Multi-source heterogeneous data analysis method and system for CNC machine tools based on edge-cloud collaboration

Through the multi-source heterogeneous data analysis method of CNC machine tools with edge cloud collaboration, edge nodes are used for lightweight processing and in-depth analysis in the cloud, and the model is dynamically updated, solving the problem of low data analysis efficiency and accuracy of high-end CNC machine tools, achieving high efficiency, Accurate state analysis.

CN118503632BActive Publication Date: 2024-11-12JILIN UNIVERSITY
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410946368.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-07-16
Publication Date
2024-11-12
Estimated Expiration
2044-07-16

AI Technical Summary

Technical Problem

In the prior art, multi-source heterogeneous data analysis of high-end CNC machine tools has problems such as insufficient processing performance of edge computing nodes and large communication delays of cloud computing nodes, resulting in low data analysis efficiency and accuracy.

Method used

Using multi-source heterogeneous data analysis method of CNC machine tools based on edge cloud collaboration, pre-process and classify data through lightweight processing nodes in edge data, generate high-quality data sets, and use cloud data enhancement processing nodes to perform in-depth analysis and dynamic updates Data analysis model to ensure analysis accuracy.

Benefits of technology

It effectively avoids network channel blockage and data processing delays, improves data analysis efficiency and accuracy, reduces redundant data upload and analysis time, and improves the operation efficiency and state analysis accuracy of artificial intelligence algorithms.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118503632B_ABST
    Figure CN118503632B_ABST
Patent Text Reader

Abstract

The present invention is applicable to the technical field of CNC machine tool data processing, and proposes a method and system for analyzing multi-source heterogeneous data of CNC machine tools based on edge-cloud collaboration. The analysis method adopts an edge-cloud collaborative architecture, and the analysis system includes a data acquisition source, an edge data lightweight processing node, and a cloud data enhancement processing node. The data acquisition source is responsible for building a hardware system for multi-source data acquisition; the edge data lightweight processing node matches the communication protocol of the CNC machine tool multi-source heterogeneous data and collects the CNC machine tool multi-source heterogeneous data, and pre-processes the processing process data, machine tool status data, component audio data, and workpiece image data; the cloud data enhancement processing node evaluates the performance status of the data analysis model and makes a data set increment judgment based on the performance status, and dynamically updates the data analysis model. The present invention realizes high-efficiency, high-precision, and intelligent analysis of the machine tool status through the coordinated cooperation of the edge data lightweight processing node and the cloud data enhancement processing node.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of CNC machine tool data processing, and in particular to a CNC machine tool multi-source heterogeneous data analysis method and system based on edge-cloud collaboration. Background Art

[0002] At present, edge computing-based analysis methods and cloud computing-based analysis methods are mostly used for multi-source heterogeneous data analysis of high-end CNC machine tools.

[0003] Among them, the edge computing-based analysis method deploys computing nodes with general processing performance near the machine tool, connects to the machine tool through Ethernet communication, and deploys data preprocessing algorithms and data analysis algorithms in the edge computing nodes. However, high-end CNC machine tools have multi-level components and multi-level systems, and the data in their service process is large and the status analysis is particularly complex. Taking the five-axis drilling center equipped with the Huazhong CNC system as an example, there are nearly 350,000 key data items of the machine tool, and the status analysis needs to consider the influence of factors such as multi-source coupling of components and time-varying working conditions. Its algorithm is usually a high-complexity intensive computing solution code. Although the edge computing node can quickly obtain status data through Ethernet, saving a lot of data acquisition time, due to its general processing performance, it cannot quickly run high-complexity intensive computing data algorithms, which makes the edge processing node overloaded and unbalanced, thereby increasing the probability of edge node downtime and affecting data analysis efficiency and accuracy. The data analysis method based on cloud computing can solve the problem of data analysis efficiency and accuracy by relying on cloud computing nodes with rich and powerful computing performance. However, cloud computing nodes are usually far away from the industrial site and collect data through long-distance communication methods such as wireless. This means that the communication channel needs to transmit multi-source heterogeneous noise data of high-end CNC machine tools all the time, which is easy to cause channel congestion and uneven allocation of bandwidth resources, which leads to extremely long data transmission and data acquisition time, and data loss is prone to occur, which also affects the efficiency and accuracy of data analysis. For this reason, a multi-source heterogeneous data analysis method and system for CNC machine tools based on edge-cloud collaboration is proposed. Summary of the invention

