Cloud-edge collaborative belt conveyor roller bearing digital twin condition monitoring method

Abstract

The present invention discloses a cloud-edge collaborative belt conveyor roller bearing digital twin state monitoring method, including: obtaining underground belt conveyor roller bearing operation data; transmitting the operation data to the edge device and the cloud center server; constructing a geometric model and a three-dimensional simulation model in the cloud center server, constructing a fault diagnosis model in the edge device, and using the geometric model, the three-dimensional simulation model and the fault diagnosis model to form a digital twin; combining measured data, updating the three-dimensional simulation model in the digital twin in real time, and using incremental learning to update the fault diagnosis model in the digital twin in real time, and using the fault diagnosis model to determine the operating state of the belt conveyor roller bearing. The present invention allocates computing resources based on the cloud-edge collaborative algorithm, realizes fast calculation of the digital twin, improves the virtual-real interaction capability of the digital twin, and finally realizes the belt conveyor roller bearing state monitoring, early warning and visual analysis based on the digital twin.

Description

Technical Field

[0001] The present invention relates to the technical field of mining equipment fault monitoring, and in particular to a cloud-edge collaborative digital twin state monitoring method for belt conveyor roller bearings. Background Art

[0002] Underground belt conveyors are the main means of transportation in mines. Roller bearings are the "joints" in underground belt conveyors. There are many of them. In addition, they bear heavy loads and work in harsh environments. Therefore, they are prone to failures during long-term operation, which seriously affects the production efficiency and economic benefits of mines. Therefore, monitoring the operating status of roller bearings in underground belt conveyors and handling operating failures in a timely manner are of great practical significance to the safe and efficient production of mines.

[0003] Digital twin refers to the construction of a mapping body in virtual space that can reflect the real condition of mechanical equipment. Based on virtual-real mapping and virtual-real evolution methods, it reflects the spatiotemporal evolution of key performance parameters of mechanical equipment. On the technical level, digital twin technology has the potential to monitor the operating status of roller bearings in real time. However, the belt conveyor roller bearing status monitoring method based on digital twin faces the following challenges:

[0004] 1) The traditional cloud computing method transmits sensor data to the central cloud server for centralized processing. The single cloud computing method will cause the underground communication network to be overloaded, causing underground data transmission delays, affecting the real-time data interaction between the digital twin virtual body and the physical entity of the roller bearing;

[0005] 2) The operation of roller bearings is characterized by time-varying working conditions and unsteady performance evolution. The construction of digital twins of roller bearings has high requirements on computing resources and real-time computing.

[0006] In summary, under the constraints of communication and computing resources, it is difficult to implement the belt conveyor roller bearing condition monitoring method based on digital twins. Summary of the invention

[0007] Purpose of the invention: In view of the above problems, the purpose of the present invention is to provide a cloud-edge collaborative digital twin state monitoring method for belt conveyor roller bearings. Based on the cloud-edge collaborative method, real-time data interaction between the digital twin and the physical entity is realized, and a digital twin of the roller bearing with iterative update function is further constructed to reflect the entire process of the roller bearing state evolution in the virtual space.

[0008] Technical solution: A cloud-edge-coordinated belt conveyor roller bearing digital twin state monitoring method of the present invention comprises the following steps:

[0009] Step 1, obtaining the operating data of the roller bearing of the underground belt conveyor through the sensor arranged on the underground belt conveyor;

[0010] Step 2: Use the ZigBee terminal node to send the operating data to the router through the ZigBee network, the router transmits it to the coordinator node through the ZigBee network, the coordinator node sends it to the edge device through the 5G network, and the edge device sends it to the cloud center server through the 5G network;

[0011] Step 3: construct a geometric model and a three-dimensional simulation model of the roller bearing in the cloud center server, construct a fault diagnosis model on the edge device, and use the geometric model, three-dimensional simulation model and fault diagnosis model of the roller bearing to form a digital twin;

[0012] Step 4: Combine the measured data to update the three-dimensional simulation model in the digital twin in real time, and further use incremental learning to update the fault diagnosis model in the digital twin in real time. Use the fault diagnosis model to determine the operating status of the belt conveyor roller bearing and issue corresponding early warning reminders.

[0013] Furthermore, the monitoring method further comprises the following steps:

[0014] Step 5: Monitor the operating status of the belt conveyor roller bearing group in real time through the remote monitoring platform.

