Nonlinear system fault diagnosis method based on sliding mode observer and DCNN

Abstract

The application provides a kind of nonlinear system fault diagnosis method based on sliding mode observer and DCNN, it is related to fault diagnosis technical field.The method first constructs the mathematical model of nonlinear system, the mathematical model considers concurrent situation of system actuator and sensor fault and nonlinear term and external interference, and sliding mode observer is designed for nonlinear system;Then define the residual error of sliding mode observer, establish residual error equation;Determine the residual error threshold of sliding mode observer;Nonlinear system is monitored using sliding mode observer, residual error information is obtained and residual error information is preprocessed based on residual error threshold to establish residual error signal data set;Finally, the residual error signal generated by sliding mode observer is diagnosed in depth using deep learning model DCNN, and the fault diagnosis of nonlinear system is realized.

Description

Technical Field

[0001] This invention relates to the field of fault diagnosis technology, and in particular to a fault diagnosis method for nonlinear systems based on sliding mode observer and DCNN. Background Technology

[0002] In the wave of industrial intelligence, the complex structure and high integration of nonlinear systems make actuator and sensor failures a significant threat to system reliability and safety. Among model-based fault diagnosis methods, sliding mode observers, with their strong robustness to system parameter variations and external disturbances, have become a powerful tool for diagnosing nonlinear system faults. Based on sliding mode variable structure control theory, they ensure the system state trajectory reaches and remains on the sliding surface, accurately estimating system state and fault information. A well-designed sliding mode observer can effectively address nonlinear characteristics, disturbances, and fault issues. The construction of residual equations is crucial for fault diagnosis; the residuals reflect the difference between the observer output and the actual output, containing fault information.

[0003] However, with the increasing complexity of nonlinear systems and the diversification of fault modes, traditional fault diagnosis methods are inadequate in feature extraction and pattern recognition. Deep convolutional neural networks (DCNNs), with their powerful feature extraction and pattern recognition capabilities, bring new opportunities to fault diagnosis, as they can automatically learn deep features of residual signals and identify complex fault modes. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a nonlinear system fault diagnosis method based on sliding mode observer and DCNN to address the shortcomings of the prior art, thereby solving the problem that it is difficult to achieve accurate fault diagnosis in nonlinear systems under the conditions of actuator and sensor faults and unknown interference.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows:

[0006] On the one hand, this invention provides a fault diagnosis method for nonlinear systems based on sliding mode observers and DCNNs, comprising:

[0007] Construct a mathematical model for the nonlinear system that considers concurrent failures of the system's actuators and sensors, as well as nonlinear terms and external disturbances;

[0008] Design a sliding mode observer for nonlinear systems;

[0009] Define the sliding mode observer residuals and establish the residual equations; the sliding mode observer residuals include actuator fault residuals and sensor fault residuals;

[0010] Determine the residual threshold for the sliding mode observer;

[0011] A sliding mode observer is used to monitor the nonlinear system, obtain residual information, and preprocess the residual information based on the residual threshold to establish a residual signal dataset;

[0012] Based on the relevant parameters of the sliding mode observer, a deep learning model DCNN is used to perform in-depth diagnosis of the residual signal generated by the sliding mode observer, thereby realizing fault diagnosis of nonlinear systems.

[0013] Furthermore, the mathematical model of the nonlinear system is shown in the following formula:

[0014] ;

[0015] in, Represents the state vector of a nonlinear system. for The derivative; Represents the input vector of a nonlinear system; Represents the output vector of a nonlinear system; Represents the actuator fault vector. Represents the sensor fault vector; Indicates unknown bounded interference. Represents nonlinear continuous terms; This is the state gain matrix; To control the input gain matrix; This is the output gain matrix; For actuator fault matrix; This is the external interference gain matrix.

[0016] Furthermore, the sliding mode observer is shown in the following formula:

[0017] ;

[0018] in, This represents the state estimate of a nonlinear system. for The derivative; Output estimates for nonlinear systems; This represents the estimated value of the actuator failure. This represents the estimated sensor fault value; All are sliding mode observer gain matrices; These are all sliding mode control terms, as shown in the following formulas:

[0019] ;

[0020] in, All are sliding surfaces; , All are positive matrices.

[0021] Furthermore, the actuator fault residual and the sensor fault residual are represented by the following formula:

[0022] ;

[0023] The residual equation is further established as shown in the following formula:

[0024] ;

[0025] in, These are the actuator fault residuals and the sensor fault residuals, respectively.

