Analog circuit early micro-failure diagnosis device, method and system

Abstract

The application provides a kind of analog circuit early small fault diagnosis method, device and system, the method comprises: obtaining the input voltage data and output voltage data of analog circuit for the resistance to be diagnosed, as encoding input data, input to bidirectional space-time attention fusion network fault diagnosis model, obtain fault diagnosis result;Bidirectional space-time attention fusion network fault diagnosis model includes: auto-encoder feature extraction network, for based on encoding input data, by encoding and decoding data features are extracted;Convolutional attention module, for extracting data features as input, respectively in channel dimension and spatial dimension sequentially inferring attention mapping, obtain refined features;Bidirectional GRU network, for by bidirectional semantic learning to refined features, obtain context semantics;Fusion inference module, for according to context semantics, carry out fusion inference, obtain fault diagnosis result.The application can carry out high-precision fault diagnosis to complex hardware circuit small fault.

Description

Technical Field

[0001] This invention relates to the field of tunnel construction technology, and in particular to a device, method and system for diagnosing early minor faults in analog circuits. Background Technology

[0002] Analog circuits are widely used in various sophisticated instruments and equipment, performing functions such as signal generation, amplification, transmission, and conversion. Faults in analog circuits will affect the performance of these instruments and equipment, leading to reduced functionality or even malfunction. In contrast, early, minor faults in hardware circuit components often show little difference between their early fault parameter values ​​and normal values, making accurate identification of these early faults difficult.

[0003] In recent years, with the in-depth research on analog circuit fault diagnosis, self-encoders have been widely used to diagnose analog circuit faults. For example, existing technologies have proposed an analog circuit fault diagnosis method and system based on an improved limit self-encoder. However, when faced with hardware circuits with numerous components and complex network structures, the existing relatively simple processing methods are difficult to accurately detect early minor faults of components. Therefore, it is necessary to design an early minor fault diagnosis method and system for analog circuits. Summary of the Invention

[0004] This invention proposes a method for early-stage minor fault diagnosis in analog circuits, enabling high-precision fault diagnosis of minor faults in complex hardware circuits. The method includes:

[0005] Obtain the input voltage and output voltage data of the analog circuit for the resistor to be diagnosed;

[0006] The input voltage data and output voltage data are used as encoded input data and input into the bidirectional spatiotemporal attention fusion network fault diagnosis model to obtain the fault diagnosis result.

[0007] The fault diagnosis model for bidirectional spatiotemporal attention fusion networks includes:

[0008] Autoencoder feature extraction networks are used to extract data features based on encoded input data through encoding and decoding.

[0009] The convolutional attention module is used to infer attention mappings sequentially in the channel dimension and spatial dimension, respectively, using the extracted data features as input, to obtain refined features;

[0010] A bidirectional GRU network is used to refine features and obtain contextual semantics through bidirectional semantic learning.

[0011] The fusion inference module is used to perform fusion inference based on contextual semantics to obtain fault diagnosis results.

[0012] This invention provides an early-stage minor fault diagnosis device for analog circuits, used for high-precision fault diagnosis of minor faults in complex hardware circuits. The device includes:

[0013] The voltage data acquisition module is used to acquire the input voltage data and output voltage data of the analog circuit for the resistor to be diagnosed.

[0014] The fault diagnosis module is used to input the input voltage data and output voltage data as encoded input data into the bidirectional spatiotemporal attention fusion network fault diagnosis model to obtain fault diagnosis results;

[0015] The fault diagnosis model for bidirectional spatiotemporal attention fusion networks includes:

[0016] Autoencoder feature extraction networks are used to extract data features based on encoded input data through encoding and decoding.

[0017] The convolutional attention module is used to infer attention mappings sequentially in the channel dimension and spatial dimension, respectively, using the extracted data features as input, to obtain refined features;

[0018] A bidirectional GRU network is used to refine features and obtain contextual semantics through bidirectional semantic learning.

[0019] The fusion inference module is used to perform fusion inference based on contextual semantics to obtain fault diagnosis results.

[0020] This invention proposes an early-stage micro-fault diagnosis system for analog circuits, used for high-precision fault diagnosis of micro-faults in complex hardware circuits. The system includes: an early-stage micro-fault diagnosis device for analog circuits, a microcontroller module, and a power supply module; wherein...