[0004] The purpose of the present invention is to provide a multi-source heterogeneous data analysis method and system for CNC machine tools based on edge-cloud collaboration, aiming to solve the problems raised in the above background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] The multi-source heterogeneous data analysis method for CNC machine tools based on edge-cloud collaboration includes the following steps:

[0007] Collect multi-source heterogeneous data of CNC machine tools, including machining process data, machine status data, component audio data and workpiece image data;

[0008] Lightweight processing of multi-source heterogeneous data of CNC machine tools, including:

[0009] Matching the communication protocol of multi-source heterogeneous data of CNC machine tools and collecting multi-source heterogeneous data of CNC machine tools; specifically including: identifying the corresponding communication port according to the category of multi-source heterogeneous data of CNC machine tools; matching the device connection parameters required by the corresponding communication port and establishing a handshake connection; starting to cyclically collect CNC machine tool related data; classifying the collected data into processing process data, machine tool status data, component audio data and workpiece image data;

[0010] Preprocessing of processing data, machine tool status data, component audio data and workpiece image data; specifically including: verifying data compliance; determining the data source of compliant data; encapsulating processing data into a data set based on a data dictionary template; using data cleaning methods to delete duplicate values ​​and abnormal values; using corresponding denoising and interference removal methods to clean machine tool status data and component audio data, and extracting data features; using OpenCV library algorithms to clean workpiece image data; finally, processing process data, machine tool status data, component audio data and workpiece image data are lightweight processed according to the specified method to generate a data set, which is stored and uploaded;

[0011] Analyze and predict the multi-source heterogeneous data of CNC machine tools after lightweight processing, including:

[0012] Evaluate the performance status of the data analysis model and make a data set increment judgment based on the performance status; specifically include: matching the processing data, machine tool status data, component audio data and workpiece image data to the corresponding data analysis model in sequence to generate analysis results; compare the analysis results with the real-time collected data; if the error between the analysis results and the real-time data exceeds a threshold, the corresponding data analysis model is regarded as being in a performance state degradation state, and the data set used to prove that it is in a performance state degradation state is encapsulated into a new data set as an incremental data set for updating the corresponding data analysis model;

[0013] Dynamically update the data analysis model; specifically, for performance degradation, use incremental learning to train a new model; for model unsuitability, use transfer learning to match model parameters that adapt to new tasks, and train a model that meets the data analysis requirements of new tasks; finally, display the CNC machine tool status analysis results calculated based on the dynamically updated data analysis model.

[0014] The multi-source heterogeneous data analysis system for CNC machine tools based on edge-cloud collaboration includes:

[0015] Data acquisition source, used to collect multi-source heterogeneous data of CNC machine tools, including:

[0016] CNC system, used to collect processing data;

[0017] Sensors, used to collect machine tool status data;

[0018] Noise monitors to collect component audio data;

[0019] Industrial cameras, used to collect workpiece image data;

[0020] Edge data lightweight processing node, used for lightweight processing of multi-source heterogeneous data of CNC machine tools, including:

[0021] The data acquisition module is used to match the communication protocol of multi-source heterogeneous data of CNC machine tools and collect multi-source heterogeneous data of CNC machine tools; specifically includes: identifying the corresponding communication port according to the category of multi-source heterogeneous data of CNC machine tools; matching the device connection parameters required by the corresponding communication port and establishing a handshake connection; starting to cyclically collect CNC machine tool related data; classifying the collected data into processing data, machine tool status data, component audio data and workpiece image data;

[0022] The data preprocessing module is used to preprocess the processing data, machine tool status data, component audio data and workpiece image data; specifically, it includes: verifying data compliance; determining the data source of compliant data, encapsulating the processing data into a data set based on the data dictionary template, using data cleaning methods to delete duplicate values ​​and abnormal values, using corresponding denoising and interference removal methods to clean the machine tool status data and component audio data, and extracting data features, and using OpenCV library algorithms to clean the workpiece image data; finally, the processing data, machine tool status data, component audio data and workpiece image data are lightweight processed according to the specified method to generate a data set, which is stored and uploaded;

