A method and system for online monitoring of multi-task machining processes based on cloud-edge collaboration

Abstract

The present invention discloses a method and system for online monitoring of a multi-task machining process based on cloud-edge collaboration, including a multi-source sensing terminal, an edge computing platform and a cloud server. The multi-source sensing terminal uses sensing equipment to obtain multi-source heterogeneous signals such as spindle vibration and acoustic emission generated during the machining process, and transmits them to the edge computing platform controller; the edge computing platform periodically collects and processes the signals, and combines the intelligent prediction model to synchronously monitor cutting vibration, tool wear, tool breakage and machining surface quality, and uploads the monitoring data to the cloud server; the cloud server stores the monitoring data of each edge computing platform in the cloud database, and uses global data to regularly update the intelligent prediction model to improve the model generalization ability. It has multi-state monitoring, cloud storage and model update functions, good real-time monitoring and high accuracy, breaking the information islands between different machine tools and equipment, and meeting the needs of intelligent machining monitoring.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent manufacturing and relates to an online monitoring method and system for a multi-task machining process based on cloud-edge collaboration. Background Art

[0002] With the continuous improvement of the comprehensive performance requirements of equipment in the fields of energy, power, aerospace, etc., complex thin-walled structural parts with light weight and high specific strength are widely used, such as aviation transparent structural parts, engine casings, integral blisks, gas wave cooling hubs, etc. However, due to the large size, thin wall thickness, weak rigidity, and the common use of difficult-to-process or heat-sensitive materials such as titanium alloy, high-strength steel, high-temperature alloy, and oriented organic glass, cutting chatter, tool failure, surface quality degradation and other problems are very likely to occur during the processing, resulting in difficulty in ensuring processing accuracy, a sharp drop in production efficiency, and even damage to the machine tool.

[0003] The online monitoring system of the machining process can sense the cutting vibration state, tool health state and machining surface quality in real time, and provide guidance for adaptive control of the cutting process, thereby achieving high-quality and efficient machining of complex thin-walled structural parts. Traditional machining monitoring systems predict the machining state in real time through local computing platforms. However, in the case of massive monitoring data, the storage capacity of such systems is limited, and it is difficult to share monitoring data and prediction models between different machine tools. In recent years, with the development of the Internet of Things, cloud computing and edge computing technologies, machining monitoring systems based on cloud-edge collaboration have been proposed one after another. For example, invention patent CN115034504A involves a tool wear state prediction system and method based on cloud-edge collaborative training, which monitors the tool wear state in real time through a prediction model deployed on the edge side, and updates and trains the prediction model in the cloud to improve the accuracy of tool wear state monitoring. To sum up, the existing cloud-edge collaborative processing monitoring systems usually have relatively simple functions and only monitor a certain cutting state. However, in actual production, the processing quality of the workpiece is affected by many factors such as the cutting vibration state and the tool health state. There is currently no processing monitoring system with multi-state monitoring capabilities; in addition, the above-mentioned system only uses the cloud server to update and train the processing prediction model, but fails to give full play to its powerful cloud storage function. It has not yet formed a processing monitoring big data platform, and it is difficult to meet the needs of intelligent processing monitoring. Summary of the invention

[0004] In view of the shortcomings of the existing machining process monitoring system, the present invention provides a multi-task machining process online monitoring method and system based on cloud-edge collaboration to achieve synchronous monitoring of multiple machining states and information sharing between different machine tool equipment.

[0005] The technical solution of the present invention:

[0006] An online monitoring system for multi-task machining processes based on cloud-edge collaboration includes a multi-source sensing terminal, an edge computing platform, and a cloud server, wherein:

[0007] The multi-source sensing terminal is deployed on the machine tool equipment to obtain the multi-source heterogeneous signals generated during the processing and transmit them to the edge computing platform;

[0008] The edge computing platform is built on the basis of CompactRIO embedded controller and host computer, and is used for periodic collection and processing of multi-source heterogeneous signals, real-time monitoring of the cutting process in combination with intelligent prediction models, and uploading the monitoring data to the cloud server;

[0009] The cloud server is used to store the monitoring data sent by the edge computing platform, as well as update and send the intelligent prediction model.

