Robot fault diagnosis method, device, equipment and storage medium

Abstract

The application discloses a robot fault diagnosis method and device, equipment and a storage medium, and relates to the field of industrial intelligent manufacturing. The method comprises the following steps: acquiring torque information of a robot; inputting the torque information into a trained fault diagnosis model to obtain fault type information output by the fault diagnosis model; and the fault diagnosis model is an LSTM-Resnet hybrid deep learning model. The application improves the efficiency of determining robot fault information.

Description

Technical Field

[0001] This application relates to the field of intelligent industrial manufacturing, and in particular to a method, apparatus, equipment, and storage medium for diagnosing robot faults. Background Technology

[0002] In related technologies, the application of industrial robots will expand from traditional manufacturing to other manufacturing sectors. Industrial robots can operate day and night with high efficiency; therefore, how to maintain their high-efficiency operation is a major concern for researchers. Currently, a significant factor affecting the operation of industrial robots is malfunction.

[0003] Current methods for resolving industrial robot malfunctions include: investing significant human resources in daily, weekly, and monthly inspections and maintenance of the robots, creating a comprehensive equipment operation status checklist based on these records, and compiling an industrial robot maintenance manual to summarize the robot's parameters under fault conditions. Data analysis reveals the frequency of malfunctions for each device, as well as patterns and causes, accumulating practical experience for future troubleshooting; alternatively, developing a remote robot service platform allows for remote monitoring and real-time data recording of industrial robots in the factory workshop, assessing their operational status and providing customers with corresponding technical support for industrial robot malfunction repair and routine maintenance; and utilizing a self-developed TCP / IP network to communicate with industrial robots involved in factory production, assisting customer maintenance personnel in resolving robot malfunctions.

[0004] However, the above methods for solving industrial robot malfunctions are too inefficient at determining robot malfunction information.

[0005] Application content

[0006] The main objective of this application is to provide a fault diagnosis method, apparatus, device, and storage medium, which aims to solve the technical problem of low efficiency in determining robot fault information.

[0007] Firstly, to achieve the above objectives, this application provides a robot fault diagnosis method, the method comprising:

[0008] Obtain the robot's torque information;

[0009] The torque information is input into the trained fault diagnosis model to obtain the fault type information output by the fault diagnosis model; the fault diagnosis model is an LSTM-Resnet hybrid deep learning model.

[0010] Optionally, after obtaining the robot's torque information, the method further includes:

[0011] The preset sample segmentation frame is controlled to translate multiple times in the torque vibration information scale according to the preset movement scale, and multiple sample torque data are extracted from the torque information; the preset movement scale is smaller than the torque vibration information scale of the torque information and smaller than the scale of the preset sample segmentation frame;

[0012] The torque information is input into a trained fault diagnosis model to obtain fault type information output by the fault diagnosis model; wherein, the fault diagnosis model is an LSTM-ResNet hybrid deep learning model, including:

[0013] The sample torque data is input into the trained fault diagnosis model to obtain the fault type information output by the fault diagnosis model; wherein, the fault diagnosis model is an LSTM-Resnet hybrid deep learning model.

[0014] Optionally, the fault diagnosis model includes a preprocessing network for preprocessing torque information to obtain torque fault signals;

[0015] The preprocessing network consists of a first convolutional layer, a first pooling layer, a second convolutional layer, and a second pooling layer connected in sequence.

[0016] The first and second convolutional layers are both one-dimensional convolutional layers with the Inception structure, and the first and second pooling layers are both batch normalized layers.

[0017] Optionally, the preprocessing network is also used to denoise and normalize the torque information to obtain the torque fault signal.

[0018] Optionally, the fault diagnosis model includes a feature extraction network for extracting features from the torque fault signal to obtain torque vibration signal features;

[0019] The feature extraction network is an FB-LSTM ResNet network, and the feature extraction layer in the feature extraction network is a bidirectional LSTM meta-feature extraction layer.

[0020] Optionally, the fault diagnosis model includes a fault diagnosis network for fault diagnosis of torque vibration signal characteristics to obtain fault type information;

[0021] The fault diagnosis module consists of a global pooling layer and an extreme learning machine layer connected in sequence;

[0022] A global pooling layer is used to flatten the torque vibration signal to obtain a flattened torque vibration signal.

