Rotary transformer fault grading and intelligent diagnosis method and device

Abstract

The invention provides a rotary transformer fault grading and intelligent diagnosis method and device, and belongs to the technical field of fault detection, and the method comprises the steps: calculating a motor rotor position based on the structure and working principle of a rotary transformer, and taking the motor rotor position as a rotary transformer position output signal; establishing a fault grading system, dividing faults into a plurality of different grades, and defining typical fault types corresponding to the grades; sine and cosine output signals, rotary transformer position output signals and position angle error signals of the rotary transformer under different operation conditions are collected, and a sample database containing multi-channel images, multiple working conditions and multiple fault types is constructed; constructing a convolutional neural network model, and utilizing the sample database to train the ability of the model to identify the fault level and type; and realizing online fault identification of the rotary transformer by using the trained convolutional neural network model. The method not only improves the diagnosis precision and classification capability of the abnormal state of the rotary transformer, but also enhances the fault tolerance and robustness of the system in a complex environment.

Description

Technical Field

[0001] This invention relates to the field of fault detection technology, and in particular to a method and apparatus for fault classification and intelligent diagnosis of rotary transformers. Background Technology

[0002] Resolvers are widely used position feedback devices in permanent magnet synchronous motor systems, providing real-time rotor position signals to the control system for high-precision closed-loop control. In complex environments such as high temperature, vibration, humidity, and oil fumes, resolvers are prone to signal degradation, phase shift, and inductive imbalance. Their failure modes are diverse, their characteristics complex, and accurate diagnosis is challenging. Once a resolver fails, it often leads to a decline in motor control performance, and may even result in serious consequences such as system malfunction or shutdown. Currently, existing resolver fault detection methods mostly rely on traditional threshold judgment or abnormal signal monitoring, lacking systematic fault modeling and classification mechanisms. This makes it impossible to accurately identify various complex failure states, resulting in low diagnostic accuracy and practicality, and failing to meet the high reliability requirements of real-world industrial scenarios. Summary of the Invention

[0003] This invention provides a method and apparatus for fault classification and intelligent diagnosis of rotary transformers, which solves the problems of existing technologies being prone to diverse faults under complex operating conditions and lacking accurate classification and efficient diagnosis mechanisms.

[0004] In a first aspect, the present invention provides a method for fault classification and intelligent diagnosis of a rotary transformer, comprising: calculating the rotor position of a motor based on the structure and working principle of the rotary transformer, and using it as the position output signal of the rotary transformer; establishing a fault classification system, dividing faults into multiple different levels, and defining typical fault types corresponding to each level; collecting the sine and cosine output signals, the position output signal of the rotary transformer, and the position angle error signal of the rotary transformer under different operating conditions, and constructing a sample database containing multi-channel images, multiple operating conditions, and multiple fault types; constructing a convolutional neural network model, and using the sample database to train the model's ability to identify fault levels and types; and using the trained convolutional neural network model to realize online fault identification of the rotary transformer.

[0005] According to the method for fault classification and intelligent diagnosis of a resolver provided by the present invention, the rotor position of the motor is calculated based on the structure and working principle of the resolver and used as the resolver position output signal. The method includes: applying a sinusoidal excitation signal of a certain frequency to the primary winding, so that the two secondary windings respectively induce amplitude-modulated sinusoidal and cosine signals; and determining the resolver position output signal based on the amplitude-modulated sinusoidal and cosine signals.

[0006] According to the rotary transformer fault classification and intelligent diagnosis method provided by the present invention, multiple different levels include mild faults, moderate faults, and severe faults; among them, mild faults include four types: reverse polarity of sine channel, reverse polarity of cosine channel, reverse coupling of dual channels, and reverse direction of excitation source; moderate faults include five types: phase offset decoding error, asymmetric channel amplitude response, phase symmetry disturbance, high-frequency coupling noise interference, and channel baseline drift; severe faults include one type: output signal freezing-type functional failure.