[0021] The power supply module is used to power microcontroller modules and analog circuits;

[0022] The microcontroller module is used to: generate PWM waveforms, filter the PWM model to generate an output reference voltage, and input it to the analog circuit.

[0023] In this circuit, the input voltage and output voltage data of the resistor to be diagnosed change with the resistance value of the resistor to be diagnosed.

[0024] This invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the above-described method for diagnosing early minor faults in analog circuits.

[0025] This invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the above-described method for diagnosing early minor faults in analog circuits.

[0026] This invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for diagnosing early minor faults in analog circuits.

[0027] This invention also provides a computer program product, which includes a computer program that, when executed by a processor, implements the above-described method for diagnosing early minor faults in analog circuits.

[0028] In this embodiment of the invention, input voltage data and output voltage data of the analog circuit for the resistor to be diagnosed are obtained; the input voltage data and output voltage data are used as encoded input data and input to a bidirectional spatiotemporal attention fusion network fault diagnosis model to obtain a fault diagnosis result; wherein, the bidirectional spatiotemporal attention fusion network fault diagnosis model includes: an autoencoder feature extraction network, used to extract data features based on encoded input data through encoding and decoding; a convolutional attention module, used to infer attention mappings sequentially in the channel dimension and spatial dimension respectively, using the extracted data features as input, to obtain refined features; a bidirectional GRU network, used to obtain contextual semantics from the refined features through bidirectional semantic learning; and a fusion inference module, used to perform fusion inference based on the contextual semantics to obtain a fault diagnosis result. Compared with the prior art, which only uses an autoencoder for analog circuit fault diagnosis, this embodiment of the invention uses a bidirectional spatiotemporal attention fusion network fault diagnosis model. First, a convolutional attention module is added to the autoencoder feature extraction network to enhance the expression of data features by the feature extraction network. Second, the addition of a bidirectional GRU module enables accurate evaluation of contextual semantics, resulting in better model performance and improved fault judgment. Attached Figure Description

[0029] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In the drawings:

[0030] Figure 1 This is a flowchart of the method for diagnosing early minor faults in analog circuits in an embodiment of the present invention;

[0031] Figure 2 This is a schematic diagram of the bidirectional spatiotemporal attention fusion network fault diagnosis model in an embodiment of the present invention;

[0032] Figure 3 This is a schematic diagram of the convolutional block attention module in an embodiment of the present invention;

[0033] Figure 4 This is a schematic diagram of the channel attention module in an embodiment of the present invention;

[0034] Figure 5 This is a schematic diagram of the spatial attention module in an embodiment of the present invention;

[0035] Figure 6 This is a structural diagram of a bidirectional GRU network in an embodiment of the present invention;

[0036] Figure 7 This is a schematic diagram of an early minor fault diagnosis device for analog circuits in an embodiment of the present invention;

[0037] Figure 8 This is a schematic diagram of an early minor fault diagnosis system for analog circuits in an embodiment of the present invention;

[0038] Figure 9 This is a structural diagram of some operational amplifier circuits in an embodiment of the present invention;

[0039] Figure 10 This is a schematic diagram of a computer device in an embodiment of the present invention. Detailed Implementation

[0040] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings. Here, the illustrative embodiments of the present invention and their descriptions are used to explain the present invention, but are not intended to limit the present invention.

[0041] In the description of this specification, the terms "comprising," "including," "having," and "containing" are open-ended terms, meaning that they include but are not limited to. The terms "an embodiment," "a specific embodiment," "some embodiments," and "for example," etc., refer to specific features, structures, or characteristics described in connection with that embodiment or example that are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, or characteristics described can be combined in any suitable manner in one or more embodiments or examples. The order of steps involved in the various embodiments is used to illustrate the implementation of this application, and the order of steps is not limited and can be adjusted appropriately as needed.

[0042] Figure 1This is a flowchart of an early minor fault diagnosis method for analog circuits in an embodiment of the present invention. The method includes:

[0043] Step 101: Obtain the input voltage data and output voltage data of the analog circuit for the resistor to be diagnosed;

[0044] Step 102: The input voltage data and output voltage data are used as encoded input data and input into the bidirectional spatiotemporal attention fusion network fault diagnosis model to obtain the fault diagnosis result;

[0045] The fault diagnosis model for bidirectional spatiotemporal attention fusion networks includes:

[0046] Autoencoder feature extraction networks are used to extract data features based on encoded input data through encoding and decoding.

