Intelligent fault diagnosis platform of end-edge-cloud full link based on virtual rack

Abstract

The invention relates to an end-edge-cloud full-link intelligent fault diagnosis platform based on a virtual rack, and the platform comprises a virtual rack system which is used for carrying out the dynamic modeling and fault type modeling of a tested system or part, and simulating the response characteristics of the tested system or part under a set boundary condition and a fault state, converting the response characteristic into an electrical signal; the edge calculation system is used for carrying out preprocessing and feature extraction on the collected electric signals; a user of the cloud intelligent fault early warning system carries out fault diagnosis, reliability analysis or residual life estimation according to the features extracted by the edge computing system. According to the platform, an end-edge-cloud full-link technology is verified based on a virtual rack model, so that the experiment cost can be remarkably reduced, and the early-stage test and deployment efficiency of a diagnosis algorithm can be improved; a modularized structure is adopted, algorithms and models in the modules can be modified and replaced according to actual needs, communication modes among the modules can be configured according to needs, and the system has good expandability and adaptability.

Description

Technical Field

[0001] The present application relates to the technical field of fault diagnosis, and in particular to an end-edge-cloud full-link intelligent fault diagnosis platform based on a virtual test bench. Background Art

[0002] As artificial intelligence algorithms represented by machine learning and deep learning are widely used in equipment fault diagnosis and early warning, in order to ensure the accuracy of fault diagnosis, the size of the model and the complexity of the algorithm are increasing, while the resources and computing frequency of embedded systems are limited. Therefore, deploying intelligent diagnosis algorithms on cloud servers and realizing communication between the tested equipment and cloud servers through edge computing systems, thereby building an intelligent diagnosis system based on end-edge-cloud collaboration has become a research hotspot and technical trend in the field of PHM (Prognostics and Health Management).

[0003] In terms of the end-edge-cloud full-link intelligent fault diagnosis platform, academia and industry have conducted extensive and in-depth research. The document (announcement number is CN116106005A) discloses an end-edge-cloud collaborative fault diagnosis framework. By deploying the TinyML model on the end node, the end device computing power is effectively utilized, the computing and communication costs after the model deployment are reduced, and the contradiction between fault diagnosis accuracy and delay is solved. The document (announcement number is CN115186883A) discloses an industrial equipment health status monitoring system and method based on edge-cloud collaborative computing, and provides an implementation process from edge-end data reading to cloud-end fault diagnosis. The document (announcement number is CN112101532A) discloses an adaptive multi-model driven equipment fault diagnosis method based on edge-cloud collaboration, which divides the deep learning-based fault diagnosis model between the edge and the cloud at the layer granularity, and proposes a cross-operating condition diagnosis method based on edge-cloud collaboration. By training a general operating condition model on the cloud and sending it to the edge, and then the edge diagnoses the data of its personalized operating condition, the delay problem of edge-cloud collaborative fault diagnosis is effectively reduced. The document (announcement number CN113567159A) proposed a method based on edge-cloud collaboration for the status monitoring and fault diagnosis of scraper conveyors. By building a distributed neural network model DDNN consisting of an edge-side network model and a cloud-side network model, it effectively solved the problem that the neural network model cannot be directly used for IoT devices.

[0004] Although certain effects have been achieved, the above inventions all have certain limitations. For example, in the above-mentioned public research results, the focus is mostly on data processing at the edge and intelligent diagnostic algorithms in the cloud, while the "end" in the "end-edge-cloud" framework is rarely involved, and existing studies have assumed that the "end" is the object to be tested in actual engineering. However, in most application scenarios, building an "end-edge-cloud" system based on the actual object to be tested or test bench is not only expensive and has a long debugging cycle, but is also not conducive to quickly verifying the diagnostic effect of the intelligent algorithm in the cloud or edge-cloud collaboration; in addition, the operating conditions and boundary conditions of the actual equipment are often limited to a few typical conditions for safety reasons, which is not conducive to fully testing the performance of the diagnostic algorithm in the "end-edge-cloud" framework. Summary of the invention

[0005] Based on this, it is necessary to provide an end-edge-cloud full-link intelligent fault diagnosis platform based on a virtual test bench to address the above technical problems.