[0010] The multi-source sensing terminal includes an acceleration sensor, a sound pressure sensor, an acoustic emission sensor, a preamplifier, an acoustic vibration input module and a voltage input module, and the acquired multi-source heterogeneous signals include a spindle vibration signal, a sound signal and an acoustic emission signal.

[0011] The edge computing platform includes a signal acquisition module, a signal preprocessing module, a feature extraction module, a feature fusion module, a multi-task monitoring module, a data storage module, a human-computer interaction module and a cloud upload module; the signal acquisition module is used to collect the spindle vibration signal, sound signal and acoustic emission signal obtained by the multi-source sensing terminal according to a preset sampling frequency, and transmit the above signals to the signal preprocessing module; the signal preprocessing module is used to perform low-pass filtering on the collected spindle vibration signal and sound signal, and band-pass filtering on the collected acoustic emission signal; the feature extraction module is used to extract time domain, frequency domain and time-frequency domain features from the filtered multi-source heterogeneous signals; the feature fusion module is used to reduce the dimensionality of the extracted time domain, frequency domain and time-frequency domain features; the multi-task monitoring module includes a vibration Monitoring submodule, tool wear monitoring submodule, tool breakage monitoring submodule and processing quality monitoring submodule. The above submodules monitor cutting vibration, tool wear, tool breakage and processing surface roughness according to their respective intelligent prediction models; the data storage module is used to store the preprocessed multi-source heterogeneous signals, extracted signal features and monitoring results to the host computer, which is convenient for the playback of historical monitoring data and tracing of processing anomalies; the human-computer interaction module includes a parameter setting interface and a monitoring visualization interface. The parameter setting interface is used for the configuration of various monitoring parameters and information with the cloud server, and the monitoring visualization interface is used for the real-time display of signals, features and monitoring results; the cloud upload module is used to upload the feature data and monitoring results stored in the host computer to the cloud server through the MQTT protocol.

[0012] The signal acquisition module is deployed in the FPGA chassis of the CompactRIO embedded controller, the signal preprocessing module and the feature extraction module are deployed in the RT controller of the CompactRIO embedded controller; the feature fusion module, the multi-task monitoring module, the data storage module, the human-computer interaction module and the cloud upload module are deployed in the host computer; the FPGA chassis and the RT controller perform data transmission through the DMA FIFO mode, and the CompactRIO embedded controller and the host computer perform data transmission through the network stream protocol.

[0013] The CompactRIO embedded controller and host computer are both developed based on the LabVIEW software platform, and the producer / consumer model is used for LabVIEW program design to improve the program operation efficiency and ensure the real-time performance of the processing monitoring system; the signal acquisition module is the producer, and the signal preprocessing module, feature extraction module, feature fusion module, multi-task monitoring module, data storage module, human-computer interaction module and cloud upload module are all consumers.

[0014] The intelligent prediction models of the vibration monitoring submodule, tool wear monitoring submodule, tool breakage monitoring submodule and processing quality monitoring submodule adopt an online sequence extreme learning machine or a support vector machine. The model inputs are all features after dimensionality reduction by a feature fusion module, and the outputs are vibration status, tool wear status, tool breakage status and processing surface roughness respectively; each intelligent prediction model is implemented in the MATLAB Script node of LabVIEW.

[0015] The cloud server includes a cloud database module and a cloud model update module; the cloud database module includes a knowledge database and a feature database, the knowledge database is used to store the intelligent prediction models of each sub-module of the multi-task monitoring module, and the feature database is used to store the feature data and monitoring results sent by each edge computing platform; the cloud model update module is used to update and train the intelligent prediction models of each sub-module of the multi-task monitoring module after the global data of the feature database reaches a set threshold, and store the updated model parameters in the knowledge database of the cloud database module, and send them to each edge computing platform at the same time.

