VR fault diagnosis system and method based on digital twinning

Abstract

The invention discloses a VR fault diagnosis system and method based on digital twinning. A data management platform, a digital twinning diagnosis platform and a VR visual display platform are integrated; the data management platform is used for managing digital twin data; the digital twin data comprises real-time experimental data, historical experimental data and three-dimensional time domain signal data information; the digital twinborn diagnosis platform is used for equipment control management and data fault diagnosis prediction; the equipment control management comprises the steps of responding to a front-end control request, submitting a background feedback request, interacting with a data management platform and setting a fault diagnosis trigger triggering condition; and the VR visual display platform realizes equipment three-dimensional visualization and flow type control management through VR equipment. According to the invention, virtual visualization service requirements of current accelerator equipment can be effectively improved, a set of integrated intelligent diagnosis system is constructed by using a digital twinning technology, a VR technology and a deep learning algorithm, and the real-time management and control efficiency of the equipment and the convenience of equipment maintenance are improved.

Description

Technical Field

[0001] The present invention relates to the field of digital intelligence and fault diagnosis of ion source accelerator equipment, and in particular to a VR fault diagnosis system and method based on digital twins. Background Art

[0002] In the new era of rapid development of digitalization and intelligence, the increasingly mature digital twin technology has gradually been applied to various fields, especially in the industrial field that lacks digitalization and intelligence, showing its unique value. At the same time, with the advancement of VR technology, digital twins have shown unique advantages and further promoted the development of automation in the industrial field. However, the virtual visualization and data management system of ion source accelerator equipment is not perfect at present, lacking a more intuitive and easy-to-understand visualization management system and a complete equipment intelligent maintenance solution. Summary of the invention

[0003] In order to overcome the deficiencies of existing system intelligent management and equipment fault diagnosis technology, the present invention provides a VR fault diagnosis system and method based on digital twins. Through the interaction between 3D visualization tools and VR devices, the TensorFlow framework is used to improve the identification, monitoring and voice control of fault-prone parts, thereby improving the intelligent management of existing equipment. The deep learning algorithm is used to solve the problems of high equipment maintenance costs and lack of fault diagnosis prediction.

[0004] To achieve the above object, the technical solution adopted by the present invention is as follows:

[0005] A VR fault diagnosis system based on digital twins, including a data management platform, a digital twin diagnosis platform and a VR visual display platform;

[0006] The data management platform is used to manage digital twin data; the digital twin data includes real-time experimental data, historical experimental data and three-dimensional time domain signal data information; the real-time experimental data and historical experimental data are collected by the data management platform through an automatic control device and transferred to a database;

[0007] The digital twin diagnosis platform is used for equipment control management and data fault diagnosis prediction; the equipment control management includes responding to front-end control requests, submitting back-end feedback requests, interacting with the data management platform, and setting fault diagnosis trigger conditions; the data fault diagnosis prediction includes data preprocessing, AI algorithm analysis, and fault result feedback;

[0008] The VR visual display platform uses VR glasses to manipulate information, uses virtual reality interactive plug-ins to achieve three-dimensional visual interaction and display of information, and uses the TensorFlow framework to achieve identification and monitoring of fault-prone parts and information submission; the TensorFlow framework can be set with a voice recognition function to realize voice transmission of control requests.

[0009] Furthermore, when the real-time experimental data and historical experimental data are transferred, the multi-source data are processed by graphical programming software and Python program and then transferred and managed uniformly.

[0010] Furthermore, the three-dimensional time domain signal data is collected by setting the collection parameters and the collection time period of each information through the VR device; the three-dimensional time domain signal data information includes basic equipment information, bearing status information, base status information, pump body status information, and intake pipe status information.

[0011] Furthermore, the fault diagnosis trigger is realized through an event monitoring function; the three-dimensional visual display includes a maintenance reminder of the fault location when the user uses a VR device to control the device, a 3D visualization simulation model of the accelerator device, and a simulation of the experimental operation environment of the accelerator device.

[0012] The present invention also provides a fault diagnosis method of a VR fault diagnosis system based on digital twins, comprising the following steps:

[0013] Step S1: Scan the accelerator physical device and set key part acquisition parameters and acquisition time period according to the geometric connection relationship and physical characteristics of the device;

[0014] Step S2: The voice control data is submitted in segments, and the data storage status of the data management platform is obtained through the event monitoring function;

[0015] Step S3: The data management platform transfers the data, triggers the fault diagnosis trigger of the digital twin diagnosis platform, and starts data fault diagnosis prediction;

[0016] Step S4: setting a maintenance plan according to the prediction results, updating the 3D visualization simulation model of the accelerator equipment in real time, and simulating the fault state of the accelerator equipment;

[0017] Step S5: Match the maintenance plan with the fault location, and the VR visual display platform displays the fault location fault information and maintenance plan in real time;

[0018] Step S6: The user maintains the accelerator equipment according to the maintenance plan, and the maintenance results and equipment status are fed back to the digital twin diagnosis platform in real time. At the same time, the 3D visualization simulation model of the accelerator equipment is updated again.