[0026] Furthermore, the specific method for determining the residual threshold of the sliding mode observer is as follows:

[0027] 1) In nonlinear systems, a sliding mode observer is used to monitor the system; in this case, actuator faults and sensor faults should be in a state of non-occurrence, i.e. State estimates are obtained through a sliding mode observer. Actuator fault estimate Sensor fault estimate Then, the actuator fault residual is calculated. and sensor fault residual Record the actuator fault residuals and sensor fault residuals over a period of time to obtain the actuator fault residual sequence and sensor fault residual sequence;

[0028] 2) For actuator fault residuals Find the maximum value in its sequence. For sensor fault residuals Find the maximum value in its sequence. ;

[0029] 3) Repeat steps 1)-2), simulate N external random disturbances to the system operating state, and obtain the actuator fault residuals under each simulation. maximum value and sensor fault residual maximum value , ;

[0030] 4) Actuator fault residuals obtained from N simulations maximum value and sensor fault residual maximum value The actuator fault residual threshold is obtained. and sensor fault residual threshold As shown in the formula below:

[0031] ;

[0032] .

[0033] Furthermore, the specific method for establishing the residual signal dataset is as follows:

[0034] First, a sliding mode observer is used to monitor the nonlinear system, collecting data including actuator fault residuals. and sensor fault residual The actual residual signal is used to determine if the nonlinear system has a fault when the residual signal exceeds the residual threshold; and the residual signal is timestamped to ensure that the signal is strictly aligned in the time dimension.

[0035] Then, the residual signal is standardized to create a residual signal dataset;

[0036] Finally, the residual signal dataset obtained after the above processing is divided into training set and test set.

[0037] Furthermore, the deep learning model DCNN includes an input layer, a multi-layer convolutional module, a deep convolutional layer group, an attention mechanism module, a multi-scale feature fusion module, a fully connected layer module, and an output layer arranged sequentially.

[0038] The input layer can receive multi-dimensional time-series data and preprocess the input residual signal data to eliminate the dimensional differences between different features.

[0039] The multi-layer convolutional module includes multiple consecutive convolutional layers, which initially extract local temporal features from the residual signal;

[0040] The deep convolutional layer group constructs a deep convolutional module to further mine the complex features in the residual signal; multiple residual blocks are set after the convolutional module, and each residual block contains two convolutional layers;

[0041] The attention mechanism module performs average pooling and max pooling operations on the feature maps output by each convolutional layer of the deep convolutional layer group in the channel dimension to obtain two different feature descriptions.

[0042] The multi-scale feature fusion module upsamples or downsamples the feature maps output from different convolutional layers to make them have the same size before performing feature fusion to obtain fused features.

[0043] The fully connected layer module includes three fully connected layers, which map the fused features obtained by the multi-scale feature fusion module, adjust the dimensions, and judge them with the residual threshold. The fault diagnosis result is determined based on the judgment result.

[0044] Secondly, this application proposes an electronic device, including: one or more processors, and a memory for storing instructions, which, when executed by the one or more processors, cause the one or more processors to perform the nonlinear system fault diagnosis method based on sliding mode observer and DCNN.

[0045] Thirdly, this application proposes a computer-readable storage medium storing executable instructions that, when executed, cause a processor to perform the aforementioned nonlinear system fault diagnosis method based on sliding mode observer and DCNN.

[0046] Fourthly, this application proposes a computer program product, including a computer program or instructions, which, when executed by a processor, implements the aforementioned nonlinear system fault diagnosis method based on sliding mode observer and DCNN.

[0047] The beneficial effects of adopting the above technical solution are as follows: The nonlinear system fault diagnosis method based on sliding mode observer and DCNN provided by the present invention (1) combines the strong robustness of sliding mode observer with the powerful feature extraction and pattern recognition capabilities of deep learning model DCNN, which can be directly applied to various practical nonlinear systems, effectively breaking through the technical bottleneck of traditional fault diagnosis methods in handling complex nonlinearity, multiple small faults and interference coexistence, and greatly expanding the application scenarios and practical value of this technology in nonlinear systems in different industries. (2) Based on system characteristic analysis, targeted actuator fault residuals and sensor fault residuals are generated through sliding mode observer, separating complex fault information. At the same time, the residual signal is intelligently diagnosed using deep learning model DCNN, without the need for complex manual feature extraction and rule setting. This method of combining hierarchical processing and intelligent diagnosis not only simplifies the fault diagnosis process and reduces manual intervention, but also improves the efficiency and convenience of diagnosis, and can quickly and accurately complete the diagnosis task when system actuator and sensor faults occur simultaneously. (3) By adopting a strategy that combines sliding mode observation with deep learning diagnosis, the sliding mode observer can effectively suppress the influence of external unknown disturbances on the system state estimation, providing relatively pure residual signals for subsequent fault diagnosis; the deep learning model DCNN can accurately extract fault features from the residual signals and accurately identify the fault type and degree. The synergistic effect of the two greatly improves the accuracy and reliability of the fault diagnosis results, provides more solid data support for the system's fault handling and fault-tolerant control, and effectively ensures the stable operation of the nonlinear system under complex working conditions. Attached Figure Description