[0016] A multi-task machining process online monitoring method based on cloud-edge collaboration includes the following steps:

[0017] S1. Program deployment and settings: deploy the LabVIEW program of the edge computing platform, set the sampling frequency, data storage path, cloud server information, etc. in the human-computer interaction module;

[0018] S2. Multi-source heterogeneous signal acquisition; the signal acquisition module acquires the spindle vibration signal, sound signal and acoustic emission signal acquired by the multi-source sensing terminal according to the set sampling frequency, and transmits the acquired signal to the signal preprocessing module;

[0019] S3. Signal filtering and multi-domain feature extraction: The signal preprocessing module filters the collected three-axis spindle vibration signal, sound signal and acoustic emission signal, and the filtered data is input into the feature extraction module to extract time domain features, frequency domain features and time-frequency domain features. On this basis, the filtered multi-source heterogeneous signals and the extracted multi-domain signal features are transmitted to the host computer through the network flow protocol;

[0020] S4. Feature fusion and multi-task processing monitoring; the host computer inputs the read feature data into the feature fusion module for dimensionality reduction, and then inputs the obtained low-dimensional features into the multi-task monitoring module to predict the cutting vibration state, tool wear state, tool damage state and processing surface roughness;

[0021] S5. Data visualization and local storage: The filtered signals, extracted features and monitoring results are transmitted to the human-computer interaction module and the data storage module for real-time display and stored in a text file under the specified path;

[0022] S6. Repeat steps S2 to S5 until the processing monitoring process is completed;

[0023] S7. Data upload to the cloud; after the processing monitoring is completed, the cloud upload module reads the text file used to store feature data and monitoring results in the host computer, sends the read data to the cloud server through the MQTT protocol, and deletes the corresponding text file after the data is successfully uploaded to free up the disk space of the host computer;

[0024] S8. Cloud database storage: The cloud server will receive the feature and prediction data sent by each edge computing platform and store it in the specified feature database;

[0025] S9. Model cloud update and distribution; the cloud model update module regularly accesses the feature database. When the global data increment reaches the set threshold, the intelligent prediction models of the vibration monitoring submodule, tool wear monitoring submodule, tool breakage monitoring submodule and processing quality monitoring module are trained respectively, and the updated model parameters are stored in the knowledge database, and distributed to the host computer of each edge computing platform through the MQTT protocol.

[0026] Beneficial effects of the present invention: The present invention is oriented to the field of intelligent manufacturing technology, and has constructed a set of online monitoring systems for multi-task machining processes, and based on this, has proposed a multi-task machining process monitoring method based on cloud-edge collaboration. The present invention utilizes an edge computing platform to achieve synchronous monitoring of cutting vibration, tool wear, tool breakage, and machining surface quality, and has high computing efficiency and good real-time monitoring. At the same time, the cloud server constructed can realize efficient storage of massive monitoring data and iterative updates of intelligent prediction models, eliminating information islands between different machine tools and equipment, and further improving the accuracy of machining monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 It is the architecture diagram of the multi-task processing monitoring system of the present invention.

[0028] Figure 2 This is a schematic diagram of the hardware equipment installation and connection of the multi-task processing monitoring system.

[0029] Figure 3 Schematic diagram of data collection and processing of edge controller.

[0030] Figure 4 This is a flow chart of a multi-task machining process monitoring method based on cloud-edge collaboration.

[0031] In the figure: 1 machine tool equipment; 11 machine tool spindle; 12 machine tool worktable; 2 multi-source sensing terminal; 21 three-axis acceleration sensor; 22 acoustic emission sensor; 23 sound pressure sensor; 24 preamplifier; 25 acoustic vibration input module; 26 voltage input module; 3 edge computing platform; 31 CompactRIO embedded controller; 32 host computer; 4 cloud server; 5 workpiece fixture; 6 magnetic suction fixture; 7 magnetic table stand; 8 DC power supply. DETAILED DESCRIPTION

[0032] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution in the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is only a part of the embodiment of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. Example

[0033] like Figure 1As shown, a multi-task machining process online monitoring system based on cloud-edge collaboration includes a multi-source sensing terminal 2, an edge computing platform 3 and a cloud server 4. The multi-source sensing terminal is deployed on the machine tool equipment 1, including a three-axis acceleration sensor, a sound pressure sensor, an acoustic emission sensor, a preamplifier, an acoustic vibration input module and a voltage input module, which are used to obtain the three-way spindle vibration signal, sound signal and acoustic emission signal during the machining process; the edge computing platform is built based on the CompactRIO embedded controller and the host computer, including a signal acquisition module, a signal preprocessing module, a feature extraction module, a feature fusion module, a multi-task monitoring module, a data storage module, a human-computer interaction module and a cloud upload module, which are used for the periodic acquisition and processing of multi-source heterogeneous signals and the real-time monitoring of the cutting process in combination with the intelligent prediction model, and uploading the monitoring data to the cloud server; the cloud server includes a cloud database module and a cloud model update module. The cloud database module includes a knowledge database and a feature database, which are used to store the monitoring data sent by the edge computing platform, as well as the update and distribution of the intelligent prediction model.

