Parameter optimization control system and method based on ZB45 packaging machine visual inspection equipment

Abstract

The invention discloses a parameter optimization control system and method based on visual inspection equipment of a ZB45 packaging machine, and the whole control system employs an industrial control computer, builds a local area network cloud server, collects the related detection data and operation parameters of each visual inspection device, carries out the deep learning and twinborn detection of the collected data, and achieves the optimal control of the parameters of the visual inspection equipment. Aiming at the defects of information isolated island, time consumption in adjustment and optimization, poor precision and the like existing in a traditional visual detection solution, intelligent development is carried out, and a unified monitoring platform and a parameter adjustment and optimization intelligent analysis system of each visual detector are established. After the method is implemented, the parameter tuning duration of each visual inspection device can be reduced, and the effective operation rate of the device is improved; according to the method, the defect detection precision is improved, missed detection and false detection of defective cigarettes are reduced, the functions of state monitoring, parameter optimization, self-tuning adjustment, remote control and the like of each detection device are realized, and a foundation is laid for intelligent construction in the field of cigarette detection devices.

Description

Technical Field

[0001] The invention relates to the technical field of quality detection of cigarette production equipment, and in particular to a parameter optimization control system and method based on visual detection equipment of a ZB45 packaging machine. Background Art

[0002] Visual inspection devices are widely used in cigarette production lines. These devices use high-resolution cameras and image processing algorithms to conduct real-time inspections of the appearance of cigarettes, and can accurately identify stains, scratches, damage and other defects on cigarettes. At the same time, visual inspection devices can also inspect the packaging quality of cigarettes to ensure that the packaging is complete and beautiful. With the rapid development of artificial intelligence, big data analysis technology and digital technology, cigarette production quality inspection equipment has also ushered in new changes, moving towards higher detection accuracy, automation, more complete detection algorithms, and intelligent solutions.

[0003] At present, the visual inspection equipment widely used in domestic cigarette factory packaging equipment includes empty and missing cigarette detection, No. 3 and No. 5 wheel detection, small package wire detection, etc., which are distributed in multiple key stations of the winding and packaging production line. Visual inspection equipment mostly adopts rule-based detection, which requires operators to identify features based on experience, set detection areas, algorithm types, detection parameters, etc. The operating parameters of the visual inspection equipment need to match the detection target. In order for the inspection equipment to achieve the ideal effect, the parameters must be continuously optimized and adjusted according to the usage. This will result in a long time-consuming parameter tuning, which not only affects production efficiency, but also easily causes material waste and quality risks. Summary of the invention

[0004] The purpose of the present invention is to overcome the deficiencies in the prior art and provide a parameter optimization control system and method based on the visual inspection equipment of the ZB45 packaging machine. In view of the deficiencies of information islands, time-consuming tuning, and poor precision in traditional visual inspection solutions, the present invention can reduce the time for parameter tuning and improve the effective operation rate of the equipment; improve the accuracy of defect detection and reduce the occurrence of missed detection and false detection of defective cigarettes. Implement the transformation of related electronic / electrical equipment, use emerging technologies such as industrial Internet, machine vision, detection twins, and deep learning, build an efficient and reliable parameter optimization control system and optimization method for visual inspection equipment of packaging machines, realize the functions of status monitoring, parameter optimization, self-tuning adjustment, remote control, etc. of each detection equipment, and lay the foundation for the intelligent construction of cigarette inspection equipment in this field.

[0005] The present invention discloses a parameter optimization control system and method based on the visual inspection equipment of ZB45 packaging machine, which targets two major drawbacks of poor operating performance of the visual inspection equipment: the operating parameters need to be continuously optimized and adjusted during the production process, the parameter tuning takes a long time, and the detection area, algorithm type, and detection parameters are manually set by the rule detection method. The number of detection area settings is limited, and analysis and research are carried out on the situation where defects are missed in uncovered areas; the visual inspection equipment includes a composite cigarette detector, a three- to five-wheel detector, and a small package pull-wire detector (multiple visual inspection equipment include a short detector, a missing cigarette detector, a three- and five-wheel detector, and a small package pull-wire detector).