[0048] Figure 1 The flowchart shows the nonlinear system fault diagnosis method based on sliding mode observer and DCNN provided in Embodiment 1 of the present invention.

[0049] Figure 2 This is the fault diagnosis result of the nonlinear system actuator provided in Embodiment 1 of the present invention;

[0050] Figure 3 The fault diagnosis result of the nonlinear system sensor provided in Embodiment 1 of the present invention. Detailed Implementation

[0051] The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and are not intended to limit the scope of the invention.

[0052] In this embodiment, a fault diagnosis method for nonlinear systems based on sliding mode observers and DCNNs is described, such as... Figure 1 As shown, it includes the following steps:

[0053] Step 1: Construct a mathematical model of the nonlinear system, which takes into account concurrent failures of the system actuators and sensors, as well as nonlinear terms and external disturbances.

[0054] The mathematical model of the nonlinear system is:

[0055] ;

[0056] in, Represents the state vector of a nonlinear system. for The derivative; Represents the input vector of a nonlinear system; Represents the output vector of a nonlinear system; Represents the actuator fault vector. Represents the sensor fault vector; Indicates unknown bounded interference. Represents nonlinear continuous terms; This is the state gain matrix; To control the input gain matrix; This is the output gain matrix; For actuator fault matrix; This is the external interference gain matrix.

[0057] Step 2: Design a sliding mode observer for the nonlinear system;

[0058] The sliding mode observer is shown in the following formula:

[0059] ;

[0060] in, This represents the state estimate of a nonlinear system. for The derivative; Output estimates for nonlinear systems; This represents the estimated value of the actuator failure. This represents the estimated sensor fault value; All are sliding mode observer gain matrices; These are all sliding mode control terms, as shown in the following formulas:

[0061] ;

[0062] in, All are sliding surfaces; , All are sufficiently large positive matrices.

[0063] Step 3: Define the sliding mode observer residuals and establish the residual equations;

[0064] The sliding mode observer residuals are defined, including actuator fault residuals and sensor fault residuals, as shown in the following formula:

[0065] ;

[0066] The residual equation is further established as shown in the following formula:

[0067] ;

[0068] in, These are the actuator fault residuals and sensor fault residuals, respectively.

[0069] Step 4: Determine the residual threshold for the sliding mode observer;

[0070] 1) When a nonlinear system is subjected to random external disturbances, a sliding mode observer is used to monitor the system; at this time, actuator faults and sensor faults should be in a state of non-occurrence (i.e., The state estimate is obtained through a sliding mode observer. Actuator fault estimate Sensor fault estimate Then, the actuator fault residual is calculated. and sensor fault residual Record the actuator fault residuals and sensor fault residuals over a period of time to obtain the actuator fault residual sequence and sensor fault residual sequence;

[0071] 2) For actuator fault residuals Find the maximum value in its sequence. For sensor fault residuals Find the maximum value in its sequence. ;

[0072] 3) Repeat steps 1)-2), simulate N external random disturbances to the system operating state, and obtain the actuator fault residuals under each simulation. maximum value and sensor fault residual maximum value , ;

[0073] 4) Actuator fault residuals obtained from N simulations maximum value and sensor fault residual maximum value The actuator fault residual threshold is obtained. and sensor fault residual threshold As shown in the formula below:

[0074] ;

[0075] .

[0076] Step 5: Use a sliding mode observer to monitor the nonlinear system, obtain residual information, and preprocess the residual information based on the residual threshold to establish a residual signal dataset;

[0077] First, a sliding mode observer is used to monitor the nonlinear system, collecting data including actuator fault residuals. and sensor fault residual The actual residual signal is used to determine if the nonlinear system has a fault when the residual signal exceeds the residual threshold; and the residual signal is timestamped to ensure that the signal is strictly aligned in the time dimension.