[0047] The convolutional attention module is used to infer attention mappings sequentially in the channel dimension and spatial dimension, respectively, using the extracted data features as input, to obtain refined features;

[0048] A bidirectional GRU network is used to refine features and obtain contextual semantics through bidirectional semantic learning.

[0049] The fusion inference module is used to perform fusion inference based on contextual semantics to obtain fault diagnosis results.

[0050] In one embodiment, the input voltage data and output voltage data are used as encoded input data, including:

[0051] The discrete data forms of the input voltage data and output voltage data are converted into two-dimensional matrix forms; the two-dimensional matrix forms are then used as coded input data.

[0052] Figure 2 This is a schematic diagram of the bidirectional spatiotemporal attention fusion network fault diagnosis model in an embodiment of the present invention. This bidirectional spatiotemporal attention fusion network fault diagnosis model is the core part of the method of the present invention. In view of the problem that the data samples of analog circuits are small, which leads to the decline in the generalization ability of existing models, the present invention proposes a bidirectional spatiotemporal attention fusion network fault diagnosis model.

[0053] See Figure 2 The autoencoder feature extraction network includes multiple convolutional layers and multiple pooling layers;

[0054] The encoding and decoding process of an autoencoder feature extraction network is represented as follows:

[0055]

[0056] Where, x m For encoding input data, h m (xm R represents the encoded data. m The data is decoded and represented as extracted data features. To encode network parameters, To decode network parameters, For the coding network part, This is the decoding network part. In formula (1), m∈{v,c}, the autoencoder feature extraction network can extract the input voltage data (x... ν ), Output voltage data (x c Input it into the network for learning.

[0057] The Convolutional Block Attention Module (CBAM) is a simple yet effective attention module for feedforward convolutional neural networks. Given an intermediate feature map as input, this module infers the attention map sequentially along two independent dimensions (channel and spatial).

[0058] Figure 3 This is a schematic diagram of the convolutional block attention module in an embodiment of the present invention. The convolutional block attention module includes:

[0059] The channel attention module is used to extract data features from the autoencoder feature extraction network output, infer attention mappings sequentially along the channel dimension, and obtain the channel attention mapping output.

[0060] The first isotope multiplication module is used to perform isotope multiplication on the decoded representation data and the channel attention mapping output to obtain intermediate features;

[0061] The spatial attention module is used to infer attention mappings sequentially in the spatial dimension for intermediate features, and obtain spatial attention mapping outputs.

[0062] The second isotope multiplication module is used to perform isotope multiplication on the spatial attention mapping output and intermediate features to obtain refined features.

[0063] Figure 4 This is a schematic diagram of the channel attention module in an embodiment of the present invention. Figure 5 This is a schematic diagram of the spatial attention module in an embodiment of the present invention. See also... Figure 4 and Figure 5 The formulas for the two modules can be compiled.

[0064] The channel attention module is represented as follows:

[0065] F' = M c (R m )=σ2(MLP(AvgPool(R m ))+MLP(MaxPool(Rm )))(2)

[0066] The spatial attention module is represented as follows:

[0067] F” = M s (F')=σ2(f 7×7 ([AvgPool(F');MaxPool(F')])) (3)

[0068] Among them, R m For data features, F' is the channel attention mapping output; F” is the spatial attention mapping output, MLP is the multilayer perceptron mapping, M c To infer attention mappings sequentially along the channel dimension, M s To infer attention mappings sequentially in the spatial dimension, AvgPool is the average pooling operation, and MaxPool is the max pooling operation. 7×7 The diagram shows a convolutional layer with a kernel size of 7×7, where σ2 is the Sigmoid activation function.

[0069] GRU networks are widely used in learning and predicting the semantics of text or context. However, a one-way GRU network can only predict the current information based on the semantic information of the previous time, while a two-way GRU (Bi-GRU) can combine the semantic information of the past and the future to analyze and diagnose faults. Figure 6 This is a structural diagram of a bidirectional GRU network in an embodiment of the present invention. The bidirectional GRU network is composed of two stacked GRU modules, and is represented as follows:

[0070]

[0071] in, To perform semantic learning from past time states to the current time state. To perform semantic learning from a future time state to the current time state, the arrows indicate the direction of time movement, h t-1 As the current learning state, h t The contextual semantics output by the bidirectional GRU network, f is the semantic learning function, and x is the semantic contextual semantics output by the network. t These are the refined features output by the attention module of the convolutional block. and In hidden state, and This is the weight matrix. and This is a paranoid trait.