[0007] According to the fault classification and intelligent diagnosis method for rotary transformers provided by the present invention, the sine and cosine output signals, resolver position output signals, and position angle error signals of the rotary transformer under different operating conditions are collected to construct a sample database containing multi-channel images, multiple operating conditions, and multiple fault types. The method includes: sampling the sine and cosine output signals, resolver position output signals, and position angle error signals along a unified time axis; converting the three types of signals into grayscale images of the same preset size; and each sample group contains time-aligned three-channel image data.

[0008] According to the rotary transformer fault classification and intelligent diagnosis method provided by the present invention, the convolutional neural network model includes: an input layer for receiving grayscale images of sine and cosine output signals, rotary transformer position output signals, and position angle error signals; a feature extraction module comprising three levels of convolutional blocks, each level consisting of a convolutional layer, a ReLU activation function, and a max pooling layer, with the output feature map size decreasing progressively; a fully connected layer flattening the feature map and connecting it to neuron nodes for feature fusion; and an output layer employing a softmax function to output fault level classification results and fault type classification results in parallel.

[0009] According to the fault classification and intelligent diagnosis method for rotary transformers provided by the present invention, online fault identification of rotary transformers is realized by using a trained convolutional neural network model, including: integrating the convolutional neural network model as a software module into a motor control chip or a host computer fault monitoring module to monitor the fault level classification results and fault type classification results in real time; and triggering different fault-tolerant control mechanisms and / or alarm strategies according to the determined fault level classification results.

[0010] Secondly, the present invention also provides a rotary transformer fault classification and intelligent diagnostic device, comprising: The first processing module is used to calculate the motor rotor position based on the structure and working principle of the rotary transformer, and use it as the rotary transformer position output signal. The second processing module is used to establish a fault classification system, divide faults into multiple different levels, and define the typical fault types corresponding to each level. The third processing module is used to collect the sine and cosine output signals, resolver position output signals and position angle error signals of the rotary transformer under different operating conditions, and to build a sample database containing multi-channel images, multiple operating conditions and multiple fault types. The fourth processing module is used to construct a convolutional neural network model and train the model to identify fault levels and types using the sample database. The fifth processing module is used to realize online fault identification of rotary transformers using the trained convolutional neural network model.

[0011] Thirdly, the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the rotary transformer fault classification and intelligent diagnosis method as described above.

[0012] Fourthly, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the rotary transformer fault classification and intelligent diagnosis method as described above.

[0013] The method and apparatus for fault classification and intelligent diagnosis of rotary transformers provided by this invention have the following beneficial effects: This invention constructs a systematic resolver fault level classification mechanism. Based on factors such as fault manifestation, signal degradation characteristics, and repair difficulty, fault states are divided into different levels (e.g., mild, moderate, and severe), and corresponding identification standards are established in conjunction with typical failure modes. Simultaneously, based on data-driven artificial intelligence algorithms, an intelligent classification model suitable for this classification system is constructed, capable of automatically identifying faults of different levels during operation, achieving accurate detection of complex resolver fault states.

[0014] This invention, by introducing a multi-level fault modeling and intelligent identification strategy, not only improves the diagnostic accuracy and classification capability of abnormal states of resolvers, but also enhances the fault tolerance and robustness of the system in complex environments, and has good prospects for engineering applications. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in this 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 some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0016] Figure 1 This is a flowchart illustrating the fault classification and intelligent diagnosis method for rotary transformers provided by the present invention. Figure 2 This is a schematic diagram illustrating the working principle and position demodulation process of the resolver provided by the present invention; Figure 3 This is a fault classification and grading structure diagram of a resolver provided by the present invention; Figure 4 This is a schematic diagram of the fault identification and diagnosis results based on the trained CNN model of the present invention; Figure 5 This is a schematic diagram of the structure of the rotary transformer fault classification and intelligent diagnosis device provided by the present invention; Figure 6 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0018] It should be noted that, in the description of the embodiments of the present invention, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element. Those skilled in the art can understand the specific meaning of the above terms in the present invention according to the specific circumstances.

[0019] The terms "first," "second," etc., used in this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such terms can be used interchangeably where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class, without limiting the number of objects; for example, a first object can be one or more. Furthermore, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0020] The following is combined Figures 1-6 This invention describes the method and apparatus for fault classification and intelligent diagnosis of rotary transformers provided in embodiments of the present invention.