[0023] Cloud data enhancement processing nodes are used to analyze and predict multi-source heterogeneous data of CNC machine tools after lightweight processing, including:

[0024] The data increment module is used to evaluate the performance status of the data analysis model and make a data set increment judgment based on the performance status; specifically, it includes: matching the processing data, machine tool status data, component audio data and workpiece image data to the corresponding data analysis model in sequence to generate analysis results; comparing the analysis results with the real-time collected data; if the error between the analysis results and the real-time data exceeds a threshold, the corresponding data analysis model is regarded as being in a performance state degradation state, and the data set used to prove that it is in a performance state degradation state is encapsulated into a new data set as an incremental data set for updating the corresponding data analysis model;

[0025] The new model training module is used to dynamically update the data analysis model. Specifically, it includes: for performance degradation, incremental learning is used to train a new model. For model unsuitability, transfer learning is used to match model parameters that adapt to new tasks, and a model that meets the data analysis requirements of new tasks is trained. Finally, the CNC machine tool status analysis results calculated based on the dynamically updated data analysis model are displayed.

[0026] Furthermore, in the step of identifying corresponding communication ports according to the categories of multi-source heterogeneous data of CNC machine tools, the collection process of machining process data matches the communication port provided by the CNC system, the collection process of machine tool status data and component audio data matches the communication port provided by the NI controller, and the collection process of workpiece image data matches the communication port provided by the industrial camera.

[0027] Furthermore, in the step of starting the loop to collect CNC machine tool related data, the loop collection of the CNC system data is started by calling the data collection function of the CNC system dynamic link library, the loop collection of the sensor and noise monitor data is started by identifying the corresponding sensor signal channel, and the loop collection of the industrial camera data is started by calling the data collection function of the camera dynamic link library.

[0028] Furthermore, in the step of using corresponding denoising and de-interference methods to clean the machine tool status data and component audio data, and extracting data features, Gaussian filtering is used to remove noise and interference from vibration data, and the vibration amplitude and time domain features are extracted; Gaussian filtering is used to remove noise and interference from temperature data, and the average temperature, maximum temperature, and minimum temperature are extracted; a filter is used to filter the current signal, and the amplitude and spectrum characteristics of the current are extracted; a wavelet denoising method is used to remove noise from noise data, and the spectrum diagram and power spectral density of the noise are extracted.

[0029] Furthermore, the data increment module has built-in data analysis models, including a time series prediction model, a neural network prediction model, a support vector machine prediction model and a random forest prediction model.

[0030] Furthermore, the new model training module embeds online stochastic gradient descent as an incremental learning model, and the new model training module embeds a domain adaptation algorithm as a transfer learning model.

[0031] Compared with the prior art, the present invention has the following beneficial effects:

[0032] This multi-source heterogeneous data analysis method and system for CNC machine tools based on edge-cloud collaboration,

[0033] 1. In the present invention, the edge data lightweight processing node enables high-end CNC machine tool multi-source heterogeneous data to be processed in an extremely short communication link, avoiding network channel congestion and increasing data processing and analysis delays due to a large amount of uplink of multi-noise and multi-redundant data; the edge data acquisition module is equipped with technologies that can process multiple protocols and data types according to the needs of different types of data acquisition sources, thereby improving the flexibility of the system; the edge data preprocessing module has a rich and lightweight embedded data processing algorithm, and flexibly selects the appropriate data processing algorithm according to different data types to generate noise-free, non-redundant high-quality data sets, thereby avoiding the data processing pressure of the cloud data enhancement processing node. The cloud data enhancement processing node uses more powerful computing resources to deeply analyze the multi-source heterogeneous data of high-end CNC machine tools, ensuring the accuracy of data analysis; the cloud data set increment module and the new model training module continuously monitor the model performance, ensuring that the data analysis model maintains the latest and highest accuracy, providing accurate information for decision support, and making the system more adaptable to time-varying working conditions.

