An electrode scrap automatic collecting method for a lithium battery automatic winding machine

Abstract

The present application relates to a kind of pole piece waste automatic collection method for lithium battery automatic winding machine in the field of information technology, comprising: vacuum suction nozzle array is set below cutting device, and is connected to central vacuum pump, according to the material characteristics and size of cutting waste, set appropriate vacuum degree and suction time, realize the automatic suction and transportation of cutting waste by controlling the start-stop and suction force size of vacuum pump;Flow sensor is installed in vacuum suction pipeline, the speed and pressure of suction airflow are monitored in real time, according to the preset flow curve, judge whether the suction process is normal, if there is suction force deficiency or airflow disorder abnormal situation, then promptly adjust vacuum pump parameter;Image acquisition and analysis are carried out to the waste suction, the material type and shape feature of waste are identified and classified using convolution neural network algorithm, and the quantity and weight of waste are estimated by combining suction time and vacuum degree data, provide basis for subsequent waste recovery and processing.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of information technology, in particular to the production technology of lithium battery cells, and more particularly to a method for automatically collecting scrap of pole pieces for a lithium battery automatic winding machine. BACKGROUND

[0002] In a lithium battery automatic winding machine, electrode sheet cutting and scrap collection are two key process links. Electrode sheet cutting requires precise control of cutting shape and size to ensure that the electrode sheet meets the requirements of the subsequent winding process. A large amount of cutting scrap is generated during the cutting process, which, if not properly handled, can affect the normal operation of the equipment and even cause equipment failure. At the same time, the presence of cutting scrap also occupies equipment space and affects production efficiency. Therefore, automatic collection and disposal of scrap need to be achieved at the same time as cutting.

[0003] Traditional scrap collection methods usually use manual cleaning or simple mechanical collection, which is inefficient and prone to pollution. In order to achieve efficient and pollution-free scrap collection, a vacuum suction method can be used to automatically extract cutting scrap to a designated collection device using strong suction. However, vacuum suction requires higher sealing and airflow control of the equipment, and requires precise control of suction time and suction force to avoid air leakage or insufficient suction during the suction process. In addition, while achieving automated scrap collection, the collection process needs to be monitored in real time to ensure that the collection effect meets expectations. The monitoring system needs to be able to accurately identify the type and quantity of scrap and automatically adjust the suction parameters according to the collection progress. At the same time, the monitoring system also needs to be able to timely detect and alarm abnormal situations such as suction channel blockage, full load of scrap collection device, etc. so as to be handled in time to ensure continuous and efficient operation of the equipment. SUMMARY

[0004] The present application provides a method for automatically collecting scrap of pole pieces for a lithium battery automatic winding machine, which comprises the following steps:

[0005] Firstly, the electrode sheet cutting process is monitored in real time by a machine vision system, the cutting contour image is obtained, the cutting edge is extracted by an edge detection algorithm, and whether the cutting accuracy meets the requirements is judged according to the preset electrode sheet size parameters. If the accuracy deviation exceeds the threshold, a warning signal is issued and the cutting parameters are automatically adjusted.

[0006] Secondly, a vacuum nozzle array is arranged below the cutting device and connected to a central vacuum pump. According to the material properties and size of the cutting scrap, appropriate vacuum degree and suction time are set, and by controlling the start-stop of the vacuum pump and the suction force, automatic suction and transportation of the cutting scrap are realized.

[0007] Third step, install flow sensor in vacuum suction pipeline, real-time monitor suction airflow speed and pressure, according to preset flow curve, judge whether suction process is normal, if abnormal condition such as insufficient suction or airflow turbulence appears, then adjust vacuum pump parameter in time, ensure that suction effect is stable;

[0008] Fourth step, carry out image acquisition and analysis to suctioned waste, adopt convolution neural network algorithm to identify and classify material type and shape feature of waste, and combine suction time and vacuum degree data, estimate waste quantity and weight, provide basis for subsequent waste recycling and processing;

[0009] Fifth step, set sensor in waste collection device, sensor includes liquid level sensor, pressure sensor and weight sensor, real-time monitor state of collection device, when device is about to be full, send early warning signal, remind timely replacement or cleaning collection device, avoid suction channel blockage or waste overflow;

[0010] Sixth step, establish data model of waste collection process, comprehensively consider cutting material, size, speed and other parameters, and vacuum suction time, frequency, suction force and other parameters, adopt support vector machine algorithm to predict and optimize collection efficiency, and according to actual collection data, continuously train and improve model;

[0011] Seventh step, set monitoring interface of waste collection in system control software, real-time display running state of cutting device, suction device and collection device, and show change trend of key parameters in the form of curve chart and column chart, when abnormal data appears, system automatically marks and generates alarm information, and give possible cause analysis and processing suggestion, assist operator to carry out problem diagnosis and maintenance.