[0021] Figure 1 This is a flowchart illustrating the fault classification and intelligent diagnosis method for rotary transformers provided by the present invention, as shown below. Figure 1 As shown, including but not limited to the following steps: Step 101: Based on the structure and working principle of the rotary transformer, calculate the position of the motor rotor and use it as the position output signal of the rotary transformer.

[0022] Optionally, by applying a sinusoidal excitation signal of a certain frequency to the primary winding, the two secondary windings respectively induce amplitude-modulated sinusoidal and cosine signals; based on the amplitude-modulated sinusoidal and cosine signals, the resolver position output signal is determined.

[0023] Specifically, Figure 2 This is a schematic diagram illustrating the working principle and position demodulation process of the resolver provided by the present invention, as shown below. Figure 2 As shown, this step, based on the structure and working principle of the resolver, constructs a fundamental signal model for subsequent fault classification and diagnosis. A resolver typically includes one excitation winding (primary winding) and two spatially orthogonal receiving windings (secondary windings). By applying a sinusoidal excitation signal of a certain frequency to the primary winding, the two secondary windings respectively induce amplitude-modulated sinusoidal and cosine signals, reflecting the mechanical angular difference between the motor rotor and stator. The specific process is as follows: (1) Signal excitation and modulation signal acquisition. A sinusoidal excitation signal of a certain frequency is applied to the primary winding of the resolver:

[0024] in, V sinin For high-frequency excitation signals, V and ω These represent the injected amplitude and frequency, respectively. The two secondary windings respectively induce the corresponding sine and cosine modulation signals:

[0025] in, K For coupling coefficients, V sinout and V cosout These are sine and cosine modulated signals, respectively. θ This indicates the rotor position.

[0026] (2) Demodulation and Filtering. To extract rotation angle information from the modulated signal, the signal needs to be demodulated and filtered. Demodulation uses synchronous shaping. When the excitation signal is positive, the negative half-cycle of the modulated signal is flipped to positive; conversely, when the carrier is negative, the positive half-cycle of the modulated signal is flipped to negative. High-frequency components in the modulated signal are filtered out by integration.

[0027] in, V * sinout and V * cosout The output signal after filtering out high-frequency signals, φ sin( ωt ).

[0028] (3) Rotor angle calculation. Based on the sine and cosine components after filtering out high-frequency signals, the rotation angle is calculated using the arctangent function:

[0029] The resolver output signal obtained through the above decoding process will serve as the basis for subsequent fault identification and classification. It should be noted that the aforementioned resolver structure and signal processing method are known background technologies upon which this invention relies, used to establish the basis for fault modeling, and are not innovative aspects of this invention; therefore, they are not the focus of protection in this embodiment.

[0030] Step 102: Establish a fault classification system, divide faults into multiple different levels, and define the typical fault types corresponding to each level.

[0031] Figure 3 This is a fault classification and grading structure diagram of a resolver provided by the present invention, as shown below. Figure 3 As shown, in step 102, in order to achieve systematic identification and diagnosis of complex fault phenomena in resolver position sensors, a fault classification system based on three dimensions—impact level, compensability, and detection difficulty—is proposed. This system divides resolver faults into three levels: mild, moderate, and severe, corresponding to different system impacts and handling strategies, and further refines them into ten specific typical fault types.

[0032] (1) Minor faults (four types in total). Minor faults usually occur during the system power-on or debugging phase, mainly caused by hardware wiring errors. Although they interfere with the position information, they can be eliminated by manual inspection before startup or by simply modifying the wiring to restore normal operation. These faults include: sine channel polarity reverse connection fault, cosine channel polarity reverse connection fault, dual-channel reverse coupling fault, and excitation source directionality reverse connection fault.

[0033] a. Light-Level Fault 1: Phase-Sine Reversal (L1-PSR) – The polarity of the sine channel is reversed compared to the expected direction, causing the sine modulation signal to be out of phase, manifested as an overall reverse offset in the decoding angle:

[0034] b. Light-Level Fault 2: Phase-Cosine Reversal (L2-PCR) – The cosine channel polarity is reversed, causing the cosine component to be out of phase.