[0072] In one embodiment, the fusion inference module includes:

[0073] The splicing and fusion module is used to perform vector splicing operations on the input voltage semantics and output voltage semantics in the context semantics of the bidirectional GRU network output to obtain the spliced ​​fusion vector;

[0074] The inference network module is used to perform fault inference based on the fusion vector and obtain the fault inference results.

[0075] Data fusion can increase network confidence and improve the accuracy of network fault diagnosis. This invention embodiment performs vector concatenation on the input voltage semantics and output voltage semantics to achieve fusion. After obtaining the contextual semantics, the concatenation and fusion are performed, as specifically shown below:

[0076]

[0077] in, This represents the vector concatenation operation; Fusion represents the concatenated fused vector. The calculated vector is then input into the inference network module (MLP) for fusion and fault inference, as shown below:

[0078] Inf = MLP(Fusion; θ) MLP (6)

[0079] In the above formula: Inf represents the inference result; θ MLP To infer network parameters.

[0080] In one embodiment, the loss function of the bidirectional spatiotemporal attention fusion network fault diagnosis model during training is expressed as follows:

[0081]

[0082] Among them, L focal (P) represents the task loss, L sim For similarity loss, For reconstruction loss, β and η are weighting factors of the loss function;

[0083] L focal (P)=-α(1-P) γ log(P) (8)

[0084] Where P is the probability that the network classifies a sample as normal, and α is the weighting factor; (1-P) γ To indicate the adjustment factor, γ is an adjustable focusing parameter;

[0085] L sim =CMD K (M v M c (9)

[0086] Among them, M vM c These are the input voltage semantics and output voltage semantics in the context semantics of the bidirectional GRU network output, respectively.

[0087]

[0088] Where, x m To encode input data, R m For data features, This represents the squared L2 norm.

[0089] In specific implementation, L focal (P) is also known as focal loss, L focal (P) adds an adjustment factor to the cross-entropy loss function. Since Focal loss can adaptively adjust the influence of samples on parameters, it is used as the target loss to guide learning. The hyperparameters can be set to α = 0.25; γ = 2.

[0090] Since autoencoder networks complete the splicing and fusion operation at the MLP layer, it is difficult to effectively learn the interaction relationship between data. Therefore, in order to obtain comprehensive semantic information, this embodiment of the invention constructs a similarity loss function for constraining the consistency of contextual expression. The distance distribution between data is measured by the center distance deviation (CMD), which can effectively reduce the computational cost of the network.

[0091] This invention also proposes an early minor fault diagnosis device for analog circuits, see [link to relevant documentation]. Figure 7 ,include:

[0092] Voltage data acquisition module 701 is used to acquire input voltage data and output voltage data of analog circuit for the resistor to be diagnosed;

[0093] The fault diagnosis module 702 is used to input the input voltage data and output voltage data as encoded input data into the bidirectional spatiotemporal attention fusion network fault diagnosis model to obtain fault diagnosis results;

[0094] The fault diagnosis model for bidirectional spatiotemporal attention fusion networks includes:

[0095] Autoencoder feature extraction networks are used to extract data features based on encoded input data through encoding and decoding.

[0096] The convolutional attention module is used to infer attention mappings sequentially in the channel dimension and spatial dimension, respectively, using the extracted data features as input, to obtain refined features;

[0097] A bidirectional GRU network is used to refine features and obtain contextual semantics through bidirectional semantic learning.

[0098] The fusion inference module is used to perform fusion inference based on contextual semantics to obtain fault diagnosis results.

[0099] This invention also proposes an early minor fault diagnosis system for analog circuits, see [link to relevant documentation]. Figure 8 The system includes an early minor fault diagnosis device for analog circuits, a microcontroller module (MCU), and a power supply module; among which,

[0100] The power supply module is used to power microcontroller modules and analog circuits;

[0101] The microcontroller module is used to: generate PWM waveforms, filter the PWM model to generate an output reference voltage, and input it to the analog circuit.