[0006] A virtual test bench-based end-edge-cloud full-link intelligent fault diagnosis platform, the platform includes: a virtual test bench system, an edge computing system, and a cloud-based intelligent fault warning system.

[0007] The virtual bench system is used to perform dynamic modeling and fault type modeling on the system or component under test, simulate the response characteristics of the system or component under test under set boundary conditions and fault states, and convert the response characteristics into electrical signals.

[0008] The edge computing system is used to preprocess and extract features of the collected electrical signals, and transmit the extracted features to the cloud-based intelligent fault warning system. It is also used to receive the results of fault diagnosis, reliability analysis or remaining life estimation transmitted by the cloud-based intelligent fault warning system, so that users can analyze the diagnostic results of the tested system or components at the edge, improve the edge data preprocessing, feature extraction methods and parameter configurations according to the diagnostic results, and further optimize the cloud-based diagnostic results.

[0009] The cloud-based intelligent fault warning system is used to perform fault diagnosis, reliability analysis or remaining life estimation on the system or component under test based on the features extracted by the edge computing system, and transmit the results of the fault diagnosis, reliability analysis or remaining life estimation to the edge computing system.

[0010] In one embodiment, the virtual rig system includes: a dynamic model, a fault model, and a hardware interface.

[0011] The dynamic model is used to establish the kinematic and dynamic equations of the system or component under test in each degree of freedom according to Newton's laws; the kinematic and dynamic equations are used to characterize the response characteristics of the system under different boundary conditions.

[0012] Fault models are used to model wear-out and aging faults of the system or component under test.

[0013] The hardware interface is used to convert the response characteristics of the system or component under test under set conditions and fault conditions simulated by the dynamic model and fault model into electrical signals, and transmit the electrical signals to the edge computing system through the TCP / IP protocol.

[0014] In one embodiment, the edge computing system includes a hardware module composed of an embedded system or an edge computing component and a signal processing algorithm module running on the hardware module; the signal processing algorithm module includes a signal preprocessing submodule, a feature extraction submodule, and a feature data / file transmission submodule.

[0015] The signal preprocessing submodule is used to preprocess the collected electrical signals.

[0016] The feature extraction submodule is used to extract fault features from the time-frequency domain of the preprocessed signal.

[0017] The feature data / file transmission submodule is used to transmit the fault feature data extracted by the feature extraction module to the cloud-based intelligent fault warning system.

[0018] In one embodiment, the cloud-based intelligent fault warning system includes a fault diagnosis hardware module composed of a cloud computing platform or a high-performance server and a fault diagnosis algorithm module running on the fault diagnosis hardware module; the fault diagnosis algorithm module includes: a fault diagnosis and warning sub-module, a visualization processing sub-module, and a fault data / file transfer sub-module.

[0019] The fault diagnosis and early warning submodule is used to reason and analyze the fault feature data sent by the edge computing system using machine learning methods, deep learning methods, or a combination of machine learning and deep learning methods, to achieve fault classification, reliability assessment, and remaining life estimation of the system or component under test.

[0020] The visualization processing submodule is used to package the process results and final results of the fault diagnosis and early warning submodule into a visualized chart or file.

[0021] The fault data / file transfer submodule is used to transfer the visualized charts or files back to the edge computing system through the SFTP protocol.

[0022] The above-mentioned end-edge-cloud full-link intelligent fault diagnosis platform based on virtual bench, in the end-edge-cloud full-link intelligent fault diagnosis platform based on virtual bench, the platform includes: a virtual bench system, an edge computing system, and a cloud-based intelligent fault warning system, wherein the virtual bench system is used to perform dynamic modeling and fault type modeling on the system or component under test, and simulate the response characteristics of the system or component under set boundary conditions and fault states, and convert the response characteristics into electrical signals; the edge computing system is used to preprocess and extract features of the collected electrical signals, and transmit the extracted features to the cloud-based intelligent fault warning system; the cloud-based intelligent fault warning system is used to perform fault diagnosis, reliability analysis or remaining life estimation on the system or component under test based on the features extracted by the edge computing system. The platform can verify the end-edge-cloud full-link technology based on the virtual bench model, which can significantly reduce the experimental cost and improve the efficiency of the preliminary testing and deployment of the diagnostic algorithm; the platform adopts a modular structure, and the algorithms and models in the modules can be modified and replaced according to actual needs, and the communication methods between modules can also be configured as needed, with good scalability and adaptability. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 The present invention is a block diagram of a full-link intelligent fault diagnosis platform for end-edge-cloud based on a virtual test bench in one embodiment;