[0035] c. Light-Level Fault 3: Dual Channel Coupling (L3-DCC) – The sine and cosine channels are simultaneously reversed, causing the two components to synchronously reverse. However, because the relative angle relationship remains unchanged, there is a constant offset in the angle decoding.

[0036] d. Light-Level Fault 4: Excitation Direction Reversal (L4-EDR) – The excitation signal of the resolver is reversed, interfering with the synchronous shaping process.

[0037] The above four types of faults interfere with the decoding system, but do not cause signal loss. They can be quickly repaired by adjusting the hardware wiring, and are therefore classified as mild.

[0038] (2) Moderate faults (five types in total): Moderate faults are characterized by the resolver still being able to output valid sine and cosine signals, but the signal shape exhibits structural distortion, resulting in periodic errors or linear drift in the decoding angle, which needs to be corrected through algorithm compensation. These faults include: phase offset type angle decoding error faults, channel amplitude response asymmetry faults, phase symmetry structural disturbance faults, induced harmonics, and channel baseline drift type bias anomaly faults.

[0039] a. Medium-Level Fault 1: Phase-AngleDeviation (M1-PAD) causes the calculated angle to deviate constantly from the actual rotor angle:

[0040] in, This is the actual (desired) angle output value of the refractive index. This represents the error between the output value of the resolver angle and the actual angle value.

[0041] b. Medium-Level Fault 2: Voltage Imbalance Asymmetry (M2-VIA) – Inconsistent amplitudes between sine and cosine signals lead to increased nonlinearity in angle fluctuations.

[0042] c. Medium-Level Fault 3: Quadrature Symmetry Disturbance (M3-QSD): The sine and cosine signals deviate from the ideal 90° orthogonality, resulting in periodic distortion of the decoding angle.

[0043] d. Medium-Level Fault 4 (M4-HDI): In high-speed rotation or high-frequency interference situations, high-frequency harmonic components are superimposed on the induced signal of the secondary winding, which exacerbates the jitter of the filtered angle output.

[0044] e. Medium-Level Fault 5 (M5-BOD): The sine and cosine signals exhibit a DC component offset, disrupting signal symmetry and causing shifts and distortions in the angle mapping curve.

[0045] in, α This is the amplitude offset. β This is the phase offset. and The amplitudes of the third and fifth high-frequency harmonic components, Δ, are respectively. u sin and Δ u cos These are the offsets for the sine and cosine channels, respectively.

[0046] The above five types of faults can be corrected to a certain extent through digital compensation algorithms, but this will increase the system's processing burden and affect system stability. Therefore, they are classified as medium-level faults.

[0047] (3) Severe fault (one type): A severe fault refers to a completely abnormal turntable signal, in which effective angle information cannot be obtained, and the system needs to rely on a backup sensor or switch control modes to maintain operation. This type of fault includes: output signal freezing type functional failure fault caused by line damage or disconnection.

[0048] Heavy-Level Fault 1 (H1-LOF): The resolver output signal becomes constant, stagnant, or exhibits abnormal jumps. This manifests as a loss of periodicity in the sine and cosine waveforms, potentially caused by broken windings, detached solder joints, loose connections, or chip damage. Consequently, angle calculations become impossible.

[0049] in, for t The desired angle output value of the rotation at any given time. for t -1 is the actual angle output value of the resolver at time 1.

[0050] Severe failure is essentially a functional failure, requiring immediate shutdown or entry into fault-tolerant mode. It represents the most serious type of resolver failure.

[0051] By constructing the above-mentioned hierarchical system, clear discrimination labels are provided for subsequent intelligent detection models, thereby improving the diagnostic system's discriminative ability and interpretability.

[0052] Step 103: Collect the sine and cosine output signals, resolver position output signals, and position angle error signals of the rotary transformer under different operating conditions, and construct a sample database containing multi-channel images, multiple operating conditions, and multiple fault types.

[0053] Optionally, the sine and cosine output signals, the resolver position output signal, and the position angle error signal are sampled along a unified time axis; the three types of signals are converted into grayscale images of the same preset size; each set of samples contains time-aligned three-channel image data.