[0012] The technical scheme provided by the embodiment of the application can include the following beneficial effects:

[0013] The application discloses a method for automatically collecting electrode scrap of a lithium battery automatic winding machine, and the method realizes real-time monitoring of a cutting process through machine vision, extracts a cutting contour by using an edge detection algorithm, and judges whether the cutting precision meets the requirements. A vacuum nozzle array is arranged below a cutting device, a suitable vacuum degree and suction time are set according to the characteristics of the scrap, and automatic suction and conveying of the scrap are realized. The application also uses a flow sensor to monitor the suction airflow, and timely adjusts the parameters of a vacuum pump to ensure the stability of the suction effect. The suctioned scrap is subjected to image analysis and identification classification, and the quantity and weight of the scrap are estimated. A plurality of sensors are arranged in the collecting device, the state is monitored in real time, and a warning signal is sent. The application establishes a scrap collection data model, uses a machine learning algorithm to predict and optimize the collection efficiency. The system control software sets a monitoring interface, displays the running state of each device in real time, automatically marks the abnormality, and gives analysis suggestions. The application realizes intelligent automatic collection of electrode scrap, and improves the production efficiency and the scrap recycling rate. DETAILED DESCRIPTION

[0014] In order to make the purpose, technical scheme and advantages of the application clearer, the application will be described in detail below with specific examples.

[0015] The application discloses a method for automatically collecting electrode scrap of a lithium battery automatic winding machine, and the method realizes real-time monitoring of a cutting process through machine vision, extracts a cutting contour by using an edge detection algorithm, and judges whether the cutting precision meets the requirements. A vacuum nozzle array is arranged below a cutting device, a suitable vacuum degree and suction time are set according to the characteristics of the scrap, and automatic suction and conveying of the scrap are realized. The application also uses a flow sensor to monitor the suction airflow, and timely adjusts the parameters of a vacuum pump to ensure the stability of the suction effect. The suctioned scrap is subjected to image analysis and identification classification, and the quantity and weight of the scrap are estimated. A plurality of sensors are arranged in the collecting device, the state is monitored in real time, and a warning signal is sent. The application establishes a scrap collection data model, uses a machine learning algorithm to predict and optimize the collection efficiency. The system control software sets a monitoring interface, displays the running state of each device in real time, automatically marks the abnormality, and gives analysis suggestions. The application realizes intelligent automatic collection of electrode scrap, and improves the production efficiency and the scrap recycling rate.

[0016] The first step is to monitor the cutting process of the electrode sheet in real time through a machine vision system, obtain a cutting contour image, extract the cutting edge by using an edge detection algorithm, judge whether the cutting precision meets the requirements according to preset electrode sheet size parameters, and if the precision deviation exceeds a threshold value, send a warning signal and automatically adjust the cutting parameters.

[0017] The first step includes: obtaining real-time image data in the cutting process of the electrode sheet, pre-processing the image data to improve the quality of the image data; for the pre-processed image data, extracting a cutting contour by using an edge detection algorithm to obtain a cutting edge image; according to preset electrode sheet size parameters, performing size measurement on the cutting edge image to obtain the actual size of the cut electrode sheet; comparing the actual size with the preset size parameters to calculate a size deviation value; judging whether the size deviation value exceeds a preset threshold value, if yes, triggering a warning signal; obtaining the size deviation value, using a machine learning algorithm to adaptively adjust the cutting parameters to obtain adjusted cutting parameters; feeding back the adjusted cutting parameters to the cutting device to guide the cutting process, so that the cutting precision continuously meets the preset requirements.

[0018] Exemplary, through the high-resolution industrial camera to collect the real-time image data in the electrode piece cutting process, the image resolution reaches 2048x2048 pixels, the sampling frequency is 60 frames / s. The collected image data is preprocessed, the median filter algorithm is used to remove image noise, the signal-to-noise ratio is improved by more than 3dB, and then the histogram equalization algorithm is used to enhance the image contrast, so that the cutting contour is clearer. For the preprocessed image, the Canny edge detection algorithm is used to extract the cutting contour, and by setting the high and low threshold values T1=50, T2=100, the cutting edge is accurately located in the binary image, and a clear cutting edge image is obtained. According to the preset electrode piece target size parameters, the pixel calibration method is used to measure the size of the cutting edge image, and by calibrating that 1 pixel corresponds to an actual physical size of 0.5mm, the length and width dimensions of the cut electrode piece are calculated, and the measurement accuracy is ±1mm. The actual size measured is compared with the preset target size, the length and width deviation values are calculated, and it is judged whether the cutting accuracy is within the tolerance range of ±2mm. If the cutting accuracy deviation exceeds the preset threshold value, the PLC controller sends an alarm signal to notify the operator to pause the production line for inspection, and at the same time the deviation data is uploaded to the MES system through the industrial Ethernet. The control system uses the BP neural network algorithm, takes the cutting speed, pressure and other parameters as input, and takes the cutting accuracy deviation value as output, and trains and optimizes the network through sample data. When the accuracy deviation value exceeds ±2mm, the cutting speed and pressure parameters are adjusted through back propagation, such as reducing the cutting speed from 60m / min to 55m / min and increasing the pressure from 2MPa to 2MPa. After 10 iterations of optimization, the cutting accuracy deviation value is controlled within ±1.5mm. The adjusted cutting parameters are sent to the PLC controller of the cutting device through Profibus bus, the speed of the servo motor and the pressure of the cylinder are adjusted in real time, and the cutting accuracy is continuously monitored through the machine vision system, forming a closed-loop feedback control, realizing automatic adjustment and intelligent management of the cutting process, ensuring that the electrode piece cutting accuracy is stable within ±1mm, and improving the battery production efficiency and quality consistency.