[0006] In order to achieve the above object, the technical solution adopted by the present invention is: A parameter optimization control system based on the visual inspection equipment of the ZB45 packaging machine, which includes a high-performance industrial computer, multiple visual inspection equipment of the ZB45 packaging machine, a human-computer interaction module and a data communication system between the modules. The industrial computer contains system software. The industrial computer uploads various types of data collected by various visual inspection equipment to the terminal device through the communication protocol to realize the status monitoring, data analysis, detection twin, deep learning and remote control applications of various equipment; multiple visual inspection equipment include empty detectors, missing branch detectors, third and fifth wheel detectors, and small package wire detectors; the system software solution adopts a top-down hierarchical design approach for software development from the perspective of management, upgrade and maintenance, and is divided into different hierarchical structures, including acquisition interaction layer, integrated processing layer, modeling analysis layer and decision control layer; efficiently integrate the real-time data of various visual inspection equipment to realize the status monitoring, data analysis, detection twin, deep learning and remote control of various equipment; Among them, status monitoring: through free access and rapid networking of various visual inspection equipment, the parameter optimization control system can centrally manage the networking data source modules, realize the working status and real-time data monitoring of various visual inspection equipment, and query the history of fault information, judge the operating status of each visual inspection equipment in real time, and provide timely feedback when a fault occurs; Data analysis: Through the storage and statistics of production test data, multi-dimensional comparison and scientific calculation and analysis of test algorithm parameters can be achieved, so that test data can further play an important role in assisting monitoring equipment status and scientifically improving test performance, providing key data support for quality traceability and process optimization of industrial production; Among them, the detection twin: through the twin replication of historical and real-time data and algorithm models in the parameter optimization control system, a digital mirror symmetrical to the actual detection process is constructed to achieve high-fidelity parameter synchronization, algorithm testing and twin application of field status and detection behavior, reduce the cost of long-term tracking and optimization of detection equipment, realize rapid prediction of parameter adjustment, and provide scientific guidance for parameter tuning of detection equipment; Deep learning: By creating a defect detection model based on a convolutional neural network, training and evaluating the model, and deploying it on existing inspection equipment, deep learning technology is deeply integrated with visual inspection technology to significantly improve the accuracy and stability of defect detection and reduce the false-kick rate and missed-kick rate; Remote control: through dedicated image encoding and decoding algorithms and high-speed network transmission protocols, the human-machine interface of the visual inspection equipment in the system can be viewed in real time to achieve remote management and control.

[0007] Furthermore, the human-machine interaction module: designs an integrated control cabinet and embeds the human-machine interface in the display. The layout and component configuration inside the control cabinet are the key to ensuring the stable operation of the industrial automation system, including industrial computers, screen modules, switching power supplies and fans, to achieve efficient space utilization, optimized thermal management and reliable system performance, providing a solid hardware foundation for system parameter control.

[0008] Furthermore, the deep learning algorithm selects image segmentation model defect recognition, and the algorithm based on target detection labels the positive and negative samples according to supervised learning. The algorithm completes feature learning, so that the model accuracy rate reaches 0.99, the loss function is less than 0.01, and the model inference time is less than 50ms.

[0009] Furthermore, the deep learning model training method selects LAN cloud server training, the LAN cloud server upgrades hardware resources, configures the environment, builds model training software on the cloud server, and performs model training by automatically collecting samples after establishing a remote connection with the detector during the production process or in production intervals.

[0010] Furthermore, the data analysis adopts K-means spatial clustering analysis algorithm, based on the partitioning algorithm, the distance between initial cluster centers is as far as possible, and the algorithm is run by importing photoelectric detection results, adjusting algorithm parameters, and the algorithm satisfies the silhouette coefficient>0.8, Davis-Balding index<0.5, and Kalinsky-Harabas index>500.

[0011] Furthermore, the detection twin adopts cloud-based synchronous twin, and the detection twin function is uniformly configured on the cloud server. During the production process or production intervals, a digital mirror is established by remotely connecting to the detector and synchronizing parameters, and samples are collected remotely to perform detection twin.

[0012] Furthermore, the human-computer interaction interface adopts a 20-24 inch screen module, which specifically includes an LCD screen module, a touch driver board assembly, and a display driver board assembly. It has an aspect ratio of 16:9, a resolution of 1920*1080, and the interface is optionally powered by 12V according to demand; the main control processor adopts a high-performance RCX series industrial computer.

