, MARK positions are synchronized, real-time quality trend analysis is carried out, equipment problems are predicted, and maintenance suggestions are automatically generated.

Benefits of technology

It improves the accuracy and stability of detection, reduces false detection rates and missed detection rates, can adapt to environmental changes in real time, predict and prevent equipment problems and quality abnormalities, and reduces production costs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118654575B_ABST
    Figure CN118654575B_ABST
Patent Text Reader

Abstract

The present invention discloses a method and system for detecting NG product marking paper by using a camera instead of a color mark sensor, which relates to the field of automatic control and adjustment, including initializing the camera system, collecting environmental data and adjusting dynamic thresholds; using an intelligent camera to capture label images in real time, calculating label parameters, synchronizing MARK positions, and calculating label relative positions; judging label parameters according to the adjusted dynamic thresholds, and performing real-time quality trend analysis on the judgment results; predicting possible equipment problems or quality anomalies based on historical data and real-time quality trends, and performing automatic closed-loop adjustments. When potential problems are predicted, the system automatically generates maintenance suggestions and notifies relevant personnel. The present invention significantly improves the accuracy and stability of detection through an intelligent camera system and an adaptive algorithm, and realizes the prediction and prevention of equipment problems and quality anomalies through the analysis of historical data and real-time quality trends.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of automatic control and regulation, and in particular to a method and system for detecting NG product marking paper using a camera instead of a color mark sensor. Background Art

[0002] In the manufacturing process of lithium battery pole pieces, die cutting and slitting machines are one of the key equipment, responsible for cutting battery pole pieces into the required shape and size. In the production process, machine vision inspection gives automation equipment "eyes" and "brains", making it more intelligent, improving production efficiency and product quality. The application of visual inspection technology in factory quality inspection is particularly important, and it can promptly detect various defective products in the production process. However, existing visual inspection technology and equipment still have many shortcomings, especially in label inspection, there are many technical bottlenecks and challenges.

[0003] At present, the lithium battery industry mainly relies on color mark sensors for label detection. This sensor determines the presence of a label by detecting the color mark on the label and records it. However, the color mark sensor can only detect the presence of a label, but cannot further detect the label's pasting position, whether it is skewed, or identify the type of label paper. This technical limitation brings many problems. For example, when the label paper is pasted in the incorrect or skewed position, the label may fall off and be rolled into the pole piece, causing production line failure. In addition, the offset of the marking position may also affect the overall quality of the product, increasing production and maintenance costs.

[0004] The limitations of color mark sensors are mainly reflected in the following aspects: First, color mark sensors can only detect a single presence or absence signal and cannot provide detailed information about the location and status of the label. This lack of information makes it difficult to quickly locate and solve problems on the production line when they occur. Secondly, color mark sensors have poor adaptability to environmental changes. For example, changes in light, dust, and vibration will affect their detection accuracy. Thirdly, color mark sensors perform poorly when faced with complex label types and cannot distinguish between different types of label paper, which is a major defect for defective product detection that requires precise identification and classification. Finally, the adjustment and maintenance of color mark sensors are relatively cumbersome, requiring frequent calibration and manual intervention, which increases the downtime and maintenance costs of the production line. Summary of the invention

[0005] In view of the problems existing in the inherent defects of the above-mentioned color mark sensor, the present invention is proposed.

[0006] Therefore, the problem to be solved by the present invention is how to provide a new and more intelligent detection method.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0008] In the first aspect, an embodiment of the present invention provides a method for detecting NG product marking paper by using a camera instead of a color mark sensor, which includes initializing a camera system, collecting environmental data and adjusting dynamic thresholds; using an intelligent camera to capture label images in real time, calculating label parameters, synchronizing MARK positions, and calculating label relative positions; judging label parameters based on the adjusted dynamic thresholds, and performing real-time quality trend analysis on the judgment results; predicting possible equipment problems or quality anomalies based on historical data and real-time quality trends, and performing automatic closed-loop adjustments. When potential problems are predicted, the system automatically generates maintenance suggestions and notifies relevant personnel.

[0009] As a preferred solution of the method of using a camera to replace a color mark sensor to detect NG product marking paper of the present invention, the camera system initialization includes: replacing the original color mark sensor with a smart camera; setting the initial parameters of the smart camera, including the image length Y, so that it is an integer multiple of the laser head distance X:

[0010] X=n×Y

[0011] Wherein, n is a multiple relationship; initialize the dynamic threshold system and set the initial threshold of each detection parameter.