[0078] Then, the residual signal is standardized (e.g., Z-score standardization) to create a residual signal dataset;

[0079] Finally, the residual signal dataset obtained after the above processing is divided into a training set and a test set.

[0080] Step 6: Based on the obtained parameters of the sliding mode observer, and with the help of the powerful feature extraction and pattern recognition capabilities of the deep learning model DCNN, the residual signal generated by the sliding mode observer is analyzed in depth to achieve fault diagnosis of the nonlinear system.

[0081] The deep learning model DCNN includes an input layer, a multi-layer convolutional module, a deep convolutional layer group, an attention mechanism module, a multi-scale feature fusion module, a fully connected layer module, and an output layer.

[0082] Considering the temporal correlation and multi-dimensional characteristics of the residual signal generated by the sliding mode observer, the input layer is designed to receive multi-dimensional time-series data. The input residual signal data is preprocessed to eliminate dimensional differences between different features, thereby accelerating the convergence speed of model training.

[0083] The design incorporates multiple consecutive convolutional layers to form a multi-layer convolutional module. Each convolutional layer group consists of two convolutional layers. The first convolutional layer uses a 5×1 kernel with a stride of 1; the second convolutional layer uses 128 3×1 kernels with a stride of 1. Through these two convolutional layers, local temporal features in the residual signal are initially extracted.

[0084] Deep convolutional layer group: Construct deeper convolutional modules to further mine complex features in the residual signal; after the convolutional modules, set multiple residual blocks, each containing two convolutional layers.

[0085] Each convolutional layer in the multi-layer convolutional module and the deep convolutional layer group is followed by an activation function, with Parametric ReLU selected to enhance the non-linear expressive power of the model.

[0086] Attention Mechanism Module: A temporal attention mechanism is introduced after the deep convolutional layer group to highlight the importance of different time points in the residual signal. In the specific implementation, the feature maps output by each convolutional layer of the deep convolutional layer group are first subjected to average pooling and max pooling operations in the channel dimension to obtain two different feature descriptions.

[0087] Multi-scale feature fusion module: To fully utilize features extracted from different levels, a multi-scale feature fusion module is designed. It upsamples or downsamples feature maps output from different convolutional layers to make them have the same size before fusing them to obtain fused features.

[0088] The fully connected layer module consists of three fully connected layers. These layers map the fused features obtained from the multi-scale feature fusion module, adjust the dimensions, and compare them with a residual threshold to determine the fault diagnosis result. The first fully connected layer has 512 neurons and uses Dropout (with a dropout rate of 0.5) to prevent overfitting. The second fully connected layer has 256 neurons. The final fully connected layer determines the number of neurons and the activation function based on the requirements of the fault diagnosis task.

[0089] The output layer outputs the fault diagnosis results of the nonlinear system.

[0090] In this embodiment, the deep learning model DCNN is trained using a training set and tested using a test set to obtain the deep learning model DCNN with optimal parameters.

[0091] Through the above steps, the present invention can achieve accurate fault diagnosis of nonlinear systems under the condition of multiple minor faults occurring simultaneously, providing strong support for ensuring the reliable operation of nonlinear systems.

[0092] This embodiment uses a robotic arm as an example to illustrate the effectiveness of the proposed nonlinear system fault diagnosis method based on sliding mode observer and DCNN.

[0093] In this embodiment, the mathematical model parameters of the nonlinear system are:

[0094]

[0095] The fault diagnosis results of the actuator in the nonlinear system are shown in Figures 2 and 3. Figure 2 As can be seen, this method can accurately detect the timing and severity of actuator failures; from Figure 3 The sensor fault diagnosis results shown in the figure demonstrate that effective observation can be achieved even for sensor faults.

[0096] As demonstrated by the results of the above embodiments, the nonlinear system fault diagnosis method based on sliding mode observer and DCNN proposed in this invention exhibits significant effectiveness in nonlinear systems such as robotic arms. By designing a sliding mode observer and using it to generate residual signals containing rich fault information, and then employing a designed deep learning model (DCNN) to perform deep diagnosis on the residual signals, the system can quickly and accurately diagnose multiple minor faults.