[0034] 2. The present invention is based on the demand for improving the intelligence of high-end CNC machine tools, and provides large-scale data services that integrate the acquisition, processing and analysis of multi-source heterogeneous data. It adopts an edge-cloud collaborative method to break the information island between CNC machine tools and computing resources. At the same time, it also solves the problems of limited computing resources, low computing efficiency, and high data analysis model update latency caused by traditional edge computing methods, thereby achieving the purpose of high-efficiency, high-precision and intelligent status analysis of high-end CNC machine tools.

[0035] 3. The present invention realizes the rapid collection, rapid preprocessing and rapid analysis of large-scale multi-source heterogeneous data in the status analysis of high-end CNC machine tools during their service. Through edge-cloud collaboration, it effectively avoids the uploading of about 50% redundant data, reduces data transmission time by about 30%, reduces data analysis time by about 30%, improves the operating efficiency of artificial intelligence algorithms by about 20%, and improves the status analysis and prediction accuracy based on artificial intelligence algorithms by about 15%. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 It is a system architecture diagram of the present invention.

[0037] Figure 2 The data acquisition module in the present invention has the following working flow chart.

[0038] Figure 3 This is a workflow diagram of the data preprocessing module in the present invention.

[0039] Figure 4 This is a workflow diagram of the data set increment module in the present invention.

[0040] Figure 5This is a workflow diagram of the new model training module in the present invention.

[0041] In the figure: 1-data collection source; 2-edge data lightweight processing node; 3-cloud data enhancement processing node; 4-data collection module; 5-data preprocessing module; 6-data set increment module; 7-new model training module. DETAILED DESCRIPTION

[0042] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0043] The specific implementation of the present invention is described in detail below in conjunction with specific embodiments.

[0044] An embodiment of the present invention provides a method for analyzing multi-source heterogeneous data of CNC machine tools based on edge-cloud collaboration, comprising the following steps:

[0045] S1, collect multi-source heterogeneous data of CNC machine tools, including machining process data, machine tool status data, component audio data and workpiece image data;

[0046] S2. Lightweight processing of multi-source heterogeneous data of CNC machine tools, including:

[0047] S21, matching the communication protocol of multi-source heterogeneous data of CNC machine tools and collecting multi-source heterogeneous data of CNC machine tools; specifically including: identifying the corresponding communication port according to the category of multi-source heterogeneous data of CNC machine tools; matching the device connection parameters required by the corresponding communication port, and establishing a handshake connection; starting to cyclically collect CNC machine tool related data; classifying the collected data into processing data, machine tool status data, component audio data and workpiece image data;

[0048] S22, preprocessing the processing data, machine tool status data, component audio data and workpiece image data; specifically including: verifying data compliance; determining the data source of the compliant data, encapsulating the processing data into a data set based on a data dictionary template, using a data cleaning method to delete duplicate values ​​and abnormal values, using corresponding denoising and interference removal methods to clean the machine tool status data and component audio data, and extracting data features, and using the OpenCV library algorithm to clean the workpiece image data; finally, the processing data, machine tool status data, component audio data and workpiece image data are lightweight processed according to the specified method to generate a data set, and the data set is stored and uploaded;

[0049] S3. Analyze and predict the multi-source heterogeneous data of CNC machine tools after lightweight processing, including:

[0050] S31, evaluating the performance status of the data analysis model and making a data set increment judgment based on the performance status; specifically including: matching the processing data, machine tool status data, component audio data and workpiece image data to the corresponding data analysis model in sequence to generate analysis results; comparing the analysis results with the real-time collected data; if the error between the analysis results and the real-time data exceeds a threshold, the corresponding data analysis model is regarded as being in a performance state decay state, and the data set used to prove that it is in a performance state decay state is encapsulated into a new data set as an incremental data set for updating the corresponding data analysis model;

[0051] S32. Dynamically update the data analysis model; specifically, for performance degradation, use incremental learning to train a new model; for model unsuitability, use transfer learning to match model parameters that adapt to new tasks, and train a model that meets the data analysis requirements of new tasks; finally, display the CNC machine tool status analysis results calculated based on the dynamically updated data analysis model.