[0015] Furthermore, in step 1, the sensors include temperature sensors, vibration sensors, bearing force sensors, rotation speed sensors and load sensors; the operation data include roller bearing temperature data, roller bearing vibration data, pressure data of the roller bearing surface, roller rotation speed and load.

[0016] Furthermore, before step 2, the method further includes: establishing multiple ZigBee networks with tree topology, each ZigBee network including a coordinator node, multiple routers and multiple ZigBee terminal nodes, and each ZigBee terminal node including a single chip microcomputer and a ZigBee communication module.

[0017] Furthermore, in step 2, the process of using the ZigBee terminal node to send the operation data to the router via the ZigBee network includes:

[0018] The sensor is used to convert the operation data from physical quantity into electrical analog signal, and the electrical analog signal is sent to the single chip microcomputer. The single chip microcomputer converts the electrical analog signal into digital signal and transmits it to the router through the ZigBee communication module.

[0019] Furthermore, in step 3, a three-dimensional simulation model of the roller bearing is constructed through the roller bearing geometric model and operation data. The three-dimensional simulation model includes a temperature field model, a vibration modal model and a stress distribution model.

[0020] Further, in step 3, after building the fault diagnosis model on the edge device, the method further includes: training the fault diagnosis model;

[0021] The training process is:

[0022] The 3D simulation model is run in real time on the cloud center server. The roller speed information and load information of the belt conveyor during real-time operation are input into the 3D simulation model. The 3D simulation model is used to generate virtual temperature signals, force signals and vibration signals at different speeds and loads, and a simulation signal library is established. D 1 , and stored in the cloud center server;

[0023] Using the Simulation Signal Library D 1 Train the fault diagnosis model. The mathematical expression of the fault diagnosis model is:

[0024] ,

[0025] in, f is the fault diagnosis result. x The data samples of roller bearing temperature, vibration, roller speed and load are θ i are the fault diagnosis model parameters, F It is a fault diagnosis model.

[0026] Furthermore, there are two edge devices, which are placed on the edge sides of the belt conveyor, namely the first edge device and the second edge device, and the two edge devices are used to calculate the tasks generated by the fault diagnosis model.

[0027] Furthermore, two edge devices are used to calculate the tasks generated by the fault diagnosis model, including:

[0028] First, according to the computing resources on the first edge device, all or part of the tasks are offloaded to the device for calculation. If there are any remaining computing tasks, the remaining computing tasks are offloaded to the second edge device through the 5G network, and the results are returned to the first edge device after the calculation is completed.

[0029] Secondly, if the first edge device does not receive the calculation results returned by the second edge device after completing the calculation task, the remaining calculation tasks in the second edge device are partially unloaded to the first edge device, so that the first edge device and the second edge device complete all remaining calculation tasks at the same time.

[0030] Furthermore, in step 4, updating the three-dimensional simulation model and the fault diagnosis model in the digital twin in real time includes:

[0031] Based on the actual measurement results of roller bearing temperature, vibration, force, roller speed and load, the simulation signal library is updated in real time using the 3D simulation model in the cloud center server. D 1 ;

[0032] Collect the bearing physical temperature signal, vibration signal and force signal of the belt conveyor under normal operation and establish a physical signal library D 2 , stored in the cloud center server;

[0033] Using the Pearson correlation coefficient, the simulation signal library is calculated on the cloud central server D 1 With the entity signal library D 2 The similarity of the corresponding signals is calculated, and the similarity calculation result is transmitted to the edge device through 5G; if the similarity is greater than the threshold, the fault diagnosis model is used for fault identification, otherwise it is judged that a new type of fault has occurred.

[0034] Furthermore, when a new type of fault occurs, the collected new type of fault signal is stored in the new category database D new , D new Stored on edge devices, and the physical signal library D 2 Simulation Signal Library D 1 The old category data in the dataset is used as the example set, and the iCaRL (Incremental Classifier and Representation Learning) incremental learning method is used to D new Update the fault diagnosis model incrementally and calculate the updated fault diagnosis model F new ; The loss function of incremental update is as follows:

[0035] ,

[0036] in, L is the loss function during training, and the hyperparameter The value range is \left [ {0,1} \right ] , L new The updated fault diagnosis model F new In the new category database D new The loss function on LKD The original fault diagnosis model F and the updated fault diagnosis model F new Distillation loss on the example set.