[0054] Specifically, based on the "three-level, ten-category" resolver fault classification system proposed in this invention, a sample database for intelligent identification tasks is constructed, and a convolutional neural network with resolver signal structure perception capability is designed for automated identification of fault levels and types.

[0055] To fully characterize the output features of the resolver sensor under different fault conditions, the constructed sample data employs a multi-signal fusion input structure. Each sample set consists of the following three signal channels: a. Waveform images of sine and cosine signals; b. Decoded spin converter position output signal; c. The error signal between the resolver output angle and the ideal angle, i.e., the position angle error signal.

[0056] The three types of signals are sampled along a unified time axis and constructed as model input units in the form of two-dimensional images (grayscale images), with a uniform size of 128×128 pixels. Each group of three images constitutes a complete sample, maintaining the correspondence between time series, and ensuring that the CNN model can discover the structural correlations and temporal features between signals.

[0057] To improve the robustness of the model, the sample data collection covers a variety of operating states, including various fault scenarios under different speeds, loads, and temperature rises. Transient and stability faults occurring at different time points are also sampled, ultimately forming a multi-dimensional sample database covering three levels and ten types.

[0058] Step 104: Construct a convolutional neural network model and use the sample database to train the model's ability to identify fault levels and types.

[0059] The constructed convolutional neural network model consists of an input layer, a feature extraction module, a fully connected module, and an output module. It has strong spatial feature extraction capabilities for revolved signals, and its structure is as follows: a. Input layer I1, the input is a grayscale image of size 128×128×1, accepts sine and cosine, position and error signal image input.

[0060] b. Feature Extraction Module (Layers C2-C4): C2 convolutional block: contains one 3×3 convolutional layer and one 2×2 max-pooling layer, outputting a 16×64×64 feature map, extracting low-level edge and local waveform features; C3 convolutional block: 3×3 convolution + 2×2 max-pooling, outputting a 32×32×32 image, aggregating periodic waveform information; C4 convolutional block: 3×3 convolution + 2×2 max-pooling, outputting a 64×16×16 image, learning high-level spatial and interference pattern features. All convolutional layers use ReLU as the activation function to enhance nonlinear modeling capabilities, and max-pooling effectively extracts main features while suppressing interference.

[0061] c. Fully connected layer, which uses a flattening operation to flatten the convolution result into a one-dimensional vector and connect it to 128 neuron nodes for high-level feature fusion.

[0062] d. Output layer: The output layer uses the softmax activation function and supports single output (classifying into ten fault types) or multi-task output (parallel output of fault level (3 types) and fault type (10 types) to improve the model's discrimination ability) depending on the task settings.

[0063] Figure 4This is a schematic diagram illustrating the fault identification and diagnosis results based on the trained CNN model of this invention, as shown below. Figure 4 As shown, a CNN model is trained based on the constructed sample database to achieve online diagnosis and accurate classification of resolver faults. This step aims to utilize the constructed fault sample database to perform supervised training on the CNN model, enabling it to identify different levels and types of faults in the resolver position sensor. The trained model is then integrated into the motor control system to achieve online fault detection.

[0064] CNN model training method. Image samples under different working conditions in the sample database are selected as training data. Three-channel images (sine and cosine output signals, resolver position output signals, and position error signals) are used as model inputs, and the corresponding fault levels and fault types are used as model output labels. Supervised learning is used for training.

[0065] During training, the neural network first extracts features and generates predicted outputs through forward propagation. Then, based on the difference between the true labels and the predicted results, it calculates the loss function and adjusts the network parameters through the backpropagation algorithm.

[0066] During forward propagation, the input layer of layer I1:

[0067] in, M =128, N =128, D =1 represents the height, width, and number of channels of the image.

[0068] C2, C3, and C4 convolutional layers:

[0069] in, For convolution operations, σ (·) represents the activation function ReLU. Pool (·) represents the maximum pooling operation. W l (d) for d Channel convolution kernel, D l-1 For input channel number, D l Number of output channels X l For the first l Output features of convolutional layers For the first l -1st convolutional layer d Output characteristics of each channel.

[0070] F5 fully connected layer:

[0071] in, f l-1 For the output vector of the previous layer, W l and b l These are the weight matrix and the bias, respectively.

