Equipment fault diagnosis model establishment method and equipment fault diagnosis method

Abstract

The present application relates to a method, apparatus, computer equipment, storage medium and computer program product for establishing a device fault diagnosis model. The method comprises: obtaining a device digital twin model of a device fault injection test bench, and an initial device fault diagnosis model based on a convolutional neural network; obtaining device twin simulation data after different fault injections under different working conditions, and obtaining corresponding real physical data of the device; constructing a fault diagnosis data set; obtaining a time-frequency spectrum of the fault diagnosis data set; obtaining an ideal label corresponding to a preset time-frequency spectrum; inputting the time-frequency spectrum and the ideal label into an initial device fault diagnosis model based on a convolutional neural network for learning and training, and obtaining a target device fault diagnosis model. This method can accurately perform device fault diagnosis; in addition, the present application also provides an accurate device fault diagnosis method, apparatus, computer equipment, computer-readable storage medium and computer program product.

Description

Technical Field

[0001] The present application relates to the field of Internet of Things technology, and in particular to a method, apparatus, computer equipment, storage medium and computer program product for establishing a device fault diagnosis model, as well as a method, apparatus, computer equipment, storage medium and computer program product for device fault diagnosis. Background Art

[0002] With the development of Internet of Things information technology, the importance of diagnosing equipment faults has become increasingly prominent. Take the gearbox as an example. As one of the most critical components in rotating machinery, it plays an important role in regulating the input and output speed and torque. Fault diagnosis of it can improve the manufacturing and production quality of the entire set of equipment.

[0003] Currently, the means of analyzing equipment failures generally involves using signal analysis or machine learning to generate a fault diagnosis model, which is then used to diagnose the equipment failure. However, although equipment fault diagnosis can be achieved with the help of signal analysis or machine learning methods, the fault diagnosis models generated by these methods are of low accuracy and are insufficient for accurate fault diagnosis. Summary of the Invention

[0004] Based on this, it is necessary to provide an accurate equipment fault diagnosis model establishment method, device, computer equipment, computer-readable storage medium and computer program product to address the above technical problems; in addition, an accurate equipment fault diagnosis method, device, computer equipment, computer-readable storage medium and computer program product are also provided.

[0005] In a first aspect, the present application provides a method for establishing a device fault diagnosis model. The method comprises:

[0006] Obtain a digital twin model of the device fault injection testbed and an initial device fault diagnosis model based on a convolutional neural network;

[0007] According to the device digital twin model, different faults are injected into the device under different working conditions to obtain device twin simulation data after the different faults are injected under different working conditions, and the device real physical data corresponding to the different faults injected into the device under different working conditions by the device fault injection test bench is obtained;

[0008] Combining the real physical data of the device and the twin simulation data of the device to construct a fault diagnosis data set;

[0009] Performing wavelet analysis on the fault diagnosis data set to obtain a time-frequency spectrum of the fault diagnosis data set;

[0010] Get the ideal label corresponding to the preset time-frequency spectrum;

[0011] The time-frequency spectrum and the ideal labels corresponding to the preset time-frequency spectrum are input into the initial equipment fault diagnosis model based on the convolutional neural network for learning and training to obtain a target equipment fault diagnosis model.

[0012] In one embodiment, obtaining a device digital twin model of a device fault injection test bench includes:

[0013] Obtain the geometric parameters, material parameters and kinematic relationships of the equipment fault injection test bench;

[0014] A device digital twin model of a device fault injection test bench is established based on the geometric parameters, the material parameters, and the kinematic relationship.

[0015] In one embodiment, according to the device digital twin model, different faults are injected into the device under different working conditions, and the device twin simulation data after the different faults are injected under different working conditions are obtained, and the real physical data of the device corresponding to the different faults injected into the device under different working conditions by the device fault injection test bench are obtained. The data includes:

[0016] According to the device digital twin model, different faults are injected into the device under different working conditions to obtain initial twin simulation data of the device after the different faults are injected under different working conditions;

[0017] Obtain the real physical data of the equipment corresponding to the different fault injections performed on the equipment under different working conditions by the equipment fault injection test bench;

[0018] Compare the real physical data of the device and the initial twin simulation data of the device to obtain the device twin simulation data.

[0019] In one embodiment, comparing the real physical data of the device with the initial twin simulation data of the device to obtain the device twin simulation data includes:

[0020] Get the preset gap threshold;

[0021] Comparing the real physical data of the device with the initial twin simulation data of the device to obtain a data difference between the real physical data of the device and the initial twin simulation data of the device;

[0022] When the data difference exceeds a preset gap threshold, the initial twin simulation data of the device is updated according to the real physical data of the device to obtain the device twin simulation data.

[0023] In one embodiment, inputting the time-frequency spectrum and the ideal labels corresponding to the preset time-frequency spectrum into the initial equipment fault diagnosis model based on the convolutional neural network for learning and training to obtain the target equipment fault diagnosis model includes:

[0024] Obtain activation functions for the initial equipment fault diagnosis model based on convolutional neural networks;

[0025] Inputting the time-frequency spectrum into the convolutional layer of the initial equipment fault diagnosis model based on the convolutional neural network, and obtaining linearly separable spectrum feature information according to the activation function;

[0026] Inputting the linearly separable graph feature information into the pooling layer of the initial equipment fault diagnosis model based on the convolutional neural network, performing pooling processing on the linearly separable graph feature information, and obtaining pooled graph feature information;

[0027] According to the pooled graph feature information and the ideal label corresponding to the preset time-frequency graph, the initial equipment fault diagnosis model based on the convolutional neural network is learned and trained to obtain a target equipment fault diagnosis model.

[0028] In a second aspect, the present application also provides a device for establishing a device fault diagnosis model. The device comprises:

[0029] An initial model acquisition module, used to obtain the device digital twin model of the device fault injection test bench and the initial device fault diagnosis model based on convolutional neural networks;

[0030] A data acquisition module is used to inject different faults into the device under different working conditions according to the device digital twin model, obtain device twin simulation data after the different faults are injected under different working conditions, and obtain the real physical data of the device corresponding to the different faults injected into the device under different working conditions by the device fault injection test bench;

[0031] A data set construction module, configured to construct a fault diagnosis data set by combining the real physical data of the device and the twin simulation data of the device;

[0032] A wavelet analysis module is used to perform wavelet analysis on the fault diagnosis data set to obtain a time-frequency spectrum of the fault diagnosis data set;

[0033] The label acquisition module is used to obtain the ideal label corresponding to the preset time-frequency spectrum;

[0034] The target model establishment module is used to input the time-frequency spectrum and the ideal labels corresponding to the preset time-frequency spectrum into the equipment fault diagnosis model based on the convolutional neural network for learning and training to obtain a target equipment fault diagnosis model.