[0037] Furthermore, the fault diagnosis model is used to identify the operating status of the belt conveyor roller bearing and to issue corresponding early warning reminders, including:

[0038] When the fault diagnosis model in the edge device diagnoses a fault, a fault warning is issued; when a new type of fault is identified, an incremental warning of the new type of fault is issued.

[0039] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages:

[0040] 1. The present invention constructs a digital twin of the belt conveyor roller bearing, including a geometric model, a three-dimensional simulation model and a fault diagnosis model. The digital twin is used to realize remote real-time health status monitoring of all roller bearings of the belt conveyor, thereby improving the intelligence level of roller bearing status monitoring and ensuring timely, accurate and comprehensive presentation of the health status of the roller bearings of the underground belt conveyor;

[0041] 2. The present invention reduces the load of the underground communication network and the delay of underground data transmission by establishing a joint communication network of ZigBee and 5G, and realizes the real-time data interaction between the digital twin virtual body and the physical entity of the belt conveyor roller bearing;

[0042] 3. The present invention uses cloud-edge collaborative computing technology to reasonably allocate the computing power of the cloud center server and edge devices, arrange the fault diagnosis model on the edge device close to the roller bearing, and arrange the three-dimensional simulation model on the cloud center server, thereby improving the operation and maintenance efficiency of the belt conveyor;

[0043] 4. The present invention uses incremental learning technology to update the digital twin of the belt conveyor roller bearing in real time, thereby ensuring and improving the iterative update performance of the digital twin and improving the accuracy of fault diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is a flow chart of the digital twin state monitoring method of belt conveyor roller bearings in cloud-edge collaboration in the embodiment;

[0045] Figure 2 A schematic diagram of a communication network integrating 5G and ZigBee in an embodiment;

[0046] Figure 3 It is a flowchart of digital twin cloud-edge collaborative computing in an embodiment;

[0047] Figure 4A flowchart of updating and iterating the digital twin in an embodiment. DETAILED DESCRIPTION

[0048] In order to make the objectives, technical solutions and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments.

[0049] The cloud-edge collaborative belt conveyor roller bearing digital twin state monitoring method described in this embodiment is as follows: Figure 1 The method comprises the following steps:

[0050] Step 1, obtaining the operating data of the roller bearing of the underground belt conveyor through the sensor arranged on the underground belt conveyor;

[0051] Step 2: Use the ZigBee terminal node to send the operating data to the router through the ZigBee network, the router transmits it to the coordinator node through the ZigBee network, the coordinator node sends it to the edge device through the 5G network, and the edge device sends it to the cloud center server through the 5G network;

[0052] Step 3: construct a geometric model and a three-dimensional simulation model of the roller bearing in the cloud center server, construct a fault diagnosis model on the edge device, and use the geometric model, three-dimensional simulation model and fault diagnosis model of the roller bearing to form a digital twin;

[0053] Step 4: Combine the measured data to update the three-dimensional simulation model in the digital twin in real time, and further use incremental learning to update the fault diagnosis model in the digital twin in real time. Use the fault diagnosis model to determine the operating status of the belt conveyor roller bearing and issue corresponding early warning reminders.

[0054] In one example, the cloud-edge collaborative belt conveyor roller bearing digital twin condition monitoring method also includes the following steps:

[0055] Step 5: Monitor the operating status of the belt conveyor roller bearing group in real time through the remote monitoring platform.

[0056] Specifically, in step 1, the sensors include temperature sensors, vibration sensors, bearing force sensors and speed sensors; the operation data include temperature data, vibration data, force data, speed and load data, and the roller bearing temperature data, roller bearing vibration data, roller bearing force surface pressure data, roller speed and load data of the underground belt conveyor during operation are obtained through the temperature sensor, vibration sensor, bearing force sensor, speed sensor and load sensor respectively.

[0057] The cloud-edge-coordinated belt conveyor roller bearing digital twin state monitoring method described in this embodiment further includes before step 2:

[0058] Multiple tree-topology ZigBee networks are established. Each ZigBee network includes a coordinator node, multiple routers, and multiple ZigBee terminal nodes. Each ZigBee terminal node includes a CC2530 microcontroller and a ZigBee communication module.

