A digital twin system for real-time monitoring of tool status

Abstract

The present invention discloses a digital twin system for real-time monitoring of tool status, comprising a signal acquisition module, a data transmission module, a tool wear prediction module, and a data visualization module; the signal acquisition module uses sensors to collect static data information and dynamic data information of the machine tool operation process; the data transmission module transmits the static and dynamic data information collected by the signal acquisition module to the tool wear prediction module and the data visualization module; the tool wear prediction module uses the dynamic data information obtained from the data transmission module to train the tool wear prediction model; the data visualization module maps the cutting process information and the predicted value of tool wear obtained from the data transmission module to the constructed tool digital twin and visualizes it. The present invention has the advantages of low latency and high visualization, and can more intuitively understand the real-time cutting status of the tool, thereby better managing the tool.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent manufacturing technology, and in particular relates to a digital twin system for real-time monitoring of tool status. Background Art

[0002] In the Industry 4.0 era, intelligent manufacturing has become a key development direction for the manufacturing industry. As the foundation of the manufacturing industry, the digitization and intelligence of the cutting process are of great significance to the intelligent development of the entire manufacturing industry. When machining certain large and complex parts, if tool dullness or damage cannot be detected in time during the machining process, it may not only cause the workpiece to be scrapped, but also reduce the accuracy and service life of the machine tool. Therefore, it is necessary to monitor the tool status during the machining process.

[0003] During the cutting process, a vast amount of cutting process information is generated, such as cutting forces, vibrations, and acoustic emissions. Effectively utilizing this information is crucial for tool condition monitoring. Most current tool condition monitoring systems only collect and analyze machine tool signals, outputting a single observation. These outputs are not intuitive and require expert knowledge to quickly understand, hindering operator familiarity.

[0004] Digital twin technology digitizes physical entities, enabling real-time data exchange between the physical world and virtual models. Through real-time data collection and analysis, the behavior of physical entities can be monitored, maintained, and optimized. Applying digital twin technology to tool condition monitoring systems creates a more intuitive three-dimensional interactive interface, providing the monitoring system with more comprehensive and intuitive tool condition information.

[0005] Chinese invention patent CN 110900307 B proposes a digital twin-driven CNC machine tool tool monitoring system that enables tool status monitoring during the milling process. However, the method proposed in this patent is mainly targeted at the milling process and is not very adaptable to other processing methods. In addition, the simulation and life prediction methods used in this method require a large amount of computation, which will cause system delays. Chinese invention patent CN 113126564A proposes a digital twin-driven CNC milling tool wear online monitoring method, but the prediction model used in this method requires complex data preprocessing and requires a large amount of computation. Summary of the Invention

[0006] In response to the above problems, the present invention proposes a digital twin system for real-time monitoring of tool status, which can achieve low-latency and highly visualized tool status monitoring.

[0007] To achieve the above-mentioned object, the technical solution of the present invention is as follows: a digital twin system for real-time monitoring of tool status, comprising a signal acquisition module, a data transmission module, a tool wear prediction module, and a data visualization module;

[0008] The signal acquisition module uses sensors to collect static and dynamic data during machine tool operation. The static data includes tool brand, tool material, workpiece size, workpiece material, spindle speed, cutting speed, cutting depth, and feed rate; the dynamic data includes force signals in the x, y, and z directions, acceleration signals in the x, y, and z directions, and acoustic emission signals.

[0009] The data transmission module transmits the static and dynamic data information collected by the signal acquisition module to the tool wear prediction module and the data visualization module.

[0010] The tool wear prediction module uses the dynamic data information obtained from the data transmission module to train the tool wear prediction model, and uses the trained tool wear prediction model to predict the tool wear state and tool wear prediction value, and inputs the tool wear prediction value into the data visualization module.

[0011] The data visualization module maps the cutting process information and tool wear prediction value obtained from the data transmission module to the constructed tool digital twin and presents them visually. At the same time, it determines the next processing plan of the machine tool based on the tool wear status obtained from the tool wear prediction module.

[0012] Furthermore, the static signal transmission method:

[0013] First, set the IP address, subnet mask, and gateway on the machine tool so that the machine tool and the computer are in the same local area network.

[0014] Then use a network cable to connect the Ethernet interface of the machine tool to the Ethernet interface of the computer to connect to the network.

[0015] Finally, socket programming is used and data interaction between the machine tool and the computer is carried out based on the TCP / IP protocol to realize the transmission of the machine tool's static signals.

[0016] The dynamic signal transmission method is as follows: the signal collected by the sensor is processed by a signal amplifier and a digital-to-analog converter, and then transmitted to a computer via a USB protocol.