[0035] In a third aspect, the present application further provides a computer device. The computer device includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are performed:

[0036] Obtain a digital twin model of the device fault injection testbed and an initial device fault diagnosis model based on a convolutional neural network;

[0037] According to the device digital twin model, different faults are injected into the device under different working conditions to obtain device twin simulation data after the different faults are injected under different working conditions, and the device real physical data corresponding to the different faults injected into the device under different working conditions by the device fault injection test bench is obtained;

[0038] Combining the real physical data of the device and the twin simulation data of the device to construct a fault diagnosis data set;

[0039] Performing wavelet analysis on the fault diagnosis data set to obtain a time-frequency spectrum of the fault diagnosis data set;

[0040] Get the ideal label corresponding to the preset time-frequency spectrum;

[0041] The time-frequency spectrum and the ideal labels corresponding to the preset time-frequency spectrum are input into the initial equipment fault diagnosis model based on the convolutional neural network for learning and training to obtain a target equipment fault diagnosis model.

[0042] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the following steps:

[0043] Obtain a digital twin model of the device fault injection testbed and an initial device fault diagnosis model based on a convolutional neural network;

[0044] According to the device digital twin model, different faults are injected into the device under different working conditions to obtain device twin simulation data after the different faults are injected under different working conditions, and the device real physical data corresponding to the different faults injected into the device under different working conditions by the device fault injection test bench is obtained;

[0045] Combining the real physical data of the device and the twin simulation data of the device to construct a fault diagnosis data set;

[0046] Performing wavelet analysis on the fault diagnosis data set to obtain a time-frequency spectrum of the fault diagnosis data set;

[0047] Get the ideal label corresponding to the preset time-frequency spectrum;

[0048] The time-frequency spectrum and the ideal labels corresponding to the preset time-frequency spectrum are input into the initial equipment fault diagnosis model based on the convolutional neural network for learning and training to obtain a target equipment fault diagnosis model.

[0049] In a fifth aspect, the present application further provides a computer program product. The computer program product includes a computer program that, when executed by a processor, implements the following steps:

[0050] Obtain a digital twin model of the device fault injection testbed and an initial device fault diagnosis model based on a convolutional neural network;

[0051] According to the device digital twin model, different faults are injected into the device under different working conditions to obtain device twin simulation data after the different faults are injected under different working conditions, and the device real physical data corresponding to the different faults injected into the device under different working conditions by the device fault injection test bench is obtained;

[0052] Combining the real physical data of the device and the twin simulation data of the device to construct a fault diagnosis data set;

[0053] Performing wavelet analysis on the fault diagnosis data set to obtain a time-frequency spectrum of the fault diagnosis data set;

[0054] Get the ideal label corresponding to the preset time-frequency spectrum;

[0055] The time-frequency spectrum and the ideal labels corresponding to the preset time-frequency spectrum are input into the initial equipment fault diagnosis model based on the convolutional neural network for learning and training to obtain a target equipment fault diagnosis model.

[0056] The above-mentioned equipment fault diagnosis model establishment method, device, computer equipment, storage medium and computer program product obtain the equipment digital twin model of the equipment fault injection test bench and the initial equipment fault diagnosis model based on convolutional neural network; according to the equipment digital twin model, different faults are injected into the equipment under different working conditions to obtain the equipment twin simulation data after the different faults are injected under different working conditions, and the equipment real physical data corresponding to the different faults injected into the equipment under different working conditions by the equipment fault injection test bench are obtained; combining the real physical data of the equipment and the equipment twin simulation data, a fault diagnosis data set is constructed; the fault diagnosis data set is subjected to wavelet analysis to obtain the time-frequency spectrum of the fault diagnosis data set; the ideal label corresponding to the preset time-frequency spectrum is obtained; the time-frequency spectrum and the ideal label corresponding to the preset time-frequency spectrum are input into the initial equipment fault diagnosis model based on convolutional neural network for learning and training to obtain the target equipment fault diagnosis model. In this process, not only real data is input into the initial equipment fault diagnosis model for learning, but also the real physical data of the equipment and the equipment twin simulation data are obtained through the equipment fault injection test bench and the equipment digital twin model. The real data and simulation data are combined to construct a fault diagnosis data set for wavelet analysis and then input into the equipment fault diagnosis model based on convolutional neural network for learning. This improves the dimensionality and accuracy of the equipment fault diagnosis model based on convolutional neural network, ensuring more accurate fault diagnosis.

[0057] In a sixth aspect, the present application provides a device fault diagnosis method. The method comprises:

[0058] Obtain real physical data of the equipment to be diagnosed;

[0059] Performing wavelet analysis on the real physical data of the device to be diagnosed to obtain a time-frequency spectrum of the real physical data of the device to be diagnosed;

[0060] Inputting the time-frequency spectrum of the real physical data of the device to be diagnosed into the target device fault diagnosis model to perform fault diagnosis and obtain the type of device fault;

[0061] The target device fault diagnosis model is established using the above-mentioned device fault diagnosis model establishment method.

[0062] In a seventh aspect, the present application provides a device for diagnosing equipment faults. The device comprises:

[0063] Data acquisition module, used to obtain real physical data of the device to be diagnosed;

[0064] A spectrum acquisition module, configured to perform wavelet analysis on the real physical data of the device to be diagnosed to obtain a time-frequency spectrum of the real physical data of the device to be diagnosed;

[0065] A fault diagnosis module is used to input the time-frequency spectrum of the real physical data of the device to be diagnosed into the target device fault diagnosis model to perform fault diagnosis and obtain the type of device fault;

[0066] The target device fault diagnosis model is established using the above-mentioned device fault diagnosis model establishment method.

[0067] In an eighth aspect, the present application further provides a computer device. The computer device includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0068] Obtain real physical data of the equipment to be diagnosed;

[0069] Performing wavelet analysis on the real physical data of the device to be diagnosed to obtain a time-frequency spectrum of the real physical data of the device to be diagnosed;

[0070] Inputting the time-frequency spectrum of the real physical data of the device to be diagnosed into the target device fault diagnosis model to perform fault diagnosis and obtain the type of device fault;

[0071] The target device fault diagnosis model is established using the above-mentioned device fault diagnosis model establishment method.

[0072] In a ninth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the following steps:

[0073] Obtain real physical data of the equipment to be diagnosed;

[0074] Performing wavelet analysis on the real physical data of the device to be diagnosed to obtain a time-frequency spectrum of the real physical data of the device to be diagnosed;

[0075] Inputting the time-frequency spectrum of the real physical data of the device to be diagnosed into the target device fault diagnosis model to perform fault diagnosis and obtain the type of device fault;

[0076] The target device fault diagnosis model is established using the above-mentioned device fault diagnosis model establishment method.

[0077] In a tenth aspect, the present application further provides a computer program product. The computer program product includes a computer program that, when executed by a processor, implements the following steps:

[0078] Obtain real physical data of the equipment to be diagnosed;

[0079] Performing wavelet analysis on the real physical data of the device to be diagnosed to obtain a time-frequency spectrum of the real physical data of the device to be diagnosed;

[0080] Inputting the time-frequency spectrum of the real physical data of the device to be diagnosed into the target device fault diagnosis model to perform fault diagnosis and obtain the type of device fault;

[0081] The target device fault diagnosis model is established using the above-mentioned device fault diagnosis model establishment method.