[0024] Figure 2 A schematic diagram of a hardware interface of a virtual gantry in one embodiment;

[0025] Figure 3 1 is a fault simulation result on a virtual test bench in one embodiment, wherein (a)-(d) represent the acceleration response of the bearing under normal, inner ring fault, outer ring fault, and ball fault conditions, respectively;

[0026] Figure 4 This is a fault classification result diagram based on the “end-edge-cloud” platform in another embodiment;

[0027] Figure 5 This is a diagram of the remaining life estimation result based on the “end-edge-cloud” platform in one embodiment; DETAILED DESCRIPTION

[0028] In order to make the purpose, technical solution 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. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0029] In one embodiment, Figure 1As shown, a full-link end-edge-cloud intelligent fault diagnosis platform based on a virtual test bench is provided, and the platform includes: a virtual test bench system 10, an edge computing system 20, and a cloud-based intelligent fault warning system 30.

[0030] The virtual bench system 10 is used to perform dynamic modeling and fault type modeling on the system or component under test, simulate the response characteristics of the system or component under test under set boundary conditions and fault states, and convert the response characteristics into electrical signals.

[0031] Specifically, in order to reduce the deployment and testing costs of the "end-edge-cloud" full-link platform and improve the testing efficiency of the edge-cloud collaborative algorithm, a virtual bench model for "end-edge-cloud" full-link intelligent diagnosis is proposed.

[0032] The virtual bench model consists of three parts: dynamic model, fault model, and hardware interface. The dynamic model is responsible for simulating the response characteristics of the system under normal conditions. The fault model is used to model the common faults and aging characteristics of the system. The hardware interface is responsible for converting the output generated by the simulation model into electrical signals for real-time acquisition by the edge computing system.

[0033] The virtual bench is used to simulate the sensor output of the system under normal and fault conditions.

[0034] The edge computing system 20 is used to preprocess and extract features of the collected electrical signals, and transmit the extracted features to the cloud-based intelligent fault warning system 30; it is also used to receive the results of fault diagnosis, reliability analysis or remaining life estimation transmitted by the cloud-based intelligent fault warning system 30, so that users can analyze the diagnosis results of the tested system or components at the edge, improve the edge data preprocessing, feature extraction methods and parameter configurations according to the diagnosis results, and further optimize the cloud-based diagnosis results.

[0035] The cloud-based intelligent fault warning system 30 is used to perform fault diagnosis, reliability analysis or remaining life estimation on the system or component under test based on the features extracted by the edge computing system 20, and transmit the results of the fault diagnosis, reliability analysis or remaining life estimation to the edge computing system 20.

[0036] In the above-mentioned end-edge-cloud full-link intelligent fault diagnosis platform based on virtual bench, the platform includes: a virtual bench system, an edge computing system, and a cloud-based intelligent fault warning system, wherein the virtual bench system is used to perform dynamic modeling and fault type modeling on the system or component under test, and simulate the response characteristics of the system or component under set boundary conditions and fault conditions, and convert the response characteristics into electrical signals; the edge computing system is used to preprocess and extract features of the collected electrical signals, and transmit the extracted features to the cloud-based intelligent fault warning system; the cloud-based intelligent fault warning system is used to perform fault diagnosis, reliability analysis or remaining life estimation on the system or component under test based on the features extracted by the edge computing system. The platform can verify the end-edge-cloud full-link technology based on the virtual bench model, which can significantly reduce the experimental cost and improve the efficiency of the preliminary testing and deployment of the diagnostic algorithm; the platform adopts a modular structure, and the algorithms and models in the modules can be modified and replaced according to actual needs, and the communication method between modules can also be configured as needed, with good scalability and adaptability.