[0052] like Figure 1 As shown, a multi-source heterogeneous data analysis system for CNC machine tools based on edge-cloud collaboration is provided as an embodiment of the present invention, including:

[0053] Data acquisition source 1 refers to the hardware source for collecting key data of CNC machine tools. The configuration of data acquisition source 1 can be customized according to the user's needs for CNC machine tool acquisition. It includes:

[0054] CNC system, used to collect processing data such as spindle speed, spindle load, feed rate, motion coordinates, etc.;

[0055] Sensors are used to collect machine tool status data such as spindle temperature, vibration, current, guide rail temperature, vibration, etc.

[0056] Noise monitors, used to collect audio data from key functional components such as spindles, motors, and tools;

[0057] Industrial cameras are used to collect workpiece image data such as workpiece morphology and defects.

[0058] Edge data lightweight processing node 2 is used for lightweight processing of multi-source heterogeneous data of CNC machine tools. It is composed of a general performance CPU and a lightweight AI chip. Edge data lightweight processing node 2 has embedded data acquisition module 4 and data preprocessing module 5. Multi-source heterogeneous data of high-end CNC machine tools usually exhibits strong temporal sequence, and has many noises and redundancies. It is preliminarily processed by edge data lightweight processing node 2, and then uploaded to cloud data enhancement processing node 3 for enhanced processing, which effectively avoids the uplink of noise and redundant data. Among them:

[0059] like Figure 2As shown, it is a work flow chart of the data acquisition module 4, which is used to match the communication protocol of multi-source heterogeneous data of CNC machine tools and collect multi-source heterogeneous data of CNC machine tools; specifically, it includes:

[0060] First, when data acquisition source 1 is connected to the system, the system will identify the corresponding communication port according to the category of data acquisition source 1, where the acquisition process of machining process data matches the communication port provided by the CNC system, the machine tool status data comes from the sensor, the component audio data comes from the noise monitor, the sensor and noise monitor channels are integrated in the NI controller, and the acquisition process of the workpiece image data matches the communication port provided by the industrial camera. Therefore, the acquisition process of machining process data matches the communication port provided by the CNC system, the acquisition process of machine tool status data and component audio data matches the communication port provided by the NI controller, and the acquisition process of workpiece image data matches the communication port provided by the industrial camera;

[0061] Secondly, set the Ethernet IP addresses of the CNC system, NI controller, industrial camera and this system to be in the same field and have the same gateway number, so that the system matches the device connection parameters required by the corresponding communication port and establishes a handshake connection;

[0062] Then, the cyclic collection of CNC machine tool related data begins; the cyclic collection of CNC system data begins by calling the data collection function of the CNC system dynamic link library, the cyclic collection of sensor and noise monitor data begins by identifying the corresponding sensor signal channel, and the cyclic collection of industrial camera data begins by calling the data collection function of the camera dynamic link library;

[0063] Finally, the acquired data is classified into machining process data, machine tool status data, component audio data and workpiece image data;

[0064] like Figure 3 As shown, it is a workflow diagram of the data preprocessing module 5, which is used to preprocess the processing data, machine tool status data, component audio data and workpiece image data; specifically, it includes:

[0065] First, all data passed into the data preprocessing module 5 will be checked for data compliance. Any data message that does not match the specified data acquisition source 1 will be considered as non-compliant data and will be excluded from the data preprocessing scope.

[0066] Secondly, determine the data source of the compliant data. For the processing data, encapsulate it into a data set based on the data dictionary template. The data set has a "key-value" key-value association relationship, and then use the data cleaning method to delete duplicate values ​​and abnormal values; for the machine tool status data and component audio data, use the corresponding denoising and interference removal methods to clean the data and extract data features (for vibration data, use Gaussian filtering to remove noise and interference, and extract vibration amplitude and time domain features; for temperature data, use Gaussian filtering to remove noise and interference, and extract average temperature, maximum temperature, and minimum temperature; for current signals, use filters to filter, and extract current amplitude and spectrum features; for noise data, use wavelet denoising method to remove noise, and extract noise spectrum diagram and power spectrum density); for workpiece image data, use OpenCV library algorithm to clean the image (first, select high-resolution images according to image resolution, then use threshold segmentation technology to convert the image into a binary image, and then use morphological operations to eliminate small noise points or smooth edges in the image);

