Equipment fault diagnosis method and system based on FrFT and information entropy quantification

Abstract

The invention discloses an equipment fault diagnosis method and system based on FrFT and information entropy quantization, and the method comprises the steps: obtaining a real-time preprocessed mechanical vibration signal of electrical equipment, and carrying out the fractional Fourier transform feature extraction, and obtaining feature data; performing information entropy quantification processing according to the feature data to obtain corresponding information entropy data; according to the feature data and the corresponding information entropy data, feature fusion is carried out through a weighted average method, and fused features are obtained; according to the fusion features, dimension reduction is carried out based on a principal component analysis method, and final extraction features are obtained; and inputting the final extracted features into a pre-trained equipment fault diagnosis model to obtain an equipment fault diagnosis result. According to the method, multi-scale time-frequency features can be extracted from complex mechanical vibration signals by using fractional Fourier transform, complexity information of the signals is extracted in combination with an information entropy quantization technology, and finally, fault diagnosis is performed on dimensionality reduction features, so that the accuracy and reliability of fault diagnosis are effectively improved.

Description

Technical Field

[0001] The present invention relates to a device fault diagnosis method and system based on FrFT and information entropy quantization, and belongs to the technical field of device fault diagnosis. Background Art

[0002] During the operation of electrical equipment, mechanical vibration is often a precursor to failure. Vibration signals contain information about the equipment's operating status. Abnormal mechanical vibration signals may indicate a malfunction, such as bearing damage, gear wear, or motor imbalance. Early detection and fault diagnosis of these vibration signals are crucial tasks in electrical equipment health monitoring.

[0003] Traditional mechanical vibration signal analysis methods, such as Fourier transforms, are primarily used to process stationary signals. However, mechanical vibration signals are often non-stationary and contain multiple frequency components. Traditional frequency domain analysis methods often fail to provide sufficient time-frequency information. Furthermore, as the equipment operating environment changes, the complexity and diversity of vibration signals continue to increase. Traditional methods struggle to efficiently extract effective features, which in turn affects the results of electrical equipment fault diagnosis. Summary of the Invention

[0004] The purpose of the present invention is to overcome the shortcomings of the existing technology and provide an equipment fault diagnosis method and system based on FrFT and information entropy quantization. First, by using fractional Fourier transform, multi-scale time-frequency features can be extracted from complex mechanical vibration signals, and the complexity information of the signal is extracted by combining information entropy quantization technology. Finally, the dimensionality reduction feature is used for fault diagnosis, which effectively improves the accuracy and reliability of fault diagnosis.

[0005] To achieve the above object, the present invention is implemented by adopting the following technical solutions:

[0006] In one aspect, the present invention discloses a device fault diagnosis method based on FrFT and information entropy quantization, comprising the following steps:

[0007] Obtain real-time pre-processed mechanical vibration signals of electrical equipment;

[0008] Performing fractional Fourier transform feature extraction on the preprocessed mechanical vibration signal to obtain feature data;

[0009] Performing information entropy quantization processing on the characteristic data to obtain corresponding information entropy data;

[0010] According to the feature data and the corresponding information entropy data, feature fusion is performed by a weighted average method to obtain a fused feature;

[0011] According to the fusion features, dimension reduction is performed based on principal component analysis to obtain the final extracted features;

[0012] The final extracted features are input into a pre-trained equipment fault diagnosis model to obtain equipment fault diagnosis results.

[0013] Furthermore, the step of obtaining the real-time pre-processed mechanical vibration signal of the electrical equipment includes the following steps:

[0014] Obtain real-time original mechanical vibration signals of electrical equipment;

[0015] According to the original mechanical vibration signal, recursively decompose it based on a preset number of decomposition layers by wavelet packet decomposition to obtain low-frequency data containing all low-frequency signals;

[0016] Reconstructing the signal by inverse wavelet packet transform according to the low-frequency data to obtain a preprocessed mechanical vibration signal;

[0017] Among them, the wavelet packet decomposition includes the following steps: during the first decomposition, the original mechanical vibration signal is decomposed by applying wavelet transform to obtain the low-frequency signal and high-frequency signal of the first decomposition; in addition to the first decomposition, the high-frequency signal obtained by the previous decomposition is low-pass filtered and high-pass filtered each time to obtain the low-frequency signal and high-frequency signal of the current decomposition.