[0013] An optimization method for a parameter optimization control system based on a visual inspection device of a ZB45 packaging machine comprises the following steps: Step S1: Select a 20-24 inch screen module, which includes an LCD screen, a touch driver board assembly, and a display driver board assembly. Use a high-performance RCX series industrial computer. The two are designed into an integrated control cabinet. Install the industrial computer according to the assembly drawing, and assemble and power on the main controller for testing. Step S2: Using the Ethernet module Ha-VIS econ and the seven-category 10 Gigabit network cable, the visual inspection controller and the control system main control processor are connected through a switch, and the main control processor and the IP address of each detector are set in the same network segment. The main control processor is connected to each detector through the TCP protocol to realize network communication; Step S3: Build a LAN cloud server to realize automatic training of the detection model on the LAN cloud server, select a deep learning algorithm for defect recognition of the image detection model, and train the model by automatically collecting samples after establishing a remote connection with the detector during the production process or production interval; first, organize a total of 10,000 visual images of tobacco, first divide the data into OK and NG according to positive and negative samples, and then annotate the data through the labeling tool labelimg according to the requirements of supervised learning, introduce the YOLOV5 detection model, write relevant detection algorithms, write visual simulation software, and intuitively verify the accuracy, false rejection rate, missed rejection rate and model reasoning speed of the detection model; Step S4: Perform K-means algorithm operation, sort out a total of 10,000 optoelectronic vision pictures, first divide the 10,000 data into OK and NG according to positive and negative samples, and convert the upper and lower lines of information into two-dimensional data coordinate information according to the script, write the K-means++ algorithm based on the Python language, adjust the algorithm parameters, run the algorithm, and demonstrate and test the K-means algorithm; Step S5: Build a cloud server, uniformly configure the detection twin function in the cloud server, enable the cloud server to actively obtain the detector configuration parameter information stored in the human-machine interface, and build a digital twin with the same operating environment as the site; implement the cloud quality inspection laboratory function, adjust the algorithm parameters, and verify the parameter tuning effect through a large amount of historical and real-time data; implement the cloud server to update the parameter data to the human-machine interface and detector by brand overloading, and send the twin brand's algorithm and parameter configuration files to the designated detector. The human-machine interface and the server shake hands and confirm in turn to complete the detection upgrade of the designated brand.

[0014] The present invention is a parameter optimization control system and method based on the visual inspection equipment of the ZB45 packaging machine, which combines digital twin, deep learning defect detection technology and traditional detection parameter tuning technology. By building a cloud server, a digital mirror is established by remotely connecting with the detector and synchronizing parameters, and each visual detector is twinned to realize the functions of real-time display and collection of each visual inspection data. The collected image data set is annotated, and model training is carried out to generate a deep learning defect detection model file, which is deployed after performance evaluation, avoiding the complex feature extraction process, and the performance of image recognition far exceeds that of traditional methods. The accuracy rate is continuously optimized as the amount of data increases. For the detection of photoelectric data, cluster data analysis is performed to mine the correlation between data, identify abnormal points, and provide support for parameter optimization decisions. Through a series of technical means, a parameter optimization control system and method for each visual detector in the entire process of cigarette pack production are established, and intelligent analysis of visual detection and intelligent optimization configuration of parameters are realized, providing a scientific basis and production guarantee for complex and changeable detection environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 The hardware design principle diagram of the parameter optimization control system of the present invention; Figure 2 The software design principle diagram of the parameter optimization control system of the present invention; Figure 3 This is a physical display diagram of the parameter optimization control system of the present invention. DETAILED DESCRIPTION

[0016] In order to make the technical solution and advantages of the present invention clearer, the technical solution of the present invention will be further described in detail in detail with reference to the accompanying drawings. It can be understood that the specific embodiments described here are only partial embodiments of the present invention, which are only used to explain the present invention, not to limit the present invention. It should be noted that, for the convenience of description, only the parts / structures related to the present invention are shown in the accompanying drawings, and other related parts can refer to the general design. In the absence of conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other to obtain new embodiments.

[0017] Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. In addition, unless otherwise defined, the technical terms or scientific terms used in the description of the present invention should be the common meanings understood by ordinary technicians in the field to which the present invention belongs.

[0018] The present invention will be further described in detail below in conjunction with the accompanying drawings.