[0023] The Extreme Learning Machine layer is used to output fault category information based on the flattened torque vibration signal.

[0024] Optionally, the Extreme Learning Machine is also used to output optimal parameters.

[0025] After inputting the torque information into the trained fault diagnosis model, the method also includes:

[0026] Based on the optimal parameters, the fault diagnosis model is updated to obtain the updated fault diagnosis model.

[0027] Secondly, this application also provides a robot fault diagnosis device, the device comprising:

[0028] The acquisition module is used to acquire the robot's torque information;

[0029] The input module is used to input torque information into the trained fault diagnosis model to obtain the fault type information output by the fault diagnosis model; wherein, the fault diagnosis model is an LSTM-Resnet hybrid deep learning model.

[0030] Thirdly, this application provides a robot fault diagnosis device, including: a processor, a memory, and a robot fault diagnosis program stored in the memory, wherein the robot fault diagnosis program is executed by the processor to implement the steps of the robot fault diagnosis method of the first aspect.

[0031] Fourthly, this application provides a computer-readable storage medium storing a robot fault diagnosis program, which, when executed by a processor, implements the robot fault diagnosis method as described in the first aspect.

[0032] This application proposes a fault diagnosis method for robots, which involves acquiring the robot's torque information; inputting the torque information into a trained fault diagnosis model to obtain fault type information output by the fault diagnosis model; wherein the fault diagnosis model is an LSTM-Resnet hybrid deep learning model.

[0033] Therefore, by inputting the robot's torque information into the LSTM-Resnet hybrid deep learning model, the robot's fault information can be quickly obtained, improving the efficiency of determining robot fault information. Attached Figure Description

[0034] Figure 1 This is a schematic diagram of the architecture of the robot fault diagnosis system according to the robot fault diagnosis method of this application;

[0035] Figure 2 A schematic diagram of the structure of a robot fault diagnosis device that forms the hardware operating environment for the robot fault diagnosis method of this application;

[0036] Figure 3This is a flowchart illustrating a first embodiment of a robot fault diagnosis method provided in this application;

[0037] Figure 4 This is a flowchart illustrating a second embodiment of a robot fault diagnosis method provided in this application;

[0038] Figure 5 A schematic diagram of the structure of a preprocessing network for a robot fault diagnosis method provided in this application;

[0039] Figure 6 A schematic diagram of the signal acquisition and processing module of a robot fault diagnosis method provided in this application;

[0040] Figure 7 A schematic diagram of a fault feature extraction module for a robot fault diagnosis method provided in this application;

[0041] Figure 8 A schematic diagram of a fault diagnosis module for a robot fault diagnosis method provided in this application;

[0042] Figure 9 This is a schematic diagram of the robot fault diagnosis device of this application.

[0043] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0044] It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit this application.

[0045] In related technologies, the application of industrial robots will expand from traditional manufacturing to other manufacturing sectors. Industrial robots can operate day and night with high efficiency; therefore, how to maintain their high-efficiency operation is a major concern for researchers. Currently, a significant factor affecting the operation of industrial robots is malfunction.

[0046] Current methods for resolving industrial robot malfunctions include: investing significant human resources in daily, weekly, and monthly inspections and maintenance of the robots, creating a comprehensive equipment operation status checklist based on these records, and compiling an industrial robot maintenance manual to summarize the robot's parameters under fault conditions. Data analysis reveals the frequency of malfunctions for each device, as well as patterns and causes, accumulating practical experience for future troubleshooting; alternatively, developing remote robot service platforms to remotely monitor and record real-time data from industrial robots in the factory workshop, assessing their operational status and providing customers with corresponding technical support for industrial robot malfunction repair and routine maintenance; and utilizing self-developed TCP / IP networks to communicate with industrial robots involved in factory production, assisting customer maintenance personnel in resolving robot malfunctions.

[0047] However, the above methods for solving industrial robot malfunctions are too inefficient at determining robot malfunction information.

[0048] This application proposes a fault diagnosis method for robots, which involves acquiring the robot's torque information; inputting the torque information into a trained fault diagnosis model to obtain fault type information output by the fault diagnosis model; wherein the fault diagnosis model is an LSTM-Resnet hybrid deep learning model.

[0049] Therefore, by inputting the robot's torque information into the LSTM-Resnet hybrid deep learning model, the robot's fault information can be quickly obtained, improving the efficiency of determining robot fault information.