[0067] Finally, the machining process data, machine tool status data, component audio data and workpiece image data are lightweight processed according to the specified method to generate a noise-free, non-redundant high-quality data set, and the data set is stored (the edge data lightweight processing node 2 also has a data storage function. For the multi-source heterogeneous data of high-end CNC machine tools, the data pattern exhibits strong correlation and strong temporal characteristics. Therefore, the edge data lightweight processing node 2 stores the machine tool static data and low refresh frequency data in a relational database. SQL Server and MySQL can be used as reference databases. For high refresh frequency strong temporal data, it is stored in a time series database. InfluxDB can be used as a reference database) and uploaded.

[0068] The cloud data enhancement processing node 3 is used to analyze and predict the multi-source heterogeneous data of the CNC machine tool after lightweight processing. It is composed of a high-performance CPU and an intensive solving AI chip. The cloud data enhancement processing node 3 has an embedded data set increment module 6 and a new model training module 7. The cloud data enhancement processing node 3 receives high-quality data sets from the edge data lightweight processing node 2 through the data interface (the data interface is used to receive data sets from the edge data lightweight processing node 2. The data interface adopts the MQTT communication protocol. The cloud data enhancement processing node 3 is used as the client and the edge data lightweight processing node 2 is used as the server. The client subscribes to data requirements and the server publishes related data. When the edge data lightweight processing node 2 keeps generating new data, the cloud data enhancement processing node 3 will continue to receive data). It dynamically updates the data analysis model and intelligently generates CNC machine tool status analysis results that meet actual working conditions. Among them:

[0069] like Figure 4As shown, it is a workflow diagram of the data set increment module 6. The data set from the edge data lightweight processing node 2 is uploaded to the cloud data enhancement processing node 3 based on the MQTT protocol, and is first imported into the data set increment module 6 of the cloud data enhancement processing node 3. The data set increment module 6 evaluates the performance status of the data analysis model based on real-time data, and makes a data set increment judgment based on the performance status; specifically, it includes:

[0070] First, the processing data, machine tool status data, component audio data, and workpiece image data are matched with corresponding data analysis models in sequence to generate analysis results;

[0071] Secondly, the analysis results are compared with the real-time collected data; if the error between the analysis results and the real-time data exceeds a threshold, the data analysis model is considered to be in a state of performance degradation and needs to be updated in a timely manner;

[0072] Furthermore, the data set that is used to prove that it is in a state of performance degradation is encapsulated into a new data set as an incremental data set for updating the corresponding data analysis model;

[0073] like Figure 5 As shown in FIG. 1 , it is a workflow diagram of the new model training module 7. The new model training module 7 is responsible for dynamically updating the data analysis model under the influence of time-varying working conditions. Time-varying working conditions can lead to two phenomena: performance degradation of the data analysis model and model inapplicability. Performance degradation is manifested as a high error between the data analysis results and the real-time collected data; model inapplicability is manifested as the system generates a new data analysis task, and the current model is applicable to the new task. For different phenomena, the new model training module 7 adopts different solutions to ensure the accuracy of the model, including:

[0074] For the performance degradation phenomenon, the new model training module 7 embeds online random gradient descent as an incremental learning model. This algorithm is suitable for data streams with strong time series, identifies different data patterns from the data stream, and then updates the data analysis model; for the model inapplicability phenomenon, the new model training module 7 embeds a domain adaptive algorithm as a transfer learning model. This algorithm can significantly reduce the distribution difference between the source task and the target task, but the training process may be relatively complex and requires the deployment of a large amount of computing resources. Therefore, it is deployed in the cloud data enhancement processing node 3, and the model parameters adapted to the new task are matched through the domain adaptive algorithm to train a model that meets the data analysis requirements of the new task. The high-end CNC machine tool status analysis results calculated according to the dynamically updated data analysis model are displayed on the visualization interface (the cloud data enhancement processing node 3 opens WebSocket for transmitting data analysis results to the Web visualization terminal for data visualization. WebSocket provides the ability to perform full-duplex communication on a single TCP connection. The TCP connection created by WebSocket will be maintained during the active period of both communicating parties, which can provide better real-time performance).