[0019] like Figure 1-3As shown, a parameter optimization control system based on the visual inspection equipment of the ZB45 packaging machine includes a high-performance industrial computer (including system software), multiple visual inspection equipment of the ZB45 packaging machine, a human-computer interaction module and a data communication system between modules. The industrial computer uploads various types of data collected by each visual inspection equipment to the terminal device through the communication protocol, so as to realize applications such as equipment status monitoring, data analysis, detection twins, deep learning and remote control; multiple visual inspection equipment include empty head detector, missing branch detector, third and fifth wheel detectors, and small package pulling wire detector.

[0020] From the perspective of management, upgrade and maintenance, the system software solution adopts a top-down hierarchical design approach for software development and is divided into different hierarchical structures, including acquisition and interaction layer, integrated processing layer, modeling and analysis layer, and decision-making control layer; it efficiently integrates the real-time data of various visual inspection equipment to realize applications such as equipment status monitoring, data analysis, detection twins, deep learning, and remote control.

[0021] Among them, status monitoring: through free access and rapid networking of various visual inspection equipment, the parameter optimization control system can centrally manage the networking data source modules, realize the working status and real-time data monitoring of various visual inspection equipment, and query the history of fault information, judge the operating status of each visual inspection equipment in real time, and provide timely feedback when a fault occurs; Data analysis: Through the storage and statistics of production test data, multi-dimensional comparison and scientific calculation and analysis of test algorithm parameters can be achieved, so that test data can further play an important role in assisting monitoring equipment status and scientifically improving test performance, providing key data support for quality traceability and process optimization of industrial production; Among them, the detection twin: through the twin replication of historical and real-time data and algorithm models in the parameter optimization control system, a digital mirror symmetrical to the actual detection process is constructed to achieve high-fidelity parameter synchronization, algorithm testing and twin application of field status and detection behavior, reduce the cost of long-term tracking and optimization of detection equipment, realize rapid prediction of parameter adjustment, and provide scientific guidance for parameter tuning of detection equipment; Deep learning: By creating a defect detection model based on a convolutional neural network, training and evaluating the model, and deploying it on existing inspection equipment, deep learning technology is deeply integrated with visual inspection technology to significantly improve the accuracy and stability of defect detection and reduce the false-kick rate and missed-kick rate; Remote control: through dedicated image encoding and decoding algorithms and high-speed network transmission protocols, the human-machine interface of the visual inspection equipment in the system can be viewed in real time to achieve remote management and control.

[0022] Among them, the human-machine interaction module: design an integrated control cabinet and embed the human-machine interface into the display. The layout and component configuration inside the control cabinet are the key to ensuring the stable operation of the industrial automation system, including industrial computers, screen modules, switching power supplies and fans, to achieve efficient space utilization, optimized thermal management and reliable system performance, providing a solid hardware foundation for system parameter control.

[0023] Among them, the deep learning algorithm selects image segmentation model defect recognition, and the algorithm based on target detection labels the positive and negative samples according to supervised learning. The algorithm completes feature learning, so that the model accuracy rate reaches 0.99, the loss function is less than 0.01, and the model inference time is less than 50ms.

[0024] The deep learning model training method chooses LAN cloud server training. The LAN cloud server upgrades hardware resources, configures the environment, builds model training software on the cloud server, and performs model training by automatically collecting samples after establishing a remote connection with the detector during the production process or during production breaks.

[0025] The data analysis uses the K-means spatial clustering analysis algorithm, which is based on the partitioning algorithm. The distance between the initial cluster centers is as far as possible. By importing the photoelectric detection results, adjusting the algorithm parameters, and running the algorithm, the algorithm meets the requirements of silhouette coefficient>0.8, Davis-Balding index<0.5, and Kalinsky-Harabas index>500.

[0026] The detection twin adopts cloud-based synchronous twin, and the detection twin function is uniformly configured on the cloud server. During the production process or production intervals, a digital mirror is established by remotely connecting to the detector and synchronizing parameters, and samples are collected remotely to perform detection twin.

[0027] The human-computer interaction interface uses the Weierke 21.5-inch screen module, which specifically includes an LCD screen module, a touch driver board assembly, and a display driver board assembly. It has an aspect ratio of 16:9 and a resolution of 1920*1080. The interface can be equipped with 12V power supply as required. The main control processor uses the Chaoen RCX-2330 industrial computer.

