CN116032969B - Cloud-edge cooperative intelligent numerical control workshop self-regulation system and control method - Google Patents

Abstract

The invention discloses a cloud-edge collaborative intelligent CNC workshop self-regulation system and control method, which belongs to the field of intelligent manufacturing technology. This invention proposes a cloud-edge collaborative intelligent CNC workshop self-regulation model based on edge computing technology. On the basis of analyzing the CNC workshop operating mechanism and regulation characteristics, it builds an intelligent CNC workshop manufacturing unit edge sensing-cloud collaborative operation-intelligent regulation framework, completed the deployment of two-level collaborative interaction scenarios between the cloud and edge, and improved the computing performance of edge sensing nodes by designing the judgment logic of the rule engine; secondly, the long-short-term memory neural network was used to train and modify the cloud-edge processing model, and the cloud-edge processing model was formulated. Collaborative production logic completes the self-regulation of intelligent CNC workshop equipment; finally, through application cases, it is verified that the proposed cloud-edge collaboration framework has the characteristics of easy collaboration, easy regulation, and low delay, and provides technical support for realizing intelligent production in CNC workshops.

Description

A cloud-edge collaborative intelligent CNC workshop self-regulation system and control method

Technical field

The invention relates to a cloud-edge collaborative intelligent CNC workshop self-regulation system and control method, and belongs to the field of intelligent manufacturing technology.

Background technique

The continuous development of Internet of Things technology has accelerated the informatization and intelligence process of manufacturing workshops, helping to enhance the independent innovation and market competitiveness of manufacturing enterprises. Intelligent Manufacturing (IM) is the deep integration of manufacturing and Internet of Things information technology. It is the only way for enterprises to upgrade and optimize. It is also the key to my country's promotion of "Made in China 2025" to become the forefront of the world's manufacturing industry. The intelligent CNC workshop is an important embodiment of the IM production process. The continuous increase in the number and types of access equipment in the intelligent CNC workshop will generate a massive amount of real-time process data. How to efficiently utilize these data has become the key to the data processing technology of the intelligent CNC workshop. At present, most enterprises upload the real-time industrial data generated to cloud servers for centralized and unified processing. However, the process of importing and exporting massive data to the cloud center is very complicated, and problems such as insufficient bandwidth and large delays affect the direct interaction between the cloud center and the equipment. Therefore, the introduction of edge computing technology can better solve the above problems, that is, deploy edge gateways at the edge of the network close to devices or data sources, integrate core capabilities such as network, computing, and storage, and provide edge intelligent services nearby to meet different business fields. key needs.

In the face of massive heterogeneous multi-source data in intelligent CNC workshops, although edge computing technology can meet the needs of real-time computing, it cannot replace the high computing performance of cloud servers. It focuses on the perception of manufacturing resources in intelligent CNC workshops, hierarchical processing of data and cloud computing. A lot of research has been carried out on server construction and other related work. Lu Youlong et al. proposed a smart factory technology architecture with a data center and five industrial business layers, and discussed a dynamically optimized big data analysis method for the big data-driven manufacturing process, which is beneficial to solving the problem of multi-source heterogeneous data analysis in smart factories. Layer processing and data application issues. TAO et al. proposed cloud manufacturing access resource classification, manufacturing resource-aware access architecture and typical cases; Li Hai et al. proposed a machine tool equipment resource selection method based on multi-criteria decision-making for cloud manufacturing environment access resources. However, the above methods have better expanded the application of cloud service platforms in the industrial Internet of Things, but have not yet solved the need for real-time response on the device side. Therefore, cloud-edge collaboration technology based on edge computing is an inevitable trend in future development. At present, There are few studies on integrating cloud-edge collaboration into workshop production, and there is even less research on cloud-edge collaboration for self-regulation of intelligent CNC workshop production.

Contents of the invention

The present invention provides a cloud-edge collaborative intelligent CNC workshop self-regulation system and control method. It processes multi-source heterogeneous data in the CNC workshop by coupling edge computing and cloud computing to process multi-source heterogeneous data in the CNC workshop. By abstracting the process equipment of the CNC workshop As an edge node, the CNC workshop control center serves as a cloud application server to build a cloud-edge collaboration framework, which solves the problem of hierarchical scheduling and processing of multi-source heterogeneous data in the CNC workshop in the cloud center and edge nodes, and uses edge nodes to interact collaboratively with CNC equipment. The edge equipment responds quickly and solves the problem of self-regulation of manufacturing and production in CNC workshops.