[0050] The following embodiments of this application will describe the multimedia playback system used in the technical implementation of this application:

[0051] Reference Figure 1 , Figure 1 This is a schematic diagram of the architecture of a robot fault diagnosis system provided in an exemplary embodiment. Figure 1 As shown, the robot fault diagnosis system may include a server 11, a network 12, and a robot fault diagnosis device 13.

[0052] Server 11 can be a physical server containing an independent host, or it can be a virtual server hosted in a host cluster. During operation, server 11 can run server-side programs for a specific application to implement the relevant business functions of that application. For example, when the robot fault diagnosis device 13 acquires the robot's torque information, server 11 can act as the server for the application that acquires the robot's torque information, supporting the robot fault diagnosis device 13 in completing the task of acquiring the robot's torque information.

[0053] Network 12 may include various types of wired or wireless networks. In one embodiment, network 12 may include the Public Switched Telephone Network (PSTN) and the Internet. The robot fault diagnosis device 13 can interact with server 11 through network 12.

[0054] The robot fault diagnosis device 13 may include electronic devices such as workstations, smartphones, tablets, laptops, PDAs (Personal Digital Assistants), etc., and one or more embodiments of this specification are not intended to limit this.

[0055] Reference Figure 2 , Figure 2 This is a schematic diagram of the structure of a robot fault diagnosis device in the hardware operating environment involved in the embodiments of this application.

[0056] like Figure 2 As shown, the robot fault diagnosis device may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include a display screen or an input unit such as a keyboard; optionally, the user interface 1003 may also include a standard wired interface or a wireless interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface). The memory 1005 may be a high-speed random access memory (RAM) or a stable non-volatile memory (NVM), such as a disk drive. The memory 1005 may also optionally be a storage device independent of the aforementioned processor 1001.

[0057] Those skilled in the art will understand that Figure 2 The structure shown does not constitute a limitation on the playback terminal and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0058] like Figure 2As shown, the memory 1005, which serves as a storage medium, may include an operating system, a data storage module, a network communication module, a user interface module, and a robot fault diagnosis program.

[0059] exist Figure 2 In the robot fault diagnosis device shown, the network interface 1004 is mainly used for data communication with the network server; the user interface 1003 is mainly used for data interaction with the user; the processor 1001 and the memory 1005 in the robot fault diagnosis device of this application can be set in the robot fault diagnosis device. The robot fault diagnosis device calls the robot fault diagnosis program stored in the memory 1005 through the processor 1001 and executes the robot fault diagnosis method provided in the embodiment of this application.

[0060] Based on, but not limited to, the hardware structure of the robot fault diagnosis device described above, this application provides a first embodiment of robot fault diagnosis. (Refer to...) Figure 3 , Figure 3 A flowchart illustrating the first embodiment of applying for robot fault diagnosis is shown.

[0061] It should be noted that although the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than that shown here.

[0062] In this embodiment, the robot fault diagnosis method includes:

[0063] S10, obtain the robot's torque information;

[0064] The execution subject in this embodiment is a robot fault diagnosis device.

[0065] It is important to understand that robot fault diagnosis equipment includes torque sensors, which are used to obtain the robot's torque information.

[0066] S20, the torque information is input into the trained fault diagnosis model to obtain the fault type information output by the fault diagnosis model; wherein, the fault diagnosis model is an LSTM-Resnet hybrid deep learning model.

[0067] It's important to understand that the robot fault diagnosis equipment inputs torque information into a trained fault diagnosis model to obtain the fault type information output by the model. This fault diagnosis model is an LSTM-ResNet hybrid deep learning model.

[0068] In this embodiment, current methods for locating industrial robot faults include manually locating the robot fault, manually determining the robot fault based on the robot's monitoring data, and manually investigating the location of the robot fault after it occurs. However, this application can quickly obtain the robot's fault information by inputting the robot's torque information into the LSTM-Resnet hybrid deep learning model, thus improving the efficiency of determining robot fault information.

[0069] Based on the above embodiments, this application also provides a second embodiment of a robot fault diagnosis method. Figure 4 This is a flowchart illustrating a second embodiment of a robot fault diagnosis method provided in this application.