[0019] Furthermore, in step S1, the geometric connection relationship of the equipment includes the shape of each component of the equipment and the cross-sectional connection method, and the physical properties include the maximum stress point of the component and the maximum internal pressure.

[0020] Furthermore, in step S3, the data fault diagnosis prediction uses a deep residual shrinkage network to process the three-dimensional time domain signal; when processing the three-dimensional time domain signal, its input signal includes a time dimension, a space dimension and a feature dimension; assuming that the input signal ,in, is the field of real numbers, is the number of time domain signal acquisition channels of the accelerator device, is the length of the acquisition time, is the feature dimension, which represents the input signal as a whole is a tensor with three dimensions, and each element in the tensor is a real number; in the deep residual contraction network, the three-dimensional convolution layer The output of the layer convolution is:

[0021] ;

[0022] in, and Respectively The weights and biases of the convolution kernels of the layers, represents a three-dimensional convolution operation, represents the activation function, For the The output of the layer;

[0023] In the residual block, the input is set to , the output is:

[0024] ;

[0025] in, Representative The features after convolution mapping of the layer are added with residual connection;

[0026] To suppress noise, soft threshold shrinkage is performed on the features of each channel;

[0027] Assume The feature map of the layer is , then the contraction operation is:

[0028] ;

[0029] in, is a dynamically adjusted threshold, determined by the mean or standard deviation of each channel. It is a sign function used to obtain the sign information of the input feature. When the input feature is greater than 0, the return value is 1, when the input feature is less than 0, the return value is -1, and when the input feature is equal to 0, the return value is 0. In the shrinkage operation, it is used to retain the positive and negative signs of the eigenvalues ​​so that the signs of the shrunken values ​​are consistent with those of the original values.

[0030] After multiple residual blocks and contraction operations, the final output of the network is:

[0031] ;

[0032] in, represents the input signal of the network, is the parameter set of the entire deep residual shrinkage network, including the convolution kernel weights , bias term , threshold parameter . The representative neural network model is the deep residual shrinkage network, whose main functions are to extract features of input signals, suppress noise and effectively represent learning signals.

[0033] Furthermore, in step S4, the simulation of the fault state of the accelerator device includes deformation, force distribution and internal pressure state of the faulty component.

[0034] Furthermore, in step S5, the fault information and maintenance plan will dynamically display the information of the corresponding scanned part as the VR device scans.

[0035] Furthermore, in step S6, the maintenance results and real-time feedback of the device status are submitted through voice control via the TensorFlow framework, and the background updates the 3D visualization simulation model of the accelerator device according to the submitted data.

[0036] Beneficial effects:

[0037] The present invention can effectively improve the current virtual visualization business needs of accelerator equipment, use digital twin technology and deep learning algorithms to build an integrated intelligent diagnosis system, continuously improve the training of AI algorithm models according to the experimental data of each device, improve the accuracy of fault prediction, improve the efficiency of equipment fault warning, ensure the safe operation of equipment experiments, and closely connect various scattered devices.

[0038] The present invention integrates digital twin technology, VR technology and industrial automation control technology to establish a three-dimensional visualization platform for accelerator equipment, truly restores the experimental operation status and experimental environment of the equipment, and realizes mobile control management through VR equipment, allowing users to control equipment experiments and monitor equipment operation status anytime and anywhere, thereby improving the real-time performance of equipment management and control and reducing equipment maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 This is a flow chart of a VR fault diagnosis method based on digital twins of the present invention. DETAILED DESCRIPTION

[0040] In order to make the purpose, technical solutions 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 intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0041] A VR fault diagnosis system based on digital twins of an embodiment of the present invention includes a data management platform, a digital twin diagnosis platform and a VR visual display platform. The data management platform is used to manage digital twin data; the digital twin data includes real-time experimental data, historical experimental data and three-dimensional time domain signal data information; the real-time experimental data and historical experimental data are collected by the data management platform through an automated control device and transferred to a database; the digital twin diagnosis platform is used for equipment control management and data fault diagnosis prediction; equipment control management includes responding to front-end control requests, submitting background feedback requests, interacting with the data management platform and setting fault diagnosis trigger conditions; data fault diagnosis prediction includes data preprocessing, AI algorithm analysis and fault result feedback; the VR visual display platform performs information manipulation through VR glasses, uses virtual reality interactive plug-ins to realize three-dimensional visual interaction and display of information, realizes fault-prone part identification monitoring and information submission through the TensorFlow framework, and sets a voice recognition function to realize voice transmission control requests.