[0034] In this embodiment, refer to Figure 2 The hardware equipment of the system is installed and connected. To avoid interference with the processing process, the three-axis acceleration sensor 21 is installed on the spindle 11 of the machine tool 1 by glue, and is used to obtain the vibration signal of the spindle in the three directions of X, Y, and Z during the processing; the sound pressure sensor 23 is adsorbed on the workbench 12 of the machine tool 1 by the magnetic table seat 7, and is used to obtain the sound signal during the processing; the acoustic emission sensor 22 is installed on the workpiece fixture 5 by the magnetic fixture 6, and is used to obtain the acoustic emission signal generated during the processing; the three-axis acceleration sensor 21 and the sound pressure sensor 23 are connected to the acoustic vibration input module 25 through the signal cable, and the acoustic emission sensor 22 is connected to the input end of the preamplifier 24, and the output end of the preamplifier 24 is connected to the voltage input module 26. The CompactRIO embedded controller 31 is powered by a DC power supply 8, and the acoustic vibration input module 25 and the voltage input module 26 are directly inserted into the card slot of the CompactRIO embedded controller 31, and the CompactRIO embedded controller 31 is connected to the host computer 32 through a network cable.

[0035] The CompactRIO embedded controller is equipped with an FPGA chassis and an RT controller. The CompactRIO embedded controller and the host computer transmit data through a network stream protocol, and the FPGA chassis and the RT controller transmit data through a DMA FIFO mode. The signal acquisition module is deployed in the FPGA chassis, the signal preprocessing module and the feature extraction module are deployed in the RT controller, and the signal fusion module, the multi-task monitoring module, the data storage module, the human-computer interaction module and the cloud upload module are deployed in the host computer. The CompactRIO embedded controller and the host computer are both developed based on the LabVIEW software platform, and the producer / consumer mode is used for LabVIEW program design to improve the program running efficiency and ensure the real-time performance of the processing monitoring system. The signal acquisition module is a producer, and the signal preprocessing module, the feature extraction module, the feature fusion module, the multi-task monitoring module, the data storage module, the human-computer interaction module and the cloud upload module are all consumers.

[0036] like Figure 3 As shown in the figure, the signal acquisition module deployed in the FPGA chassis collects the three-way spindle vibration and sound signals obtained by the acoustic vibration input module according to the set sampling frequency, and writes them into the first DMA FIFO buffer area in a loop, and collects the acoustic emission signal obtained by the voltage input module, and writes it into the second DMA FIFO buffer area in a loop; the signal preprocessing module deployed in the RT controller cyclically reads the data in the first DMA FIFO buffer area and the second DMA FIFO buffer area, and uses a 4th-order Butterworth low-pass filter with a cutoff frequency of 3kHz to filter the collected three-way spindle vibration signal and sound signal, and uses a 4th-order Butterworth bandpass filter with a passband range of 50kHz to 300kHz to filter the collected acoustic emission signal, and only retains the acoustic emission signal components related to the cutting state; the feature extraction module extracts time domain features including mean, standard deviation, root mean square, peak-to-peak value, skewness, and kurtosis from the filtered vibration signal, sound signal, and acoustic emission signal; the extracted frequency domain features include: power spectrum mean, center of gravity frequency, and frequency standard deviation; the extracted time-frequency domain features include: wavelet energy ratio and wavelet energy entropy.

[0037] The feature fusion module uses the kernel principal component analysis algorithm to reduce the dimension of the received signal feature data and selects the principal component with a cumulative contribution rate greater than 80% to enhance the feature representation capability and improve the accuracy and efficiency of processing state recognition.