[0028] The deep learning defect detection technology adopted in the present invention can enable the composite cigarette detector, three-five wheel detection, and small package pull line detection (multiple visual inspection equipment including empty head detector, missing cigarette detector, three- and five-wheel detector, and small package pull line detector) to avoid complex feature extraction processes, avoid over-reliance on manual experience and subjective judgment, and improve the ability to adapt to different environments through model training of collected images and deployment of production models; as learning continues to deepen, the accuracy of detection will continue to be optimized and improved.

[0029] The human-computer interaction module of the present invention adopts a touch industrial computer with expandable communication interface types and channel numbers. Through digital twin technology, it realizes data synchronization with all historical data, real-time data, detection parameters, etc. of the compound cigarette detector, three- to five-round detection, and small package pulling wire detection, and realizes high-fidelity parameter synchronization and algorithm testing of on-site status and detection behavior, reducing the cost of long-term tracking and optimization of detection equipment.

[0030] An optimization method for a parameter optimization control system based on a visual inspection device of a ZB45 packaging machine comprises the following steps: Step S1: Select a 20-24 inch Weike screen module, which includes an LCD screen, a touch driver board assembly, and a display driver board assembly. Use a Chaowen RCX-2330 industrial computer. The two are designed into an integrated control cabinet. Install the industrial computer according to the assembly drawing, and assemble and power on the main controller for testing. Step S2: Using the Ethernet module Ha-VIS econ and the seven-category 10 Gigabit network cable, the visual inspection controller and the control system main control processor are connected through a switch, and the main control processor and the IP address of each detector are set in the same network segment. The main control processor is connected to each detector through the TCP protocol to realize network communication; Step S3: Build a LAN cloud server to realize automatic training of the detection model on the LAN cloud server, select a deep learning algorithm for defect recognition of the image detection model, and train the model by automatically collecting samples after establishing a remote connection with the detector during the production process or production interval; first, organize a total of 10,000 visual images of tobacco, first divide the data into OK and NG according to positive and negative samples, and then annotate the data through the labeling tool labelimg according to the requirements of supervised learning, introduce the YOLOV5 detection model, write relevant detection algorithms, write visual simulation software, and intuitively verify the accuracy, false rejection rate, missed rejection rate and model reasoning speed of the detection model; Step S4: Perform K-means algorithm operation, sort out a total of 10,000 optoelectronic vision pictures, first divide the 10,000 data into OK and NG according to positive and negative samples, and convert the upper and lower lines of information into two-dimensional data coordinate information according to the script, write the K-means++ algorithm based on the Python language, adjust the algorithm parameters, run the algorithm, and demonstrate and test the K-means algorithm; Step S5: Build a cloud server, uniformly configure the detection twin function in the cloud server, enable the cloud server to actively obtain the detector configuration parameter information stored in the human-machine interface, and build a digital twin with the same operating environment as the site; implement the cloud quality inspection laboratory function, adjust the algorithm parameters, and verify the parameter tuning effect through a large amount of historical and real-time data; implement the cloud server to update the parameter data to the human-machine interface and detector by brand overloading, and send the twin brand's algorithm and parameter configuration files to the designated detector. The human-machine interface and the server shake hands and confirm in turn to complete the detection upgrade of the designated brand.

[0031] Through the above steps, each visual inspection device can be freely accessed and quickly networked. The system can efficiently integrate the real-time data of each visual inspection device. By storing and counting the inspection data and performing calculation and analysis of the inspection algorithm, the monitoring function of the detector is optimized and the performance of the detector is scientifically improved. Through the twin replication of multiple visual inspection devices, the parameter synchronization, algorithm testing and twin application of the parameter control system and the on-site visual detector are realized, which reduces the parameter tuning time. The deep learning inspection model is deployed on the existing inspection equipment, and the deep learning technology is deeply integrated with the traditional visual inspection technology, which significantly improves the accuracy and stability of defect detection and reduces the false kick rate and missed kick rate.