[0019] Second step, a vacuum nozzle array is arranged below the cutting device and connected to a central vacuum pump. According to the material properties and size of the cutting waste, the appropriate vacuum degree and suction time are set, and by controlling the start-stop and suction force of the vacuum pump, the automatic suction and transportation of the cutting waste are realized.

[0020] The second step includes: obtaining the structural characteristics and layout information of the cutting device, designing a distribution scheme for the vacuum suction nozzle array based on the information, which can cover all positions in the cutting area; establishing a material property database including material density, hardness and viscosity parameters; using image recognition technology to obtain the size information of the cutting waste; dynamically adjusting the suction force and distribution of the vacuum suction nozzle array according to the material property database and the size information; controlling the start-stop and operating power of the central vacuum pump through a PID control algorithm, and achieving accurate control of the vacuum degree according to the vacuum degree set value and suction time parameters; setting a vacuum pipeline system between the vacuum suction nozzle array and the central vacuum pump, minimizing pressure loss by optimizing the pipeline diameter, material and layout of the vacuum pipeline system; combining material properties, waste size and vacuum degree factors to establish a mathematical model of the cutting waste suction process; training the mathematical model using simulation optimization and machine learning algorithms to obtain optimized suction strategies and control parameters; setting a servo positioning system between the vacuum suction nozzle array and the cutting device, and controlling the servo positioning system to dynamically adjust the position and angle of the vacuum suction nozzle array according to real-time information of the cutting trajectory and waste generation position.

[0021] For example, according to the 2m×5m cutting range of the cutting device, an 8×10 vacuum suction nozzle array is arranged below it, with a nozzle diameter of 10mm and a spacing of 150mm, and the layout is optimized through finite element analysis to ensure that the vacuum degree in the cutting area reaches-60kPa. Vacuum adsorption tests are conducted on 100 common materials to establish a mapping database of material hardness, density, and required vacuum degree and adsorption time. Convolutional neural network algorithm is used for semantic segmentation of cutting waste images to obtain real-time waste size distribution, and the vacuum degree of each nozzle is dynamically adjusted according to the distribution characteristics, with an adsorption efficiency of more than 95% for waste with a diameter of 2-50mm. Fuzzy PID control algorithm is used to adjust the speed of the vacuum pump in real time according to the set-60kPa vacuum degree and 2s adsorption time, with a pressure fluctuation of less than ±5kPa. The pipe diameter and layout of the pipeline system are optimized, and a layered structure with an inner diameter of 20-100mm is used, reducing pressure loss by 30% and increasing conveying speed by 50%. A multi-physical field coupled mathematical model of the adsorption process is established, genetic algorithm is used to optimize the simulation parameters, and reinforcement learning algorithm is used to continuously optimize the control strategy, with an adsorption efficiency improvement of 10%. An electric servo positioning system is set between the nozzle array and the cutting device, and the cutting trajectory is detected in real time by a laser sensor, the nozzle pose is adjusted 50ms in advance according to the trajectory speed and acceleration information, and dynamic tracking of the waste adsorption point is realized, with a positioning accuracy of ±1mm.

[0022] Third step, install flow sensor in vacuum suction pipeline, real-time monitor suction airflow speed and pressure, according to preset flow curve, judge whether suction process is normal, if suction abnormality such as insufficient suction or airflow turbulence occurs, adjust vacuum pump parameters in time to ensure stable suction effect.

[0023] The third step includes: obtaining a preset flow curve, which is used as a reference standard for judging whether the suction process is normal; real-time collecting speed and pressure data of the suction airflow through the flow sensor arranged in the vacuum suction pipeline; comparing the real-time collected speed and pressure data with the preset flow curve to obtain a deviation value; if the deviation value exceeds a preset threshold range, it is determined that the suction process is abnormal, and an abnormal processing procedure is triggered; in the abnormal processing procedure, the adjustment direction and amplitude of the vacuum pump parameters are determined according to the size and sign of the deviation value; the adjustment parameters corresponding to the adjustment direction and amplitude are sent to the vacuum pump control module; the vacuum pump control module adjusts the working state of the vacuum pump in real time according to the received adjustment parameters; the adjusted suction airflow speed and pressure are continuously monitored to obtain an adjusted deviation value; if the adjusted deviation value does not return to the preset normal range, the abnormal processing procedure is returned to execute; if the adjusted deviation value returns to the preset normal range, the adaptive optimization of the suction process is completed, and a signal indicating that the suction process adjustment is completed is output.