[0018] Furthermore, the fractional Fourier transform feature extraction includes the following data:

[0019] According to the preprocessed mechanical vibration signal, fractional Fourier transform is performed based on a plurality of preset angle parameters to obtain a plurality of signal features;

[0020] Obtaining feature data according to multiple signal features;

[0021] The expression of the fractional Fourier transform is as follows:

[0022]

[0023] Where, Indicates angle-based The signal characteristics at time t obtained by performing fractional Fourier transform; Represents the process of fractional Fourier transform; represents the preprocessed mechanical vibration signal at time t; Represents the time variable The differential of Indicates about time The kernel function of the fractional Fourier transform of .

[0024] Furthermore, the time The kernel function of the fractional Fourier transform of The expression is as follows:

[0025]

[0026] Where, Indicates the preset angle for fractional Fourier transform; represents the imaginary part; Indicates the center frequency of the preprocessed mechanical vibration signal; Indicates the time of the pre-processed mechanical vibration signal.

[0027] Furthermore, the information entropy quantization process includes the following steps:

[0028] For each signal feature in the feature data, calculating the corresponding information entropy value and quantizing it to obtain a quantized information entropy value;

[0029] Obtaining information entropy data according to the quantized information entropy values corresponding to all signal features in the feature data;

[0030] The calculation expression of the information entropy value is as follows:

[0031]

[0032] Where, Indicates that the nth angle parameter is used The signal characteristics at time t obtained by performing fractional Fourier transform; Indicates signal characteristics The corresponding information entropy value; Indicates signal characteristics Probability density on the mth segment; Indicates signal characteristics characteristic length.

[0033] Furthermore, obtaining the fusion features includes the following steps:

[0034] For each signal feature in the feature data, assign a weight according to the corresponding quantized information entropy value;

[0035] According to all signal features and corresponding weights in the feature data, feature fusion is performed by weighted average method to obtain a high-dimensional fusion feature;

[0036] The expression of the fusion feature is as follows:

[0037]

[0038] Where, represents fusion features; Shows the nth angle parameter The quantitative information entropy value of Indicates that the nth angle parameter is used The signal characteristics at time t are obtained by performing fractional Fourier transform.

[0039] Furthermore, obtaining the final extracted features includes the following steps:

[0040] According to the fusion features, data standardization is performed to obtain a standard feature matrix;

[0041] According to the standard characteristic matrix, a covariance matrix is obtained;

[0042] Performing eigenvalue decomposition on the covariance matrix to obtain eigenvalues and eigenvectors corresponding to the covariance matrix; wherein an eigenvalue represents the importance of a principal component, and the corresponding eigenvector represents the direction of the corresponding principal component;

[0043] Select the eigenvectors with the largest eigenvalues in the preset number of eigenvalues as the principal component matrix;

[0044] According to the standard feature matrix and the principal component matrix, the final extracted features are obtained.

[0045] Furthermore, the expression of the final extracted features is as follows:

[0046]

[0047] Where, Represents the final extracted features; represents the standard feature matrix; Represents the principal component matrix composed of the eigenvectors with the first k largest eigenvalues, where k represents the preset number of principal components.

[0048] Furthermore, obtaining the equipment fault diagnosis result includes the following steps:

[0049] Performing standardization and time series slicing processing according to the final extracted features to obtain slice data;

[0050] According to the slice data, tensor format conversion and virtual batch dimension embedding are performed to obtain preprocessed data;

[0051] The preprocessed data is input into a pre-trained equipment fault diagnosis model to obtain an equipment fault diagnosis result; wherein, the equipment fault diagnosis model adopts a block loading technology based on a memory mapping mechanism and a parallel computing method based on multi-threaded reasoning.

[0052] In another aspect, the present invention discloses an equipment fault diagnosis system based on FrFT and information entropy quantization, which is applicable to the above-mentioned equipment fault diagnosis method based on FrFT and information entropy quantization, including:

[0053] A data acquisition module is used to obtain real-time pre-processed mechanical vibration signals of electrical equipment;

[0054] A feature extraction module, configured to perform fractional Fourier transform feature extraction based on the preprocessed mechanical vibration signal to obtain feature data;

[0055] An information entropy module is used to perform information entropy quantization processing based on the feature data to obtain corresponding information entropy data;

[0056] A feature fusion module is used to perform feature fusion based on the feature data and the corresponding information entropy data by a weighted average method to obtain a fused feature;

[0057] A feature dimensionality reduction module is used to perform dimensionality reduction based on the fusion features and the principal component analysis method to obtain the final extracted features;

[0058] The fault diagnosis module is used to input the final extracted features into a pre-trained equipment fault diagnosis model to obtain equipment fault diagnosis results.