[0075] The above are only preferred embodiments of the present invention. It should be pointed out that, for those skilled in the art, several modifications and improvements can be made without departing from the concept of the present invention. These should also be regarded as the protection scope of the present invention. These will not affect the effect of the implementation of the present invention and the practicality of the patent.

Claims

1. A multi-source heterogeneous data analysis method for CNC machine tools based on edge-cloud collaboration, characterized in that: The following steps are involved: Collect multi-source heterogeneous data of CNC machine tools, including machining process data, machine status data, component audio data and workpiece image data; Lightweight processing of multi-source heterogeneous data of CNC machine tools, including: Matching the communication protocol of multi-source heterogeneous data of CNC machine tools and collecting multi-source heterogeneous data of CNC machine tools; specifically including: identifying the corresponding communication port according to the category of multi-source heterogeneous data of CNC machine tools; matching the device connection parameters required by the corresponding communication port and establishing a handshake connection; starting to cyclically collect CNC machine tool related data; classifying the collected data into processing process data, machine tool status data, component audio data and workpiece image data; Preprocessing of processing data, machine tool status data, component audio data and workpiece image data; specifically including: verifying data compliance; determining the data source of compliant data; encapsulating processing data into a data set based on a data dictionary template; using data cleaning methods to delete duplicate values ​​and abnormal values; using corresponding denoising and interference removal methods to clean machine tool status data and component audio data, and extracting data features; using OpenCV library algorithms to clean images for workpiece image data; generating a data set after lightweight processing of processing data, machine tool status data, component audio data and workpiece image data, and storing and uploading the data set; The machine tool status data and component audio data are cleaned by using corresponding denoising and interference removal methods, and in the step of extracting data features, Gaussian filtering is used to remove noise and interference for vibration data, and the vibration amplitude and time domain features are extracted; Gaussian filtering is used to remove noise and interference for temperature data, and the average temperature, maximum temperature, and minimum temperature are extracted; a filter is used to filter the current signal, and the amplitude and spectrum features of the current are extracted; a wavelet denoising method is used to remove noise for noise data, and the spectrum diagram and power spectrum density of the noise are extracted; Analyze and predict the multi-source heterogeneous data of CNC machine tools after lightweight processing, including: Evaluate the performance status of the data analysis model and make a data set increment judgment based on the performance status; specifically include: matching the processing data, machine tool status data, component audio data and workpiece image data to the corresponding data analysis model in sequence to generate analysis results; compare the analysis results with the real-time collected data; if the error between the analysis results and the real-time data exceeds the threshold, the corresponding data analysis model is regarded as being in a state of performance degradation, and the data set used to prove that it is in a state of performance degradation is encapsulated into a new data set as an incremental data set for updating the corresponding data analysis model; if the system generates a new data analysis task, and the parameter input of the current model does not match the required parameters of the new task, the corresponding data analysis model is regarded as being in a state of model inapplicability; Dynamically update the data analysis model; specifically, for performance degradation, use incremental learning to train a new model; for model unsuitability, use transfer learning to match model parameters that adapt to new tasks, and train a model that meets the data analysis requirements of new tasks; finally, display the CNC machine tool status analysis results calculated based on the dynamically updated data analysis model.