[0037] In one embodiment, the virtual test bench system includes: a dynamic model, a fault model and a hardware interface; the dynamic model is used to establish the kinematic and dynamic equations of the system or component under test in each degree of freedom according to Newton's law; the kinematic and dynamic equations are used to characterize the response characteristics of the system under different boundary conditions; the fault model is used to model the wear faults and aging faults of the system or component under test; the hardware interface is used to convert the response characteristics of the system or component under test simulated by the dynamic model and the fault model under set conditions and fault conditions into electrical signals, and transmit the electrical signals to the edge computing system through the TCP / IP protocol.

[0038] Specifically, the construction and testing of the virtual bench includes system dynamics modeling, system fault modeling, hardware interface design, model integration and testing of the virtual bench, etc. In order to facilitate the explanation of the implementation plan, bearings are taken as an example to introduce the implementation steps in detail. Other systems or components under test can also refer to this method and process for implementation.

[0039] (1) System dynamics modeling

[0040] The system dynamics model is the kinematic and dynamic equations of the system in each degree of freedom, which is used to characterize the response characteristics of the system under different boundary conditions. In general, the kinematic and dynamic equations of the system in each degree of freedom are established based on Newton's law. In this embodiment, the 5-degree-of-freedom dynamic equation of the bearing is used as an example to illustrate the system dynamics modeling, where the 5 degrees of freedom include: the horizontal and vertical degrees of freedom of the inner and outer rings of the bearing, and the vertical degrees of freedom of the shock absorber. Based on Newton's second law, the following equation can be established:

[0041]

[0042] Where: m s , m p , m r are the masses of the bearing inner ring, bearing outer ring and shock absorber respectively; k s , k p , k r are the stiffness of the bearing inner ring, bearing outer ring and shock absorber respectively; c s , c p , c r are the damping of the bearing inner ring, bearing outer ring and shock absorber respectively; x s , x p ,y s ,y p are the displacements of the inner and outer rings of the bearing in the x and y directions respectively, b is the displacement of the shock absorber in the y direction; F x and F y are the external forces in the x and y directions respectively; f x , f y are the contact forces between the ball and the raceway in the x and y directions respectively.

[0043] The contact force f between the ball and the raceway in the x and y directions x 、f y The contact force f between the ball and the raceway in the x and y directions can be determined by the Herz contact theory. x 、f y for:

[0044]

[0045] Among them, k b is the stiffness of the bearing ball, δ j is the deformation of the jth ball, φ j is the angular position of the jth ball, γ j is a switching function, such as δ j >0,γ j =1, otherwise, γ j =0.

[0046] (2) System Fault Modeling

[0047] Common mechanical system fault types include wear faults and aging faults. The principles of wear fault modeling and aging fault modeling are introduced below.

[0048] 1) Wear failure modeling

[0049] From the above formulas (1) and (2), it can be seen that the Herz contact force between the ball and the raceway is mainly determined by the contact deformation, and the deformation of each ball can be calculated from the relative displacement of the inner and outer raceways at that position and the clearance of the bearing. The deformation of the jth ball is:

[0050] δ j =(x s -x p )cosφ j +9y s -y p )sinφ j -c (3)

[0051] When a bearing fails, the location, size, and number of the faults will further affect the deformation between the ball and the raceway. Therefore, the ball deformation of the faulty bearing can be modeled as:

[0052] δ′ j =(x s -x p )cosφ j +(y s -y p )sinφ j -c-β j c d (4)

[0053] Among them, c is the bearing clearance, β j Affected by the fault location, c d is affected by the size and shape of the fault. Therefore, by further j and c d The wear failure is modeled by modeling.

[0054] 2) Aging Failure Modeling

[0055] For aging failures of mechanical systems, a common modeling method is to use an exponential aging function. The specific method is shown in formula (5):

[0056]

[0057] Where h(t) is the health index of the bearing, θ and β are coefficients that characterize the aging rate, and ∈ represents the noise of the normal distribution, with a mean of 0 and a variance of σ. 2 , Based on this aging model and combined with the collected historical data, a model can be fitted for a specific bearing to achieve modeling of its aging characteristics.

[0058] (3) Hardware interface

[0059] The virtual bench system, which consists of a normal dynamics model and an aging fault model, can simulate the response characteristics of the system under set boundary conditions and fault conditions. In order to transmit the output signal of the virtual bench to the edge computing system, it is also necessary to add a hardware interface to the virtual bench system model. For the sake of convenience, this article takes the development of a virtual bench model using the current mainstream modeling software Matlab / Simulink and the construction of an edge computing system based on Raspberry Pi as examples to introduce the design process of the hardware interface.