[0012] As a preferred solution of the method of using a camera instead of a color mark sensor to detect NG product marking paper according to the present invention, the synchronization of the MARK position and the calculation of the relative position of the label include the following contents: when the laser head of the device cuts the MARK hole, a mark signal is sent to the smart camera; when the smart camera receives the mark signal, the current frame number is recorded as the received MARK frame number; when the label is detected, the current frame number is recorded as the detected label frame number; the calculation formula for calculating the position of the label from the MARK hole is as follows: the position of the label from the MARK hole = X-[(detected label frame number-received MARK frame number)×Y+Z], wherein Y is the image length; X is the distance from the laser head; and Z is the correction factor or offset.

[0013] As a preferred solution of the method of using a camera instead of a color mark sensor to detect NG product marking paper according to the present invention, the real-time quality trend analysis of the judgment results includes the following steps: receiving the judgment results of each detection in real time and setting an analysis time window; performing statistical calculations on each parameter within different time windows, including: mean value, standard deviation, maximum value, minimum value and rate of change, and calculating the proportion of each type of judgment results; using a time series analysis method to predict short-term trends; applying abnormal pattern judgment rules to the calculation results, marking and recording the detected abnormal patterns.

[0014] As a preferred solution of the method of using a camera instead of a color mark sensor to detect NG product marking paper of the present invention, the method of using a time series analysis method to predict short-term trends includes: checking whether the time series is stable, and if not, performing a differential operation:

[0015] ,

[0016] Among them, B is the lag operator, d is the number of differences, and xt is the value of the time series before time t; the autoregressive AR term p and the moving average MA term q of the model are determined using the autocorrelation function ACF and partial autocorrelation function PACF graphs; select p, d and q parameters to fit the ARIMA model:

[0017] ,

[0018] in, and are AR and MA polynomials respectively, is the error term, yt is the value of the time series at time t; check the autocorrelation plot and other diagnostic plots of the residuals to ensure that the residuals are white noise; if not, adjust the model parameters and refit; use the fitted ARIMA model to predict future values:

[0019] ,

[0020] Among them, p is the number of autoregressive terms; q is the number of moving average terms.

[0021] As a preferred solution of the method for detecting NG product marking paper by using a camera instead of a color mark sensor according to the present invention, the method includes the following steps: extracting historical production data from a database, integrating short-term trend data, and standardizing the data; combining the ARIMA prediction results with the machine learning model, adding the prediction results of the ARIMA model as new features to the input of the machine learning model, calculating the difference between the ARIMA prediction value and the actual value as an additional feature, and extracting the parameters p, d, and q of the ARIMA model as indicator features of model complexity; assuming that the original feature is X and the new feature vector is Xnew: Xnew = [ X , x ^ t + k ARIMA , e t + k , p , d , q ] ,in, is the difference between the ARIMA predicted value and the actual value; the expanded feature vector Xnew is used to train and predict the machine learning model; a meta-model is constructed, with the ARIMA prediction results and the machine learning model prediction results as input, and the meta-model is trained to optimize the weights of the two predictions, realize the dynamic weight adjustment mechanism, and automatically adjust the weights according to the prediction accuracy of different time scales; the residual of the ARIMA model is used as a new time series feature, the moving average and standard deviation of the ARIMA model prediction value are calculated, the stability of the prediction is captured, the seasonal component of the ARIMA model is extracted, and it is input into the machine learning model as a periodic feature; for short-term predictions, the ARIMA model results are mainly used; for medium- and long-term predictions, the weight of the machine learning model is gradually increased, a rolling prediction mechanism is realized, the prediction starting point is regularly updated, and the real-time nature of the prediction is maintained.

[0022] As a preferred solution of the method of detecting NG product marking paper using a camera instead of a color mark sensor according to the present invention, the method of predicting possible equipment problems or quality anomalies based on historical data and real-time quality trends also includes:

[0023] Set multi-level warning thresholds, including minor anomalies, moderate anomalies, and severe anomalies.