[0082] The aforementioned equipment fault diagnosis method, apparatus, computer device, storage medium, and computer program product obtain actual physical data of the equipment to be diagnosed; perform wavelet analysis on the actual physical data of the equipment to be diagnosed to obtain a time-frequency spectrum of the actual physical data of the equipment to be diagnosed; and input the time-frequency spectrum of the actual physical data of the equipment to be diagnosed into a target equipment fault diagnosis model for fault diagnosis to determine the type of equipment fault. The target equipment fault diagnosis model is established using the aforementioned equipment fault diagnosis model establishment method. By inputting the actual physical data of the equipment to be diagnosed into an accurate target equipment fault diagnosis model, accurate fault diagnosis of the equipment to be diagnosed is ensured. BRIEF DESCRIPTION OF THE DRAWINGS

[0083] Figure 1 A diagram showing an application environment of a method for establishing a device fault diagnosis model and a method for diagnosing device faults in one embodiment;

[0084] Figure 2 A schematic flow chart of a method for establishing a device fault diagnosis model in one embodiment;

[0085] Figure 3 This is a diagram of the equipment fault injection test bench structure for a gearbox in a specific application example;

[0086] Figure 4 A flowchart for establishing a device twin simulation database in another specific application example;

[0087] Figure 5 A schematic flow chart of a method for establishing a device fault diagnosis model in another embodiment;

[0088] Figure 6 This is a step diagram of the input layer, convolution layer, pooling layer, and output layer in a convolutional neural network in another specific application example;

[0089] Figure 7 A flowchart for establishing a target device fault diagnosis model and performing fault diagnosis using the target device fault diagnosis model in another specific application example;

[0090] Figure 8 1 is a flow chart of a device fault diagnosis method according to an embodiment;

[0091] Figure 9A structural block diagram of an apparatus for establishing a device fault diagnosis model in one embodiment;

[0092] Figure 10 is a structural block diagram of a device fault diagnosis apparatus in one embodiment;

[0093] Figure 11 FIG. 1 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0094] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0095] The device fault diagnosis model establishment method and device fault diagnosis method provided in the embodiments of the present application can be applied to Figure 1In the application environment shown, the terminal 102 communicates with the server 104 via a network. The data storage system can store data that the server 104 needs to process. The data storage system can be integrated on the server 104 or placed on the cloud or other network servers. Terminal 102 sends a device fault diagnosis model establishment request to server 104. The device fault diagnosis model establishment request carries a device digital twin model of a device fault injection test bench and an initial device fault diagnosis model based on a convolutional neural network. Server 104 receives the device fault diagnosis model establishment request sent by terminal 102, extracts the device digital twin model of the device fault injection test bench and the initial device fault diagnosis model based on a convolutional neural network in the device fault diagnosis model establishment request; server 104 injects different faults into the device under different working conditions according to the device digital twin model, obtains device twin simulation data after the different faults are injected under different working conditions, and obtains the real physical data of the device corresponding to the different faults injected into the device under different working conditions by the device fault injection test bench; constructs a fault diagnosis data set by combining the real physical data of the device and the device twin simulation data; performs wavelet analysis on the fault diagnosis data set to obtain a time-frequency spectrum of the fault diagnosis data set; obtains an ideal label corresponding to a preset time-frequency spectrum; inputs the time-frequency spectrum and the ideal label corresponding to the preset time-frequency spectrum into the initial device fault diagnosis model based on the convolutional neural network for learning and training to obtain a target device fault diagnosis model. Furthermore, upon receiving a device fault diagnosis request, server 104 extracts the actual physical data of the device to be diagnosed included in the request; performs wavelet analysis on the actual physical data of the device to be diagnosed to obtain a time-frequency spectrum of the actual physical data of the device to be diagnosed; and inputs the time-frequency spectrum of the actual physical data of the device to be diagnosed into a target device fault diagnosis model to perform fault diagnosis and determine the type of device fault. Furthermore, server 104 transmits the type of device fault to terminal 102. Terminal 102 may include, but is not limited to, various personal computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices may include smart speakers, smart TVs, smart air conditioners, smart in-vehicle devices, and the like. Portable wearable devices may include smart watches, smart bracelets, head-mounted devices, and the like. Server 104 may be implemented as a standalone server or as a server cluster consisting of multiple servers.

[0096] In one embodiment, Figure 2 As shown in the figure, a method for establishing a device fault diagnosis model is provided, which is applied to Figure 1 Taking the server 104 in the example as an example, the following steps are included:

[0097] S100: Obtain a device digital twin model of the device fault injection test bench and an initial device fault diagnosis model based on a convolutional neural network.

[0098] Among them, the equipment fault injection test bench is a test bench built in the real physical world, which can obtain test results by injecting different faults; digital twins are a technology that makes full use of physical models, sensor updates, operation history and other data, integrates multi-disciplinary, multi-physical quantity, multi-scale, and multi-probability simulation processes, and completes mapping in virtual space, thereby reflecting the full life cycle process of the corresponding physical equipment; the equipment digital twin model refers to the equipment digital twin model obtained by mapping the equipment fault injection test bench in the physical world using digital twin technology; the convolutional neural network is a type of feedforward neural network that includes convolution calculations and has a deep structure. It has representational learning capabilities and can perform translation-invariant classification of input information according to its hierarchical structure. It includes a multi-layer network structure. In each layer of the network, there are several two-dimensional planes, and in each plane there are several independent neurons.

[0099] Specifically, taking the gearbox as an example, the gearbox equipment fault injection test bench includes the equipment part to be diagnosed, the power part and the optional part. The structure diagram is as follows: Figure 3 As shown, the device under test consists of a two-stage planetary gearbox and a two-stage parallel gearbox. The power unit consists of a motor and a controller. The optional unit is an optional device consisting of a magnetic powder brake and a torque assembly. Furthermore, the device under test can be configured into various transmission combinations, such as configuring the planetary gearbox as a single stage, configuring the parallel gearbox as a single stage, and swapping the large and small gears of the parallel gearbox to achieve different transmission ratios (speed increase or speed decrease). After establishing the gearbox device fault injection test bench, a 1:1 three-dimensional model of the gearbox device fault injection test bench is created to construct a device digital twin model of the gearbox device fault injection test bench. Furthermore, an initial device fault diagnosis model based on a convolutional neural network must be obtained. At this point, the terminal sends a device fault diagnosis model establishment request to the server. The device fault diagnosis model establishment request carries the device digital twin model of the device fault injection test bench and the initial device fault diagnosis model based on a convolutional neural network. The server receives the device fault diagnosis model establishment request sent by the terminal and extracts the device digital twin model of the device fault injection test bench and the initial device fault diagnosis model based on a convolutional neural network from the device fault diagnosis model establishment request.