The technical solution of the present invention is:

According to one aspect of the present invention, a cloud-edge collaborative intelligent CNC workshop self-regulation system is provided, including an intelligent CNC workshop cloud application center, an intelligent CNC workshop edge sensing node, and an intelligent CNC workshop terminal device; the intelligent CNC workshop cloud application The center is equipped with a cloud database, data application module, and edge application module. The cloud database is used to store non-sensitive data uploaded by edge sensing nodes; the data application module is used for data preprocessing of non-sensitive data in the cloud database and training of prediction models. Correction; the edge application module is used to issue the prediction model and processing program of the edge sensing node; the edge sensing node of the intelligent CNC workshop includes an equipment service layer, a core data layer, and a support service layer; the equipment service layer communicates with the intelligent CNC workshop through a communication protocol The terminal device establishes communication for data collection and command interaction; the core data layer includes the core data module, data management module, edge database and command control module. The core data module is used to display the data collected by the terminal device, and the data management module is used to distinguish sensitive type data and non-sensitive data. The edge database is used to store sensitive data. The command control module receives commands from the support service layer rule engine module and processing program module to control terminal devices; the support service layer includes application service modules and rule engines. Module, algorithm module, processing program module. The application service module is used for transmission and interaction of cloud edge data. The rule engine module is used to set rules. The algorithm module is used to receive the prediction model issued by the cloud application center for real-time prediction. The processing program module It is used to receive the processing program issued by the cloud application center to process the corresponding process of the workpiece.

The server-client file transfer method is used to realize the prediction model interaction between the intelligent CNC workshop cloud application center and the intelligent CNC workshop edge sensing node.

According to another aspect of the present invention, a control method for a cloud-edge collaborative intelligent CNC workshop self-regulation system is also provided. The method is applied to any one of the above described cloud-edge collaborative intelligent CNC workshop self-regulation systems. ,include:

The terminal equipment of the intelligent CNC workshop establishes communication with the edge sensing node of the intelligent CNC workshop through the communication protocol of the equipment service layer, and collects the data to the core data module for data display of the edge sensing node; the core data module transmits the data to the data management module for data processing Processing, the data is divided into sensitive data and non-sensitive data; sensitive data is stored in the edge database for further application of edge-aware nodes. At the same time, sensitive data is also passed into the rule engine module according to the preset rules. Make judgments. When the detected data exceeds the limited range, the rules are directly triggered for command control and action response of the CNC equipment. When the detected data meets the limited range, the prediction model is entered for real-time prediction of the processing data, and analysis and comparison are performed. , in the case of inconsistent comparison results between the predicted value and the real value, the predicted value shall be passed to the processing program module for program adjustment and then passed to the command control module. The command control module communicates with the terminal device through the communication protocol of the equipment service layer according to the processing program module. Carry out command control to achieve self-regulation of terminal equipment in the intelligent CNC workshop; for non-sensitive data, establish communication with the cloud application center of the intelligent CNC workshop through the application service module, and transfer the non-sensitive data to the cloud database for storage, and the cloud application center data The application module preprocesses the non-sensitive data in the cloud database and trains and corrects the prediction model. Then the edge application module of the cloud application center passes the modified and updated prediction model to the edge sensing node algorithm module for periodic updates of the prediction model. , through this cloud-edge collaborative interaction mechanism, the edge sensing node prediction model can be updated in real time.

The beneficial effects of the present invention are: the present invention proposes a cloud-edge collaborative intelligent CNC workshop self-regulation model based on edge computing technology. Based on the analysis of the CNC workshop operating mechanism and control characteristics, it builds an intelligent CNC workshop manufacturing unit edge sensing -Cloud collaborative operation-intelligent control framework, completed the deployment of two-level collaborative interaction scenarios between cloud and edge, and improved the computing performance of edge sensing nodes by designing the judgment logic of the rule engine; secondly, the long and short-term memory neural network was used to develop the cloud-edge processing model training and correction, and formulated cloud-edge collaborative production logic to complete the self-regulation of intelligent CNC workshop equipment; finally, through application cases, it was verified that the proposed cloud-edge collaboration framework has the characteristics of easy collaboration, easy regulation, and low delay. Chemical production provides technical support.