[0017] Furthermore, the tool wear prediction model includes a feature extraction part, a damage detection part, a feature fusion part and a state classification part. The feature extraction part is composed of a one-dimensional convolutional neural network containing six convolutional layers, three pooling layers and one fully connected layer. Its function is to extract features from the dynamic data information obtained from the data acquisition module.

[0018] The operation process of the convolution layer is as follows:

[0019]

[0020] in: Represents the output result after the lth layer of convolution, Represents the output data of the i-th channel of the l-1th layer, * is the convolution symbol, is the weight matrix of the i-th channel of the l-th layer of the convolutional layer, Represents the bias term of the i-th channel of the l-th convolutional layer.

[0021] The operation process of the pooling layer is as follows:

[0022]

[0023] in: Represents the output result after the l-th layer pooling, Represents the weight matrix of the i-th channel of the l-th layer of the pooling layer, represents the bias term of the i-th channel of the l-th layer of the pooling layer, and g represents the maximum pooling function.

[0024] The operation process of the fully connected layer is as follows:

[0025]

[0026] in: Represents the output result of the lth fully connected layer, represents the weight matrix of the ith channel of the lth layer of the fully connected layer, represents the bias term of the i-th channel of the l-th layer of the fully connected layer, and s represents the Sigmoid activation function.

[0027] The damage detection part determines whether the tool is chipped based on whether the extracted signal characteristics exceed the preset threshold. If the signal characteristics exceed the threshold, the tool wear status is output as broken.

[0028] The feature fusion part uses the deep generalized canonical correlation analysis method to achieve feature fusion.

[0029] The state classification part uses a softmax classifier to classify the tool wear conditions based on the fused feature vector. The classifier first calculates the score of each category based on the fused feature vector:

[0030] r j =xT+e

[0031] Among them, r j represents the score of the j-th category, x represents the fused feature vector, T represents the weight vector of the classifier, and e is the bias vector.

[0032] Then calculate the probability of each category:

[0033] Among them, p j is the probability of the jth category, and K is the total number of categories.

[0034] After calculating the probabilities for all classes, the classifier selects the class with the highest probability as the output.

[0035] Furthermore, the visual presentation includes presentation of spindle speed, real-time cutting force, real-time vibration and tool wear area, and the tool wear area is calculated based on the tool wear prediction value obtained by the tool wear prediction module.

[0036] Furthermore, the tool wear status obtained from the tool wear prediction module is used to determine the next machining plan for the machine tool. Specifically, the tool wear status is classified into four types: initial wear, stable wear, blunting, and chipping. When the tool status is initial wear or stable wear, the machine tool continues machining according to the original machining plan; when the tool status is blunt, the system issues a tool change prompt; when the tool status is chipping, the system controls the machine tool to shut down.

[0037] In summary, the beneficial effects of the present invention are:

[0038] 1. The present invention proposes a digital twin system for real-time monitoring of tool status, which combines digital twin technology with machine learning technology. Compared with existing systems, it has the advantages of low latency and high visualization. It can more intuitively understand the real-time cutting status of the tool, thereby better managing the tool.

[0039] 2. The present invention identifies tool breakage through a threshold comparison method and predicts tool wear values ​​based on the idea of ​​classification problems, thereby improving the accuracy of prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 This is a system structure diagram of the present invention.

[0041] Figure 2Schematic diagram of a data acquisition system in an embodiment of the present invention.

[0042] Figure 3 This is a flowchart of tool wear prediction in an embodiment of the present invention.

[0043] In the figure: 1 is a three-jaw chuck, 2 is a workpiece, 3 is a tool holder, 4 is a tool, 5 is an acceleration sensor, and 6 is a dynamometer. DETAILED DESCRIPTION

[0044] The present invention will be further described in detail below with reference to the following examples. It should be understood that the specific examples described herein are only used to illustrate the present invention and are not intended to limit the present invention.

[0045] like Figure 1 As shown in the figure, a digital twin system for real-time monitoring of tool status includes a signal acquisition module, a data transmission module, a tool wear prediction module, and a data visualization module:

[0046] The main function of the signal acquisition module is to use sensors to collect static data and dynamic data information during the operation of the machine tool. Among them, the static data information includes tool brand, tool material, workpiece size, workpiece material, spindle speed, cutting speed, cutting depth and feed rate; the dynamic data information includes force signals in the three directions of x, y, and z, acceleration signals in the three directions of x, y, and z, and acoustic emission signals. In this embodiment, the components of the signal acquisition module are as follows: Figure 2 As shown, the workpiece 2 is clamped by a three-jaw chuck 1, the tool holder 3 is mounted on a dynamometer 6, the tool 4 is mounted on the tool holder 3, and the acceleration sensor 5 is mounted at the front end of the tool 4.