[0100] S200, according to the device digital twin model, different faults are injected into the device under different working conditions, and the device twin simulation data after the different faults are injected under different working conditions are obtained, and the device real physical data corresponding to the different faults injected into the device under different working conditions by the device fault injection test bench is obtained.

[0101] Among them, different working conditions can be the working conditions of the part of the equipment to be tested in the digital twin model of the equipment at different speeds and loads; taking the digital twin model of the equipment of the gearbox as an example, different faults can include standard fault gear and bearing faults; standard fault gears include missing teeth, broken teeth, tooth root cracks, wear and eccentricity faults, and helical gear kits and faulty parts are optional; bearing faults include inner ring faults, outer ring faults, rolling element faults and combined faults, and planetary bearings or parallel bearings are optional; fault injection is a reliability verification technology that deliberately introduces faults into the system through controlled experiments and observes the behavior when the fault exists in the system.

[0102] Specifically, the equipment fault injection test bench injects different faults into the equipment under different working conditions, obtains the real physical data of the equipment corresponding to the different fault injections under different working conditions, and sends the real physical data of the equipment to the server; the server obtains the real physical data of the equipment corresponding to the different fault injections into the equipment under different working conditions by the equipment fault injection test bench, and according to the equipment digital twin model of the equipment fault injection test bench, injects different faults into the equipment in the equipment digital twin model under different working conditions, and obtains the equipment twin simulation data after the injection of different faults under different working conditions.

[0103] Furthermore, the obtained device twin simulation data will be stored in the device twin simulation database. The flowchart of establishing the device twin simulation database is as follows: Figure 4 As shown, the constraints refer to constraints such as kinematic relationships and geometric parameters.

[0104] S300 combines the real physical data of the equipment and the equipment twin simulation data to build a fault diagnosis dataset.

[0105] Specifically, the real physical data of the device is the real physical data obtained after the fault injection test on the device fault injection test bench in the physical world. The device twin simulation data is the twin simulation data obtained by constructing the device digital twin model and testing the device digital twin model in the space where the device digital twin model is located. The server combines the real physical data of the device and the device twin simulation data to obtain a fault diagnosis data set, that is, the fault diagnosis data set includes the real physical data of the device and the device twin simulation data.

[0106] S400 , performing wavelet analysis on the fault diagnosis data set to obtain a time-frequency spectrum of the fault diagnosis data set.

[0107] Among them, wavelet analysis is a time-frequency localization analysis method with a fixed window size, changeable shape, and changeable time window and frequency window. It has higher frequency resolution and lower time resolution in the low-frequency part, and higher time resolution and lower frequency resolution in the high-frequency part, which conforms to the characteristics that low-frequency signals change slowly and high-frequency signals change rapidly. It actually maps a one-dimensional time signal to a two-dimensional time scale in order to clearly see the changes in signal frequency in a small time.

[0108] Specifically, the server performs wavelet analysis on the fault diagnosis dataset to extract the time-frequency spectrum of the device's real physical data and the device's twin simulation data, thereby obtaining the time-frequency spectrum of the fault diagnosis dataset.

[0109] S500: Obtain an ideal label corresponding to a preset time-frequency spectrum.

[0110] The ideal label corresponding to the preset time-frequency spectrum refers to the ideal fault type label corresponding to the preset time-frequency spectrum of the fault diagnosis dataset.

[0111] Specifically, after obtaining the time-frequency spectrum of the fault diagnosis dataset, the server also needs to obtain the ideal fault type label corresponding to the preset time-frequency spectrum.

[0112] S600: Input the time-frequency spectrum and the ideal label corresponding to the preset time-frequency spectrum into an initial equipment fault diagnosis model based on a convolutional neural network for learning and training to obtain a target equipment fault diagnosis model.

[0113] Specifically, the time-frequency spectrum and the ideal labels corresponding to the preset time-frequency spectrum are input into the initial equipment fault diagnosis model based on the convolutional neural network for learning and training, so that the initial equipment fault diagnosis model based on the convolutional neural network learns the time-frequency spectrum of the test data and the fault type labels corresponding to the time-frequency spectrum of the test data, thereby obtaining the target equipment fault diagnosis model. The target equipment fault diagnosis model can obtain the fault type corresponding to the time-frequency spectrum of the test data of the equipment to be diagnosed through previous learning when the time-frequency spectrum of the test data of the equipment to be diagnosed is input next time.

[0114] The above-mentioned method for establishing an equipment fault diagnosis model obtains an equipment digital twin model of an equipment fault injection test bench and an initial equipment fault diagnosis model based on a convolutional neural network; according to the equipment digital twin model, different faults are injected into the equipment under different working conditions to obtain equipment twin simulation data after different faults are injected under different working conditions, and the equipment real physical data corresponding to the different faults injected into the equipment under different working conditions by the equipment fault injection test bench are obtained; a fault diagnosis data set is constructed by combining the real physical data of the equipment and the equipment twin simulation data; the fault diagnosis data set is subjected to wavelet analysis to obtain a time-frequency spectrum of the fault diagnosis data set; the ideal label corresponding to the preset time-frequency spectrum is obtained; the time-frequency spectrum and the ideal label corresponding to the preset time-frequency spectrum are input into the initial equipment fault diagnosis model based on the convolutional neural network for learning and training to obtain a target equipment fault diagnosis model. In this process, not only real data is input into the initial equipment fault diagnosis model for learning, but also the real physical data of the equipment and the equipment twin simulation data are obtained through the equipment fault injection test bench and the equipment digital twin model. The real data and simulation data are combined to construct a fault diagnosis data set for wavelet analysis and then input into the equipment fault diagnosis model based on convolutional neural network for learning. This improves the dimensionality and accuracy of the equipment fault diagnosis model based on convolutional neural network, ensuring more accurate fault diagnosis.

[0115] In one embodiment, obtaining a device digital twin model of a device fault injection testbed includes:

[0116] Obtain the geometric parameters, material parameters, and kinematic relationships of the equipment fault injection test bench; and establish a device digital twin model of the equipment fault injection test bench based on the geometric parameters, material parameters, and kinematic relationships.

[0117] Among them, geometric parameters generally include the size, geometric shape (macro and micro), and the relative positional relationships between points, lines, and surfaces of the equipment fault injection test bench; material parameters refer to the parameters of the materials used in the equipment; kinematic relationships refer to the relationship between the position, velocity, angular velocity, acceleration, angular acceleration and other characteristics of the research object.