[0024] For example, first, a preset flow curve is needed as a reference standard, which can be obtained by statistical analysis of historical data, for example, a quadratic curve y=ax 2 +bx+c is obtained by using the least squares method to fit historical data as a standard curve, where x is time and y is flow. Then install a flow sensor such as a turbine flowmeter on the vacuum suction pipeline to real-time collect speed and pressure data of the suction airflow, and the sampling frequency can be set to 10 Hz. Substitute the collected speed and pressure data into the standard curve equation to calculate the deviation Ay of the actual flow value from the standard flow value. Set the deviation threshold to ±5% of the standard flow, if Ay exceeds the threshold, it is determined that the suction is abnormal. When handling the abnormality, the vacuum degree can be judged to be too high or too low according to the sign of Ay, and then multiplied by a proportional coefficient k as the adjustment amplitude of the vacuum pump parameters, and kAy is added to the current vacuum pump parameters to obtain the corrected parameter value, which is sent to the vacuum pump control module for real-time adjustment. Continuously collect the adjusted speed and pressure data to calculate the deviation value until Ay returns to within ±5%, and the adaptive optimization is completed. Through this method, real-time monitoring and automatic adjustment of the suction process can be realized to ensure the consistency and stability of the suction effect.

[0025] The fourth step is to collect and analyze images of the collected waste, identify and classify the material type and shape characteristics of the waste using a convolutional neural network algorithm, and estimate the quantity and weight of the waste by combining the suction time and vacuum data, to provide a basis for subsequent waste recycling and processing.

[0026] The fourth step includes: obtaining a waste image; preprocessing the waste image to obtain a preprocessed waste image; inputting the preprocessed waste image into a pre-trained convolutional neural network model, extracting features of the preprocessed waste image through the convolutional neural network model, and identifying the material type and shape characteristics of the waste; obtaining the suction time and vacuum data when the waste corresponding to the waste image is sucked; according to the material type and shape characteristics of the waste identified by the convolutional neural network model, combining the suction time and vacuum data, and using a regression algorithm to estimate the quantity and weight of the waste; comparing the estimated quantity and weight of the waste with a preset threshold value, if the quantity or weight of the waste exceeds the threshold value, it is determined as recyclable waste, otherwise it is determined as non-recyclable waste; if it is determined as recyclable waste, the waste is sent to a recycling processing device for recycling processing; if it is determined as non-recyclable waste, the waste is sent to a waste treatment device for harmless treatment; uploading the result data of the recycling processing and harmless treatment to a data management platform; optimizing and updating the convolutional neural network model according to the result data.

[0027] For example, the waste suction device sucks the waste through a vacuum suction cup, and records the suction time and vacuum data through a pressure sensor. The suction time is accurate to 1 second, and the vacuum degree is measured in the range of 0-100 kPa with a resolution of 1 kPa. The sucked waste is sent to the image collection device through the conveyor belt, and the industrial camera is used for multi-angle and multi-spectrum imaging of the waste to obtain color images and infrared images of the waste. The collected images are preprocessed, and the median filter algorithm is used to remove image noise and improve the signal-to-noise ratio by more than 3 dB; the histogram equalization algorithm is used to enhance the image contrast and improve the image quality. The preprocessed waste image is input into the pre-trained ResNet-50 convolutional neural network model to extract a 512-dimensional feature vector of the waste image, and the support vector machine algorithm is used to identify the material type of the waste with an accuracy of more than 95%; the Hough transform algorithm is used to extract the shape characteristics of the waste, such as straight lines and circles. According to the material type and shape characteristics of the waste, combined with the suction time and vacuum data, a multiple linear regression algorithm is used to estimate the quantity and weight of the waste, and the determination coefficient R 2Reaching 9 or above. The estimated waste quantity and weight are compared with preset thresholds. The waste quantity threshold is set at 100 pieces, and the weight threshold is set at 1kg. Any waste exceeding the threshold is considered recyclable. Based on the recyclability judgment, the recyclable waste is sent to the recycling and processing device by a robotic arm for crushing, cleaning, and sorting. Non-recyclable waste is sent to a high-temperature incinerator for harmless treatment, with the incineration temperature reaching above 800°C. The waste processing result data is uploaded to the data management platform via the Internet of Things. Big data analysis technology is used to mine and analyze the waste data and optimize the waste management process. At the same time, transfer learning technology is used to optimize and update the convolutional neural network model to improve the accuracy of waste identification and classification. The optimized model's recognition accuracy has increased by more than 2%.