Description of the drawings

Figure 1 is a schematic diagram of the architecture of the cloud-edge collaborative intelligent CNC workshop self-regulation system of the present invention;

Figure 2 is a flow chart of the control method of the cloud-edge collaborative intelligent CNC workshop self-regulation system of the present invention;

Figure 3 is a self-regulation flow chart of optional application cases of the present invention;

Figure 4 is a schematic diagram of the control method architecture of an optional embodiment of the present invention.

Detailed ways

The invention will be further described below with reference to the accompanying drawings and examples, but the content of the invention is not limited to the described scope.

As shown in Figure 1, a cloud-edge collaborative intelligent CNC workshop self-regulation system includes an intelligent CNC workshop cloud application center, an intelligent CNC workshop edge sensing node, and an intelligent CNC workshop terminal device; the intelligent CNC workshop cloud application center is equipped with a cloud database , data application module, edge application module, can also include a visualization system. The cloud database is used to store non-sensitive data uploaded by edge sensing nodes; the data application module is used for data preprocessing and prediction model of non-sensitive data in the cloud database. Training correction; the edge application module is used to issue the prediction model and processing program of the edge sensing node; the visualization system is used for the visual display of the intelligent CNC workshop; the edge sensing node of the intelligent CNC workshop includes the equipment service layer, core data layer, and support service layer. The equipment service layer establishes communication with various communication protocols and intelligent CNC workshop terminal equipment for data collection and command interaction; the core data layer includes core data module, data management module, edge database and command control module. The core data module is used for terminal equipment. For the display of collected data, the data management module is used to distinguish sensitive data from non-sensitive data. The edge database is used to store sensitive data of edge sensing nodes. Some data that does not need to be uploaded to the cloud application center are directly stored to alleviate the problem of cloud Storage pressure, the command control module receives commands from the support service layer rule engine module and processing program module to control terminal equipment; the support service layer includes application service module, rule engine module, algorithm module, processing program module, and the application service module is used for cloud edge For data transmission and interaction, the rule engine module is used to set rules to ensure that CNC workshop equipment operates within a limited range. The algorithm module is used to receive the prediction model issued by the cloud application center and make real-time predictions. The processing program module is used to receive data from the cloud. Apply the processing procedures issued by the center to process the workpiece in the corresponding process.

Optionally, the intelligent CNC workshop terminal equipment includes CNC equipment, sensors and early warning devices. The CNC equipment includes CNC machine tools, detection devices, robotic arms, AGV trolleys, etc.

Optionally, the server-client file transfer method is used to realize the prediction model interaction between the intelligent CNC workshop cloud application center and the intelligent CNC workshop edge sensing node; specifically, in the embodiment of the present invention, the LSTM constructed by the present invention The prediction model is based on the Tensorflow architecture. The Tensorflow model mainly includes trained network parameter graphics and parameter values. After training a neural network, the model is saved in HDF5 file format for direct use in the future. Therefore, when building a prediction model, the built LSTM prediction network completes training and introduces the model.save (LSTM.h5) command to save it. Through this method, the trained model is saved as a whole. Later loading and application no longer need to define the network and compile the model. That is, the model configuration information such as the structure, weight, loss function, optimizer and status of the neural network has been saved. The saved model exists in the format of LSTM.h5, and the edge application module uses python to establish the cloud application center server and edge sensing node client to implement the delivery and reception of the LSTM prediction model. This is specifically implemented by the SSHClient command of the Paramiko module in python. .