[0059] like Figure 2 As shown, the sensor will obtain the operating data of the roller bearing of the underground belt conveyor and send it to the single-chip microcomputer. The single-chip microcomputer sends the operating data to the router through the ZigBee communication module. One router is connected to multiple terminal nodes (in this example, one router corresponds to three terminal nodes). The router is connected to the coordinator node through the 5G smart gateway. The router sends the operating data to the coordinator node through the 5G smart gateway under the ZigBee network. The coordinator node sends the data to the edge device through the 5G network, and the edge device sends the data to the cloud center server through the 5G network. Each coordinator node is connected to the corresponding edge device, and each edge device is connected to the 5G network, and all edge devices are connected to the same cloud center server. Through the fusion networking of the tree-like ZigBee network and the 5G network, the data obtained by the ZigBee terminal node is transmitted to the edge device, and the communication between the edge device and the cloud center server is maintained.

[0060] In step 2, the process of using the ZigBee terminal node to send the operation data to the router through the ZigBee network includes:

[0061] The sensor is used to convert the operation data from physical quantity into electrical analog signal, and the electrical analog signal is sent to the single chip microcomputer. The single chip microcomputer converts the electrical analog signal into digital signal and transmits it to the router through the ZigBee communication module.

[0062] In step 3, a three-dimensional simulation model of the roller bearing is constructed through the roller bearing geometric model and operation data. The three-dimensional simulation model includes a temperature field model, a vibration mode model and a stress distribution model.

[0063] Specifically, the geometric model of the roller bearing is constructed using a geometric 3D scanner and SolidWorks in the cloud center server. G v , and further construct a 3D simulation model of the roller bearing of the underground belt conveyor based on other operating information obtained R v ,The 3D simulation model includes a temperature field model, a vibration mode model and a stress distribution model.

[0064] The digital twin is constructed using geometric models, 3D simulation models, and fault diagnosis models, as shown in the following formula:

[0065] ,

[0066] In the formula, VE For digital twins, VE include G v , R v and F Model, F is a fault diagnosis model. In the example, the fault diagnosis model F ResNet32 network can be used. Utilize the 3D simulation model integrated in the cloud center server R v Fault diagnosis model on edge devices F , which can realize data interaction between the cloud and the edge, map the operating status of the roller bearings of the underground belt conveyor in real time, including temperature field, stress distribution, vibration mode and fault information, and build a digital twin of the roller bearings of the underground belt conveyor; the cloud center server is the computing center of the entire system, which is used for dynamic simulation of roller bearings and realizes remote health status monitoring of all roller bearings of the underground belt conveyor through a visualization platform; the edge device is close to the roller bearings, and the signal communication delay is low, so it can timely diagnose the faults of the signals generated by the roller bearings, and perform incremental learning on the fault diagnosis model when new types of faults occur.

[0067] In step 3, after the fault diagnosis model is built on the edge device, the following steps are further included: training the fault diagnosis model;

[0068] The training process is:

[0069] The 3D simulation model is run in real time on the cloud center server. The roller speed information and load information of the belt conveyor during real-time operation are input into the 3D simulation model. The 3D simulation model is used to generate virtual temperature signals, force signals and vibration signals under different speeds and loads, and a simulation signal library is established. D 1 , and stored in the cloud center server;

[0070] Using the Simulation Signal Library D 1 Train the fault diagnosis model. The mathematical expression of the fault diagnosis model is:

[0071] ,

[0072] in, f is the fault diagnosis result. x It is the operating data sample of roller bearing temperature, vibration, roller speed and load, etc. θi are the fault diagnosis model parameters, F It is a fault diagnosis model.

[0073] There are two edge devices, which are placed on the edge sides of the belt conveyor, namely the first edge device and the second edge device. The two edge devices are used to calculate the tasks generated by the fault diagnosis model.

[0074] like Figure 3 to Figure 4 As shown, in step 4, the real-time updating of the three-dimensional simulation model and the fault diagnosis model in the digital twin includes:

[0075] Based on the actual measurement results of roller bearing temperature, vibration, force, roller speed and load, the simulation signal library is updated in real time using the 3D simulation model in the cloud center server. D 1 ;

[0076] Collect the bearing physical temperature signal, force signal and vibration signal of the belt conveyor under normal operation and establish a physical signal library D 2 , stored in the cloud center server;

[0077] Using the Pearson correlation coefficient, the simulation signal library is calculated on the cloud central server D 1 With the entity signal library D 2 The similarity of the corresponding signal in the 5G network is calculated, and the similarity calculation result is transmitted to the edge device through 5G. If it is greater than the threshold, the fault diagnosis model is used to judge the fault, otherwise it is judged that a new type of fault has occurred. In the example, the threshold can be set to 0.6.