[0032] The present invention is a parameter optimization control system and method based on the visual inspection equipment of the ZB45 packaging machine, which combines digital twin, deep learning defect detection technology and traditional detection parameter tuning technology. By building a cloud server, a digital mirror is established by remotely connecting with the detector and synchronizing parameters, and each visual detector is twinned to realize the functions of real-time display and collection of each visual inspection data. The collected image data set is annotated, and model training is carried out to generate a deep learning defect detection model file, which is deployed after performance evaluation, avoiding the complex feature extraction process, and the performance of image recognition far exceeds that of traditional methods. The accuracy rate is continuously optimized as the amount of data increases. For the detection of photoelectric data, cluster data analysis is performed to mine the correlation between data, identify abnormal points, and provide support for parameter optimization decisions. Through a series of technical means, a parameter optimization control system and method for each visual detector in the entire process of cigarette pack production are established, and intelligent analysis of visual detection and intelligent optimization configuration of parameters are realized, providing a scientific basis and production guarantee for complex and changeable detection environments.

[0033] The parameter optimization control system and method based on the visual inspection equipment of the ZB45 packaging machine of the present invention have the following beneficial technical effects: by establishing a collaborative system of visual inspection equipment, parameter optimization control of the visual inspection equipment of the production line-level packaging machine is realized, and the traditional visual inspection application is promoted to enter a new stage of refined performance parameter tuning, thereby effectively improving production efficiency and cigarette product quality, reducing production costs, and actively promoting the upgrade and improvement of cigarette packaging production line quality inspection.

[0034] The above-mentioned implementation modes are for explanation of the present invention rather than limitation of the present invention. It can be understood that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principle and spirit of the present invention. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A parameter optimization control system based on the visual inspection equipment of the ZB45 packaging machine, which includes a high-performance industrial computer, multiple visual inspection equipment of the ZB45 packaging machine, a human-computer interaction module, and a data communication system between the modules. The industrial computer contains system software. The industrial computer uploads various types of data collected by various visual inspection equipment to the terminal device through the communication protocol to realize the status monitoring, data analysis, detection twin, deep learning and remote control applications of various equipment; multiple visual inspection equipment include empty detectors, missing detectors, third and fifth wheel detectors, and small package wire detectors; The system software solution adopts a top-down hierarchical design approach to software development from the perspective of management, upgrade and maintenance, and is divided into different hierarchical structures, including the collection and interaction layer, the integrated processing layer, the modeling and analysis layer, and the decision-making and control layer; Efficiently integrate the real-time data of various visual inspection equipment to realize equipment status monitoring, data analysis, inspection twins, deep learning and remote control applications; Among them, status monitoring: through free access and rapid networking of various visual inspection equipment, the parameter optimization control system can centrally manage the networking data source modules, realize the working status and real-time data monitoring of various visual inspection equipment, and query the history of fault information, judge the operating status of each visual inspection equipment in real time, and provide timely feedback when a fault occurs; Data analysis: Through the storage and statistics of production test data, multi-dimensional comparison and scientific calculation and analysis of test algorithm parameters can be achieved, so that test data can further play an important role in assisting monitoring equipment status and scientifically improving test performance, providing key data support for quality traceability and process optimization of industrial production; Among them, the detection twin: through the twin replication of historical and real-time data and algorithm models in the parameter optimization control system, a digital mirror symmetrical to the actual detection process is constructed to achieve high-fidelity parameter synchronization, algorithm testing and twin application of field status and detection behavior, reduce the cost of long-term tracking and optimization of detection equipment, realize rapid prediction of parameter adjustment, and provide scientific guidance for parameter tuning of detection equipment; Deep learning: By creating a defect detection model based on a convolutional neural network, training and evaluating the model, and deploying it on existing inspection equipment, deep learning technology is deeply integrated with visual inspection technology to significantly improve the accuracy and stability of defect detection and reduce the false-kick rate and missed-kick rate; Remote control: through dedicated image encoding and decoding algorithms and high-speed network transmission protocols, the human-machine interface of the visual inspection equipment in the system can be viewed in real time to achieve remote management and control.

2. A parameter optimization control system based on the visual inspection equipment of the ZB45 packaging machine as claimed in claim 1, characterized in that: Among them, the human-machine interaction module: design an integrated control cabinet and embed the human-machine interface into the display. The layout and component configuration inside the control cabinet are the key to ensuring the stable operation of the industrial automation system, including industrial computers, screen modules, switching power supplies and fans, to achieve efficient space utilization, optimized thermal management and reliable system performance, providing a solid hardware foundation for system parameter control.