[0047] The main function of the data transmission module is to transmit the static and dynamic signals collected by the signal acquisition module to the tool wear prediction module and the data visualization module.

[0048] Furthermore, the static signal transmission method mentioned is that, first, the IP address, subnet mask and gateway are set on the machine tool so that the machine tool and the computer are in the same local area network. Then, use a network cable to connect the Ethernet interface of the machine tool to the Ethernet interface of the computer to connect to the network. Finally, use socket programming and TCP / IP protocol to perform data interaction between the machine tool and the computer to realize the transmission of the static signal of the machine tool. The dynamic signal transmission method mentioned is that the collected three-axis force signal is processed by the signal amplifier and the digital-to-analog converter, and then transmitted to the computer via the USB protocol. The collected acceleration signal is processed by the digital-to-analog converter and then transmitted to the computer via the USB protocol.

[0049] The main function of the tool wear prediction module is to input the model obtained from the data transmission module into the tool wear prediction model, and input the tool wear status and tool wear prediction value output by the model into the data visualization module.

[0050] Furthermore, the tool wear prediction model mentioned above includes four parts: feature extraction, damage detection, feature fusion and state classification. The prediction process is as follows: Figure 3 As shown in the figure, the feature extraction part consists of a one-dimensional convolutional neural network containing six convolutional layers, three pooling layers and one fully connected layer. Its main function is to extract features from the sensor data obtained from the data acquisition module.

[0051] The operation process of the convolution layer is as follows:

[0052]

[0053] in: Represents the output result after the lth layer of convolution, Represents the output data of the i-th channel of the l-1th layer, * is the convolution symbol, is the weight matrix of the i-th channel of the l-th layer of the convolutional layer, Represents the bias term of the i-th channel of the l-th convolutional layer.

[0054] The operation process of the pooling layer is as follows:

[0055]

[0056] in: Represents the output result after the l-th layer pooling, Represents the weight matrix of the i-th channel of the l-th layer of the pooling layer, represents the bias term of the i-th channel of the l-th layer of the pooling layer, and g represents the maximum pooling function.

[0057] The operation process of the fully connected layer is as follows:

[0058]

[0059] in: Represents the output result of the lth fully connected layer, represents the weight matrix of the ith channel of the lth layer of the fully connected layer, represents the bias term of the i-th channel of the l-th layer of the fully connected layer, and s represents the Sigmoid activation function.

[0060] The damage detection component determines whether the tool is chipped based on whether the extracted signal features exceed a preset threshold. If the signal features exceed the threshold, the tool wear status is output as broken. The feature fusion component uses deep generalized canonical correlation analysis to achieve feature fusion.

[0061] The state classification part uses the softmax classifier to classify the tool wear conditions based on the fused feature vector. The classifier first calculates the score of each category based on the fused feature vector:

[0062] r j =xT+e

[0063] Among them, r j represents the score of the j-th category, x represents the fused feature vector, T represents the weight vector of the classifier, and e is the bias vector.

[0064] Then calculate the probability of each category:

[0065]

[0066] Among them, p j is the probability of the jth category, and K is the total number of categories.

[0067] After calculating the probabilities for all classes, the classifier selects the class with the highest probability as the output.

[0068] In this embodiment, the designed softmax classifier divides tool wear into 200 categories, numbered 1-200, where each increase in the number represents a 2 μm increase in tool wear value VB. When tool wear is numbered 1-50, the tool wear status output is initial wear; when tool wear is numbered 51-150, the tool wear status output is stable wear; and when tool wear is numbered 151-200, the tool wear status output is blunt.

[0069] The main function of the data visualization module is to map the cutting process information and tool wear prediction value obtained from the data transmission module to the constructed tool digital twin and present them visually, and at the same time determine the next processing plan of the machine tool based on the tool wear status obtained from the tool wear prediction module.

[0070] Furthermore, the mentioned visualization presentation specifically includes the presentation of spindle speed, real-time cutting force, real-time vibration and tool wear area, wherein the tool wear area can be calculated based on the tool wear prediction value obtained by the tool wear prediction module.

[0071] Furthermore, the tool wear status, as determined by the tool wear prediction module, determines the next machining plan for the machine tool. Specifically, tool wear status is categorized into four types: initial wear, stable wear, blunting, and chipping. When the tool is in initial wear or stable wear, the machine tool continues machining according to the original plan. When the tool is blunt, the system prompts a tool change; when the tool is chipping, the system shuts down the machine tool.