[0042] When real-time experimental data and historical experimental data are transferred, multi-source data are processed through graphical programming software and Python programs for unified transfer and management; three-dimensional time domain signal data can be collected by setting the collection parameters and the collection time period of each information through VR equipment; three-dimensional time domain signal data includes basic equipment information, bearing status information, base status information, pump body status information, and intake pipe status information; fault diagnosis triggers are realized through event monitoring functions; three-dimensional visual display includes maintenance reminders of faulty parts when users use VR equipment to control equipment, 3D visualization simulation model of accelerator equipment, and simulation of accelerator equipment experimental operation environment.

[0043] like Figure 1 As shown, the VR fault diagnosis method based on digital twins in an embodiment of the present invention includes the following steps:

[0044] Step S1: Scan the accelerator physical device and set key part acquisition parameters and acquisition time period according to the geometric connection relationship and physical characteristics of the device;

[0045] Step S2: Submit the voice control data in segments, and obtain the storage status of the data on the data management platform through the event monitoring function;

[0046] Step S3: The data management platform transfers the data, triggers the fault diagnosis trigger of the digital twin diagnosis platform, and starts data fault diagnosis prediction;

[0047] Step S4: setting a maintenance plan according to the prediction results, updating the 3D visualization simulation model of the accelerator equipment in real time, and simulating the fault state of the accelerator equipment;

[0048] Step S5: Match the maintenance plan with the fault location, and the VR visual display platform displays the fault location fault information and maintenance plan in real time;

[0049] Step S6: The user maintains the accelerator equipment according to the maintenance plan, and the maintenance results and equipment status are fed back to the digital twin diagnosis platform in real time. At the same time, the 3D visualization simulation model of the accelerator equipment is updated again.

[0050] Specifically, in step S1, the geometric connection relationship of the equipment includes the shape of each component of the equipment and the cross-sectional connection method, and the physical properties include the maximum stress point of the component and the maximum internal pressure.

[0051] Specifically, in step S3, the data fault diagnosis prediction uses a deep residual shrinkage network to process the three-dimensional time domain signal. When processing the three-dimensional time domain signal, its input signal includes a time dimension, a space dimension and a feature dimension. The input signal ,in, is the field of real numbers, is the number of time domain signal acquisition channels of the accelerator device, is the length of the acquisition time, is the feature dimension, which represents the input signal as a whole is a tensor with three dimensions, and each element in the tensor is a real number; in the deep residual contraction network, the three-dimensional convolution layer The output of the convolution layer for:

[0052] ;

[0053] in, and Respectively The weights and biases of the convolution kernels of the layers, represents a three-dimensional convolution operation, represents the activation function, For the The output of the layer;

[0054] In the residual block, the input is set to , the output is:

[0055] ;

[0056] in, Representative The features after convolution mapping of the layer are added with residual connection;

[0057] In order to suppress noise, soft threshold shrinkage is performed on the features of each channel. The feature map of the layer is , then the contraction operation is:

[0058] ;

[0059] in, is a dynamically adjusted threshold, determined by the mean or standard deviation of each channel. It is a sign function used to obtain the sign information of the input feature. When the input feature is greater than 0, the return value is 1, when the input feature is less than 0, the return value is -1, and when the input feature is equal to 0, the return value is 0. In the shrinkage operation, it is used to retain the positive and negative signs of the eigenvalues ​​so that the signs of the shrunken values ​​are consistent with those of the original values.

[0060] After multiple residual blocks and contraction operations, the final output of the network is obtained for:

[0061] ;

[0062] in, represents the input signal of the network, is the parameter set of the entire deep residual shrinkage network, including the convolution kernel weights , bias term , threshold parameter . It stands for Deep Residual Contraction Network, and its main functions are to extract features from input signals, suppress noise, and learn effective representations of signals.

[0063] Specifically, in step S4, the simulation of the fault state of the accelerator device includes deformation of the faulty component, force distribution and internal pressure state.

[0064] Specifically, in step S5, the fault information and maintenance plan will dynamically display the information of the corresponding scanned part as the VR device scans.

[0065] Specifically, in step S6, the maintenance results and real-time feedback of the device status can be submitted through voice control via the TensorFlow framework, and the background updates the 3D visualization simulation model of the accelerator device according to the submitted data.

Claims

1. A VR fault diagnosis system based on digital twins, characterized in that: Including data management platform, digital twin diagnosis platform and VR visual display platform; The data management platform is used to manage digital twin data; the digital twin data includes real-time experimental data, historical experimental data and three-dimensional time domain signal data information; the real-time experimental data and historical experimental data are collected by the data management platform through an automatic control device and transferred to a database; The digital twin diagnosis platform is used for equipment control management and data fault diagnosis prediction; the equipment control management includes responding to front-end control requests, submitting back-end feedback requests, interacting with the data management platform, and setting fault diagnosis trigger conditions; the data fault diagnosis prediction includes data preprocessing, AI algorithm analysis, and fault result feedback; The VR visual display platform uses VR glasses to manipulate information, uses virtual reality interactive plug-ins to achieve three-dimensional visual interaction and display of information, and uses the TensorFlow framework to achieve identification and monitoring of fault-prone parts and information submission; the TensorFlow framework can be set with a voice recognition function to realize voice transmission of control requests.