[0038] The multi-task monitoring module includes a vibration monitoring submodule, a tool wear monitoring submodule, a tool breakage monitoring submodule and a processing quality monitoring submodule. The above submodules monitor cutting vibration, tool wear, tool breakage and processing surface roughness according to their respective intelligent prediction models; in this embodiment, the intelligent prediction models of the vibration monitoring submodule, the tool wear monitoring submodule, the tool breakage monitoring submodule and the processing quality monitoring submodule all adopt support vector machines, and the initial model parameters are obtained by training the MATLABLibsvm toolbox. Each intelligent prediction model is implemented in the MATLAB Script node of LabVIEW, and the model inputs are all features after dimensionality reduction by the feature fusion module, and the outputs are respectively vibration state (stable state, early chatter state, severe chatter state), tool wear state (initial wear, normal wear, severe wear), tool breakage state (normal state, broken state) and processing surface roughness value.

[0039] The data storage module stores data with consistent timestamps into the same text file. The file is named "data type" + "date". The data to be stored and its timestamp are first converted into strings, and then the above strings are spliced ​​with spaces as delimiters. Finally, the spliced ​​strings are stored in a text file under the specified path. In this embodiment, the three-axis vibration signal and the sound signal are stored in the same file, the signal characteristics and the monitoring results are stored in the same file, and the acoustic emission signal is stored separately.

[0040] The human-computer interaction module includes a parameter setting interface and a monitoring visualization interface. The parameter setting interface is used for configuring various monitoring parameters and cloud server information, and the monitoring visualization interface is used for real-time display of multi-source heterogeneous signals, features and monitoring results; in this embodiment, the parameter setting interface and the monitoring visualization interface are freely switched through the tabs at the top. The parameter setting interface is constructed using input controls, Boolean buttons and Boolean indicator lights. The input controls include acoustic vibration signal sampling frequency input controls, acoustic emission signal sampling frequency input controls, data storage path input controls, cloud server IP address, port number, account number, and password input controls. The Boolean button is used for cloud server connection, and the Boolean indicator light is used to display the cloud server connection status; the processing visualization interface is provided with a signal display unit, a feature display unit, and a monitoring result display unit side by side. The signal display unit and the feature display unit are constructed using a waveform chart control, which is used to display the collected three-way vibration signals, sound signals and acoustic emission signals and their corresponding time domain features, frequency domain features, and time-frequency domain features in real time. The monitoring result display unit uses a Boolean indicator light to indicate the cutting status, and the waveform chart control displays the processing roughness prediction value. The state label corresponding to the lit Boolean indicator light is the predicted cutting state.

[0041] The cloud upload module reads the text file used to store feature data for monitoring results, converts the read string into a json string, and publishes the json string to the cloud server through the constructed MQTT server.

[0042] The cloud server is built on the Alibaba Cloud platform. The cloud database module is built using the MySQL database, including a knowledge database and a feature database. The cloud server subscribes to the data published by the host computer of each edge computing platform, and uses SQL insert statements to store feature data and monitoring results in the specified table of the feature database. The table is named "machine number" + "date". SQL statements are written to create new data tables so that daily machine tool monitoring data is stored in the specified table; the knowledge database stores the model parameters of the support vector machine of each sub-module of the multi-task monitoring module in the specified table, which is named "sub-module name"; the cloud model update module regularly accesses the feature database to integrate the feature data of each machine tool equipment. When the global data increment exceeds the set threshold, the model update function is activated to train the support vector machine model of each sub-module, and the updated model parameters are stored in the corresponding table of the knowledge database and sent to the host computer of each edge computing platform through the MQTT protocol. Example

[0043] A multi-task machining process online monitoring method based on cloud-edge collaboration, using the multi-task machining process online monitoring system based on cloud-edge collaboration in Example 1, and the specific implementation steps are as follows:

[0044] S1. Program deployment and settings: deploy the LabVIEW program of the edge computing platform, set the sampling frequency, data storage path, cloud server information, etc. in the human-computer interaction module;

[0045] In this embodiment, the sampling frequency of the acoustic vibration signal is set to 10240 Hz, the sampling frequency of the acoustic emission signal is set to 1 MHz, the data storage path is named "Date" + "Data Type", and the cloud server IP address, port number, account, and password are set according to the built Alibaba Cloud server. Click the cloud server connection button. When the Boolean indicator light is on, it means the connection is successful. Otherwise, check whether the cloud server information settings are correct, modify and reconnect until the connection is successful.