[0097] In terms of model parameter processing, this method can efficiently set and optimize relevant parameters, enabling collaborative work between modules and greatly improving the convenience and usability of the algorithm. The obtained and optimized model parameters can be directly applied to robotic arm systems in practical engineering, providing strong support for the stable operation of the system. It has broad application prospects and significant value for improving the reliability and safety of modern nonlinear control systems.

[0098] Example 2:

[0099] This embodiment proposes an electronic device, including: one or more processors, and a memory, wherein the memory is used to store instructions, and when the instructions are executed by the one or more processors, the one or more processors execute the nonlinear system fault diagnosis method based on sliding mode observer and DCNN.

[0100] The electronic device may be a mobile phone, computer, or tablet computer, etc., and includes a memory and a processor. The memory stores a computer program, which, when executed by the processor, implements the nonlinear system fault diagnosis method based on sliding mode observer and DCNN as described in the embodiments. It is understood that the electronic device may also include input / output (I / O) interfaces and communication components.

[0101] The processor is used to execute all or part of the steps in the nonlinear system fault diagnosis method based on sliding mode observer and DCNN as described in the above embodiments. The memory is used to store various types of data, which may include, for example, instructions for any application or method in the electronic device, as well as application-related data.

[0102] The processor can be implemented as an Application Specific Integrated Circuit (ASIC), Digital Signal Processor (DSP), Programmable Logic Device (PLD), Field Programmable Gate Array (FPGA), controller, microcontroller, microprocessor, or other electronic components, and is used to execute the nonlinear system fault diagnosis method based on sliding mode observer and DCNN described in the above embodiments.

[0103] Example 3:

[0104] This embodiment proposes a computer-readable storage medium that stores executable instructions. When these instructions are executed, if they are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium.

[0105] The computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the nonlinear system fault diagnosis method based on sliding mode observer and DCNN described in the various embodiments of this application.

[0106] The aforementioned storage media include: flash memory, hard disks, multimedia cards, card-type memory (e.g., SD (Secure Digital Memory Card) or DX (Memory Data Register, MDR) memory), random access memory (RAM), static random-access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic storage, disks, optical discs, servers, APP (Application) application stores, and other media capable of storing program verification codes. These media store computer programs, which, when executed by a processor, can implement the various steps of the aforementioned nonlinear system fault diagnosis method based on sliding mode observers and DCNN.

[0107] Example 4:

[0108] This embodiment proposes a computer program product, including a computer program or instructions, which, when executed by a processor, implements the aforementioned nonlinear system fault diagnosis method based on sliding mode observer and DCNN.

[0109] Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or part of the technical solution, can be embodied in the form of a computer program product.

[0110] The various embodiments in this application are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0111] The above description is merely a preferred embodiment of the present invention and does not constitute any limitation on the present invention. Any simple modifications, alterations, or equivalent structural changes made to the above embodiments based on the technical essence of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A fault diagnosis method for nonlinear systems based on sliding mode observer and DCNN, characterized in that, include: Construct a mathematical model for the nonlinear system that considers concurrent failures of the system's actuators and sensors, as well as nonlinear terms and external disturbances; Design a sliding mode observer for nonlinear systems; Define the sliding mode observer residuals and establish the residual equations; Sliding mode observer residuals include actuator fault residuals and sensor fault residuals; Determine the residual threshold for the sliding mode observer; A sliding mode observer is used to monitor the nonlinear system, obtain residual information, and preprocess the residual information based on the residual threshold to establish a residual signal dataset; Based on the relevant parameters of the sliding mode observer, a deep learning model DCNN is used to perform in-depth diagnosis of the residual signal generated by the sliding mode observer, thereby realizing fault diagnosis of nonlinear systems.

2. The nonlinear system fault diagnosis method based on sliding mode observer and DCNN according to claim 1, characterized in that, The mathematical model of the nonlinear system is shown in the following formula: ； in, Represents the state vector of a nonlinear system. for The derivative; Represents the input vector of a nonlinear system; Represents the output vector of a nonlinear system; Represents the actuator fault vector. Represents the sensor fault vector; Indicates unknown bounded interference. Represents nonlinear continuous terms; This is the state gain matrix; To control the input gain matrix; This is the output gain matrix; For actuator fault matrix; This is the external interference gain matrix.

3. The nonlinear system fault diagnosis method based on sliding mode observer and DCNN according to claim 2, characterized in that, The sliding mode observer is shown in the following formula: ； in, This represents the state estimate of a nonlinear system. for The derivative; Output estimates for nonlinear systems; This represents the estimated value of the actuator failure. This represents the estimated sensor fault value; All are sliding mode observer gain matrices; These are all sliding mode control terms, as shown in the following formulas: ； in, All are sliding surfaces; , All are positive matrices.