[0059] Compared with the prior art, the present invention has the following beneficial effects:

[0060] The present invention's equipment fault diagnosis method and system based on FrFT and information entropy quantization, first, extracts features from the pre-processed mechanical vibration signal using fractional Fourier transform, and improves recognition accuracy through multi-angle, different-scale time-frequency analysis; second, based on information entropy quantization and feature fusion, it can effectively describe the signal's time-frequency characteristics, complexity information, and multi-scale performance, which is suitable for further fault diagnosis and signal classification; then, the final extracted features are obtained based on principal component analysis, effectively solving the "dimensionality disaster" problem caused by high-dimensional features, and providing an efficient, low-redundancy data foundation for subsequent pattern recognition, visualization analysis, and real-time monitoring; finally, fault diagnosis is performed based on a pre-trained equipment fault diagnosis model, effectively improving the accuracy and reliability of fault diagnosis. This method has strong adaptability and can be widely used in health monitoring and fault diagnosis of electrical equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 This is a flow chart of the device fault diagnosis method based on FrFT and information entropy quantization provided in Example 1 of the present invention. DETAILED DESCRIPTION

[0062] The present invention will be further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and are not intended to limit the scope of protection of the present invention.

[0063] Example 1

[0064] This embodiment 1 provides a device fault diagnosis method based on FrFT and information entropy quantization, including the following steps:

[0065] Obtain real-time pre-processed mechanical vibration signals of electrical equipment;

[0066] According to the pre-processed mechanical vibration signal, fractional Fourier transform feature extraction is performed to obtain feature data;

[0067] According to the characteristic data, information entropy quantization processing is performed to obtain the corresponding information entropy data;

[0068] According to the feature data and the corresponding information entropy data, feature fusion is performed through the weighted average method to obtain the fusion feature;

[0069] According to the fusion features, the dimension reduction is performed based on the principal component analysis method to obtain the final extracted features;

[0070] The final extracted features are input into the pre-trained equipment fault diagnosis model to obtain the equipment fault diagnosis results.

[0071] The technical concept of the present invention is as follows: first, the pre-processed mechanical vibration signal is subjected to fractional Fourier transform feature extraction, and the recognition accuracy is improved through multi-angle and different-scale time-frequency analysis; second, based on information entropy quantification and feature fusion, the signal's time-frequency characteristics, complexity information and multi-scale performance can be effectively described, which is suitable for further fault diagnosis and signal classification; then, the final extracted features are obtained based on the principal component analysis method, which effectively solves the "dimensionality disaster" problem caused by high-dimensional features and provides an efficient and low-redundancy data foundation for subsequent pattern recognition, visual analysis and real-time monitoring; finally, fault diagnosis is performed based on a pre-trained equipment fault diagnosis model, effectively improving the accuracy and reliability of fault diagnosis. This method has strong adaptability and can be widely used in health monitoring and fault diagnosis of electrical equipment.

[0072] like Figure 1 As shown, step 1: obtaining a real-time pre-processed mechanical vibration signal of the electrical equipment.

[0073] The specific steps include:

[0074] 1.1. Obtain the real-time original mechanical vibration signal of electrical equipment.

[0075] Specifically, a vibration sensor is used to collect real-time original mechanical vibration signals of electrical equipment.

[0076] 1.2. According to the original mechanical vibration signal, recursive decomposition is performed based on a preset number of decomposition layers through wavelet packet decomposition to obtain low-frequency data containing all low-frequency signals.

[0077] Wavelet transform can decompose the signal into low-frequency and high-frequency parts of different scales, and use the mother wave function to obtain wavelet bases of different scales through translation and scaling operations.

[0078] Wavelet packet decomposition is an extension of the wavelet transform. It not only decomposes the low-frequency part, but also further decomposes the high-frequency part. In this way, each layer generates more frequency bands, thus achieving a more refined time-frequency analysis of the signal.

[0079] Specifically, wavelet packet decomposition includes the following steps:

[0080] During the first decomposition, the wavelet transform is applied to decompose the original mechanical vibration signal to obtain the low-frequency signal and high-frequency signal of the first decomposition. The specific expression is as follows:

[0081]

[0082] Where, Represents the original mechanical vibration signal; represents the low-frequency signal obtained by the first decomposition; Represents the high-frequency signal obtained by the first decomposition.