[0046] S2. Multi-source heterogeneous signal acquisition; the signal acquisition module acquires the spindle vibration signal, sound signal and acoustic emission signal acquired by the multi-source sensing terminal according to the set sampling frequency, and transmits the acquired signal to the signal preprocessing module;

[0047] In this embodiment, the signal acquisition module collects three-way spindle vibration signals, sound signals and acoustic emission signals at a set frequency, and cyclically writes the three-way spindle vibration signals and sound signals into the first DMA FIFO buffer area, while cyclically writing the acoustic emission signals into the second DMA FIFO buffer area, and the time interval between two adjacent writes is 50ms.

[0048] S3. Signal filtering and multi-domain feature extraction; the signal preprocessing module filters the collected three-way main shaft vibration signal, sound signal and acoustic emission signal, and the filtered data is input into the feature extraction module to extract time domain features, frequency domain features and time-frequency domain features, and on this basis, the filtered multi-source heterogeneous signal and the extracted multi-domain signal features are transmitted to the host computer through the network flow protocol; in this embodiment, the signal preprocessing module inputs the data read from the first DMAFIFO buffer area into the designed 4th order Butterworth low-pass filter with a cutoff frequency of 3kHz, and at the same time inputs the data read from the second DMA FIFO buffer area into the designed 4th order Butterworth bandpass filter with a passband range of 50kHz to 300kHz, and the above-mentioned filtered multi-source heterogeneous signal is input into the feature extraction module to extract the mean, standard deviation, root mean square, peak-to-peak value, skewness, kurtosis, power spectrum mean, center of gravity frequency, frequency standard deviation, wavelet energy ratio and wavelet energy entropy, and finally, the CompactRIO embedded controller sends the filtered signal and the extracted feature data to the host computer through the network flow protocol.

[0049] S4. Feature fusion and multi-task processing monitoring; the host computer inputs the read feature data into the feature fusion module for dimensionality reduction, and then inputs the obtained low-dimensional features into the multi-task monitoring module to predict the cutting vibration state, tool wear state, tool damage state and processing surface roughness;

[0050] In this embodiment, the host computer monitors the signals and feature data sent by CompactRIO in real time, writes the filtered signal data into two queues, and transmits them to the human-computer interaction module and the data storage module respectively; writes the feature data into three queues, two of which are transmitted to the human-computer interaction module and the data storage module respectively, and the other is input into the feature fusion module, and a 4-dimensional feature vector is obtained after dimensionality reduction by kernel principal component analysis, which is input into the MATLABScript node of each submodule of the multi-task monitoring module, and the current processing state and surface roughness are predicted by using the pre-trained support vector machine model, and the prediction results are written into two queues and input into the human-computer interaction module and the data storage module respectively;

[0051] S5. Data visualization and local storage: The filtered signals, extracted features and monitoring results are transmitted to the human-computer interaction module and the data storage module for real-time display and stored in a text file under the specified path;

[0052] In this embodiment, the human-computer interaction module reads the signals, features and predicted data in the queue, and displays them in real time on the monitoring visualization interface with the help of a waveform icon control; taking the acoustic vibration signal as an example, the specific storage process is as follows: first, the timestamp, three-way spindle vibration signal, and sound signal are converted into strings, and then they are concatenated into a group of strings with spaces as delimiters, and then stored in a specified text file. The first column of the text file is the timestamp, and the second to fourth columns correspond to the spindle vibration signals in the X, Y, and Z directions, respectively, and the fifth column is the sound signal.

[0053] S6. Repeat steps S2 to S5 until the processing monitoring process is completed;

[0054] S7. Data upload to the cloud; after the processing monitoring is completed, the cloud upload module reads the text file used to store feature data and monitoring results in the host record, sends the read data to the cloud server through the MQTT protocol, and deletes the corresponding text file after the data is successfully uploaded to free up the host computer disk space;

[0055] In this embodiment, the cloud upload module calls the text reading function to read the feature and prediction data text file, converts the read string line by line into a json string, and publishes the json string to the cloud server through the constructed MQTT server.