[0078] Furthermore, when a new type of fault occurs, the collected new type of fault signal is stored in the new category database D new , D new Stored on edge devices. D 2 Simulation Signal Library D 1 The old category data in the dataset is used as the example set, and the iCaRL (Incremental Classifier and Representation Learning) incremental learning method is used to D new Update the fault diagnosis model incrementally and calculate the updated fault diagnosis model F new ; The loss function of incremental update is as follows:

[0079] ,

[0080] in, L is the loss function during training, and the hyperparameter The value range is \left [ {0,1} \right ] , L new The updated fault diagnosis model F new In the new category database D new The loss function on L KD The original fault diagnosis model F and the updated fault diagnosis model F new Distillation loss on the example set.

[0081] Furthermore, edge computing has the characteristics of real-time and low latency, which can ensure the real-time fault diagnosis and life prediction of belt conveyors. The tasks generated by the fault diagnosis model are calculated using two edge devices, including:

[0082] First, according to the first edge device I 1 The computing resources on the device are used to offload all or part of the tasks to the device for calculation. If there are any remaining computing tasks, the remaining computing tasks are offloaded to the second edge device through the 5G network. I 2 , and returns the result to the first edge device after the calculation is completed I 1 .

[0083] Secondly, if the first edge device I 1 After completing the computing task, the second edge device did not receive I 2 The calculation result returned will be sent to the second edge device I 2 The remaining computing tasks are partially offloaded to the first edge device I 1 , making the first edge device I 1 With the second edge device I 2 At the same time, complete all remaining computing tasks.

[0084] Furthermore, the fault diagnosis model is used to identify the operating status of the belt conveyor roller bearing and to issue corresponding early warning reminders, including:

[0085] When the fault diagnosis model in the edge device diagnoses a fault, a fault warning is issued; when a new type of fault is identified, an incremental warning of the new type of fault is issued.

[0086] Furthermore, a remote monitoring platform for the roller bearings of the underground belt conveyor is established on the cloud center server to monitor the operating status of the roller group in real time.

[0087] Through the remote monitoring platform, the network and real-time operating status of each bearing in the underground belt conveyor can be viewed; the historical fault information of each bearing and the historical fault information of the entire underground belt conveyor can be viewed; the digital twin simulation models and simulation data of various fault types, various health status data in the physical signal library can be viewed, and signal data can be manually added, deleted and modified.

Claims

1. The digital twin state monitoring method of belt conveyor roller bearings in cloud-edge collaboration is characterized by: The steps include: Step 1, obtaining the operating data of the roller bearing of the underground belt conveyor through the sensor arranged on the underground belt conveyor; Step 2: Use the ZigBee terminal node to send the operating data to the router through the ZigBee network, the router transmits it to the coordinator node through the ZigBee network, the coordinator node sends it to the edge device through the 5G network, and the edge device sends it to the cloud center server through the 5G network; Step 3: construct a geometric model and a three-dimensional simulation model of the roller bearing in the cloud center server, construct a fault diagnosis model on the edge device, and use the geometric model, three-dimensional simulation model and fault diagnosis model of the roller bearing to form a digital twin; Step 4: Combine the measured data to update the three-dimensional simulation model in the digital twin in real time, and further use incremental learning to update the fault diagnosis model in the digital twin in real time. Use the fault diagnosis model to determine the operating status of the belt conveyor roller bearing and issue corresponding early warning reminders. In step 3, after the fault diagnosis model is built on the edge device, the following steps are further included: training the fault diagnosis model; The training process is: The 3D simulation model is run in real time on the cloud center server. The roller speed information and load information of the belt conveyor during real-time operation are input into the 3D simulation model. The 3D simulation model is used to generate virtual temperature signals, force signals and vibration signals at different speeds and loads, and a simulation signal library is established. D 1. And stored in the cloud center server; Using the Simulation Signal Library D 1. Train the fault diagnosis model. The mathematical expression of the fault diagnosis model is: ， in, f is the fault diagnosis result. x The data samples are the roller bearing temperature, vibration, roller speed and load. θ i are the fault diagnosis model parameters, F It is a fault diagnosis model; In step 4, the real-time updating of the 3D simulation model and the fault diagnosis model in the digital twin includes: Based on the actual measurement results of roller bearing temperature, vibration, force, roller speed and load, the simulation signal library is updated in real time using the 3D simulation model in the cloud center server. D 1; Collect the bearing physical temperature signal, vibration signal and force signal of the belt conveyor under normal operation and establish a physical signal library D 2. Store to the cloud center server; Using the Pearson correlation coefficient, the simulation signal library is calculated on the cloud central server D 1 and physical signal library D 2, and transmit the similarity calculation result to the edge device through the 5G network; if the similarity is greater than the threshold, the fault diagnosis model is used for fault identification; if the similarity is less than the threshold, it is determined that a new type of fault has occurred.