[0083] The number of decomposition layers is set to L. Each decomposition process will produce low-frequency and high-frequency components. Except for the first decomposition, each time the high-frequency signal obtained from the previous decomposition is low-pass filtered and high-pass filtered respectively to obtain the low-frequency and high-frequency signals of the current decomposition.

[0084] The expression of the second decomposition is as follows:

[0085]

[0086] Where, represents the low-frequency signal obtained by the second decomposition; Represents the high-frequency signal obtained by the second decomposition.

[0087] The expression of the third decomposition is as follows:

[0088]

[0089] Where, represents the low-frequency signal obtained by the third decomposition; Represents the high-frequency signal obtained by the third decomposition.

[0090] And so on, until the predetermined number of decomposition levels is reached.

[0091] In wavelet packet decomposition, the decomposition of each layer is completed by filtering operation, and the signal Through a low-pass filter and high-pass filter Decompose it, the formula is as follows:

[0092] Low-pass filtering for low-frequency signals:

[0093]

[0094] Where, represents the low-frequency signal obtained by the lth decomposition; represents a low-pass filter; represents convolution; Represents the signal to be decomposed, such as the original mechanical vibration signal , the high-frequency signal obtained by the first decomposition .

[0095] High-pass filtering for high-frequency signals:

[0096]

[0097] Where, represents the high-frequency signal obtained by the lth decomposition; Represents a high-pass filter.

[0098] Through recursive decomposition, the high-frequency part of each layer will be further low-pass and high-pass filtered, so that the decomposition result of the first layer will produce 2 l frequency bands to form a complete frequency band structure.

[0099] 1.3. Based on the low-frequency data, the signal is reconstructed through inverse wavelet packet transform to obtain the preprocessed mechanical vibration signal.

[0100] When it is necessary to recover the signal from the decomposition result, the retained frequency band is reconstructed through the Inverse Wavelet Packet Transform (IWPT). The mechanical vibration signal is concentrated in the low-frequency range. In order to remove the high-frequency noise, the low-frequency data in each decomposition is selected. 、 … Perform inverse wavelet packet transform to obtain the signal after wavelet packet decomposition and denoising, that is, the preprocessed mechanical vibration signal.

[0101]

[0102] in, Represents the original mechanical vibration signal, is the preprocessed mechanical vibration signal at time t, is the reconstructed wavelet basis function of the selected frequency band at time t based on the scale factor a and the translation factor b.

[0103] Step 2: Perform fractional Fourier transform feature extraction based on the preprocessed mechanical vibration signal to obtain feature data.

[0104] The fractional Fourier transform (FrFT) is an extended Fourier transform that includes not only the conventional Fourier transform but also several other specialized transforms. The basic idea of the FrFT is to transform a signal from the time domain to the frequency domain using an adjustable rotation angle. The choice of rotation angle determines the "order" of the signal transformation. When the order is 1, the FrFT is a standard Fourier transform; when the order is 0, the FrFT is a unit transform.

[0105] In practical applications, the fractional Fourier transform can provide a flexible analysis framework. It can control the transformation accuracy between the time domain and the frequency domain by adjusting the order. Therefore, in tasks such as signal processing and feature extraction, it has more degrees of freedom and expressiveness than the traditional Fourier transform.

[0106] The fractional Fourier transform is a generalization of the Fourier transform, which controls the angle parameter To define the transformation of the signal at different angles, the expression of the fractional Fourier transform is as follows:

[0107]

[0108] Where, Indicates angle-based The signal characteristics at time t obtained by performing fractional Fourier transform; Represents the process of fractional Fourier transform; represents the preprocessed mechanical vibration signal at time t; Represents the time variable The differential of Indicates about time The kernel function of the fractional Fourier transform of .

[0109] About Moment The kernel function of the fractional Fourier transform of The expression is as follows:

[0110]

[0111] Where, Indicates the preset angle for fractional Fourier transform; represents the imaginary part; Indicates the center frequency of the preprocessed mechanical vibration signal; Indicates the time of the pre-processed mechanical vibration signal.