[0072] O6 Output Layer:

[0073] in, y The output features of the output layer W out and b out These are the weight matrix and bias of the output layer, respectively. f These are the input features for the output layer.

[0074] During backpropagation, the overall loss is defined using the cross-entropy loss function:

[0075] in, k This represents the total number of resolver fault types. y i For the first in the neural network i Output characteristics of fault types This is the predicted output value.

[0076] Output layer gradient function:

[0077] Based on the chain rule, the gradient of the loss function with respect to each parameter of the network is calculated layer by layer:

[0078] in, P For parameters of any layer, η This is the learning rate.

[0079] Step 105: Use the trained convolutional neural network model to achieve online fault identification of the rotary transformer.

[0080] After model training is complete, the trained CNN model is embedded into the resolver output processing module as a parallel analysis module for position signal output, used to monitor the signal pattern of the resolver output in real time. Once the model identifies abnormal sine and cosine signal patterns or position error characteristics, the level and specific type of fault can be determined in real time.

[0081] In practical deployments, CNN models can be integrated as software modules into motor control chips or host computer fault monitoring modules. The fault levels output by the model can be used to trigger different fault-tolerant control mechanisms or alarm strategies, improving system reliability and safety.

[0082] In summary, this invention proposes a precise fault classification and detection method for resolver position sensors, which has the following significant advantages: a. Scientific fault classification system: Combining the difficulty of fault detection with the impact on the control system, a "three-level, ten-category" fault classification system is proposed to reasonably distinguish between mild, moderate and severe faults; b. Highly adaptable sample construction method: Integrates multi-channel resolver signals to cover various operating conditions and achieve comprehensive modeling of different feature patterns; c. Intelligent recognition model design: Design a CNN model with signal structure perception capability, taking into account both spatial feature extraction and fault mode recognition; d. Online deployment and diagnostic capabilities: The model is embedded in the resolver output module to achieve rapid online fault identification, which is suitable for various industrial servo system application scenarios.

[0083] The fault detection method proposed in this invention combines theoretical and engineering practicality, providing an effective guarantee for the high reliability of resolver sensors.

[0084] Figure 5 This is a schematic diagram of the structure of the rotary transformer fault classification and intelligent diagnostic device provided by the present invention, as shown below. Figure 5 As shown, the device includes: The first processing module is used to calculate the motor rotor position based on the structure and working principle of the rotary transformer, and use it as the rotary transformer position output signal. The second processing module is used to establish a fault classification system, divide faults into multiple different levels, and define the typical fault types corresponding to each level. The third processing module is used to collect the sine and cosine output signals, resolver position output signals and position angle error signals of the rotary transformer under different operating conditions, and to build a sample database containing multi-channel images, multiple operating conditions and multiple fault types. The fourth processing module is used to construct a convolutional neural network model and train the model to identify fault levels and types using the sample database. The fifth processing module is used to realize online fault identification of rotary transformers using the trained convolutional neural network model.

[0085] It should be noted that the rotary transformer fault classification and intelligent diagnosis device provided in this embodiment of the invention can execute the rotary transformer fault classification and intelligent diagnosis method described in any of the above embodiments during specific operation, which will not be elaborated in this embodiment.

[0086] Figure 6 This is a schematic diagram of the structure of the electronic device provided by the present invention, such as... Figure 6 As shown, the electronic device may include a processor 610, a communications interface 620, a memory 630, and a communications bus 640. The processor 610, communications interface 620, and memory 630 communicate with each other via the communications bus 640. The processor 610 can call logical instructions from the memory 630 to execute a method for classifying and intelligently diagnosing rotary transformer faults.

[0087] Furthermore, the logical instructions in the aforementioned memory 630 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0088] On the other hand, the present invention also provides a computer program product, the computer program product including a computer program stored on a non-transitory computer-readable storage medium, the computer program including program instructions, and when the program instructions are executed by a computer, the computer is able to execute the rotary transformer fault classification and intelligent diagnosis method provided in the above embodiments.

[0089] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform the rotary transformer fault classification and intelligent diagnosis method provided in the above embodiments.