[0070] In this embodiment, after obtaining the robot's torque information, the method further includes:

[0071] S101, according to a preset movement scale, the preset sample segmentation frame is controlled to translate multiple times in the torque vibration information scale to extract multiple sample torque data from the torque information; the preset movement scale is smaller than the torque vibration information scale of the torque information and smaller than the scale of the preset sample segmentation frame;

[0072] It should be understood that the preset movement scale can be set by the user according to the actual situation. The torque vibration information scale is the length information of the torque information, and the preset movement scale is smaller than the torque vibration information scale of the torque information and smaller than the scale of the preset sample segmentation box.

[0073] Based on a preset movement scale and torque vibration information scale, the preset sample segmentation box is translated within this scale, also known as the vibration signal window translation method. This method maximizes the total number of torque information samples, avoiding overfitting and improving model performance. Since interval sampling cannot represent all vibration signals, the robot fault diagnosis equipment uses the vibration signal window translation method to segment torque information samples. The resulting expanded sample torque data maximizes the representation of all torque data, which is beneficial for exploring the learning potential of neural networks.

[0074] The vibration signal window translation method for segmenting torque information samples preserves the continuity between time-series vibration signals and avoids the limitation of equal-interval sampling in representing all vibration signals. Robot fault diagnosis equipment employs the vibration signal window translation method to segment signals and obtain sample torque data.

[0075] Specifically, in one example, such as for a torque vibration information scale of L... z Regarding torque information, using the same standard as L... yThe vibration signal window is translated by a preset sample segmentation box, where the step size of each translation, i.e., the movement scale, is p. Therefore, after each translation, the preset sample segmentation box can extract a segment of sample torque data from the torque information.

[0076] At this point, it is understandable that the number of sample torque data that can be obtained is E, and E sample torque data constitute a sample set. E is calculated according to Formula 1.

[0077] Formula 1 is:

[0078]

[0079] Here, ⊙ represents rounding down to the nearest integer.

[0080] Understandably, since the preset movement scale is smaller than the torque vibration information scale of the torque information and smaller than the scale of the preset sample segmentation frame, the sample torque data intercepted during two adjacent translations overlap, and the scale of the overlapping signal segment is L. y -p. At this point, for this torque information, after processing using the vibration signal window translation method described above, the sample torque data in the obtained sample set increases to α times. α can be calculated using Formula 2:

[0081] Formula 2 is:

[0082]

[0083] The position of the i-th sample torque data in the torque signal can be represented as:

[0084]

[0085] In this embodiment, the torque information is input into a trained fault diagnosis model to obtain fault type information output by the fault diagnosis model; wherein, the fault diagnosis model is an LSTM-ResNet hybrid deep learning model, and the adaptability becomes:

[0086] S102, The sample torque data is input into the trained fault diagnosis model to obtain the fault type information output by the fault diagnosis model; wherein, the fault diagnosis model is an LSTM-Resnet hybrid deep learning model.

[0087] It is important to understand that the robot fault diagnosis equipment inputs the expanded sample torque data into the trained fault diagnosis model to obtain the fault type information output by the fault diagnosis model; the fault diagnosis model is an LSTM-Resnet hybrid deep learning model.

[0088] In this embodiment, the robot fault diagnosis equipment uses a vibration signal window translation method to maximize the representation of all torque data, which is beneficial for exploring the learning potential of neural networks. Simultaneously, segmenting torque information samples using the vibration signal window translation method preserves the continuity between temporal vibration signals and avoids the limitation of equal-interval sampling in representing all vibration signals.

[0089] As one specific implementation method, see [reference]. Figure 5 , Figure 5 This is a schematic diagram of the structure of a preprocessing network for a robot fault diagnosis method provided in this application. The fault diagnosis model includes a preprocessing network for preprocessing torque information to obtain a torque fault signal.

[0090] The preprocessing network consists of a first convolutional layer, a first pooling layer, a second convolutional layer, and a second pooling layer connected in sequence.

[0091] The first and second convolutional layers are both one-dimensional convolutional layers with the Inception structure, and the first and second pooling layers are both batch normalized layers.