[0118] Specifically, since the device digital twin model is a twin model obtained by 1:1 mapping of the device fault injection test bench, in order to achieve 1:1 mapping, it is necessary to obtain the geometric parameters, material parameters and kinematic relationships of the device fault injection test bench, that is, to obtain the size of the device fault injection test bench, geometric shape (macro and micro), the relative position relationship between points, lines and surfaces, the parameters of the materials used in the device fault injection test bench, and the relationship between the position, velocity, angular velocity, acceleration, angular acceleration and other characteristics between all devices in the device fault injection test bench; based on the geometric parameters, material parameters and kinematic relationships, a 1:1 mapping is performed to establish the device digital twin model of the device fault injection test bench.

[0119] In this embodiment, by obtaining the geometric parameters, material parameters and kinematic relationships of the equipment fault injection test bench, the basic characteristics of the equipment fault injection test bench can be comprehensively obtained, and a structural 1:1 mapping of the equipment fault injection test bench can be achieved to obtain a device digital twin model of the equipment fault injection test bench.

[0120] In one embodiment, based on the device digital twin model, different faults are injected into the device under different working conditions to obtain device twin simulation data after the different faults are injected under different working conditions. The real physical data of the device corresponding to the different faults injected into the device under different working conditions by the device fault injection test bench is obtained, including:

[0121] According to the digital twin model of the equipment, different faults are injected into the equipment under different working conditions to obtain the initial twin simulation data of the equipment after different faults are injected under different working conditions.

[0122] Specifically, the server injects different faults under different working conditions into the equipment digital twin model based on the 1:1 establishment of the equipment fault injection test bench, and obtains real-time display of stress and deformation cloud maps of the equipment under different fault states, and real-time response with the preset meshing force time domain and frequency domain curves, to obtain the initial twin simulation data of the equipment after different fault injections under different working conditions. The initial twin simulation data of the equipment includes stress, deformation, meshing force and cumulative damage of the equipment under test.

[0123] Obtain the real physical data of the equipment corresponding to the different fault injections performed on the equipment under different working conditions by the equipment fault injection test bench.

[0124] Specifically, taking the gearbox equipment fault injection test bench as an example, the server adjusts the speed of the motor through the controller and uses the magnetic powder brake to adjust the load to achieve different working conditions and inject various faults. It further uses sensor modules and other data perception devices as the main collection tools to collect the real physical data of the equipment. For example, acceleration sensors can be arranged at each key node of the gearbox to collect vibration signals, and other corresponding sensors can be used to collect speed, torque signals and signals representing the operating status of the gearbox, etc., to obtain the real physical data of the equipment corresponding to the different faults injected into the equipment under different working conditions by the equipment fault injection test bench. Finally, the real physical data of the equipment is sent to the server, and the server can obtain the real physical data of the equipment corresponding to the different faults injected into the equipment under different working conditions by the equipment fault injection test bench.

[0125] Furthermore, the server can establish a wired connection with the host computer through the data acquisition card of the data sensing device, so as to realize real-time monitoring of the operating status of the device corresponding to the real physical data of the device on the device fault injection test bench.

[0126] Compare the real physical data of the device and the initial twin simulation data of the device to obtain the twin simulation data of the device.

[0127] Specifically, a comparison is made from the dimensions of the device's real physical data and the device's initial twin simulation data, and the device's initial twin simulation data is verified and adjusted so that the device twin simulation data obtained after verification and adjustment of the device's initial twin simulation data is an accurate mapping of the device's real physical data.

[0128] Furthermore, the real physical data of the equipment and the operating status corresponding to the real physical data of the equipment are input into the equipment digital twin model. The equipment digital twin model simulates the operation of the equipment according to the operating status corresponding to the real physical data of the equipment. Combined with the finite element simulation analysis of Admas (Automatic Dynamic Analysis of Mechanical Systems, virtual prototype analysis software), the initial twin simulation data of the equipment is extracted from the equipment twin simulation database to obtain the initial twin simulation data of the equipment corresponding to the simulated operating status in the equipment digital twin model. The equipment twin simulation data is obtained based on the comparison results of the initial twin simulation data of the equipment corresponding to the simulated motion state and the real physical data of the equipment.

[0129] In this embodiment, by using the real physical data of the device and the operating status corresponding to the real physical data of the device to verify the initial twin simulation data of the device, the device twin simulation data in the space where the device digital twin model is located and the motion status corresponding to the device twin simulation data can be mapped to the real physical data of the device in the physical world and the operating status corresponding to the real physical data of the device, thereby realizing the accurate mapping of the space where the device digital twin model is located to the physical world in data.

[0130] In one embodiment, the device twin simulation data is obtained by comparing the actual physical data of the device with the initial twin simulation data of the device, including:

[0131] Obtain a preset gap threshold; compare the device's actual physical data with the device's initial twin simulation data to obtain the data difference between the device's actual physical data and the device's initial twin simulation data.

[0132] Among them, the preset gap threshold refers to the limit value of the difference between the real physical data of the device and the initial twin simulation data of the device.

[0133] Specifically, the limit value of the difference between the device's real physical data and the device's initial twin simulation data is obtained. Once this limit value is exceeded, it is determined that the device's initial twin simulation data cannot accurately map the device's real physical data; then the device's real physical data and the device's initial twin simulation data are compared to obtain the data difference between the device's real physical data and the device's initial twin simulation data.

[0134] When the data difference exceeds the preset gap threshold, the initial twin simulation data of the device is updated according to the real physical data of the device to obtain the device twin simulation data.

[0135] When the data difference between the device's real physical data and the device's initial twin simulation data exceeds the limit value of the difference between the device's real physical data and the device's initial twin simulation data, it is determined that the device's initial twin simulation data cannot accurately map the device's real physical data, and the device's initial twin simulation data needs to be updated based on the device's real physical data. The updated device's initial twin simulation data is the device's twin simulation data.

[0136] In this embodiment, by obtaining a preset gap threshold, the initial twin simulation data of the device can be updated, thereby obtaining device twin simulation data that is accurately mapped to the real physical data of the device.

[0137] In one embodiment, Figure 5 As shown, S600 includes:

[0138] S610: Obtain an activation function of an initial equipment fault diagnosis model based on a convolutional neural network.

[0139] Among them, the activation function is a function that runs on the neurons of the artificial neural network and is responsible for mapping the input of the neuron to the output.

[0140] Specifically, an activation function of an initial equipment fault diagnosis model based on a convolutional neural network is obtained. The activation function can be a bitangent function Tanh, a logistic regression function Sigmoid, a rectified linear unit function ReLU, and a Leaky ReLU activation function. Among them, the disadvantages of the Sigmoid function and the Tanh function are that the derivative value will approach 0 when the absolute value of the independent variable is large, and the gradient will disappear when the number of neural network layers increases. At this time, the bottom layer of the network cannot be fully trained and the overall accuracy cannot be improved; the derivative value of the ReLU function is always 1 when the independent variable is greater than 0, which can alleviate the gradient disappearance phenomenon and ensure that the error will not decay during the back propagation process, but the ReLU function is prone to cause neurons to "die" when the gradient is 0; the Leaky ReLU activation function improves the shortcomings of the ReLU function on the basis of the ReLU function, that is, by ensuring that the activation function still has a small non-zero slope in the negative range, the problem of neuron "death" caused by ReLU is alleviated.