[0056] S8. Cloud database storage: The cloud server will receive the feature and prediction data sent by each edge computing platform and store it in the specified feature database;

[0057] In this embodiment, the cloud server uses python scripts to implement MQTT subscription, reception and MySQL database storage. Whenever a new message is published to the subscribed MQTT topic, the subscribed data is stored in the specified table of the feature database through SQL insert statements.

[0058] S9. Model cloud update and distribution; the cloud model update module regularly accesses the feature database. When the global data increment reaches the set threshold, the intelligent prediction models of the vibration monitoring submodule, tool wear monitoring submodule, tool breakage monitoring submodule and processing quality monitoring module are trained respectively, and the updated model parameters are stored in the knowledge database, and distributed to the host computer of each edge computing platform through the MQTT protocol;

[0059] In this embodiment, the cloud server uses a python script to implement the timed reading of the feature database, the update of the support vector machine model of each monitoring sub-module, the storage of model parameters, and MQTT publishing; the feature database is read once every 12 hours, the feature data of each machine tool equipment is integrated, and it is determined whether the global data increment exceeds the set threshold. When the threshold is exceeded, the model update function is activated, and the updated parameters are stored in the corresponding table of the knowledge database through SQL replacement statements. At the same time, the parameters are converted into json format and published to the host computer of each edge device through the MQTT protocol.

Claims

1. A multi-task processing online monitoring system based on cloud-edge collaboration, characterized in that: The multi-task processing online monitoring system includes a multi-source sensing terminal, an edge computing platform and a cloud server, among which: The multi-source sensing terminal is deployed on the machine tool equipment to obtain the multi-source heterogeneous signals generated during the processing and transmit them to the edge computing platform; The edge computing platform is built on the basis of CompactRIO embedded controller and host computer, and is used for periodic collection and processing of multi-source heterogeneous signals, real-time monitoring of the cutting process in combination with intelligent prediction models, and uploading the monitoring data to the cloud server; The cloud server is used to store the monitoring data sent by the edge computing platform, as well as update and issue the intelligent prediction model; The edge computing platform includes a signal acquisition module, a signal preprocessing module, a feature extraction module, a feature fusion module, a multi-task monitoring module, a data storage module, a human-computer interaction module and a cloud upload module; the signal acquisition module is used to collect the spindle vibration signal, sound signal and acoustic emission signal obtained by the multi-source sensing terminal according to a preset sampling frequency, and transmit the above signals to the signal preprocessing module; the signal preprocessing module is used to perform low-pass filtering on the collected spindle vibration signal and sound signal, and band-pass filtering on the collected acoustic emission signal; the feature extraction module is used to extract time domain, frequency domain and time-frequency domain features from the filtered multi-source heterogeneous signals; the feature fusion module is used to reduce the dimensionality of the extracted time domain, frequency domain and time-frequency domain features; the multi-task monitoring module includes a vibration monitoring submodule, a tool wear monitoring submodule, and a tool wear monitoring submodule. The measuring module, the tool breakage monitoring submodule and the processing quality monitoring submodule respectively monitor cutting vibration, tool wear, tool breakage and processing surface roughness according to their respective intelligent prediction models; the data storage module is used to store the multi-source heterogeneous signals preprocessed by the signal preprocessing module, the signal features extracted by the feature extraction module and the monitoring results of the multi-task monitoring module to the host computer, so as to facilitate the playback of historical monitoring data and the tracing of processing anomalies; the human-computer interaction module includes a parameter setting interface and a monitoring visualization interface. The parameter setting interface is used for the configuration of various monitoring parameters and information with the cloud server, and the monitoring visualization interface is used for the real-time display of signals, features and monitoring results; the cloud upload module is used to upload the feature data and monitoring results stored in the host computer to the cloud server through the MQTT protocol.

2. The multi-task processing online monitoring system according to claim 1 is characterized in that: The multi-source sensing terminal includes an acceleration sensor, a sound pressure sensor, an acoustic emission sensor, a preamplifier, an acoustic vibration input module and a voltage input module, and the acquired multi-source heterogeneous signals include a spindle vibration signal, a sound signal and an acoustic emission signal.