[0092] It is important to understand that the torque fault signal is a one-dimensional sample. The robot fault diagnosis equipment performs preprocessing on the torque fault signal using a processing network, which effectively reduces signal noise. The preprocessing network consists of a first convolutional layer, a first pooling layer, a second convolutional layer, and a second pooling layer connected in sequence. Adding a batch normalization layer after the convolutional layers helps to accelerate the network's convergence rate, improve training accuracy, and thus ensure the reliability of subsequent processing.

[0093] In this embodiment, the robot fault diagnosis device preprocesses the torque fault signal using a preprocessing network, effectively reducing signal noise. Furthermore, adding a batch normalization layer after the convolutional layer accelerates the network's convergence rate, improves training accuracy, and thus ensures the reliability of subsequent processing.

[0094] As one specific implementation method, see [link / reference]. Figure 6 , Figure 6 This is a schematic diagram of the signal acquisition and processing module of a robot fault diagnosis method provided in this application.

[0095] The preprocessing network is also used to denoise and normalize the torque information to obtain the torque fault signal.

[0096] It's important to understand that denoising technology is the technique of removing noise from a signal. All signal processing devices have characteristics that make them susceptible to noise. Noise can be random noise or white noise with a uniform frequency distribution, or it can be frequency-related noise introduced by device mechanisms or signal processing algorithms.

[0097] Normalization maps torque information to the range of 0 to 1, making torque information processing more convenient and faster. Using normalization helps improve model accuracy and convergence speed.

[0098] In this embodiment, the torque fault signal is obtained by denoising and normalizing the torque information, which helps to improve the accuracy of fault diagnosis results.

[0099] As one specific implementation method, see [link / reference]. Figure 7 , Figure 7 This diagram illustrates a fault feature extraction module for a robot fault diagnosis method provided in this application. The fault diagnosis model includes a feature extraction network for extracting features from the torque fault signal to obtain torque vibration signal features.

[0100] The feature extraction network is an FB-LSTM ResNet network, and the feature extraction layer in the feature extraction network is a bidirectional LSTM meta-feature extraction layer.

[0101] It should be understood that the fault feature extraction module includes a bidirectional LSTM element, an addition layer, and a dropout layer. Here, the first torque fault information is sequentially fed into the first bidirectional LSTM element, the first dropout layer, the second bidirectional LSTM element, and the second dropout layer to obtain the first torque vibration signal feature; the second torque fault information is sequentially fed into the third bidirectional LSTM element and the third dropout layer to obtain the second torque vibration information feature; the first torque vibration signal feature and the second torque vibration information feature are then added together to obtain the torque vibration signal feature.

[0102] The FB-LSTM ResNet network is used to extract features from torque fault signals, obtaining torque vibration signal features. This network consists of a bidirectional feature extraction layer composed of LSTM elements and a fusion-type residual network. The bidirectional LSTM elements can comprehensively acquire torque vibration signal features from both the front and rear directions and adjust parameters in a timely manner. The fusion-type residual structure improves training speed and performance without increasing the number of additional parameters or computational load, and effectively handles degradation caused by layer depth.

[0103] In this embodiment, the use of the FB-LSTM ResNet network is beneficial for optimizing the data flow, thereby alleviating the gradient vanishing problem, that is, solving the problem that the output gets stuck in the function saturation region and the gradient rapidly decreases, making it impossible to learn effectively.

[0104] As one specific implementation method, see [link / reference]. Figure 8 , Figure 8This diagram illustrates a fault diagnosis module of a robot fault diagnosis method provided in this application. The fault diagnosis model includes a fault diagnosis network used to diagnose faults based on the characteristics of torque vibration signals and obtain fault type information.

[0105] The fault diagnosis module consists of a global pooling layer and an extreme learning machine layer connected in sequence;

[0106] A global pooling layer is used to flatten the torque vibration signal to obtain a flattened torque vibration signal.

[0107] The Extreme Learning Machine layer is used to output fault category information based on the flattened torque vibration signal.

[0108] It is important to understand that in existing technologies, each neuron in a fully connected layer is connected to all neurons in the previous layer to fuse the data acquired by the convolutional layer. However, the amount of data in a fully connected layer is enormous, which not only reduces the training speed of the model but also easily leads to overfitting. Therefore, this embodiment uses a global pooling layer to flatten the torque vibration signal. The resulting flattened torque vibration signal can complete the flattening process of the vibration signal and reduce the number of network parameters, thereby effectively avoiding overfitting.