[0024] On the second aspect, in order to further solve the problems existing in the inherent defects of color mark sensors, the embodiments of the present invention provide a system for detecting NG product marking paper by using a camera instead of a color mark sensor, which includes: a camera system initialization module, which is used to initialize the camera system, collect environmental data and adjust dynamic thresholds; a real-time image capture module, which is used to use a smart camera to capture label images in real time, calculate label parameters, synchronize MARK positions, and calculate label relative positions; a dynamic threshold determination module, which is used to determine label parameters according to the adjusted dynamic threshold, and perform real-time quality trend analysis on the determination results; a prediction and automatic adjustment module, which is used to predict possible equipment problems or quality abnormalities based on historical data and real-time quality trends, and perform automatic closed-loop adjustments. When potential problems are predicted, the system automatically generates maintenance suggestions and notifies relevant personnel.

[0025] In a third aspect, an embodiment of the present invention provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the method for detecting NG product marking paper by using a camera instead of a color mark sensor as described in the first aspect of the present invention is implemented.

[0026] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the method for detecting NG product marking paper by using a camera instead of a color mark sensor as described in the first aspect of the present invention is implemented.

[0027] The beneficial effects of the present invention are as follows: through environmental data collection and dynamic threshold adjustment, the system can adapt to different environmental conditions in real time, maintain high-precision detection performance, and reduce false detection rate and missed detection rate; it solves the problem that the color mark sensor can only detect the presence or absence of label paper, but cannot detect skew, position offset and the color mark sensor cannot monitor the "position of the label from the MARK hole"; the present invention significantly improves the accuracy and stability of detection through an intelligent camera system and an adaptive algorithm, and through the analysis of historical data and real-time quality trends, it realizes the prediction and prevention of equipment problems and quality anomalies. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. 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 creative work. Among them:

[0029] Figure 1 Flow chart of the method for detecting NG product marking paper using a camera instead of a color mark sensor.

[0030] Figure 2 This is a diagram of the system that uses a camera instead of a color mark sensor to detect NG product marking paper. DETAILED DESCRIPTION

[0031] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0032] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0033] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0034] Example 1

[0035] Reference Figure 1 and Figure 2, which is the first embodiment of the present invention, provides a method for detecting NG product marking paper using a camera instead of a color mark sensor, comprising the following steps:

[0036] S1: Initialize the camera system, collect environmental data and adjust the dynamic threshold.

[0037] The system initialization includes: replacing the original color mark sensor with a smart camera; setting the initial parameters of the smart camera, including the image length Y, so that it is an integer multiple of the laser head distance X (X=n×Y), where n is a multiple; initializing the dynamic threshold system and setting the initial thresholds of various detection parameters.

[0038] Environmental data collection is to monitor production environment parameters in real time, such as temperature, humidity, light, etc.

[0039] Dynamic threshold adjustment includes: Based on environmental data, the detection threshold is dynamically adjusted using an adaptive algorithm to ensure that the system maintains high accuracy under different conditions. The specific calculation is as follows:

[0040] Real-time monitoring of environmental parameters E(t), including temperature (T), humidity (H), light (L), etc., assuming the initial detection threshold is :

[0041] ,

[0042] in, is the benchmark detection threshold, , as well as They are the weight coefficients of environmental data, obtained through training with historical data; , and Environmental parameters for real-time monitoring;

[0043] Using Historical Data Train a linear regression model to determine the optimal regression coefficients:

[0044] ,

[0045] The real-time environmental data is input into the trained linear regression model to obtain the dynamically adjusted detection threshold θ(t).

[0046] S2: Use a smart camera to capture label images in real time, calculate label parameters, synchronize MARK positions, and calculate label relative positions.

[0047] Preferably, parameters such as "label leakage distance", "label tilt angle", "label presence", and "whether the label is installed upside down" are calculated by an image processing algorithm.

[0048] Furthermore, the specific calculation process is as follows:

[0049] Let the label boundary coordinates in the image be and , and the reference boundary coordinates that the label should reach and , then the label leakage distance formula is:

[0050] ,

[0051] Assume that the upper edge of the label is represented by two points in the image and , then the label tilt angle formula is:

[0052] ,

[0053] The label area is detected by the image processing algorithm. If the label area is detected, the label exists, otherwise the label does not exist. Let I be the input image and T be the label template:

[0054] ,

[0055] in, is the matching threshold, a and b are the pixel coordinates in the image, P label is the probability of label existence;

[0056] Determine whether the label is installed upside down:

[0057] ,

[0058] in, is the detected label direction; is the standard orientation of the label; is the judgment threshold.