[0060] like Figure 2 As shown in the figure, first, based on the signal type (such as AD, I / O) output by the virtual test bench, select the corresponding hardware interface module. For example, the Simulink module corresponding to the I / O signal acquisition of the Raspberry Pi 4B is GPIO. Secondly, configure the number of channels, sampling frequency, sampling accuracy and other parameters according to the requirements of the actual scene for signal acquisition. Similarly, you can configure the hardware interface of other signals that need to be collected.

[0061] (4) Virtual bench model integration and testing

[0062] After completing the virtual test bench modeling and hardware interface configuration, each hardware interface module can be connected to the corresponding output signal in the virtual test bench model to complete the integration of the virtual test bench model. Then, by setting the corresponding boundary conditions and fault parameters, run the integrated virtual test bench model and check whether the output signals of each hardware interface meet the requirements in terms of type, accuracy, frequency, etc.

[0063] In one embodiment, the edge computing system includes a hardware module composed of an embedded system or an edge computing component and a signal processing algorithm module running on the hardware module; the signal processing algorithm module includes a signal preprocessing submodule, a feature extraction submodule, and a feature data / file transmission submodule.

[0064] The signal preprocessing submodule is used to preprocess the collected electrical signals.

[0065] The feature extraction submodule is used to extract fault features from the time-frequency domain of the preprocessed signal.

[0066] The feature data / file transmission submodule is used to transmit the fault feature data extracted by the feature extraction module to the cloud-based intelligent fault warning system.

[0067] Specifically, the construction and testing of the edge computing system includes signal preprocessing module design, feature extraction module design, data / file transmission module design, edge computing system module integration and testing, etc. The implementation plan for each step will be introduced in detail below.

[0068] (1) Signal preprocessing module

[0069] The signal preprocessing module is mainly responsible for preprocessing the signals collected by the edge computing system, including signal filtering and noise reduction, singular value removal, sample segmentation and other operations. For each step, there are many mature methods in academia and industry. For example, filtering and noise reduction can use mean filtering, bandpass filtering, wavelet filtering, etc. Users only need to choose a suitable method from them. This is not the core content of this application and will not be repeated here.

[0070] (2) Feature extraction module

[0071] The feature extraction module is mainly responsible for extracting features from the time-frequency domain of the preprocessed signal to reduce the burden of data transmission from the edge to the cloud server, reduce latency, and improve efficiency. As for the feature extraction method, there are many widely used methods in the PHM field, such as the mean, standard deviation, spectrum, kurtosis and other indicators in the time-frequency domain, and the fault characteristic frequency, spectrum kurtosis, wavelet coefficients in the frequency domain. Users only need to select one or more appropriate methods from them. This is not the core content of this application and will not be repeated here.

[0072] (3) Data / file transfer module

[0073] In order to increase the rate of data transmission from the edge to the cloud, reduce time delays and data packet loss, this application uses the method of sending the extracted features rather than the original data collected by the edge system to the cloud server. The data transmission method uses the SFTP protocol. First, configure the IP address, port number, user name, password and other information of the target cloud platform in the edge computing system to establish a connection between the edge computing system and the cloud platform; then, send the extracted features to the cloud server in the form of a file through SFTP. The SFTP transmission protocol given here can be replaced with other protocols or tools that support edge-cloud communication as needed.

[0074] (4) Edge computing system module integration and testing

[0075] The signal preprocessing module, feature extraction module, and data / file transfer module are integrated, and the main function is defined to call all functional modules. By defining the main function as a periodic execution task, such as defining the execution cycle as 2ms, the entire process from signal processing to edge-cloud data transmission can be automatically processed. Finally, by sending fault data from the virtual bench to the edge computing system, the signal preprocessing, feature extraction, data / file transfer and other functions of the edge computing system are tested and verified to ensure that the data received by the cloud server meets expectations.

[0076] In one embodiment, the cloud-based intelligent fault warning system includes a fault diagnosis hardware module composed of a cloud computing platform or a high-performance server and a fault diagnosis algorithm module running on the fault diagnosis hardware module; the fault diagnosis algorithm module includes: a fault diagnosis and warning sub-module, a visualization processing sub-module, and a fault data / file transfer sub-module.