[0072] The present invention is not limited to this embodiment, and any equivalent concepts or modifications within the technical scope disclosed by the present invention are included in the protection scope of the present invention.

Claims

1. A digital twin system for real-time monitoring of tool status, characterized by: It includes signal acquisition module, data transmission module, tool wear prediction module and data visualization module; The signal acquisition module uses sensors to collect static data and dynamic data during the operation of the machine tool; the static data includes tool brand, tool material, workpiece size, workpiece material, spindle speed, cutting speed, cutting depth and feed rate; the dynamic data includes force signals in the x, y and z directions, acceleration signals in the x, y and z directions and acoustic emission signals; The data transmission module transmits the static and dynamic data information collected by the signal acquisition module to the tool wear prediction module and the data visualization module; The tool wear prediction module uses the dynamic data information obtained from the data transmission module to train the tool wear prediction model, and uses the trained tool wear prediction model to predict the tool wear state and tool wear prediction value, and inputs the tool wear prediction value into the data visualization module; The data visualization module maps the cutting process information and tool wear prediction value obtained from the data transmission module to the constructed tool digital twin and presents them visually. At the same time, it determines the next processing plan of the machine tool based on the tool wear status obtained from the tool wear prediction module.

2. The digital twin system for real-time monitoring of tool status according to claim 1, characterized in that: The static data information transmission method comprises the following steps: First, set the IP address, subnet mask, and gateway on the machine tool so that the machine tool and the computer are in the same local area network; Then use a network cable to connect the Ethernet interface of the machine tool to the Ethernet interface of the computer to connect to the network; Finally, socket programming is used to carry out data interaction between the machine tool and the computer based on the TCP / IP protocol to realize the transmission of the machine tool's static signals; The dynamic data information transmission method is as follows: the signal collected by the sensor is processed by a signal amplifier and a digital-to-analog converter, and then transmitted to a computer via a USB protocol.

3. The digital twin system for real-time monitoring of tool status according to claim 1, characterized in that: The tool wear prediction model includes a feature extraction part, a damage detection part, a feature fusion part, and a state classification part. The feature extraction part is composed of a one-dimensional convolutional neural network with six convolutional layers, three pooling layers, and one fully connected layer. Its function is to extract features from the dynamic data information obtained from the data acquisition module. The operation process of the convolution layer is as follows: in: Indicates the The output result after layer convolution, Indicates the -1st floor i Output data of each channel, is the convolution symbol, The convolutional layer Tier i The weight matrix of the channels, Represents the convolutional layer The bias term of the i-th channel of the layer; The operation process of the pooling layer is as follows: in: Indicates the The output result after layer pooling, Represents the pooling layer Tier i The weight matrix of the channels, Represents the pooling layer Tier i The bias term of each channel, represents the maximum pooling function; The operation process of the fully connected layer is as follows: in: Indicates the The output of the fully connected layer, represents the fully connected layer Tier i The weight matrix of the channels, represents the fully connected layer Tier i The bias term of each channel, s Represents the Sigmoid activation function; The damage detection part determines whether the tool is chipped based on whether the extracted signal characteristics exceed the preset threshold. If the signal characteristics exceed the threshold, the tool wear status is output as broken; The feature fusion part uses the deep generalized canonical correlation analysis method to achieve feature fusion; The state classification part uses a softmax classifier to classify the tool wear conditions based on the fused feature vector. The classifier first calculates the score of each category based on the fused feature vector: in, Representative j The scores of the categories, represents the fused feature vector, represents the weight vector of the classifier, e is the bias vector; Then calculate the probability of each category: in, is the probability of the jth category, K is the total number of categories; After calculating the probabilities for all classes, the classifier selects the class with the highest probability as the output.

4. The digital twin system for real-time monitoring of tool status according to claim 1, characterized in that: The visualization includes spindle speed, real-time cutting force, real-time vibration and tool wear area, and the tool wear area is calculated based on the tool wear prediction value obtained by the tool wear prediction module.

5. The digital twin system for real-time monitoring of tool status according to claim 1, characterized in that: The method of judging the next processing plan of the machine tool based on the tool wear status obtained from the tool wear prediction module is as follows: specifically, the tool wear status is divided into four types: initial wear, stable wear, blunting and chipping; when the tool status is initial wear and stable wear, the machine tool continues processing according to the original processing plan; when the tool status is blunt, the system issues a tool change prompt; when the tool status is chipping, the system controls the machine tool to stop.

Images