[0059] MARK position synchronization and tag position calculation include the following steps:

[0060] When the laser head of the device cuts the MARK hole, it sends a "mark signal" to the smart camera. When the smart camera receives the mark signal, it records the current frame number as the "received MARK frame number". When a label is detected, it records the current frame number as the detected label frame number. The position of the label from the MARK hole is calculated:

[0061] The distance between the tag and the MARK hole = X-[(Number of tag frames detected-Number of MARK frames received)×Y+Z]

[0062] Where Y is the image length; X is the distance from the laser head; and Z is the correction factor or offset.

[0063] S3: According to the dynamically adjusted threshold value, the calculated parameters are judged, and the judgment results are analyzed for real-time quality trends.

[0064] Preferably, each parameter to be detected is calculated in real time, and the calculated parameter value is compared with the corresponding dynamic threshold; the classification criteria for the judgment result are set, including: if the parameter value is within the threshold range, it is displayed as normal (OK); if the parameter value is close to the threshold boundary but does not exceed it, it is displayed as slightly deviated (Warning); if the parameter value exceeds the threshold range, it is displayed as unqualified (NG); and each parameter is classified according to the comparison result.

[0065] When an NG situation is detected, the alarm system is immediately triggered, sending a stop signal to the production line control system, and generating a problem description (such as "label leakage position deviation", "label position deviation", etc.) based on the specific NG parameters, and sending the problem description to the operator interface and management system.

[0066] Create a unique identifier for each determination, record the determination time, parameter values, corresponding thresholds, determination results and other information, and store the records in the database for subsequent analysis and tracing.

[0067] Input each test result into the quality trend analysis module in real time; generate real-time quality reports, including the change trend of various parameters, NG rate, etc.; display the quality trend to the operator through a visual interface to facilitate timely detection of potential problems.

[0068] Furthermore, the processing contents of the quality trend analysis module include:

[0069] Receive the judgment results of each test in real time and set the analysis time window (such as the last 1 hour, 8 hours, 24 hours, etc.).

[0070] Perform statistical calculations on each parameter in different time windows, including: mean, standard deviation, maximum and minimum values, and rate of change; calculate the proportion of various determination results, especially the NG rate; use time series analysis methods to predict short-term trends:

[0071] Test to see if the time series is stationary. If not, perform a difference operation to make it stationary:

[0072] ,

[0073] Among them, B is the lag operator, d is the number of differences, and xt is the value of the time series before time t.

[0074] Use the ACF (autocorrelation function) and PACF (partial autocorrelation function) plots to determine the model's autoregressive (AR) term p and moving average (MA) term q; select appropriate p, d, and q parameters to fit the ARIMA model:

[0075] ,

[0076] in, and are AR and MA polynomials respectively, is the error term; p is the number of autoregressive terms; q is the number of moving average terms, and yt is the value of the time series at time t.

[0077] Check the autocorrelogram and other diagnostic plots of the residuals to ensure that the residuals are white noise; if not, adjust the model parameters and refit.

[0078] Use the fitted ARIMA model to predict future values:

[0079] ,

[0080] Set abnormal mode judgment rules, such as: if the standard deviation of the NG rate in the 24-hour window exceeds 1.5%, and the NG rate in the 8-hour window shows an upward trend (rate of change > 0.1% / hour), and the ARIMA model predicts that the NG rate in the next 4 hours will exceed 5%, it will be judged as "urgent need for quality control intervention"; if the average value of a parameter in the 1-hour window is less than 10% away from the boundary of the allowable range, and the standard deviation in the 8-hour window shows an expanding trend (rate of change > 5% / hour), and the residual test of the ARIMA model shows non-white noise characteristics, it will be judged as "the parameter is about to be out of control and needs to be adjusted immediately";

[0081] If the number of differences d of the ARIMA model increases from 1 to 2 or higher within 24 hours, and the AIC (Akaike Information Criterion) value of the model continues to rise, and the parameter fluctuation range (maximum value - minimum value) within the 1-hour window exceeds 30% of the allowable range, it is judged as "the system stability has seriously declined and a comprehensive inspection is required"; if the actual value of the ARIMA model of a certain parameter falls outside the 95% confidence interval in three consecutive predictions, and the difference between the average values ​​of the 1-hour and 8-hour windows of the parameter exceeds 20%, and the NG rate of the 24-hour window is 1.5 times higher than the historical average level, it is judged as "abnormal fluctuation, there may be unidentified interference factors";