3. The multi-task processing online monitoring system according to claim 1 is characterized in that: The signal acquisition module is deployed in the FPGA chassis of the CompactRIO embedded controller, and the signal preprocessing module and the feature extraction module are deployed in the RT controller of the CompactRIO embedded controller; The feature fusion module, multi-task monitoring module, data storage module, human-computer interaction module and cloud upload module are deployed in the host computer; the FPGA chassis and the RT controller transmit data through DMA FIFO, and the CompactRIO embedded controller and the host computer transmit data through the network stream protocol.

4. The multi-task processing online monitoring system according to claim 3 is characterized in that: The CompactRIO embedded controller and host computer are both developed based on the LabVIEW software platform, and the producer / consumer model is used for LabVIEW program design to improve the program operation efficiency and ensure the real-time performance of the processing monitoring system; the signal acquisition module is the producer, and the signal preprocessing module, feature extraction module, feature fusion module, multi-task monitoring module, data storage module, human-computer interaction module and cloud upload module are all consumers.

5. The multi-task processing online monitoring system according to claim 1 is characterized in that: The intelligent prediction models of the vibration monitoring submodule, tool wear monitoring submodule, tool breakage monitoring submodule and processing quality monitoring submodule adopt an online sequence extreme learning machine or a support vector machine. The model inputs are all features after dimensionality reduction by a feature fusion module, and the outputs are vibration status, tool wear status, tool breakage status and processing surface roughness respectively; each intelligent prediction model is implemented in the MATLAB Script node of LabVIEW.

6. The multi-task processing online monitoring system according to claim 1 is characterized in that: The cloud server includes a cloud database module and a cloud model update module; the cloud database module includes a knowledge database and a feature database, the knowledge database is used to store the intelligent prediction models of each sub-module of the multi-task monitoring module, and the feature database is used to store the feature data and monitoring results sent by each edge computing platform; the cloud model update module is used to update and train the intelligent prediction models of each sub-module of the multi-task monitoring module after the global data of the feature database reaches a set threshold, and store the updated model parameters in the knowledge database of the cloud database module, and send them to each edge computing platform at the same time.

7. A multi-task processing online monitoring method based on the cloud-edge collaboration of the multi-task processing online monitoring system of claim 6, characterized in that: The steps include: S1. Program deployment and settings: deploy the LabVIEW program of the edge computing platform, set the sampling frequency, data storage path, and cloud server information in the human-computer interaction module; S2. Multi-source heterogeneous signal acquisition; the signal acquisition module acquires the spindle vibration signal, sound signal and acoustic emission signal acquired by the multi-source sensing terminal according to the set sampling frequency, and transmits the acquired signal to the signal preprocessing module; S3. Signal filtering and multi-domain feature extraction: The signal preprocessing module filters the collected spindle vibration signal, sound signal and acoustic emission signal, and the filtered data is input into the feature extraction module to extract time domain features, frequency domain features and time-frequency domain features, and the filtered multi-source heterogeneous signals and the extracted multi-domain signal features are transmitted to the host computer through the network flow protocol; S4. Feature fusion and multi-task processing monitoring; the host computer inputs the read feature data into the feature fusion module for dimensionality reduction, and then inputs the obtained low-dimensional features into the multi-task monitoring module to predict the cutting vibration state, tool wear state, tool damage state and processing surface roughness; S5. Data visualization and local storage: The filtered signals, extracted features and monitoring results are transmitted to the human-computer interaction module and the data storage module for real-time display and stored in a text file under the specified path; S6. Repeat steps S2 to S5 until the processing monitoring process is completed; S7. Data cloud upload; After the processing monitoring is completed, the cloud upload module reads the text file used to store feature data and monitoring results in the host computer, sends the read data to the cloud server through the MQTT protocol, and deletes the corresponding text file after the data is successfully uploaded to free up the disk space of the host computer; S8. Cloud database storage: The cloud server will receive the feature and prediction data sent by each edge computing platform and store it in the specified feature database; S9. Model cloud update and distribution; the cloud model update module regularly accesses the feature database. When the global data increment reaches the set threshold, the intelligent prediction models of the vibration monitoring submodule, tool wear monitoring submodule, tool breakage monitoring submodule and processing quality monitoring module are trained respectively, and the updated model parameters are stored in the knowledge database, and distributed to the host computer of each edge computing platform through the MQTT protocol.

Images