[0090] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, 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 can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0091] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for fault classification and intelligent diagnosis of a rotary transformer, characterized in that, include: Based on the structure and working principle of the rotary transformer, the position of the motor rotor is calculated and used as the position output signal of the rotary transformer. Establish a fault classification system, divide faults into multiple different levels, and define the typical fault types corresponding to each level; Collect sine and cosine output signals, resolver position output signals, and position angle error signals of the rotary transformer under different operating conditions, and construct a sample database containing multi-channel images, multiple operating conditions, and multiple fault types; Construct a convolutional neural network model and use the sample database to train the model's ability to identify fault levels and types; Online fault identification of rotary transformers is achieved using a trained convolutional neural network model.

2. The method for fault classification and intelligent diagnosis of rotary transformers according to claim 1, characterized in that, Based on the structure and working principle of the resolver, the rotor position of the motor is calculated and used as the resolver position output signal, including: By applying a sinusoidal excitation signal of a certain frequency to the primary winding, the two secondary windings respectively induce amplitude-modulated sinusoidal and cosine signals. The position output signal of the resolver is determined based on the amplitude-modulated sine and cosine signals.

3. The method for fault classification and intelligent diagnosis of rotary transformers according to claim 1, characterized in that, Several different levels include minor faults, moderate faults, and severe faults; Among them, minor faults include four categories: reverse polarity of sine channel, reverse polarity of cosine channel, reverse coupling of dual channels, and reverse direction of excitation source. Moderate faults include five types: phase-shift decoding error, channel amplitude response asymmetry, phase symmetry disturbance, high-frequency coupling clutter interference, and channel baseline drift. Severe faults include a type of functional failure such as output signal freezing.

4. The method for fault classification and intelligent diagnosis of rotary transformers according to claim 1, characterized in that, The sine and cosine output signals, resolver position output signals, and position angle error signals of the rotary transformer were collected under different operating conditions to construct a sample database containing multi-channel images, multiple operating conditions, and multiple fault types, including: Sampling of sine and cosine output signals, resolver position output signals, and position angle error signals is performed along a unified time axis. Convert the three types of signals into grayscale images of the same preset size; Each sample set contains time-aligned three-channel image data.

5. The method for fault classification and intelligent diagnosis of rotary transformers according to claim 1, characterized in that, The convolutional neural network model includes: The input layer is used to receive grayscale images of sine and cosine output signals, resolver position output signals, and position angle error signals. The feature extraction module contains three levels of convolutional blocks, each consisting of a convolutional layer, a ReLU activation function, and a max pooling layer, with the output feature map size decreasing progressively at each level. Fully connected layers flatten the feature maps and connect them to neuron nodes for feature fusion. The output layer uses the softmax function to output the fault level classification results and fault type classification results in parallel.

6. The method for fault classification and intelligent diagnosis of rotary transformers according to claim 1, characterized in that, Using a trained convolutional neural network model, online fault identification of a rotary transformer is achieved, including: The convolutional neural network model is integrated as a software module into the motor control chip or the host computer fault monitoring module to monitor the fault level classification results and fault type classification results in real time. Based on the determined fault level classification results, different fault-tolerant control mechanisms and / or alarm strategies are triggered.

7. A fault classification and intelligent diagnostic device for a rotary transformer, characterized in that, include: The first processing module is used to calculate the motor rotor position based on the structure and working principle of the rotary transformer, and use it as the rotary transformer position output signal. The second processing module is used to establish a fault classification system, divide faults into multiple different levels, and define the typical fault types corresponding to each level. The third processing module is used to collect the sine and cosine output signals, resolver position output signals and position angle error signals of the rotary transformer under different operating conditions, and to build a sample database containing multi-channel images, multiple operating conditions and multiple fault types. The fourth processing module is used to construct a convolutional neural network model and train the model to identify fault levels and types using the sample database. The fifth processing module is used to realize online fault identification of rotary transformers using the trained convolutional neural network model.

8. An electronic 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 steps of the rotary transformer fault classification and intelligent diagnosis method as described in any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the rotary transformer fault classification and intelligent diagnosis method as described in any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the rotary transformer fault classification and intelligent diagnosis method as described in any one of claims 1 to 6.

Images