[0028] Step 5. Install sensors in the waste collection device. These sensors include a liquid level sensor, a pressure sensor, and a weight sensor to monitor the status of the collection device in real time. When the device is about to be fully loaded, an early warning signal is issued to remind users to replace or clean the collection device in time to avoid blockage of the suction channel or overflow of waste.

[0029] The fifth step includes: obtaining real-time monitoring data of the liquid level sensor, pressure sensor and weight sensor installed in the waste collection device, and comprehensively judging the current state of the waste collection device based on the real-time monitoring data; judging whether the current state is full loaded based on the preset device full load threshold; if the current state exceeds the full load threshold, triggering an early warning signal; when the early warning signal is triggered, sending a prompt message to the management personnel, and the prompt message is used to prompt that the waste collection device needs to be replaced or cleaned; obtaining the time and device status data of each triggering of the early warning signal, as well as the time and operator information of the replacement or cleaning operation, and generating an operation log; regularly obtaining the operation log, using a machine learning algorithm to analyze the operation log to obtain an optimized full load threshold; setting the optimized full load threshold as the preset device full load threshold.

[0030] For example, a waste collection device is equipped with liquid level sensors, pressure sensors, and weight sensors. These sensors collect data every five seconds and transmit it to a monitoring system via a wireless network. The monitoring system utilizes a rule-based expert system to determine in real time whether the collected data is approaching a threshold based on pre-set device fullness thresholds (e.g., liquid level reaching 80%, pressure reaching 5 MPa, weight reaching 500 kg). When the monitored data reaches 90% of the threshold, the system triggers an early warning signal and automatically sends a voice and text message to the management staff, informing them that the collection device needs to be replaced or cleaned. Upon receiving the warning, the management staff uses the intelligent scheduling system to dispatch two nearby maintenance personnel to perform the replacement or cleaning operation, a process that takes no more than 30 minutes. After the operation is completed, the maintenance personnel use a handheld terminal to reset the monitoring data of the relevant sensors to their initial values ​​and enter information such as the replacement or cleaning time and the operator into the system, creating an electronic operation log. The system automatically analyzes the operation log daily and uses a decision tree algorithm to establish a correlation model between the fullness threshold and the warning trigger time. Training on a large amount of historical data optimizes the threshold setting, thereby improving the accuracy of the early warning signal. After two months of continuous optimization, the accuracy of the full load threshold increased from 85% to 95%, effectively reducing 20% ​​of unnecessary replacement or cleaning operations, and significantly improving the system's intelligence level and operation and maintenance efficiency.

[0031] Step 6. Establish a data model for the waste collection process, comprehensively considering parameters such as cutting materials, size, speed, as well as parameters such as vacuum suction time, frequency, and suction force. Use the support vector machine algorithm to predict and optimize the collection efficiency, and continuously train and improve the model based on actual collection data.

[0032] The sixth step comprises: acquiring real-time data in the running process of the collection device, the real-time data comprising cutting material, cutting size, cutting speed, suction time, suction frequency and suction suction force; judging the device state of the collection device according to the real-time data in combination with waste shape, waste weight and environmental temperature through a pre-established rule, and generating a warning signal if the parameters of the device state exceed a preset threshold range; after the warning signal is triggered, the collection device sends device state data to a data center to obtain a device state data set; a support vector machine algorithm is used to train the device state data set to generate an initial prediction model, and the initial prediction model is used to predict waste collection efficiency to obtain a predicted collection efficiency value; an actual collection efficiency value in an actual waste collection process is acquired, and a deviation between the actual collection efficiency value and the predicted collection efficiency value is determined to generate a deviation data set; a support vector machine algorithm is used to train the deviation data set to determine parameters that need to be adjusted in the initial prediction model to obtain an adjustment parameter data set; the initial prediction model is adjusted according to the adjustment parameter data set to obtain an optimized prediction model, and the optimized prediction model is used to improve the prediction accuracy of waste collection efficiency.