[0102] In this circuit, the input voltage and output voltage data of the resistor to be diagnosed change with the resistance value of the resistor to be diagnosed.

[0103] In one embodiment, the power supply module includes a digital power supply circuit and an analog power supply circuit, wherein,

[0104] The digital power supply circuit controls the output voltage to reach a first preset voltage value by controlling the resistance value of the voltage divider resistors, which is used to power the microcontroller module.

[0105] The analog power supply circuit uses the second and third preset voltage values ​​generated by the step-down circuit to power the analog circuit.

[0106] In this embodiment of the invention, the first voltage preset value is 3.3V, the second voltage preset value is +13.7V, and the third constant voltage preset value is -13.7V.

[0107] In practice, the STM32F103C8T6 is selected as the CPU, with 64K of on-chip FLASH and 20K of on-chip RAM. The main frequency is set to 60MHz to meet the processing requirements. The digital power supply circuit adopts a DC-DC converter control chip.

[0108] In one embodiment, the system further includes a communication module for communicating with other systems. The communication module can be RS485. The power module further includes a communication power supply circuit, which uses a DC-DC converter control chip. Both the communication power supply circuit and the digital power supply circuit control their output voltage by controlling the resistance value of the voltage divider resistors, so that the output voltage can reach 3.3V and 5V respectively. The 5V power supply is directly supplied to the RS485.

[0109] Figure 9The diagram shows the structure of some operational amplifier circuits in an embodiment of the present invention. The first operational amplifier uses a voltage follower to stabilize the output voltage to -2.5V to 2.5V. The second operational amplifier uses virtual short and virtual open calculations to ensure that the pin output is between -10V and 10V.

[0110] The following is a specific embodiment to illustrate the application of the methods, apparatus, and systems proposed in this invention. In this embodiment, the analog circuit is a DC-DC operational amplifier circuit, which can realize voltage conversion from -2.5V to -10V. The schematic diagram of its DC-DC operational amplifier circuit is shown below. Figure 9 As shown. In operational amplifier circuits, resistor values ​​often decrease as the circuit runs, and these resistances affect the output voltage of the operational amplifier circuit. Therefore, soft fault diagnosis of resistors is essential. Using... Figure 9 In this embodiment, R1 and R4 are used as resistors to be diagnosed, with nominal values ​​of 5.1kΩ and 100kΩ, respectively. The degradation of the resistance value is calibrated by collecting changes in the voltage values ​​of U1 and U2. Within the range of 10% to 30% resistance degradation, nine soft fault modes are defined, numbered F1 to F9, as shown in Table 1.

[0111] Table 1 Fault Mode Labeling

[0112] model R1 / R3 (kΩ) Label model R1 / R3 (kΩ) Label f1 90 / 4.59 F1 f6 80 / 3.57 F6 f2 90 / 4.08 F2 f7 70 / 4.59 F7 f3 90 / 3.57 F3 f8 70 / 4.08 F8 f4 80 / 4.59 F4 f9 70 / 3.57 F9 f5 80 / 4.08 F5

[0113] In this embodiment of the invention, the method for diagnosing early minor faults in analog circuits is implemented using Python code. The hardware environment is an AMD Ryzen 5 2600X Six-Core Processor CPU with a frequency of 3.60GHz and an NVIDIA GeForce GTX 1660 GPU. The platform versions used are Python 3.7.9 and torch 1.2.0. For each of the nine types of soft fault data, 400 coded input data points (input voltage data and output voltage data) were used for experiments, and the training and test sets were divided in a 3:1 ratio.

[0114] By selecting parameters for the bidirectional spatiotemporal attention fusion network fault diagnosis model, the model was trained. The model was then validated on a test set by saving the model with the highest classification accuracy. The average classification accuracy was 99.78% for 100 encoded input data for each category. The specific results are shown in Table 2.