2. The multi-source heterogeneous data analysis system for CNC machine tools based on edge-cloud collaboration is characterized by: include: Data acquisition source, used to collect multi-source heterogeneous data of CNC machine tools, including: CNC system, used to collect processing data; Sensors, used to collect machine tool status data; Noise monitors to collect component audio data; Industrial cameras, used to collect workpiece image data; Edge data lightweight processing node, used for lightweight processing of multi-source heterogeneous data of CNC machine tools, including: The data acquisition module is used to match the communication protocol of multi-source heterogeneous data of CNC machine tools and collect multi-source heterogeneous data of CNC machine tools; specifically includes: identifying the corresponding communication port according to the category of multi-source heterogeneous data of CNC machine tools; matching the device connection parameters required by the corresponding communication port and establishing a handshake connection; starting to cyclically collect CNC machine tool related data; classifying the collected data into processing data, machine tool status data, component audio data and workpiece image data; The data preprocessing module is used to preprocess the processing data, machine tool status data, component audio data and workpiece image data; specifically, it includes: verifying data compliance; determining the data source of compliant data, encapsulating the processing data into a data set based on the data dictionary template, using data cleaning methods to delete duplicate values ​​and abnormal values, using corresponding denoising and interference removal methods to clean the machine tool status data and component audio data, and extracting data features, and using OpenCV library algorithms to clean the workpiece image data; after lightweight processing, the processing data, machine tool status data, component audio data and workpiece image data are generated into a data set, which is stored and uploaded; The machine tool status data and component audio data are cleaned by using corresponding denoising and interference removal methods, and in the step of extracting data features, Gaussian filtering is used to remove noise and interference for vibration data, and the vibration amplitude and time domain features are extracted; Gaussian filtering is used to remove noise and interference for temperature data, and the average temperature, maximum temperature, and minimum temperature are extracted; a filter is used to filter the current signal, and the amplitude and spectrum features of the current are extracted; a wavelet denoising method is used to remove noise for noise data, and the spectrum diagram and power spectrum density of the noise are extracted; Cloud data enhancement processing nodes are used to analyze and predict multi-source heterogeneous data of CNC machine tools after lightweight processing, including: The data increment module is used to evaluate the performance status of the data analysis model and make a data set increment judgment based on the performance status; specifically, it includes: matching the processing data, machine tool status data, component audio data and workpiece image data to the corresponding data analysis model in sequence to generate analysis results; comparing the analysis results with the real-time collected data; if the error between the analysis results and the real-time data exceeds the threshold, the corresponding data analysis model is regarded as being in a state of performance degradation, and the data set used to prove that it is in a state of performance degradation is encapsulated into a new data set as an incremental data set for updating the corresponding data analysis model; if the system generates a new data analysis task, and the parameter input of the current model does not match the required parameters of the new task, the corresponding data analysis model is regarded as being in a state of model inapplicability; The new model training module is used to dynamically update the data analysis model. Specifically, it includes: for performance degradation, incremental learning is used to train a new model. For model unsuitability, transfer learning is used to match model parameters that adapt to new tasks, and a model that meets the data analysis requirements of new tasks is trained. Finally, the CNC machine tool status analysis results calculated based on the dynamically updated data analysis model are displayed.

3. The multi-source heterogeneous data analysis system for CNC machine tools based on edge-cloud collaboration according to claim 2 is characterized in that: In the step of identifying corresponding communication ports according to categories of multi-source heterogeneous data of CNC machine tools, the collection process of machining process data matches the communication port provided by the CNC system, the collection process of machine tool status data and component audio data matches the communication port provided by the NI controller, and the collection process of workpiece image data matches the communication port provided by the industrial camera.

4. The multi-source heterogeneous data analysis system for CNC machine tools based on edge-cloud collaboration according to claim 2 is characterized in that: In the step of starting the cyclic collection of CNC machine tool related data, the cyclic collection of the CNC system data is started by calling the data collection function of the CNC system dynamic link library, the cyclic collection of the sensor and noise monitor data is started by identifying the corresponding sensor signal channel, and the cyclic collection of the industrial camera data is started by calling the data collection function of the camera dynamic link library.

5. The multi-source heterogeneous data analysis system for CNC machine tools based on edge-cloud collaboration according to claim 2 is characterized in that: The data increment module has built-in data analysis models, including a time series prediction model, a neural network prediction model, a support vector machine prediction model and a random forest prediction model.

6. The multi-source heterogeneous data analysis system for CNC machine tools based on edge-cloud collaboration according to claim 2 is characterized in that: The new model training module embeds online stochastic gradient descent as an incremental learning model, and the new model training module embeds a domain adaptation algorithm as a transfer learning model.