[0141] S620: Input the time-frequency spectrum into the convolution layer of the initial equipment fault diagnosis model based on the convolutional neural network, and obtain linearly separable spectrum feature information according to the activation function.

[0142] Among them, the convolution layer refers to the part that plays the role of feature extraction in the convolutional neural network. It extracts local features of the input data in sequence through multiple convolution kernels, determines the positional relationship between local features, and after feature extraction, it will obtain "high-level" features that can better characterize the essence of the signal. At the same time, the convolution operation performed in the convolution layer can also weaken the noise interference in the original monitoring signal.

[0143] Specifically, the initial equipment fault diagnosis model based on convolutional neural network includes input layer, convolution layer, pooling layer, fully connected layer and output layer. The steps of input layer, convolution layer, pooling layer and output layer in convolutional neural network are shown in the figure below. Figure 6 As shown in the figure, the time-frequency spectrum obtained after wavelet analysis of the fault diagnosis data set is input into the convolution layer of the initial equipment fault diagnosis model based on the convolutional neural network, and the convolution layer is used to extract features of the time-frequency spectrum in a local range to obtain information that characterizes the data features in the time-frequency spectrum, namely, the spectrum feature information; at this time, the spectrum feature information is linearly inseparable multidimensional feature information, and the activation function is used to perform nonlinear transformation on the spectrum feature information, and the linearly inseparable multidimensional spectrum feature information is mapped to another space, and the multidimensional spectrum feature information is linearly separable in the space, namely, linearly separable spectrum feature information is obtained, and the linearly separable multidimensional time-frequency spectrum features are easy to classify.

[0144] S630: Input the linearly separable graph feature information into the pooling layer of the initial equipment fault diagnosis model based on the convolutional neural network, perform pooling processing on the linearly separable graph feature information, and obtain the pooled graph feature information.

[0145] Among them, the pooling layer is one of the components of the convolutional neural network. It can scale and map the data obtained by the convolution layer, reduce the dimension of the data, and reduce the complexity of subsequent calculations at the same time, thereby avoiding the occurrence of overfitting. It can also re-extract features from the feature map obtained by convolution, and play the role of a fuzzy filter. After the pooling operation, although the side length of the feature map will be reduced, the scale of the feature will not change.

[0146] Specifically, if Figure 6 As shown in the figure, the linearly separable graph feature information is input into the pooling layer of the initial equipment fault diagnosis model based on the convolutional neural network. The pooling layer will perform pooling processing on the linearly separable graph feature information to reduce the dimension of the graph feature information and obtain the pooled graph feature information to avoid overfitting.

[0147] Furthermore, the pooling methods of the pooling layer include maximum pooling and mean pooling. Maximum pooling is to find the maximum value within the pooling range of the input feature map, and mean pooling is to find the average value within the pooling range of the input feature map. Due to the translation invariance of maximum pooling, when the input feature map has a small displacement change, it can ensure that the extracted features remain unchanged. Maximum pooling makes the pooling operation more robust to displacement changes and can ignore the influence of the phase change of the input signal. Among them, the input feature map is linearly separable spectral feature information.

[0148] S640 , based on the pooled atlas feature information and the ideal label corresponding to the preset time-frequency atlas, an initial device fault diagnosis model based on a convolutional neural network is trained to obtain a target device fault diagnosis model.

[0149] Specifically, when learning and training the initial equipment fault diagnosis model based on convolutional neural network, the input feature map will obtain pooled map feature information after passing through the initial equipment fault diagnosis model based on convolutional neural network. Each pooled map feature information will have a corresponding output value, that is, each input feature map will have a corresponding output value. The corresponding output value of each pooled map feature information is compared with the ideal label corresponding to the preset time-frequency map, and the error between the two is calculated. When the error is large, the network parameters are adjusted to gradually reduce the error until the preset error threshold is reached, thereby obtaining the target equipment fault diagnosis model.

[0150] Furthermore, in the training of the initial equipment fault diagnosis model based on convolutional neural network, the step of obtaining the corresponding output value of the output layer through the fully connected layer can be: the fully connected layer of the initial equipment fault diagnosis model based on convolutional neural network flattens the dimensionality-reduced graph feature information output by the pooling layer to form a vector, and converts the flattened vector into a probability distribution with a sum of 1 using a linear classifier; finally, according to the output value of the corresponding final output layer obtained by the classifier, the output value is a real number feature vector, the dimension of the feature vector is the same as the number of classification categories, and each real value in the feature vector represents the probability of the corresponding category.

[0151] In this embodiment, through the deep learning method of the convolutional neural network in the initial equipment fault diagnosis model based on the convolutional neural network, the characteristics of the data can be autonomously learned without manual preprocessing of the data, thereby accurately establishing the target equipment fault diagnosis model.

[0152] In one embodiment, taking a gearbox as an example, a flowchart of establishing a target device fault diagnosis model and performing fault diagnosis using the target device fault diagnosis model is shown in FIG. Figure 7 As shown in the figure, first, a gearbox fault injection test bench is built to construct a physical space, and the real physical data of the equipment under various working conditions and faults are collected. The twin space is a mirror image of the physical space. A digital twin model of the gearbox equipment is constructed through simulation analysis, and multi-working condition and multi-time scale simulation tests are carried out in the digital twin model of the gearbox equipment. The obtained equipment twin simulation data is stored in the twin database for easy access at any time. At the same time, it is presented to the user in a visual form and interacted and verified with the physical data, and it is continuously optimized and iterated. Finally, a fault diagnosis data set is constructed by combining the real physical data of the equipment with the equipment twin simulation data. The feature map is extracted using wavelet time-frequency analysis and input into the equipment fault diagnosis model based on the convolutional neural network. The construction of the convolutional neural network model in the figure refers to obtaining the target equipment fault diagnosis model. During the construction, the equipment fault diagnosis model based on the convolutional neural network will be learned and trained to obtain the target equipment fault diagnosis model; the target equipment fault diagnosis model is subsequently used to realize accurate fault diagnosis and realize visual interaction.

[0153] In one embodiment, Figure 8 As shown, a device fault diagnosis method is provided, which is applied to Figure 1 Taking the server 104 in the example as an example, the following steps are included:

[0154] S700: Acquire real physical data of the device to be diagnosed.

[0155] Specifically, when the server receives a device fault diagnosis request from a device to be diagnosed, it extracts the real physical data of the device to be diagnosed carried in the device fault diagnosis request.

[0156] S800: Perform wavelet analysis on the real physical data of the device to be diagnosed to obtain a time-frequency spectrum of the real physical data of the device to be diagnosed.

[0157] Specifically, the server performs wavelet analysis on the real physical data of the device to be diagnosed, extracts time-frequency features from the real physical data of the device to be diagnosed, and obtains a time-frequency spectrum of the real physical data of the device to be diagnosed.