4. The nonlinear system fault diagnosis method based on sliding mode observer and DCNN according to claim 3, characterized in that, The actuator fault residual and sensor fault residual are given by the following formula: ； The residual equation is further established as shown in the following formula: ； in, These are the actuator fault residuals and the sensor fault residuals, respectively.

5. The nonlinear system fault diagnosis method based on sliding mode observer and DCNN according to claim 4, characterized in that, The specific method for determining the residual threshold of the sliding mode observer is as follows: 1) In nonlinear systems, a sliding mode observer is used to monitor the system; in this case, actuator faults and sensor faults should be in a state of non-occurrence, i.e. State estimates are obtained through a sliding mode observer. Actuator fault estimate Sensor fault estimate Then, the actuator fault residual is calculated. and sensor fault residual Record the actuator fault residuals and sensor fault residuals over a period of time to obtain the actuator fault residual sequence and sensor fault residual sequence; 2) For actuator fault residuals Find the maximum value in its sequence. For sensor fault residuals Find the maximum value in its sequence. ; 3) Repeat steps 1)-2), simulate N external random disturbances to the system operating state, and obtain the actuator fault residuals under each simulation. maximum value and sensor fault residual maximum value , ; 4) Actuator fault residuals obtained from N simulations maximum value and sensor fault residual maximum value The actuator fault residual threshold is obtained. and sensor fault residual threshold As shown in the formula below: ； 。 6. The nonlinear system fault diagnosis method based on sliding mode observer and DCNN according to claim 5, characterized in that, The specific method for establishing the residual signal dataset is as follows: First, a sliding mode observer is used to monitor the nonlinear system, collecting data including actuator fault residuals. and sensor fault residual The actual residual signal is used to determine if the nonlinear system has a fault when the residual signal exceeds the residual threshold; and the residual signal is timestamped to ensure that the signal is strictly aligned in the time dimension. Then, the residual signal is standardized to create a residual signal dataset; Finally, the residual signal dataset obtained after the above processing is divided into training set and test set.

7. The nonlinear system fault diagnosis method based on sliding mode observer and DCNN according to claim 6, characterized in that, The deep learning model DCNN includes, in sequence, an input layer, a multi-layer convolutional module, a group of deep convolutional layers, an attention mechanism module, a multi-scale feature fusion module, a fully connected layer module, and an output layer. The input layer can receive multi-dimensional time-series data and preprocess the input residual signal data to eliminate the dimensional differences between different features. The multi-layer convolutional module includes multiple consecutive convolutional layers, which initially extract local temporal features from the residual signal; The deep convolutional layer group constructs a deep convolutional module to further mine the complex features in the residual signal; multiple residual blocks are set after the convolutional module, and each residual block contains two convolutional layers; The attention mechanism module performs average pooling and max pooling operations on the feature maps output by each convolutional layer of the deep convolutional layer group in the channel dimension to obtain two different feature descriptions. The multi-scale feature fusion module upsamples or downsamples the feature maps output from different convolutional layers to make them have the same size before performing feature fusion to obtain fused features. The fully connected layer module includes three fully connected layers, which map the fused features obtained by the multi-scale feature fusion module, adjust the dimensions, and judge them with the residual threshold. The fault diagnosis result is determined based on the judgment result.

8. An electronic device for performing the nonlinear system fault diagnosis method based on sliding mode observer and DCNN as described in any one of claims 1-7, characterized in that, include: One or more processors, and a memory for storing instructions that, when executed by the one or more processors, cause the one or more processors to perform the described nonlinear system fault diagnosis method based on sliding mode observer and DCNN.

9. A computer-readable storage medium storing executable instructions for performing the nonlinear system fault diagnosis method based on sliding mode observer and DCNN as described in any one of claims 1-7, characterized in that, When the instruction is executed, it causes the processor to perform the nonlinear system fault diagnosis method based on sliding mode observer and DCNN.

10. A computer program product for executing the nonlinear system fault diagnosis method based on sliding mode observer and DCNN as described in any one of claims 1-7, characterized in that, This includes a computer program or instructions that, when executed by a processor, implement the aforementioned nonlinear system fault diagnosis method based on sliding mode observer and DCNN.

Images