[0109] Based on the same concept, this application also provides a third embodiment of a robot fault diagnosis method.

[0110] In this embodiment, the extreme learning machine is also used to output optimal parameters based on the flattened torque vibration signal.

[0111] In this embodiment, after step S20, the method further includes:

[0112] S30. Based on the optimal parameters, update the fault diagnosis model to obtain the updated fault diagnosis model.

[0113] It's important to understand that traditional Softmax (Flexible Maximum Transfer Function) methods, because they require pre-adjusting the layer based on the classification objective, struggle to effectively utilize features acquired from FB-LSTM ResNet. This results in poor model performance, accuracy, and robustness, and also increases training time. In contrast, Extreme Learning Machines (ELMs) require only one training iteration to obtain optimal parameters and exhibit excellent generalization capabilities.

[0114] Therefore, during each fault diagnosis process, the Extreme Learning Machine can also output the optimal parameters, and then update the fault diagnosis model based on the optimal parameters to obtain the updated fault diagnosis model.

[0115] In this embodiment, a global pooling layer is used instead of a fully connected layer. This layer can flatten the vibration signal and reduce network parameters, thereby effectively avoiding overfitting. An Extreme Learning Machine (ELM) is used to obtain relevant parameters through a set hidden layer stage, and the weights of the output layer are obtained using regularization, exhibiting good generalization ability. Furthermore, after the ELM outputs the optimal parameters, the fault diagnosis model is updated based on these optimal parameters, resulting in an updated fault diagnosis model and improving its accuracy.

[0116] Based on the same concept, this application also provides a robot fault diagnosis device. (See also...) Figure 5 , Figure 5 This is a schematic diagram of a robot fault diagnosis device provided in this application. The device specifically includes:

[0117] The acquisition module 500 is used to acquire the robot's torque information;

[0118] The input module 510 is used to input torque information into the trained fault diagnosis model to obtain fault type information output by the fault diagnosis model; wherein, the fault diagnosis model is an LSTM-Resnet hybrid deep learning model.

[0119] The technical solution of this embodiment, through the cooperation between various functional modules, obtains the torque information of the robot; inputs the torque information into the trained fault diagnosis model, and obtains the fault type information output by the fault diagnosis model; wherein, the fault diagnosis model is an LSTM-Resnet hybrid deep learning model.

[0120] Therefore, by inputting the robot's torque information into the LSTM-Resnet hybrid deep learning model, the robot's fault information can be quickly obtained, improving the efficiency of determining robot fault information.

[0121] Furthermore, embodiments of this application also propose a computer storage medium storing a robot fault diagnosis program. When the robot fault diagnosis program is executed by a processor, it implements the steps of the robot fault diagnosis method described above. Therefore, it will not be repeated here. Additionally, the beneficial effects of using the same method will not be repeated here either. For technical details not disclosed in the computer-readable storage medium embodiments of this application, please refer to the description of the method embodiments of this application. As an example, program instructions can be deployed to execute on a single computing device, or on multiple computing devices located in one location, or on multiple computing devices distributed across multiple locations and interconnected via a communication network.

[0122] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0123] It should also be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided in this application, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0124] Through the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware, or it can be implemented by special-purpose hardware including application-specific integrated circuits, special-purpose CPUs, special-purpose memory, special-purpose components, etc. Generally, any function performed by a computer program can be easily implemented by corresponding hardware, and the specific hardware structure used to implement the same function can also be diverse, such as analog circuits, digital circuits, or special-purpose circuits. However, for this application, software program implementation is more often the preferred implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a readable storage medium, such as a computer floppy disk, USB flash drive, mobile hard disk, read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of this application.