[0082] If the autoregressive term p of the ARIMA model increases from 2 to 3 or higher within 24 hours, and the moving average term q also increases accordingly, and the parameter change rate within the 1-hour window exceeds 80% of the historical maximum value, it is judged as "the system complexity has increased suddenly, and major changes may have occurred"; if the error between the ARIMA model predicted value and the actual value of a parameter shows a monotonically increasing trend in 10 consecutive predictions, and the standard deviation within the 8-hour window exceeds 3 times the standard deviation of the 1-hour window, and the change rate of the 24-hour NG rate exceeds 0.5% / day, it is judged as "the model prediction ability has decreased, and the system may undergo gradual changes";

[0083] If within the 8-hour window, the ARIMA models of more than 3 parameters need to increase the number of differences d at the same time, and the 1-hour window average values ​​of these parameters are at the edge of the allowable range (such as less than 5% from the boundary), and the ARIMA forecast of the overall NG rate shows that it will exceed the historical highest value in the next 24 hours, then it is judged as "multiple parameter linkage anomaly, the system faces the risk of collapse"; if the predicted value of the ARIMA model of a certain parameter within 24 hours shows obvious periodic fluctuations (such as significant cycles detected by fast Fourier transform), and the cycle is not related to the production shift, and the NG rates of the 1-hour and 8-hour windows both show an upward trend (the rate of change is >0.2% / hour), then it is judged as "potential periodic interference factors, requiring in-depth analysis";

[0084] If the residuals of the ARIMA model show heteroskedasticity for 24 consecutive hours (as confirmed by the ARCH test), and the standard deviation of at least three parameters in the 8-hour window increases by more than 50%, and the fluctuation range of the NG rate in the 1-hour window (maximum value - minimum value) exceeds 5%, it is judged as "the system volatility intensifies and the prediction difficulty increases"; if the parameters (p, d, q) of the ARIMA model change frequently in the 24-hour window (for example, they need to be re-determined every 4 hours), and the difference between the average parameter value in the 1-hour window and the average value in the 24-hour window exceeds 30%, and the ARIMA forecast of the NG rate shows that there will be at least two peaks in the next 8 hours (the peak is defined as more than 150% of the current NG rate), it is judged as "the system characteristics change dramatically, and the control strategy may need to be redesigned", etc.

[0085] Apply abnormal pattern determination rules to the calculation results; mark and record the detected abnormal patterns.

[0086] Based on the trend analysis results, set the warning trigger conditions. When the warning conditions are met, notify the relevant personnel through interface reminders and message push, and record all warning events, including the trigger cause, time and processing status.

[0087] S4: Based on historical data and real-time quality trends, predict possible equipment problems or quality anomalies and make automatic closed-loop adjustments. When potential problems are predicted, the system automatically generates maintenance recommendations and notifies relevant personnel.

[0088] Preferably, historical production data, including past test results, environmental parameters, equipment status, etc., are extracted from the database, and the real-time quality trend data generated in step S3 is integrated and standardized; features that may be related to equipment problems or quality anomalies are extracted, and interactions between different parameters are captured.

[0089] Combining ARIMA forecast results with machine learning models can improve forecast accuracy:

[0090] The prediction results of the ARIMA model are added as new features to the input of the machine learning model, the difference between the ARIMA prediction value and the actual value is calculated as an additional feature, and the parameters of the ARIMA model (such as p, d, q values) are extracted as indicator features of the model complexity.

[0091] Assume that the original feature is X and the new feature vector is Xnew: Xnew = [ X , x ^ t + k ARIMA , e t + k , p , d , q ] ,in, is the difference between the ARIMA predicted value and the actual value.

[0092] The expanded feature vector Xnew is used to train and predict the machine learning model. The machine learning model of the present invention uses a gradient boosting regression tree because it performs well in processing complex feature relationships and nonlinear data.

[0093] After training is complete, the gradient boosted regression tree model can be used to predict new data.

[0094] Build a meta-model with ARIMA forecast results and machine learning model forecast results as input, use techniques such as stacking or blending, train the meta-model to optimize the weights of the two forecasts, implement a dynamic weight adjustment mechanism, and automatically adjust the weights according to the forecast accuracy at different time scales.