[0158] S900: Input the time-frequency spectrum of the real physical data of the device to be diagnosed into the target device fault diagnosis model to perform fault diagnosis and obtain the type of device fault.

[0159] The target equipment fault diagnosis model is established using the above-mentioned equipment fault diagnosis model establishment method.

[0160] Specifically, the above-mentioned equipment fault diagnosis model establishment method is used to establish a target equipment fault diagnosis model. The established target equipment fault diagnosis model is a target equipment fault diagnosis model that is learned and trained based on the graph feature information after dimensionality reduction and the ideal labels corresponding to the preset time-frequency graph. Therefore, the time-frequency graph of the real physical data of the equipment to be diagnosed is input into the target equipment fault diagnosis model for fault diagnosis, and the type of equipment fault can be obtained.

[0161] The above-mentioned equipment fault diagnosis method obtains real physical data of the equipment to be diagnosed; performs wavelet analysis on the real physical data of the equipment to be diagnosed to obtain a time-frequency spectrum of the real physical data of the equipment to be diagnosed; and inputs the time-frequency spectrum of the real physical data of the equipment to be diagnosed into a target equipment fault diagnosis model for fault diagnosis to determine the type of equipment fault. The target equipment fault diagnosis model is established using the above-mentioned equipment fault diagnosis model establishment method. By inputting the real physical data of the equipment to be diagnosed into the accurate target equipment fault diagnosis model, accurate fault diagnosis of the equipment to be diagnosed is ensured.

[0162] It should be understood that, although the various steps in the flowcharts involved in the various embodiments described above are displayed in sequence according to the instructions of the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the flowcharts involved in the various embodiments described above can include multiple steps or multiple stages, and these steps or stages are not necessarily executed and completed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of steps or stages in other steps.

[0163] Based on the same inventive concept, the embodiments of the present application also provide a device for establishing a device fault diagnosis model and a device for establishing a device fault diagnosis model for implementing the aforementioned method for establishing a device fault diagnosis model. The implementation solution provided by the device is similar to the implementation solution described in the aforementioned method. Therefore, the specific limitations of the one or more device fault diagnosis model establishment devices and device fault diagnosis device embodiments provided below can be found in the above-mentioned limitations of the device fault diagnosis model establishment method and the device fault diagnosis method, and will not be repeated here.

[0164] In one embodiment, Figure 9 As shown, a device for establishing a device fault diagnosis model is provided, comprising: an initial model acquisition module 100, a data acquisition module 200, a data set construction module 300, a wavelet analysis module 400, a label acquisition module 500 and a target model establishment module 600, wherein:

[0165] The initial model acquisition module 100 is used to obtain a device digital twin model of the device fault injection test bench and an initial device fault diagnosis model based on a convolutional neural network.

[0166] The data acquisition module 200 is used to inject different faults into the equipment under different working conditions according to the equipment digital twin model, obtain the equipment twin simulation data after the different faults are injected under different working conditions, and obtain the real physical data of the equipment corresponding to the different faults injected into the equipment under different working conditions by the equipment fault injection test bench.

[0167] The data set construction module 300 is used to combine the real physical data of the equipment and the equipment twin simulation data to construct a fault diagnosis data set.

[0168] The wavelet analysis module 400 is used to perform wavelet analysis on the fault diagnosis data set to obtain a time-frequency spectrum of the fault diagnosis data set.

[0169] The label acquisition module 500 is used to obtain an ideal label corresponding to a preset time-frequency spectrum.

[0170] The target model establishment module 600 is used to input the time-frequency spectrum and the ideal label corresponding to the preset time-frequency spectrum into the equipment fault diagnosis model based on the convolutional neural network for learning and training to obtain the target equipment fault diagnosis model.

[0171] In one embodiment, the initial model acquisition module 100 is also used to obtain the geometric parameters, material parameters and kinematic relationships of the equipment fault injection test bench; and establish a device digital twin model of the equipment fault injection test bench based on the geometric parameters, material parameters and kinematic relationships.

[0172] In one embodiment, the data acquisition module 200 is also used to inject different faults into the device under different working conditions based on the device digital twin model, and obtain the initial twin simulation data of the device after the different faults are injected under different working conditions; obtain the real physical data of the device corresponding to the different faults injected into the device under different working conditions by the device fault injection test bench; compare the real physical data of the device and the initial twin simulation data of the device to obtain the device twin simulation data.

[0173] In one embodiment, the data acquisition module 200 is also used to obtain a preset gap threshold; compare the real physical data of the device and the initial twin simulation data of the device to obtain the data difference between the real physical data of the device and the initial twin simulation data of the device; when the data difference exceeds the preset gap threshold, update the initial twin simulation data of the device according to the real physical data of the device to obtain the device twin simulation data.

[0174] In one embodiment, the target model establishment module 600 is also used to obtain the activation function of the initial equipment fault diagnosis model based on the convolutional neural network; the time-frequency spectrum is input into the convolution layer of the initial equipment fault diagnosis model based on the convolutional neural network, and according to the activation function, linearly separable spectrum feature information is obtained; the linearly separable spectrum feature information is input into the pooling layer of the initial equipment fault diagnosis model based on the convolutional neural network, and the linearly separable spectrum feature information is pooled to obtain the pooled spectrum feature information; according to the pooled spectrum feature information and the ideal label corresponding to the preset time-frequency spectrum, the initial equipment fault diagnosis model based on the convolutional neural network is learned and trained to obtain the target equipment fault diagnosis model.

[0175] In one embodiment, Figure 10 As shown, a device fault diagnosis apparatus is provided, comprising: a data acquisition module 700, a map acquisition module 800 and a fault diagnosis module 900, wherein:

[0176] The data acquisition module 700 is used to acquire the real physical data of the device to be diagnosed.

[0177] The spectrum acquisition module 800 is used to perform wavelet analysis on the real physical data of the device to be diagnosed to obtain a time-frequency spectrum of the real physical data of the device to be diagnosed.

[0178] The fault diagnosis module 900 is used to input the time-frequency spectrum of the real physical data of the device to be diagnosed into the target device fault diagnosis model to perform fault diagnosis and obtain the type of device fault.

[0179] The target equipment fault diagnosis model is established using the above-mentioned equipment fault diagnosis model establishment method.

[0180] The aforementioned device fault diagnosis model establishment apparatus and the various modules within the device fault diagnosis apparatus may be implemented in whole or in part via software, hardware, or a combination thereof. Each of the aforementioned modules may be embedded in or independent of a processor within a computer device in the form of hardware, or may be stored in a memory within the computer device in the form of software, so that the processor can call and execute the corresponding operations of each of the aforementioned modules.

[0181] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 11 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O) and a communication interface. The processor, memory and input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the computer device is used to store the device digital twin model of the device fault injection test bench and the initial device fault diagnosis model data based on the convolutional neural network. The input / output interface of the computer device is used to exchange information between the processor and the external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for establishing a device fault diagnosis model is implemented.