[0115] Table 2. Classification results of the test set

[0116]

[0117] In summary, the method, apparatus, and system proposed in this invention obtain input voltage data and output voltage data of the analog circuit for the resistor to be diagnosed; the input voltage data and output voltage data are used as encoded input data and input to a bidirectional spatiotemporal attention fusion network fault diagnosis model to obtain a fault diagnosis result; wherein, the bidirectional spatiotemporal attention fusion network fault diagnosis model includes: an autoencoder feature extraction network, used to extract data features based on encoded input data through encoding and decoding; a convolutional attention module, used to infer attention mappings sequentially in the channel dimension and spatial dimension respectively, using the extracted data features as input, to obtain refined features; a bidirectional GRU network, used to obtain contextual semantics from the refined features through bidirectional semantic learning; and a fusion inference module, used to perform fusion inference based on the contextual semantics to obtain a fault diagnosis result. Compared with existing technologies that only use autoencoders for analog circuit fault diagnosis, which suffer from low classification accuracy due to unclear early minor fault sample features and reduced model generalization ability due to sample categories, this invention adopts a bidirectional spatiotemporal attention fusion network fault diagnosis model. First, a convolutional attention module is added to the autoencoder feature extraction network to enhance the feature extraction network's expression of data features. Second, a bidirectional GRU module is added to achieve accurate evaluation of contextual semantics, resulting in better model performance and improved fault judgment. Finally, a similarity constraint function is set for the convolutional autoencoder to capture independent and interactive features with spatiotemporal characteristics.

[0118] In addition, data acquisition and experiments were conducted on analog circuits in the embodiments of the present invention. The results show that a good diagnostic effect can be achieved. However, this fault diagnosis system can not only be used in analog circuits, but also applied to the fault diagnosis of components in digital circuits, providing a method for diagnosing early minor faults of hardware circuit components.

[0119] This invention also provides a computer device. Figure 10 This is a schematic diagram of a computer device in an embodiment of the present invention. The computer device 1000 includes a memory 1010, a processor 1020, and a computer program 1030 stored in the memory 1010 and executable on the processor 1020. When the processor 1020 executes the computer program 1030, it implements the above-mentioned method for diagnosing early minor faults in analog circuits.

[0120] This invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for diagnosing early minor faults in analog circuits.

[0121] This invention also provides a computer program product, which includes a computer program that, when executed by a processor, implements the above-described method for diagnosing early minor faults in analog circuits.

[0122] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program operating systems. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program operating system implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0123] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program business systems according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0124] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0125] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0126] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for diagnosing early minor faults in analog circuits, characterized in that, include: Obtain the input voltage and output voltage data of the analog circuit for the resistor to be diagnosed; The input voltage data and output voltage data are used as encoded input data and input into the bidirectional spatiotemporal attention fusion network fault diagnosis model to obtain the fault diagnosis result. The fault diagnosis model for bidirectional spatiotemporal attention fusion networks includes: Autoencoder feature extraction networks are used to extract data features based on encoded input data through encoding and decoding. The convolutional attention module is used to infer attention mappings sequentially in the channel dimension and spatial dimension, respectively, using the extracted data features as input, to obtain refined features; A bidirectional GRU network is used to refine features and obtain contextual semantics through bidirectional semantic learning. The fusion inference module is used to perform fusion inference based on contextual semantics to obtain fault diagnosis results.

2. The method as described in claim 1, characterized in that, Using the input voltage data and output voltage data as encoded input data, including: The discrete data forms of the input voltage data and output voltage data are converted into two-dimensional matrix forms; the two-dimensional matrix forms are then used as coded input data.

3. The method as described in claim 1, characterized in that, The autoencoder feature extraction network includes multiple convolutional layers and multiple pooling layers; The encoding and decoding process of an autoencoder feature extraction network is represented as follows: Where, x m For encoding input data, h m (x m R represents the encoded data. m The data is decoded and represented as extracted data features. To encode network parameters, To decode network parameters, For the coding network part, This is for decoding the network part.

4. The method as described in claim 1, characterized in that, The convolutional block attention module includes: The channel attention module is used to extract data features from the autoencoder feature extraction network output, infer attention mappings sequentially along the channel dimension, and obtain the channel attention mapping output. The first isotope multiplication module is used to perform isotope multiplication on the decoded representation data and the channel attention mapping output to obtain intermediate features; The spatial attention module is used to infer attention mappings sequentially in the spatial dimension for intermediate features, and obtain spatial attention mapping outputs. The second isotope multiplication module is used to perform isotope multiplication on the spatial attention mapping output and intermediate features to obtain refined features.