3. A parameter optimization control system based on the visual inspection equipment of the ZB45 packaging machine as described in claim 2, characterized in that: Among them, the deep learning algorithm selects image segmentation model defect recognition, and the algorithm based on target detection labels the positive and negative samples according to supervised learning. The algorithm completes feature learning, so that the model accuracy rate reaches 0.99, the loss function is less than 0.01, and the model inference time is less than 50ms.

4. A parameter optimization control system based on the visual inspection equipment of the ZB45 packaging machine as described in claim 3, characterized in that: The deep learning model training method chooses LAN cloud server training. The LAN cloud server upgrades hardware resources, configures the environment, builds model training software on the cloud server, and performs model training by automatically collecting samples after establishing a remote connection with the detector during the production process or during production breaks.

5. A parameter optimization control system based on the visual inspection equipment of the ZB45 packaging machine as claimed in claim 4, characterized in that: The data analysis uses the K-means spatial clustering analysis algorithm, which is based on the partitioning algorithm. The distance between the initial cluster centers is as far as possible. By importing the photoelectric detection results, adjusting the algorithm parameters, and running the algorithm, the algorithm meets the requirements of silhouette coefficient>0.8, Davis-Balding index<0.5, and Kalinsky-Harabas index>500.

6. A parameter optimization control system based on the visual inspection equipment of the ZB45 packaging machine as claimed in claim 5, characterized in that: The detection twin adopts cloud-based synchronous twin, and the detection twin function is uniformly configured on the cloud server. During the production process or production intervals, a digital mirror is established by remotely connecting to the detector and synchronizing parameters, and samples are collected remotely to perform detection twin.

7. A parameter optimization control system based on the visual inspection equipment of the ZB45 packaging machine as claimed in claim 6, characterized in that: The human-computer interaction interface adopts a 20-24-inch screen module, which specifically includes an LCD screen module, a touch driver board assembly, and a display driver board assembly. It has an aspect ratio of 16:9 and a resolution of 1920*1080. The interface can be equipped with 12V power supply according to needs. The main control processor adopts a high-performance RCX series industrial computer.

8. An optimization method for a parameter optimization control system based on a visual inspection device of a ZB45 packaging machine, comprising the parameter optimization control system based on a visual inspection device of a ZB45 packaging machine as claimed in claim 7, comprising the following steps: Step S1: Select a 20-24 inch screen module, which includes an LCD screen, a touch driver board assembly, and a display driver board assembly. Use a high-performance RCX series industrial computer. The two are designed into an integrated control cabinet. Install the industrial computer according to the assembly drawing, and assemble and power on the main controller for testing. Step S2: Using the Ethernet module Ha-VIS econ and the seven-category 10 Gigabit network cable, the visual inspection controller and the control system main control processor are connected through a switch, and the main control processor and the IP address of each detector are set in the same network segment. The main control processor is connected to each detector through the TCP protocol to realize network communication; Step S3: Build a LAN cloud server to realize automatic training of the detection model on the LAN cloud server, select a deep learning algorithm for defect recognition of the image detection model, and train the model by automatically collecting samples after establishing a remote connection with the detector during the production process or production interval; first, organize a total of 10,000 visual images of tobacco, first divide the data into OK and NG according to positive and negative samples, and then annotate the data through the labeling tool labelimg according to the requirements of supervised learning, introduce the YOLOV5 detection model, write relevant detection algorithms, write visual simulation software, and intuitively verify the accuracy, false rejection rate, missed rejection rate and model reasoning speed of the detection model; Step S4: Perform K-means algorithm operation, sort out a total of 10,000 optoelectronic vision pictures, first divide the 10,000 data into OK and NG according to positive and negative samples, and convert the upper and lower lines of information into two-dimensional data coordinate information according to the script, write the K-means++ algorithm based on the Python language, adjust the algorithm parameters, run the algorithm, and demonstrate and test the K-means algorithm; Step S5: Build a cloud server, uniformly configure the detection twin function in the cloud server, enable the cloud server to actively obtain the detector configuration parameter information stored in the human-machine interface, and build a digital twin with the same operating environment as the site; implement the cloud quality inspection laboratory function, adjust the algorithm parameters, and verify the parameter tuning effect through a large amount of historical and real-time data; implement the cloud server to update the parameter data to the human-machine interface and detector by brand overloading, and send the twin brand's algorithm and parameter configuration files to the designated detector. The human-machine interface and the server shake hands and confirm in turn to complete the detection upgrade of the designated brand.

Images