[0033] Exemplarily, the collection device generates process data when collecting waste, for example, a set of data is collected at a certain time: the cutting material is paperboard, the cutting size is 50 cm long and 30 cm wide, the cutting speed is 10 cm per second, the suction time is 5 seconds, the suction frequency is 2 times per second, and the suction suction is 500 pascal. At the same time, the real-time data of the current device running is obtained, combined with the waste shape being rectangular and the waste weight being 100 grams, the ambient temperature being 25 degrees Celsius, through the pre-established rule, for example, if the suction force to weight ratio is less than 4, the device state is judged to be low efficiency, at this time 500 pascal divided by 100 grams is equal to 5, which is greater than 4, the device state is normal. If at the next moment, the suction force is reduced to 350 pascal, the suction force to weight ratio is 5, which is less than 4, a warning signal is generated. After the warning signal is triggered, the collection device sends data to the data center through the message queue middleware, the data includes the device state data corresponding to each collection device, for example, the device A state data is the above data, the device B state data is that the cutting material is plastic, the cutting size is 40 cm by 25 cm, the cutting speed is 8 cm per second, the suction time is 6 seconds, the suction frequency is 5 times per second, the suction suction is 450 pascal, the waste shape is circular, the waste weight is 80 grams, the ambient temperature is 26 degrees Celsius, and the device state is normal. The collected data forms a device state data set. Through the device state data set, a support vector machine algorithm is used for model training, 80% of the data set is used as a training set and 20% is used as a test set, the kernel function is set to radial basis function, the parameter gamma is 1, and the penalty coefficient C is 10, the training is performed, and an initial prediction model is generated. The model is used to predict the waste collection efficiency, for example, a set of data is input, and the predicted collection efficiency value is 95%. The actual collection efficiency value of the actual waste collection process is obtained by calculating the waste weight collected in a unit of time and the total weight of the generated waste, and the actual collection efficiency value is 90%. According to the actual collection efficiency value 90% and the predicted collection efficiency value 95%, it is determined that the deviation between the two is 5%, and the deviation data is generated together with other deviation data to form a deviation data set. The support vector machine algorithm is used to train the deviation data set, and the parameters that need to be adjusted in the prediction model are determined through cross-validation and grid search method, for example, the gamma is adjusted to 05 and the C is adjusted to 5, and the adjusted parameter data set is obtained. The initial prediction model is adjusted by using the adjusted parameter data set, the parameters of the initial prediction model are updated, and an optimized prediction model is obtained. The optimized model is used for prediction, the predicted collection efficiency value is 92%, and the deviation from the actual value is reduced, which improves the model accuracy.

[0034] Step 7: Set up a monitoring interface for waste collection in the system control software, display the running status of the cutting device, suction device and collection device in real time, and show the trend of key parameters in the form of curve and column chart. When abnormal data appears, the system automatically marks and generates alarm information, and gives possible cause analysis and treatment suggestions to assist operators in problem diagnosis and maintenance.

[0035] Step 7 includes: obtaining real-time running state data stream of the cutting device, suction device and collection device, the real-time running state data stream being collected by built-in sensors of the equipment; judging whether the real-time running state data stream exceeds a preset threshold, if the real-time running state data stream exceeds the preset threshold, marking the real-time running state data stream as abnormal data; determining the type of the abnormal data according to a pre-established abnormal database; obtaining the cause analysis and treatment suggestion corresponding to the type of the abnormal data from a pre-established rule library; determining the parameters representing the running state of the cutting device, suction device and collection device in the real-time running state data stream as key parameters by using data visualization technology; arranging the key parameters in time sequence to obtain a key parameter time sequence; converting the key parameter time sequence into a curve chart and a column chart by using a graph drawing algorithm; associating the cause analysis and treatment suggestion with the abnormal data; displaying the associated cause analysis and treatment suggestion and abnormal data, and pushing the displayed cause analysis and treatment suggestion and abnormal data to a user terminal by using an information pushing mechanism.

[0036] For example, the cutting device, the suction device and the collecting device are provided with sensors which collect the running state data of the equipment every 1 second, including current, voltage, rotating speed, temperature and other parameters, to form a real-time equipment state data stream. The system presets the normal range threshold of each parameter, for example, the current is between 0-10A, the voltage is between 220V±5%, the rotating speed is between 1000-1500rpm, and the temperature is between 0-50℃. By comparing the difference between each parameter in the data stream and the threshold in real time, if it is found that a parameter exceeds the threshold range, for example, the current reaches 12A, the data point will be marked as abnormal data. The system analyzes the characteristics of the abnormal data based on the decision tree algorithm to determine the abnormal type, for example, the excessive current is probably caused by the high load of the equipment. According to the abnormal type, the system automatically generates an alarm information "equipment current abnormality" and sends it to the monitoring interface. The monitoring interface determines the reason for the excessive current by querying the abnormal reason knowledge base based on the association rule mining, which is pre-established, and gives the maintenance processing suggestion "check whether the equipment is aging and adjust the working strength of the equipment". At the same time, the monitoring interface receives the equipment state data stream, identifies the key parameters such as current, rotating speed and temperature by using data visualization technology, and arranges them in time sequence to form a time series. By using the line chart algorithm, the values in the time series of the key parameters are mapped to the points on the coordinate axis, and then the adjacent points are connected to form a line, so as to intuitively display the change trend of the parameters. For discrete parameters such as equipment state, the column chart algorithm is used to generate the height of the column according to the values corresponding to different states, to reflect the distribution of the equipment in each state. Finally, the monitoring interface displays the alarm information, the associated cause analysis and the processing suggestion in the form of a pop-up window, and pushes them to the mobile terminal of the maintenance personnel in the form of an instant message, to remind them to handle the abnormal situation as soon as possible to avoid affecting the production.