[0125] The above are merely preferred embodiments of this application and do not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. A robot failure diagnosis method characterized by comprising: The method comprises: acquiring torque information of a robot; inputting the torque information into a trained fault diagnosis model to obtain fault type information output by the fault diagnosis model; wherein the fault diagnosis model is an LSTM-Resnet hybrid deep learning model; After acquiring the torque information of the robot, the method further comprises: controlling a preset sample segmentation frame to move multiple times in a torque vibration information scale according to a preset movement scale, and cutting a plurality of sample torque data from the torque information; the preset movement scale is smaller than the torque vibration information scale of the torque information and smaller than the scale of the preset sample segmentation frame; the torque vibration information scale is the length information of the torque information, and the preset movement scale is the step length of each movement of the preset sample segmentation frame; inputting the sample torque data into the trained fault diagnosis model to obtain the fault type information output by the fault diagnosis model; The fault diagnosis model comprises a preprocessing network for preprocessing the torque information to obtain a torque fault signal; The preprocessing network comprises a first convolutional layer, a first pooling layer, a second convolutional layer and a second pooling layer connected in sequence; The first convolutional layer and the second convolutional layer are both one-dimensional convolutional layers of an Inception structure, and the first pooling layer and the second pooling layer are both batch normalization layers; The fault diagnosis model comprises a feature extraction network for extracting features of the torque fault signal to obtain a torque vibration signal feature; The feature extraction network is an FB-LSTM Resnet network, and the feature extraction layer in the feature extraction network is a bidirectional LSTM meta-feature extraction layer; The fault diagnosis model comprises a fault diagnosis network for diagnosing faults of the torque vibration signal feature to obtain fault type information; The fault diagnosis network comprises a global pooling layer and an extreme learning machine layer connected in sequence; The global pooling layer is configured to perform flattening processing on the torque vibration signal feature to obtain a flattened torque vibration signal; The extreme learning machine layer is configured to output fault category information according to the flattened torque vibration signal; The extreme learning machine layer is also configured to output an optimal parameter; After inputting the torque information into the trained fault diagnosis model, the method further comprises: updating the fault diagnosis model according to the optimal parameter to obtain an updated fault diagnosis model. The preprocessing network is also configured to perform denoising and normalization processing on the torque information to obtain the torque fault signal.

2. The robot failure diagnosis method according to claim 1, characterized by, The device comprises:

3. A robot failure diagnosis device characterized by comprising: an acquisition module configured to acquire torque information of a robot; an input module configured to input the torque information into a trained fault diagnosis model to obtain fault type information output by the fault diagnosis model; wherein the fault diagnosis model is an LSTM-Resnet hybrid deep learning model; ​ The input module is further configured to control the preset sample segmentation frame to move multiple times in a torque vibration information dimension according to a preset movement scale, and to obtain a plurality of sample torque data from the torque information; the preset movement scale is smaller than a torque vibration information dimension of the torque information and smaller than a dimension of the preset sample segmentation frame; the torque vibration information dimension is a length information of the torque information, and the preset movement scale is a step length of each movement of the preset sample segmentation frame. The input module is further configured to input the sample torque data into the trained fault diagnosis model to obtain fault type information output by the fault diagnosis model. The fault diagnosis model comprises a preprocessing network configured to preprocess the torque information to obtain a torque fault signal; the preprocessing network comprises a first convolutional layer, a first pooling layer, a second convolutional layer and a second pooling layer connected in sequence; the first convolutional layer and the second convolutional layer are one-dimensional convolutional layers of an Inception structure, and the first pooling layer and the second pooling layer are batch normalization layers. The fault diagnosis model comprises a feature extraction network configured to extract features of the torque fault signal to obtain torque vibration signal features; the feature extraction network is an FB-LSTM Resnet network, and a feature extraction layer in the feature extraction network is a bidirectional LSTM meta-feature extraction layer. The fault diagnosis model comprises a fault diagnosis network configured to diagnose faults of the torque vibration signal features to obtain fault type information; the fault diagnosis network comprises a global pooling layer and an extreme learning machine layer connected in sequence; the global pooling layer is configured to perform flattening processing on the torque vibration signal features to obtain flattened torque vibration signals; and the extreme learning machine layer is configured to output fault category information according to the flattened torque vibration signals. The extreme learning machine layer is further configured to output an optimal parameter; and the input module is further configured to update the fault diagnosis model according to the optimal parameter to obtain an updated fault diagnosis model.

4. A robot failure diagnosing apparatus characterized by comprising: A processor, a memory and a robot fault diagnosis program stored in the memory, wherein the robot fault diagnosis program, when executed by the processor, implements the steps of the robot fault diagnosis method according to any one of claims 1 to 2. A computer readable storage medium having a robot fault diagnosis program stored thereon, wherein the robot fault diagnosis program, when executed by a processor, implements the robot fault diagnosis method according to any one of claims 1 to 2.

5. A computer readable storage medium, characterized in that, ​

Images