[0095] Use the residuals of the ARIMA model as new time series features, calculate the moving average and standard deviation of the ARIMA model forecasts, capture the stability of the forecast, and extract the seasonal components of the ARIMA model as periodic features to input into the machine learning model.

[0096] For short-term forecasts, we mainly rely on the results of the ARIMA model. For medium- and long-term forecasts, we gradually increase the weight of the machine learning model, implement a rolling forecast mechanism, regularly update the forecast starting point, and maintain the real-time nature of the forecast.

[0097] Set multiple levels of warning thresholds, including minor anomalies: anomaly scores exceeding μ+1.5σ (μ is the average score, σ is the standard deviation); moderate anomalies: anomaly scores exceeding μ+2σ; severe anomalies: anomaly scores exceeding μ+3σ. The anomaly score is obtained by calculating the difference between the predicted value and the actual value, and is usually expressed using the standardized value of the residual.

[0098] Perform dynamic threshold adjustment: implement a sliding window mechanism, recalculate μ and σ regularly, consider seasonal factors, and set different baselines for different periods; set independent thresholds for different parameters or parameter combinations, and implement logical combinations between parameters.

[0099] Preferably, the automatic closed-loop adjustment comprises the following steps:

[0100] The system automatically makes fine adjustments based on the calculated parameters such as "the distance between the label and the position of the MARK hole", continuously monitors the adjustment effects, and continuously optimizes the prediction model.

[0101] Furthermore, the test data is recorded to achieve full life cycle traceability of the product; comprehensive reports are generated regularly, including quality trends, predictive maintenance recommendations, system performance evaluation, etc.

[0102] This embodiment also provides a system for detecting NG product marking paper using a camera instead of a color mark sensor, including:

[0103] The camera system initialization module is used to initialize the camera system and collect environmental data and adjust the dynamic threshold;

[0104] Real-time image capture module, used to use smart camera to capture label images in real time, calculate label parameters, synchronize MARK position, and calculate label relative position;

[0105] The dynamic threshold determination module is used to determine the label parameters according to the adjusted dynamic threshold and to perform real-time quality trend analysis on the determination results;

[0106] The prediction and automatic adjustment module is used to predict possible equipment problems or quality anomalies based on historical data and real-time quality trends, and perform automatic closed-loop adjustments. When potential problems are predicted, the system automatically generates maintenance recommendations and notifies relevant personnel.

[0107] This embodiment also provides a computer device, which is suitable for the case where a camera replaces a color mark sensor to detect NG product marking paper, including: a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the method of using a camera instead of a color mark sensor to detect NG product marking paper as proposed in the above embodiment.

[0108] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covered on the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0109] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method for realizing the method of replacing the color mark sensor with a camera to detect NG product marking paper as proposed in the above embodiment is implemented; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, referred to as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, referred to as EPROM), programmable read-only memory (Programmable Red-Only Memory, referred to as PROM), read-only memory (Read-Only Memory, referred to as ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0110] In summary, through environmental data collection and dynamic threshold adjustment, the present invention enables the system to adapt to different environmental conditions in real time, maintain high-precision detection performance, and reduce false detection rate and missed detection rate; it solves the problem that the color mark sensor can only detect the presence or absence of label paper but cannot detect skew, position offset and the color mark sensor cannot monitor the "position of the label from the MARK hole"; the present invention significantly improves the accuracy and stability of detection through an intelligent camera system and an adaptive algorithm, and through the analysis of historical data and real-time quality trends, it realizes the prediction and prevention of equipment problems and quality anomalies.

[0111] Example 2

[0112] Referring to Table 1, which is the second embodiment of the present invention, in order to further verify the beneficial effects of the present invention, experimental data of a method of using a camera instead of a color mark sensor to detect NG product marking paper is provided.

[0113] In order to improve the accuracy and efficiency of marking paper detection on a battery production line, a method of using a camera instead of a color mark sensor to detect NG product marking paper was implemented. The production line produces about 50,000 batteries per day. The original color mark sensor system had problems such as a high false detection rate and inability to adapt to environmental changes.

[0114] First, the system was initialized. The original color mark sensor was replaced with a high-resolution (4096x3000 pixels) smart camera with a capture frequency of 200 frames per second. The image length Y of the smart camera was set to 100 mm, and the distance X from the laser head was set to 500 mm, satisfying the relationship X=n×Y, where n=5. At the same time, the dynamic threshold system was initialized, and the initial thresholds of various detection parameters were set: the label leakage distance threshold was ±0.5 mm, the label tilt angle threshold was ±1°, and the label matching threshold was 0.85.