[0077] The fault diagnosis and early warning submodule is used to reason and analyze the fault feature data sent by the edge computing system using machine learning methods, deep learning methods, or a combination of machine learning and deep learning methods, to achieve fault classification, reliability assessment, and remaining life estimation of the system or component under test.

[0078] The visualization processing submodule is used to package the process results and final results of the fault diagnosis and early warning submodule into a visualized chart or file.

[0079] The fault data / file transfer submodule is used to transfer the visualized charts or files back to the edge computing system through the SFTP protocol.

[0080] Specifically, the construction and testing of the cloud-based intelligent fault warning platform includes four steps: fault diagnosis and warning module design, visualization processing module design, data / file transfer module design, and cloud-based intelligent fault warning platform integration and testing. The implementation plan for each step will be introduced in detail below.

[0081] (1) Fault diagnosis and warning module

[0082] The fault diagnosis and warning module is the core of the cloud-based intelligent fault warning platform. It is mainly responsible for further reasoning and analysis of the fault feature data sent by the edge computing system to the cloud server to achieve fault classification, reliability assessment, remaining life estimation and other tasks. Among them, the algorithm can adopt commonly used machine learning algorithms, such as support vector machine SVM, decision tree DT, random forest RF, etc.; it can also adopt deep learning algorithms, such as deep convolutional neural network CNN, long short-term memory network LSTM, deep autoencoder network and other DAE; or a combination of the two, such as SVM+CNN, CNN+LSTM, etc., as well as various advanced fault diagnosis algorithms, such as physical information network PINN, transformer, etc. These algorithms themselves are not the content of this application, so they will not be repeated here. Users can choose the appropriate algorithm combination according to actual diagnostic needs.

[0083] (2) Visualization processing module

[0084] The visualization processing module is mainly responsible for packaging the diagnostic process and final results into visual charts or files. Among them, the process data includes: the model used and the configured hyperparameters, the composition of the training set and the test set, the change diagram of the accuracy and loss during the model training process, and other intermediate results related to the method (such as the feature map of CNN, the intermediate variable distribution of DAE, etc.); the final result data includes: the accuracy of fault classification and the confusion matrix under different labels, the accuracy and comparison diagram of the remaining life estimation, the uncertainty boundary of the model, etc. On the one hand, the visualization processing module is responsible for collecting the above results and data and displaying them on the interface of the cloud server; on the other hand, it is responsible for packaging all results into compressed files so that the diagnostic results can be returned to the edge computing system.

[0085] (3) Data / file transfer module

[0086] The data / file transfer module of the cloud-based intelligent fault warning platform is mainly responsible for sending the packaged files organized in step 3.2 above back to the edge computing system through the SFTP protocol, so that users can analyze the diagnostic results of the system under test at the edge, thereby improving the methods and parameter configurations of edge-side data preprocessing, feature extraction and other steps based on the diagnostic results, and further optimizing the diagnostic results of the cloud. This data / file transfer method for edge-cloud intercommunication can effectively improve the effect of edge-cloud collaborative fault diagnosis and warning.

[0087] (4) Integration and testing of cloud-based intelligent fault warning platform

[0088] The fault diagnosis and warning module, visualization processing module, and data / file transfer module are integrated and tested, and all modules are encapsulated in a main function that is executed in a cycle, so as to realize the automatic processing of the whole process from fault diagnosis and warning to visualization processing of results and data / file transfer. Then, by sending fault data from the edge computing system to the cloud server, the functions of the cloud intelligent fault warning platform are tested to see if they are normal.