[0037] The above only lists some preferred embodiments of the present application, but the present application is not limited thereto, and many improvements and changes can be made. Any improvement and change made on the basis of the basic principles of the present application should be considered to fall within the protection scope of the present application.

Claims

1. A method for automatically collecting pole piece waste for a lithium battery automatic winding machine, characterized in that: The method comprises the following steps: The first step is to monitor the electrode sheet cutting process in real time through a machine vision system, obtain a cutting contour image, and use an edge detection algorithm to extract the cutting edge. Based on the preset electrode sheet size parameters, it is determined whether the cutting accuracy meets the requirements. If the accuracy deviation exceeds the threshold, a warning signal is issued and the cutting parameters are automatically adjusted. The second step is to set up a vacuum nozzle array under the cutting device and connect it to the central vacuum pump. According to the material characteristics and size of the cut waste, the appropriate vacuum degree and suction time are set. By controlling the start and stop of the vacuum pump and the suction force, the cut waste can be automatically sucked and transported. The third step is to install a flow sensor in the vacuum suction pipeline to monitor the speed and pressure of the suction airflow in real time. According to the preset flow curve, it is judged whether the suction process is normal. If there is an abnormal situation of insufficient suction or turbulent airflow, the vacuum pump parameters are adjusted in time to ensure stable suction effect. The fourth step is to collect and analyze images of the sucked waste, using a convolutional neural network algorithm to identify and classify the material type and shape characteristics of the waste. Combined with the suction time and vacuum data, the quantity and weight of the waste are estimated to provide a basis for subsequent waste recovery and treatment. Step 5: Install sensors in the waste collection device. These sensors include a liquid level sensor, a pressure sensor, and a weight sensor to monitor the status of the collection device in real time. When the device is about to be full, an early warning signal is issued to remind the user to replace or clean the collection device in time to avoid blockage of the suction channel or waste overflow. Step 6: Build a data model for the waste collection process, use support vector machine algorithms to predict and optimize collection efficiency, and continuously train and improve the model based on actual collection data; Step 7. Set up a waste collection monitoring interface in the system control software to display the operating status of the cutting device, suction device and collection device in real time, and show the changing trends of key parameters in the form of curve graphs and bar graphs. When abnormal data appears, the system automatically marks and generates an alarm message, and gives possible cause analysis and treatment suggestions to assist operators in problem diagnosis and maintenance.

2. The method for automatically collecting waste pole pieces for a lithium battery automatic winding machine according to claim 1, characterized in that: The first step comprises: Acquiring real-time image data during the electrode sheet cutting process, and preprocessing the image data to improve the quality of the image data; For the pre-processed image data, an edge detection algorithm is used to extract the cutting contour to obtain a cutting edge image; According to the preset electrode sheet size parameters, the size of the cut edge image is measured to obtain the actual size of the electrode sheet after cutting; Comparing the actual size with the preset size parameters to calculate the size deviation value; Determine whether the size deviation value exceeds a preset threshold, and if so, trigger a warning signal; Obtaining the size deviation value, and adaptively adjusting the cutting parameters using a machine learning algorithm to obtain adjusted cutting parameters; The adjusted cutting parameters are fed back to the cutting equipment to guide the cutting process so that the cutting accuracy continues to meet the preset requirements.

3. The method for automatically collecting waste pole pieces for a lithium battery automatic winding machine according to claim 1, characterized in that: The second step includes: Obtaining structural characteristics and layout information of the cutting device, and designing a distribution scheme of the vacuum nozzle array based on the information, wherein the distribution scheme can cover all positions within the cutting area; Establishing a material property database, wherein the database includes material density, hardness and viscosity parameters; Use image recognition technology to obtain the size information of the cut waste; Dynamically adjusting the suction force and distribution of the vacuum nozzle array according to the material property database and the size information; The start and stop and operating power of the central vacuum pump are controlled by PID control algorithm, and the vacuum degree is accurately controlled according to the vacuum degree setting value and suction time parameters; A vacuum piping system is provided between the vacuum nozzle array and the central vacuum pump, and pressure loss is minimized by optimizing the pipe diameter, material, and layout of the vacuum piping system; Combining the material properties, waste size and vacuum degree factors, a mathematical model of the cutting waste suction process is established; The mathematical model is trained using simulation optimization and machine learning algorithms to obtain optimized absorption strategies and control parameters; A servo positioning system is provided between the vacuum nozzle array and the cutting device, and the servo positioning system is controlled to dynamically adjust the position and angle of the vacuum nozzle array according to real-time information of the cutting trajectory and the waste generation position.