5. The method as described in claim 4, characterized in that, The channel attention module is represented as follows: F'=M c (R m )=σ2(MLP(AvgPool(R m ))+MLP(MaxPool(R m ))) The spatial attention module is represented as follows: F”=M s (F')=σ2(f 7×7 ([AvgPool(F');MaxPool(F')])) Among them, R m For data features, F' is the channel attention mapping output; F” is the spatial attention mapping output, MLP is the multilayer perceptron mapping, M c To infer attention mappings sequentially along the channel dimension, M s To infer attention mappings sequentially in the spatial dimension, AvgPool is the average pooling operation, and MaxPool is the max pooling operation. 7×7 The diagram shows a convolutional layer with a kernel size of 7×7, where σ2 is the Sigmoid activation function.

6. The method as described in claim 1, characterized in that, A bidirectional GRU network consists of two stacked GRU modules, and is represented as follows: Among them, h t To perform semantic learning from past time states to the current time state. To perform semantic learning from a future time state to the current time state, the arrows indicate the direction of time movement, h t-1 As the current learning state, h t The contextual semantics output by the bidirectional GRU network, f is the semantic learning function, and x is the semantics. t These are the refined features output by the attention module of the convolutional block. and In hidden state, and This is the weight matrix. and This is a paranoid trait.

7. The method as described in claim 1, characterized in that, The fusion inference module includes: The splicing and fusion module is used to perform vector splicing operations on the input voltage semantics and output voltage semantics in the context semantics of the bidirectional GRU network output to obtain the spliced ​​fusion vector; The inference network module is used to perform fault inference based on the fusion vector and obtain the fault inference results.

8. The method as described in claim 1, characterized in that, The loss function of the bidirectional spatiotemporal attention fusion network fault diagnosis model during training is expressed as follows: Among them, L focal (P) represents the task loss, L sim For similarity loss, For reconstruction loss, β and η are weighting factors of the loss function; L focal (P)=-α(1-P) γ log(P) Where P is the probability that the network classifies a sample as normal, and α is the weighting factor; (1-P) γ To indicate the adjustment factor, γ is an adjustable focusing parameter; L sim =CMD K (M v ,M c ) Among them, M v M c These are the input voltage semantics and output voltage semantics in the context semantics of the bidirectional GRU network output, respectively. Where, x m To encode input data, R m For data features, This represents the squared L2 norm.

9. A device for diagnosing early minor faults in analog circuits, characterized in that, include: The voltage data acquisition module is used to acquire the input voltage data and output voltage data of the analog circuit for the resistor to be diagnosed. The fault diagnosis module is used to input the input voltage data and output voltage data as encoded input data into the bidirectional spatiotemporal attention fusion network fault diagnosis model to obtain fault diagnosis results; The fault diagnosis model for bidirectional spatiotemporal attention fusion networks includes: Autoencoder feature extraction networks are used to extract data features based on encoded input data through encoding and decoding. The convolutional attention module is used to infer attention mappings sequentially in the channel dimension and spatial dimension, respectively, using the extracted data features as input, to obtain refined features; A bidirectional GRU network is used to refine features and obtain contextual semantics through bidirectional semantic learning. The fusion inference module is used to perform fusion inference based on contextual semantics to obtain fault diagnosis results.

10. A system for diagnosing early minor faults in analog circuits, characterized in that, It includes the analog circuit early minor fault diagnosis device, microcontroller module, and power supply module as described in claim 9; wherein, The power supply module is used to power microcontroller modules and analog circuits; The microcontroller module is used to: generate PWM waveforms, filter the PWM model to generate an output reference voltage, and input it to the analog circuit. In this circuit, the input voltage and output voltage data of the resistor to be diagnosed change with the resistance value of the resistor to be diagnosed.

11. The system as claimed in claim 10, characterized in that, The power supply module includes digital power supply circuits and analog power supply circuits, among which, The digital power supply circuit controls the output voltage to reach a first preset voltage value by controlling the resistance value of the voltage divider resistors, which is used to power the microcontroller module. The analog power supply circuit uses the second and third preset voltage values ​​generated by the step-down circuit to power the analog circuit.

12. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method according to any one of claims 1 to 8.

13. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method of any one of claims 1 to 8.

14. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the method according to any one of claims 1 to 8.

Images