[0089] In a verification implementation example, taking bearings as an example, a 5-DoF dynamic model and fault model of the bearings were constructed in Matlab / Simulink, and a GPIO interface for signal acquisition of the Raspberry Pi edge computing system was developed, and a complete virtual bench model was built. At the same time, in the edge computing system with Raspberry Pi as the hardware, modules such as mean filtering and time-frequency domain feature extraction were deployed. Finally, the fault features and their corresponding labels were sent to the Alibaba Cloud platform via SFTP, and two diagnostic models, support vector machines and deep convolutional neural networks (CNNs), were deployed on the cloud platform, thus completing the "end-edge-cloud" full-link intelligent diagnosis platform based on the virtual bench, and realizing the testing and verification of edge-cloud collaborative communication and intelligent diagnosis algorithms in a virtual simulation environment. Figure 3The acceleration signals under different fault conditions generated by the virtual bearing test bench are given, where (a)-(d) represent the acceleration responses of the bearing under normal, inner ring fault, outer ring fault, and ball fault conditions, respectively. Figure 4 The confusion matrix diagram of fault classification is given, where the subscripts "B", "IR", and "OR" in the labels represent the locations of the faults as ball bearing, inner race, and outer race, respectively; "7", "14", and "21" represent the sizes of the faults as 0.007 inches, 0.014 inches, and 0.021 inches; and "N_0" indicates the normal state. Figure 5 The results of the remaining life estimation are given. It can be seen that the classification accuracy under different fault types is above 99%, and the root mean square error of RUL estimation is also small. This proves the effectiveness of the entire "end-edge-cloud" platform architecture.

[0090] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0091] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. An intelligent fault diagnosis platform for the entire end-edge-cloud link based on a virtual test bench, characterized in that: The platform includes: a virtual bench system, an edge computing system, and a cloud-based intelligent fault warning system; The virtual bench system is used to perform dynamic modeling and fault type modeling on the system or component under test, and simulate the response characteristics of the system or component under test under set boundary conditions and fault conditions, and convert the response characteristics into electrical signals; The edge computing system is used to preprocess and extract features of the collected electrical signals, and transmit the extracted features to the cloud-based intelligent fault warning system; it is also used to receive the results of fault diagnosis, reliability analysis or remaining life estimation transmitted by the cloud-based intelligent fault warning system, so that users can analyze the diagnosis results of the tested system or components at the edge, improve the edge-side data preprocessing, feature extraction methods and parameter configurations according to the diagnosis results, and further optimize the cloud-based diagnosis results; The cloud-based intelligent fault warning system is used to perform fault diagnosis, reliability analysis or remaining life estimation on the system or component under test based on the features extracted by the edge computing system, and transmit the results of the fault diagnosis, reliability analysis or remaining life estimation to the edge computing system.

2. The platform according to claim 1, characterized in that The virtual test bench system includes: a dynamic model, a fault model and a hardware interface; The dynamic model is used to establish the kinematic and dynamic equations of the system or component under test in each degree of freedom according to Newton's law; the kinematic and dynamic equations are used to characterize the response characteristics of the system under different boundary conditions; The fault model is used to model wear faults and aging faults of the system or component under test; The hardware interface is used to convert the response characteristics of the system or component under test simulated by the dynamic model and the fault model under set conditions and fault conditions into electrical signals, and transmit the electrical signals to the edge computing system through the TCP / IP protocol.

3. The platform according to claim 1, characterized in that: The edge computing system includes a hardware module composed of an embedded system or an edge computing component and a signal processing algorithm module running on the hardware module; the signal processing algorithm module includes a signal preprocessing submodule, a feature extraction submodule, and a feature data / file transmission submodule; The signal preprocessing submodule is used to preprocess the collected electrical signals; The feature extraction submodule is used to extract fault features from the time-frequency domain of the preprocessed signal; The feature data / file transmission submodule is used to transmit the fault feature data extracted by the feature extraction module to the cloud-based intelligent fault warning system.

4. The platform according to claim 1, characterized in that: The cloud-based intelligent fault warning system includes a fault diagnosis hardware module composed of a cloud computing platform or a high-performance server and a fault diagnosis algorithm module running on the fault diagnosis hardware module; The fault diagnosis algorithm module includes: a fault diagnosis and early warning submodule, a visualization processing submodule, and a fault data / file transmission submodule; The fault diagnosis and early warning submodule is used to use machine learning methods, deep learning methods, or a combination of machine learning and deep learning methods to reason and analyze the fault feature data sent by the edge computing system, so as to achieve fault classification, reliability assessment, and remaining life estimation of the system or component under test; The visualization processing submodule is used to package the process results and final results of the fault diagnosis of the fault diagnosis and early warning submodule into a visualized chart or file; The fault data / file transmission submodule is used to transmit the visualized chart or file back to the edge computing system via the SFTP protocol.

Images