As shown in Figure 2, a cloud-edge collaborative intelligent CNC workshop self-regulation system control method is applied to the cloud-edge collaborative intelligent CNC workshop self-regulation system described in any one of the above, including: intelligent CNC workshop The terminal device establishes communication with the edge sensing node of the intelligent CNC workshop through the communication protocol of the equipment service layer, and collects the data to the core data module for data display on the edge sensing node; the core data module transmits the data to the data management module for data processing. The data is divided into sensitive data and non-sensitive data (the data applied in the edge sensing node is sensitive data, and the data applied in the intelligent CNC workshop cloud application center is non-sensitive data); for sensitive data, it is stored in the edge database Storage is carried out for further application of edge sensing nodes without uploading to the cloud application center storage, which relieves the storage pressure of the cloud application center. At the same time, sensitive data is also passed into the rule engine module to make judgments based on preset rules. When the detected data exceeds If the range is limited, the rules will be directly triggered for command control to respond to the action of the CNC equipment. When the detected data meets the limited range, the prediction model will be entered for real-time prediction of the processing data, and analysis and comparison will be made. For the comparison results between the predicted value and the real value In the case of inconsistencies, the predicted values shall prevail and be passed to the processing program module for program adjustment and then passed to the command control module. The command control module performs command control based on the communication protocol of the equipment service layer and the terminal equipment according to the processing program module, thereby achieving an intelligent CNC workshop. Self-regulation of the terminal equipment; if consistent, the processing program will not be adjusted; it should be noted that the initial prediction model of the edge sensing node algorithm module is first trained and distributed by the cloud data application module for prediction, and the prediction model is later modified. Corrected and updated; for non-sensitive data, establish communication with the intelligent CNC workshop cloud application center through the application service module, transfer the non-sensitive data to the cloud database for storage, and the cloud application center data application module performs on the non-sensitive data in the cloud database. Preprocess and train and correct the prediction model, and then the edge application module of the cloud application center passes the revised and updated prediction model to the edge sensing node algorithm module for periodic updates of the prediction model. Through this cloud-edge collaborative interaction mechanism, edge sensing is achieved The node prediction model is updated in real time, so that the prediction effect of the edge-aware node prediction model always maintains good results.

Furthermore, an optional cloud-edge collaboration intelligent CNC workshop self-regulation system collaboration implementation process between processes is given as follows:

The intelligent CNC workshop includes 1 material warehouse (to store workpieces to be processed/unqualified products/qualified products), 2 workpiece cache areas to be processed (to store workpieces to be processed), and 3 CNC machine tools (CNC lathes, CNC milling machines, and CNC engraving machines) ), 1 AGV trolley (responsible for material storage and material transportation between workpieces to be processed), 1 robotic arm (responsible for picking up materials and clamping workpieces between three CNC equipment), and 1 image detection device. Each terminal device in the intelligent CNC workshop deploys its own edge sensing node, and uses RaspberryPi deployed with EdgeXFoundry and related software to establish communication with the device to build an edge sensing node (EdgeAwareNode, EAN). The self-regulation system between the processes of its intelligent CNC workshop is as follows:

The cloud application center (hereinafter referred to as the "cloud") sends instructions, processing programs or prediction models to the edge sensing nodes of each CNC workshop equipment. The material storage EAN receives the pickup instructions issued by the cloud, and the stacker takes out the workpieces to be processed and transports them to the workbench. Waiting for AGV to pick up the goods. AGVEAN receives the transportation path processing program issued by the cloud, and transports the workpieces to be processed in the material warehouse to the workbench to be processed and waits for the robot arm EAN to pick up the workpieces. At the same time, the AGV moves to the completion workbench to wait for the workpieces to be processed. The robotic arm EAN receives the clamping processing program issued by the cloud, and clamps the workpiece to be processed on the workbench to the CNC lathe EAN. The CNC lathe EAN receives the CNC turning processing program and prediction model issued by the cloud and starts processing the workpiece. The sensor in the CNC lathe equipment detects whether the workpiece is clamped. If not, wait for the mechanical arm EAN to complete the clamping. The CNC lathe EAN sends instructions to the third party. The chuck is clamped and processed according to the processing program. At the same time, the prediction model issued by the cloud predicts the cutting force and workpiece roughness. The prediction results are compared and adjusted to adjust the instructions in the processing program. The prediction model is corrected by the cloud. The model is updated in real time to maintain better prediction results and ensure better quality of processed workpieces. After the CNC lathe processing is completed, the mechanical arm unloads the workpiece and transports it to the next process CNC milling machine for milling. After the milling is completed, the mechanical arm clamps the workpiece to the CNC engraving machine for engraving. The process of milling and engraving is similar to the turning process of the CNC lathe. After the CNC engraving is completed, the robot arm clamps the workpiece to the image detection device for inspection. The image detection EAN receives the image inspection program issued by the cloud for analysis and comparison to determine whether the processed workpiece is qualified. If it is not qualified, it is judged whether it can be reprocessed. Workpieces that can be reprocessed are transported by the robotic arm to the corresponding CNC equipment for reprocessing. The inspected workpiece is transported to the completion workbench by the robotic arm, and transported back to the material storage by the waiting AGV car. The material storage EAN feeds back the inventory information to the cloud application center. Through the cloud, instructions and processing programs or prediction models are sent to the edge sensing nodes of each device. Each edge sensing node completes the processing of the workpiece and the collaboration between processes, achieving self-regulation of the intelligent CNC workshop. The specific process is shown in Figure 3.