4. The method for automatically collecting waste pole pieces for a lithium battery automatic winding machine according to claim 1, characterized in that: The third step comprises: Obtaining a preset flow curve, which is used as a reference standard for determining whether the suction process is normal; The velocity and pressure data of the suction air flow are collected in real time by a flow sensor arranged in the vacuum suction pipeline; Comparing the real-time collected speed and pressure data with the preset flow curve to obtain a deviation value; If the deviation value exceeds the preset threshold range, the absorption process is determined to be abnormal, triggering the abnormality handling process; In the abnormality handling process, the adjustment direction and amplitude of the vacuum pump parameters are determined according to the magnitude and positive / negative sign of the deviation value; Sending adjustment parameters corresponding to the adjustment direction and amplitude to the vacuum pump control module; The vacuum pump control module corrects the working state of the vacuum pump in real time according to the received adjustment parameters; Continuously monitoring the adjusted suction airflow speed and pressure to obtain an adjusted deviation value; If the adjusted deviation value does not return to the preset normal range, return to execute the exception handling process; If the adjusted deviation value returns to the preset normal range, the adaptive optimization of the absorption process is completed, and a signal indicating that the absorption process adjustment is completed is output.

5. The method for automatically collecting pole piece waste for a lithium battery automatic winding machine according to any one of claims 1 to 4, characterized in that: The fourth step comprises: Obtaining waste images; preprocessing the waste image to obtain a preprocessed waste image; Inputting the pre-processed waste image into a pre-trained convolutional neural network model, extracting features of the pre-processed waste image through the convolutional neural network model, and identifying the material type and shape features of the waste; Acquiring suction time and vacuum degree data when sucking the waste corresponding to the waste image; estimating the quantity and weight of the waste using a regression algorithm based on the waste material type and shape characteristics identified by the convolutional neural network model and in combination with the suction time and vacuum degree data; Comparing the estimated waste quantity and weight with a preset threshold value, and determining that the waste is recyclable if the waste quantity or weight exceeds the threshold value, otherwise determining that the waste is non-recyclable; If it is determined to be recyclable waste, the waste will be sent to the recycling device for recycling; If it is determined to be non-recyclable waste, the waste will be sent to the waste treatment device for harmless treatment; Uploading the result data of the recycling and harmless treatment to a data management platform; The convolutional neural network model is optimized and updated according to the result data.

6. A method for automatically collecting pole piece waste for a lithium battery automatic winding machine according to any one of claims 1 to 4, characterized in that: The sixth step comprises: Acquire real-time data during the operation of the collection device, including cutting material, cutting size, cutting speed, suction time, suction frequency and suction force; Based on the real-time data, combined with the shape, weight and ambient temperature of the waste, the device status of the collection device is determined using pre-established rules, and if the parameters of the device status exceed a preset threshold range, an early warning signal is generated; After the early warning signal is triggered, the acquisition device sends the device status data to the data center to obtain a device status data set; Using a support vector machine algorithm to perform model training on the equipment status data set to generate an initial prediction model, and using the initial prediction model to predict the waste collection efficiency to obtain a predicted collection efficiency value; Obtaining an actual collection efficiency value during an actual waste collection process, determining a deviation between the actual collection efficiency value and the predicted collection efficiency value based on the actual collection efficiency value and the predicted collection efficiency value, and generating a deviation data set; Using a support vector machine algorithm to perform model training on the deviation data set, determine the parameters that need to be adjusted in the initial prediction model, and obtain an adjustment parameter data set; The initial prediction model is adjusted according to the adjustment parameter data set to obtain an optimized prediction model, and the prediction accuracy of the waste collection efficiency is improved by the optimized prediction model.

7. The method for automatically collecting waste pole pieces for a lithium battery automatic winding machine according to claim 1, characterized in that: The seventh step comprises: Acquire real-time operating status data streams of the cutting device, the suction device, and the collection device, wherein the real-time operating status data streams are collected by built-in sensors of the equipment; determining whether the real-time running status data flow exceeds a preset threshold, and if the real-time running status data flow exceeds the preset threshold, marking the real-time running status data flow as abnormal data; Determining the type of the abnormal data according to a pre-established abnormal database; According to the type of the abnormal data, obtain the cause analysis and processing suggestions corresponding to the abnormal data type from a pre-established rule base; Using data visualization technology, the parameters representing the operating states of the cutting device, the suction device, and the collection device in the real-time operating state data stream are determined as key parameters; Arranging the key parameters according to a time series to obtain a key parameter time series; Using a graphics drawing algorithm, converting the key parameter time series into a curve chart and a bar chart; Associating the cause analysis and treatment suggestions with the abnormal data; The associated cause analysis and treatment suggestions are displayed together with the abnormal data, and an information push mechanism is used to push the displayed cause analysis and treatment suggestions together with the abnormal data to the user terminal.