[0115] The environmental data acquisition system includes a temperature sensor (accuracy ±0.1°C), a humidity sensor (accuracy ±1%RH), and a light sensor (accuracy ±10lux). The system collects environmental data every 5 seconds and inputs the data into the dynamic threshold adjustment module.

[0116] The dynamic threshold adjustment uses an adaptive algorithm. First, by analyzing historical data, the weight coefficients of environmental parameters are obtained: αT=0.3, αH=0.4, αL=0.3. Then, the linear regression model is trained using the historical data of the last 7 days to obtain the optimal regression system: β0=0.95, βT=0.02, βH=0.015, βL=0.01.

[0117] During the production process, smart cameras capture label images in real time and calculate various parameters through image processing algorithms.

[0118] The MARK position synchronization uses high-precision timestamp technology. When the laser head cuts the MARK hole, it sends a "mark signal" to the smart camera to accurately record the current frame number; when the label is detected, the frame number is also recorded. Through the difference between these two timestamps, combined with the image length Y and the distance X from the laser head, the position of the label from the MARK hole is accurately calculated.

[0119] During the judgment and quality trend analysis phase, the system generates a quality report every hour, uses the ARIMA model for time series analysis, and automatically selects the optimal value for the model parameters using the AIC criterion.

[0120] At the same time, multiple complex abnormal pattern judgment rules have been implemented, such as "if the standard deviation of the NG rate within the 24-hour window exceeds 1.5%, and the NG rate within the 8-hour window shows an upward trend (rate of change > 0.1% / hour), and the ARIMA model predicts that the NG rate will exceed 5% in the next 4 hours, it will be judged that quality control intervention is urgently needed".

[0121] In order to verify the effectiveness of this method, a 7-day test was conducted, with 1,000 batteries randomly selected for testing every day. The following is part of the test results:

[0122] Table 1 Part of the experimental data of the present invention

[0123]

[0124] As can be seen from the table above, the detection accuracy of the new system has shown a steady upward trend, from 99.2% on the first day to 99.8% on the seventh day. This continuous improvement reflects the system's adaptive and learning capabilities, and demonstrates the effectiveness of dynamic threshold adjustment and ARIMA model prediction. In contrast, the accuracy of traditional color code sensor systems usually fluctuates between 97% and 98%, and the new system has significantly improved in accuracy.

[0125] The false alarm rate was reduced from 0.5% to 0.1%, and the missed detection rate was reduced from 0.3% to 0.1%. The reduction in the missed detection rate is particularly important for ensuring product quality.

[0126] By observing the data trends, we can see that the system performance shows a continuous improvement in all indicators. This trend shows that the system has the ability of self-learning and optimization, which is the embodiment of the real-time quality trend analysis and predictive maintenance functions in the present invention.

[0127] It should be noted that the above embodiments are only used to illustrate the technical solutions 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 solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for detecting NG product marking paper by using a camera instead of a color mark sensor, characterized in that: include: The camera system is initialized and environmental data is collected and dynamic thresholds are adjusted; Use smart cameras to capture label images in real time, calculate label parameters, synchronize MARK positions, and calculate label relative positions; According to the adjusted dynamic threshold, the label parameters are judged, and the judgment results are analyzed for real-time quality trends; Based on historical data and real-time quality trends, it predicts possible equipment problems or quality anomalies and makes automatic closed-loop adjustments. When potential problems are predicted, it automatically generates maintenance recommendations and notifies relevant personnel. The real-time quality trend analysis of the determination results comprises the following steps: Receive the judgment results of each test in real time and set the analysis time window; Perform statistical calculations on each detection parameter in different time windows, including: mean value, standard deviation, maximum and minimum values, and rate of change, and calculate the proportion of each type of determination result; Use time series analysis methods to predict short-term trends; Apply abnormal pattern determination rules to the calculation results, mark and record the detected abnormal patterns; The use of time series analysis methods to predict short-term trends includes: Test to see if the time series is stationary. If not, perform a difference operation: , Among them, B is the lag operator, d is the number of differences, and xt is the value of the time series before time t; Use the autocorrelation function ACF and partial autocorrelation function PACF graphs to determine the autoregressive AR term p and moving average MA term q of the model; Choose the p, d, and q parameters and fit the ARIMA model: , in, and are AR and MA polynomials respectively, is the error term, yt is the value of the time series at time t; Check the autocorrelogram and diagnostic plots of the residuals to ensure that the residuals are white noise; if not, adjust the model parameters and refit; Use the fitted ARIMA model to predict future values: 。 2. The method for detecting NG product marking paper by using a camera instead of a color mark sensor as claimed in claim 1, characterized in that: The camera system initialization includes: Replace the original color mark sensor with a smart camera; Set the initial parameters of the smart camera, including the image length Y, so that it is an integer multiple of the laser head distance X: X=n×Y Among them, n is a multiple relationship; Initialize the dynamic threshold system and set the initial threshold of the detection parameters.