2. The cloud-edge-coordinated belt conveyor roller bearing digital twin state monitoring method according to claim 1 is characterized in that: The following steps are also included: Step 5: Monitor the operating status of the belt conveyor roller bearing group in real time through the remote monitoring platform.

3. The cloud-edge-coordinated belt conveyor roller bearing digital twin state monitoring method according to claim 1 is characterized in that: In step 1, the sensors include a temperature sensor, a vibration sensor, a bearing force sensor, a rotation speed sensor and a load sensor; the operation data include roller bearing temperature data, roller bearing vibration data, pressure data of the roller bearing surface, roller rotation speed and load.

4. The cloud-edge-coordinated belt conveyor roller bearing digital twin state monitoring method according to claim 3 is characterized in that: Before step 2, the method also includes: establishing multiple ZigBee networks with tree topology, each ZigBee network includes a coordinator node, multiple routers and multiple ZigBee terminal nodes, and each ZigBee terminal node includes a single chip microcomputer and a ZigBee communication module.

5. The cloud-edge-coordinated belt conveyor roller bearing digital twin state monitoring method according to claim 4 is characterized in that: In step 2, the process of using the ZigBee terminal node to send the operation data to the router through the ZigBee network includes: The sensor is used to convert the operation data from physical quantity into electrical analog signal, and the electrical analog signal is sent to the single chip microcomputer. The single chip microcomputer converts the electrical analog signal into digital signal and transmits it to the router through the ZigBee communication module.

6. The cloud-edge-coordinated belt conveyor roller bearing digital twin state monitoring method according to claim 5 is characterized in that: In step 3, a three-dimensional simulation model of the roller bearing is constructed through the roller bearing geometric model and operation data. The three-dimensional simulation model includes a temperature field model, a vibration mode model and a stress distribution model.

7. The cloud-edge-coordinated belt conveyor roller bearing digital twin state monitoring method according to claim 6 is characterized in that: There are two edge devices, which are placed on the edge sides of the belt conveyor, namely the first edge device and the second edge device. The two edge devices are used to calculate the tasks generated by the fault diagnosis model.

8. The cloud-edge-coordinated belt conveyor roller bearing digital twin state monitoring method according to claim 7 is characterized in that: The tasks generated by the fault diagnosis model are calculated using two edge devices, including: First, according to the computing resources on the first edge device, all or part of the tasks are offloaded to the device for calculation. If there are still remaining computing tasks, the remaining computing tasks are offloaded to the second edge device through the 5G network, and the results are returned to the first edge device after the calculation is completed. Secondly, if the first edge device does not receive the calculation results returned by the second edge device after completing the calculation task, the remaining calculation tasks in the second edge device are partially unloaded to the first edge device, so that the first edge device and the second edge device complete all remaining calculation tasks at the same time.

9. The cloud-edge-coordinated belt conveyor roller bearing digital twin state monitoring method of claim 8 is characterized in that: When a new type of fault occurs, the collected new type of fault signal is stored in the new category database D new , and the entity signal library D 2 and simulation signal library D 1 is used as the example set, and the iCaRL incremental learning method is used to learn the new category database. D new Update the fault diagnosis model incrementally and calculate the updated fault diagnosis model F new ; The loss function of incremental update is as follows: ， in, L is the loss function during training, and the hyperparameter The value range is , L new The updated fault diagnosis model F new In the new category database D new The loss function on L KD The original fault diagnosis model F and the updated fault diagnosis model F new Distillation loss on the example set.

10. The cloud-edge-coordinated belt conveyor roller bearing digital twin state monitoring method according to claim 9 is characterized in that: The fault diagnosis model is used to identify the operating status of the belt conveyor roller bearing and to issue corresponding warning reminders, including: When the fault diagnosis model in the edge device diagnoses a fault, a fault warning is issued; when a new type of fault is identified, an incremental warning of the new type of fault is issued.

Images