Take the rule engine module of the robot arm EAN as an example. The rule engine module pre-sets the robot arm angle operating position limit value. When the information sensed by EAN exceeds the limit value set by the rule module, the robot arm will be activated directly through the EAN rule module. The shutdown command controls the robot arm to avoid obstacles and realize self-regulation of edge equipment. Taking robot arm axis 2 (Axis2Pos) as an example, the angle of Axis2Pos is within the range of 5°-35°. When the rule engine detects Axis2Pos <5°, the set rule 1 is triggered to perform the robot arm shutdown command; when the rule engine When the detected Axis2Pos>35° is triggered, the set rule 2 is triggered to stop the robot arm and realize self-regulation of the robot arm EAN for obstacle avoidance.

As shown in Figure 4, the CNC lathe embodiment is specifically described. The CNC lathe edge sensing node establishes communication with the CNC lathe through the equipment service layer, and collects CNC lathe equipment data, CNC lathe operation data, roughness data and cutting force data to the core. Data module, the core data module transfers data to the data management module. The data management module processes the data and distinguishes sensitive data and non-sensitive data. The sensitive data is further applied to the edge sensing nodes and is processed on the edge sensing nodes. The database is stored without uploading to the cloud application center for storage, which alleviates the storage pressure of the cloud application center. At the same time, sensitive data is also passed into the rule engine module to make judgments based on preset rules. Taking cutting force data as an example, set the cutting force Limited range. When the detected cutting force exceeds the set cutting range, the rule is triggered. The rule engine performs command regulation and controls the corresponding parameters of the CNC lathe to restore the cutting force to within the limited range. When the detected data meets the limited range, Input the data into the prediction model. The prediction model uses the LSTM neural network to predict the quality indicators of the sequential processing data of the CNC lathe. The input is the processing process parameters of the CNC lathe (such as spindle speed, feed speed, cutting depth, etc.). The output is the quality index (cutting force, roughness, etc.) for prediction, and the predicted value is analyzed and compared with the real value. If the comparison result between the predicted value and the real value is inconsistent, the predicted value shall be passed to the CNC lathe processing program for adjustment, and Passed to the command control module, the command control module controls the control parameters of the CNC lathe through the communication protocol of the equipment service layer according to the processing program, thereby achieving the prediction effect of the predicted value and realizing the self-regulation of the CNC lathe; if they are consistent, the processing program will not be performed. Adjustment; for non-sensitive data, establish communication with the cloud application center through the application service module, transfer the non-sensitive data to the cloud application center database for storage, and the cloud application center data application module preprocesses the CNC lathe data set in the database Pass the data into the prediction model for model training and correction, set the training and correction cycle of the prediction model according to the actual production and processing conditions, and then the cloud application center edge application module will pass the corrected and updated prediction model into the CNC lathe edge sensing node algorithm module for processing. Periodic updates of the prediction model (the time of model release can also be customized), through this cloud-edge collaborative interaction mechanism, the edge sensing node prediction model is updated in real time, so that the prediction effect of the edge sensing node prediction model always remains high. Good results.