3. The method for detecting NG product marking paper by using a camera instead of a color mark sensor as claimed in claim 2, characterized in that: The synchronization of MARK position and calculation of tag relative position include the following contents: When the laser head of the device cuts the MARK hole, it sends a mark signal to the smart camera; When the smart camera receives a mark signal, it records the current frame number as the received MARK frame number; When a tag is detected, the current frame number is recorded as the number of frames at which the tag is detected; The formula for calculating the distance between the label and the MARK hole is as follows: The distance between the tag and the MARK hole = X-[(Number of tag frames detected-Number of MARK frames received)×Y+Z] Where Y is the image length; X is the distance from the laser head; and Z is the correction factor or offset.

4. The method for detecting NG product marking paper by using a camera instead of a color mark sensor as claimed in claim 3, characterized in that: The method of predicting possible equipment problems or quality abnormalities based on historical data and real-time quality trends includes the following steps: Extract historical production data from the database, integrate short-term trend data, and standardize; Combine ARIMA predicted future values ​​with machine learning models, add ARIMA predicted future values ​​as new features to the input of the machine learning model, calculate the difference between ARIMA predicted future values ​​and actual values ​​as additional features, and extract the parameters p, d, and q values ​​of the ARIMA model as indicator features of model complexity; Let the new eigenvector be Xnew: ,in, The difference between the ARIMA predicted future value and the actual value; Use the new feature vector Xnew to train and predict the machine learning model; Build a meta-model, take the ARIMA predicted future values ​​and the machine learning model prediction results as input, train the meta-model to optimize the weights of the two predictions, implement a dynamic weight adjustment mechanism, and automatically adjust the weights according to the prediction accuracy at different time scales; Use the residuals of the ARIMA model as new time series features, calculate the moving average and standard deviation of the ARIMA predicted future values, capture the stability of the forecast, extract the seasonal component of the ARIMA model, and input it into the machine learning model as a periodic feature; For short-term forecasts, ARIMA is mainly used to predict future values; for medium- and long-term forecasts, the weight of the machine learning model is gradually increased to implement a rolling forecast mechanism, and the forecast starting point is regularly updated to maintain the real-time nature of the forecast.

5. The method for detecting NG product marking paper by using a camera instead of a color mark sensor as claimed in claim 4, characterized in that: The prediction of possible equipment problems or quality anomalies based on historical data and real-time quality trends also includes: Set multi-level warning thresholds, including minor anomalies, moderate anomalies, and severe anomalies.

6. A system for detecting NG product marking paper by using a camera instead of a color mark sensor, based on the method for detecting NG product marking paper by using a camera instead of a color mark sensor according to any one of claims 1 to 5, characterized in that: include: The camera system initialization module is used to initialize the camera system and collect environmental data and adjust the dynamic threshold; Real-time image capture module, used to use smart camera to capture label image in real time, calculate label parameters, synchronize MARK position, and calculate label relative position; The dynamic threshold determination module is used to determine the label parameters according to the adjusted dynamic threshold and to perform real-time quality trend analysis on the determination results; The prediction and automatic adjustment module is used to predict possible equipment problems or quality anomalies based on historical data and real-time quality trends, and perform automatic closed-loop adjustments. When potential problems are predicted, the system automatically generates maintenance recommendations and notifies relevant personnel.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method for detecting NG product marking paper using a camera instead of a color mark sensor as described in any one of claims 1 to 5 are implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for detecting NG product marking paper using a camera instead of a color mark sensor as described in any one of claims 1 to 5 are implemented.