[0182] Those skilled in the art will understand that Figure 11 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0183] In one embodiment, a computer device is further provided, including a memory and a processor. The memory stores a computer program, and the processor implements the steps in the above method embodiments when executing the computer program.

[0184] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0185] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0186] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, database or other media used in the embodiments provided in this application may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory may include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processor involved in the various embodiments provided herein may be, but are not limited to, a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic unit, a data processing logic unit based on quantum computing, and the like.

[0187] The technical features of the above embodiments can 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.

[0188] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present application. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, and these modifications and improvements fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.

Claims

1. A method for establishing a device fault diagnosis model, characterized in that: The method comprises: Obtain a digital twin model of the device fault injection testbed and an initial device fault diagnosis model based on a convolutional neural network; According to the device digital twin model, different faults are injected into the device under different working conditions to obtain device twin simulation data after the different faults are injected under different working conditions, and the device fault injection test bench obtains the real physical data of the device corresponding to the different faults injected into the device under different working conditions; the different working conditions are the working conditions of the device under test in the device digital twin model at different speeds and loads; Combining the real physical data of the device and the twin simulation data of the device to construct a fault diagnosis data set; Performing wavelet analysis on the fault diagnosis data set to obtain a time-frequency spectrum of the fault diagnosis data set; Get the ideal label corresponding to the preset time-frequency spectrum; The time-frequency spectrum and the ideal label corresponding to the preset time-frequency spectrum are input into the initial equipment fault diagnosis model based on the convolutional neural network for learning and training to obtain a target equipment fault diagnosis model; the learning and training includes: obtaining an activation function of the initial equipment fault diagnosis model based on the convolutional neural network; inputting the time-frequency spectrum into the convolution layer of the initial equipment fault diagnosis model based on the convolutional neural network, and obtaining linearly separable spectrum feature information according to the activation function; inputting the linearly separable spectrum feature information into the pooling layer of the initial equipment fault diagnosis model based on the convolutional neural network, performing pooling processing on the linearly separable spectrum feature information, and obtaining pooled spectrum feature information; based on the pooled spectrum feature information and the ideal label corresponding to the preset time-frequency spectrum, the initial equipment fault diagnosis model based on the convolutional neural network is learned and trained to obtain a target equipment fault diagnosis model.

2. The method according to claim 1, characterized in that The device digital twin model of the device fault injection test bench is obtained by: Obtain the geometric parameters, material parameters and kinematic relationships of the equipment fault injection test bench; A device digital twin model of a device fault injection test bench is established based on the geometric parameters, the material parameters, and the kinematic relationship.

3. The method according to claim 1, characterized in that According to the device digital twin model, different faults are injected into the device under different working conditions, and the device twin simulation data after the different faults are injected under different working conditions are obtained. The real physical data of the device corresponding to the different faults injected into the device under different working conditions by the device fault injection test bench are obtained, including: According to the device digital twin model, different faults are injected into the device under different working conditions to obtain initial twin simulation data of the device after the different faults are injected under different working conditions; Obtain the real physical data of the equipment corresponding to the different fault injections performed on the equipment under different working conditions by the equipment fault injection test bench; Compare the real physical data of the device and the initial twin simulation data of the device to obtain the device twin simulation data.

4. The method according to claim 3, characterized in that The comparing the real physical data of the device with the initial twin simulation data of the device to obtain the device twin simulation data includes: Get the preset gap threshold; Comparing the real physical data of the device with the initial twin simulation data of the device to obtain a data difference between the real physical data of the device and the initial twin simulation data of the device; When the data difference exceeds a preset gap threshold, the initial twin simulation data of the device is updated according to the real physical data of the device to obtain the device twin simulation data.

5. A device fault diagnosis method, characterized in that: The method comprises: Obtain the real physical data of the equipment to be diagnosed; Performing wavelet analysis on the real physical data of the device to be diagnosed to obtain a time-frequency spectrum of the real physical data of the device to be diagnosed; Inputting the time-frequency spectrum of the real physical data of the device to be diagnosed into the target device fault diagnosis model to perform fault diagnosis and obtain the type of device fault; The target device fault diagnosis model is established by using the method described in any one of claims 1 to 4.

6. A device for establishing a fault diagnosis model for equipment, characterized in that: The device comprises: An initial model acquisition module, used to obtain the device digital twin model of the device fault injection test bench and the initial device fault diagnosis model based on convolutional neural networks; A data acquisition module is used to inject different faults into the device under different operating conditions based on the device digital twin model, obtain device twin simulation data after the different faults are injected under different operating conditions, and obtain real physical data of the device corresponding to the different faults injected into the device under different operating conditions by the device fault injection test bench; different operating conditions are the operating conditions of the device under test in the device digital twin model at different speeds and loads; A data set construction module, configured to construct a fault diagnosis data set by combining the real physical data of the device and the twin simulation data of the device; A wavelet analysis module is used to perform wavelet analysis on the fault diagnosis data set to obtain a time-frequency spectrum of the fault diagnosis data set; The label acquisition module is used to obtain the ideal label corresponding to the preset time-frequency spectrum; A target model establishment module is used to input the time-frequency spectrum and the ideal label corresponding to the preset time-frequency spectrum into the equipment fault diagnosis model based on the convolutional neural network for learning and training to obtain a target equipment fault diagnosis model; the learning and training includes: obtaining the activation function of the initial equipment fault diagnosis model based on the convolutional neural network; inputting the time-frequency spectrum into the convolution layer of the initial equipment fault diagnosis model based on the convolutional neural network, and obtaining linearly separable spectrum feature information according to the activation function; inputting the linearly separable spectrum feature information into the pooling layer of the initial equipment fault diagnosis model based on the convolutional neural network, pooling the linearly separable spectrum feature information, and obtaining pooled spectrum feature information; based on the pooled spectrum feature information and the ideal label corresponding to the preset time-frequency spectrum, the initial equipment fault diagnosis model based on the convolutional neural network is learned and trained to obtain a target equipment fault diagnosis model.

7. The device according to claim 6, characterized in that The initial model acquisition module is also used to obtain the geometric parameters, material parameters and kinematic relationships of the equipment fault injection test bench; based on the geometric parameters, material parameters and kinematic relationships, establish a device digital twin model of the equipment fault injection test bench.

8. A device for diagnosing equipment faults, characterized in that: The device comprises: Data acquisition module, used to obtain real physical data of the device to be diagnosed; A spectrum acquisition module, configured to perform wavelet analysis on the real physical data of the device to be diagnosed to obtain a time-frequency spectrum of the real physical data of the device to be diagnosed; A fault diagnosis module is used to input the time-frequency spectrum of the real physical data of the device to be diagnosed into the target device fault diagnosis model to perform fault diagnosis and obtain the type of device fault; The target device fault diagnosis model is established by using the method described in any one of claims 1 to 4.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 5 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

Images