Taking the CNC lathe EAN as an example, a comparison was made between the LSTM single prediction model and the cloud-edge interaction prediction model of the present invention. The single prediction model was trained with 2,000 pieces of historical data, and the cloud-edge interaction prediction model was modified and updated based on the data uploaded by EAN. It is sent to the CNC lathe EAN every 2 hours for real-time prediction. After 24 hours, the verification results are obtained in Table 1. The mean square error value and simulation effect of the prediction model of the single cloud center and the cloud edge interaction prediction model of the present invention are The combined values are shown in Table 1:

Table 1 Comparison of experimental errors in the two modes of LSTM prediction model

From the data in Table 1 above, we can see that the cloud-edge collaborative interaction prediction model has better prediction effects, higher fitting degree, and smaller mean square error. Through this periodic cloud-edge interaction mechanism, it ensures that the LSTM prediction model changes with real-time Continuous revisions are made as data is updated to avoid the reduction in prediction effect of the prediction model as time and data increase.

The specific embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the above-mentioned embodiments. Within the scope of knowledge possessed by those of ordinary skill in the art, other modifications can be made without departing from the spirit of the present invention. various changes.

Claims (3)

1. A cloud-edge collaborative intelligent CNC workshop self-regulation system, characterized by including an intelligent CNC workshop cloud application center, an intelligent CNC workshop edge sensing node, and an intelligent CNC workshop terminal device; The cloud application center of the intelligent CNC workshop is equipped with a cloud database, a data application module, and an edge application module. The cloud database is used to store non-sensitive data uploaded by edge sensing nodes; the data application module is used to store non-sensitive data in the cloud database. Preprocessing and training correction of prediction models; the edge application module is used to issue edge sensing node prediction models and processing programs; The edge sensing node of the intelligent CNC workshop includes an equipment service layer, a core data layer, and a support service layer; the equipment service layer establishes communication with the intelligent CNC workshop terminal equipment through a communication protocol for data collection and command interaction; the core data layer includes a core data module, Data management module, edge database and command control module. The core data module is used to display data collected by terminal devices. The data management module is used to distinguish sensitive data from non-sensitive data. The edge database is used to store sensitive data. The command control module receives commands from the support service layer rule engine module and processing program module to control terminal devices; the support service layer includes application service modules, rule engine modules, algorithm modules, and processing program modules. The application service module is used for cloud edge data transmission. Interaction, the rule engine module is used to set rules, the algorithm module is used to receive the prediction model issued by the cloud application center for real-time prediction, and the processing program module is used to receive the processing program issued by the cloud application center to process the corresponding process of the workpiece; in The data applied by edge sensing nodes is sensitive data, and the data applied in the intelligent CNC workshop cloud application center is non-sensitive data. 2. The cloud-edge collaborative intelligent CNC workshop self-regulation system according to claim 1, characterized in that the server-client file transfer method is used to realize the connection between the intelligent CNC workshop cloud application center and the intelligent CNC workshop edge sensing node. Predictive model interactions. 3. A control method for a cloud-edge collaborative intelligent CNC workshop self-regulation system, characterized in that the method is applied to the cloud-edge collaborative intelligent CNC workshop self-regulation system described in any one of claims 1-2, include: The terminal equipment of the intelligent CNC workshop establishes communication with the edge sensing node of the intelligent CNC workshop through the communication protocol of the equipment service layer, and collects data to the core data module for data display of the edge sensing node; The core data module transfers data to the data management module for data processing, and divides the data into sensitive data and non-sensitive data; Sensitive data is stored in the edge database for further application of edge sensing nodes. At the same time, sensitive data is also passed into the rule engine module to make judgments based on preset rules. When the detected data exceeds the limited range, the rules are directly triggered. Carry out command control to respond to the action of the CNC equipment. When the detected data meets the limited range, enter the prediction model to perform real-time prediction of the processing data, and conduct analysis and comparison. In the case of inconsistent comparison results between the predicted value and the real value, the predicted value is The quasi-incoming processing program module adjusts the program and then passes it to the command control module. The command control module performs command control with the terminal equipment through the communication protocol of the equipment service layer according to the processing program module, thereby achieving self-regulation of the terminal equipment in the intelligent CNC workshop; For non-sensitive data, communication is established with the intelligent CNC workshop cloud application center through the application service module, and the non-sensitive data is transferred to the cloud database for storage. The cloud application center data application module preprocesses the non-sensitive data in the cloud database and The prediction model is trained and corrected, and then the edge application module of the cloud application center passes the corrected and updated prediction model to the edge sensing node algorithm module for periodic updates of the prediction model. Through this cloud-edge collaborative interaction mechanism, the edge sensing node prediction model is realized. Live Update.