2. A VR fault diagnosis system based on digital twins according to claim 1, characterized in that: When the real-time experimental data and historical experimental data are transferred, the multi-source data are processed by graphical programming software and Python program and then transferred and managed uniformly.

3. A VR fault diagnosis system based on digital twins according to claim 2, characterized in that: The three-dimensional time domain signal data is collected by setting the collection parameters and the collection time period of each information through the VR device; the three-dimensional time domain signal data information includes basic equipment information, bearing status information, base status information, pump body status information, and intake pipe status information.

4. The VR fault diagnosis system based on digital twin according to claim 1 is characterized in that: The fault diagnosis trigger is realized through an event monitoring function; the three-dimensional visual display includes a maintenance reminder of the fault location when the user uses a VR device to control the device, a 3D visualization simulation model of the accelerator device, and a simulation of the experimental operation environment of the accelerator device.

5. A fault diagnosis method based on a VR fault diagnosis system based on digital twins according to any one of claims 1 to 4, characterized in that: The steps include: Step S1: Scan the accelerator physical device and set key part acquisition parameters and acquisition time period according to the geometric connection relationship and physical characteristics of the device; Step S2: The voice control data is submitted in segments, and the data storage status of the data management platform is obtained through the event monitoring function; Step S3: The data management platform transfers the data, triggers the fault diagnosis trigger of the digital twin diagnosis platform, and starts data fault diagnosis prediction; Step S4: setting a maintenance plan according to the prediction results, updating the 3D visualization simulation model of the accelerator equipment in real time, and simulating the fault state of the accelerator equipment; Step S5: Match the maintenance plan with the fault location, and the VR visual display platform displays the fault location fault information and maintenance plan in real time; Step S6: The user maintains the accelerator equipment according to the maintenance plan, and the maintenance results and equipment status are fed back to the digital twin diagnosis platform in real time. At the same time, the accelerator equipment 3D visualization simulation model is updated again.

6. The fault diagnosis method according to claim 5, characterized in that: In step S1, the geometric connection relationship of the equipment includes the shape of each component of the equipment and the cross-sectional connection method, and the physical properties include the maximum stress point of the component and the maximum internal pressure.

7. The fault diagnosis method according to claim 5, characterized in that: In step S3, the data fault diagnosis prediction uses a deep residual shrinkage network to process the three-dimensional time domain signal; when processing the three-dimensional time domain signal, its input signal includes a time dimension, a space dimension and a feature dimension; assuming that the input signal ,in, is the field of real numbers, is the number of time domain signal acquisition channels of the accelerator device, is the length of the acquisition time, is the feature dimension, which represents the input signal as a whole is a tensor with three dimensions, and each element in the tensor is a real number; in the deep residual contraction network, the three-dimensional convolution layer The output of the layer convolution is: ； in, and Respectively The weights and biases of the convolution kernels of the layers, represents a three-dimensional convolution operation, represents the activation function, For the The output of the layer; In the residual block, the input is set to , the output is: ； in, Representative The features after convolution mapping of the layer are added with residual connection; To suppress noise, soft threshold shrinkage is performed on the features of each channel; Assume The feature map of the layer is , then the contraction operation is: ； in, is a dynamically adjusted threshold, determined by the mean or standard deviation of each channel. It is a sign function used to obtain the sign information of the input feature. When the input feature is greater than 0, the return value is 1, when the input feature is less than 0, the return value is -1, and when the input feature is equal to 0, the return value is 0. In the shrinkage operation, it is used to retain the positive and negative signs of the eigenvalues ​​so that the sign of the shrunken value is consistent with that of the original value. After multiple residual blocks and contraction operations, the final output of the network is: ； in, represents the input signal of the network, is the parameter set of the entire deep residual shrinkage network, including the convolution kernel weights , bias term , threshold parameter ; Represents a deep residual shrinkage network.

8. The fault diagnosis method according to claim 5, characterized in that: In step S4, the simulation of the fault state of the accelerator device includes deformation, force distribution and internal pressure state of the faulty component.

9. The fault diagnosis method according to claim 5, characterized in that: In step S5, the fault information and maintenance plan will dynamically display the information of the corresponding scanned part as the VR device scans.

10. The fault diagnosis method according to claim 5, characterized in that: In step S6, the maintenance results and real-time feedback of the device status are submitted through voice control via the TensorFlow framework, and the background updates the 3D visualization simulation model of the accelerator device according to the submitted data.

Images