[0112] Specifically, fractional Fourier transform feature extraction includes the following data:

[0113] According to the pre-processed mechanical vibration signal, based on multiple preset angle parameters , respectively perform fractional Fourier transform to obtain multiple signal features , N represents the number of angle parameters;

[0114] Based on multiple signal characteristics , and obtain feature data.

[0115] Step 3: Perform information entropy quantization processing based on the feature data to obtain the corresponding information entropy data.

[0116] Entropy quantification is a method based on information theory that measures the "chaos" or "randomness" of a signal or dataset. Specifically, entropy measures the "information content" of a probability distribution, representing the uncertainty of a system. In signal processing or feature extraction, entropy is often used to quantify signal complexity.

[0117] During the quantization process, information entropy is used to describe the distribution characteristics of the signal, and different parts of the signal are assigned to different quantization levels to achieve the effect of compressing or enhancing signal information.

[0118] Specifically, the information entropy quantization process includes the following steps:

[0119] For any signal feature in the feature data extracted from the fractional Fourier transform , calculate its information entropy value and measure the complexity of the signal.

[0120] The calculation expression of information entropy value is as follows:

[0121]

[0122] Where, Indicates that the nth angle parameter is used The signal characteristics at time t obtained by performing fractional Fourier transform; Indicates signal characteristics The corresponding information entropy value; Indicates signal characteristics Probability density on the mth segment; Indicates signal characteristics characteristic length.

[0123] By the information entropy By quantizing, we can get the quantitative information entropy value , which is used to describe the redundancy and noise level of the signal:

[0124]

[0125] Where, Indicates that the nth angle parameter is used The quantitative information entropy value of Indicates quantization processing.

[0126] The information entropy data is obtained according to the quantized information entropy values corresponding to all signal features in the feature data.

[0127] Step 4: Based on the feature data and the corresponding information entropy data, feature fusion is performed through the weighted average method to obtain the fused feature.

[0128] In the feature fusion stage, the signal features extracted at multiple scales are merged into a high-dimensional feature vector through the weighted averaging method, thereby enhancing the expressive power of the features.

[0129] Specifically, obtaining the fusion features includes the following steps:

[0130] For each signal feature in the feature data, a weight is assigned according to the corresponding quantized information entropy value;

[0131] According to all the signal features and corresponding weights in the feature data, feature fusion is performed through weighted average method to obtain a high-dimensional fusion feature;

[0132] Among them, the expression of fusion features is as follows:

[0133]

[0134] Where, represents fusion features; Shows the nth angle parameter The quantitative information entropy value of Indicates that the nth angle parameter is used The signal characteristics at time t are obtained by performing fractional Fourier transform.

[0135] Finally, the fusion features Represents the multi-dimensional features of mechanical vibration signals. It can effectively describe the time-frequency characteristics, complexity information and multi-scale performance of the signal, and is suitable for further fault diagnosis and signal classification.

[0136] Step 5: Based on the fusion features, perform dimensionality reduction based on principal component analysis to obtain the final extracted features.

[0137] Principal component analysis is used for dimensionality reduction to reduce feature dimensions and retain the most important information.

[0138] The final extracted features include the following steps:

[0139] 5.1. Based on the fusion features, data is standardized to obtain the standard feature matrix.

[0140] The fused features are standardized to ensure that the mean of each feature is 0 and the variance is 1. Standardization helps to avoid the influence of different dimensions in the original features.

[0141] The normalization formula is as follows:

[0142]

[0143] in, Indicates the The standardized features; Indicates the Features Indicates the The standard deviation of each feature; Indicates the The mean of the features.

[0144] 5.2. According to the standard characteristic matrix, the covariance matrix is obtained.

[0145] The core step of principal component analysis is to calculate the covariance matrix of the data, which is used to describe the correlation between various features. Suppose we have a The standard characteristic matrix of , where M is the number of samples and N is the number of features. Covariance matrix The calculation formula is as follows:

[0146]

[0147] Covariance matrix Describes the covariance between each pair of features.

[0148] 5.3. Perform eigenvalue decomposition based on the covariance matrix to obtain the eigenvalues and eigenvectors corresponding to the covariance matrix.

[0149] By performing eigenvalue decomposition on the covariance matrix, the eigenvalues and corresponding eigenvectors of the covariance matrix can be obtained. An eigenvalue represents the importance of a principal component, and the corresponding eigenvector represents the direction of the corresponding principal component.

[0150] Let the covariance matrix The characteristic value of , the corresponding eigenvector is , these eigenvectors constitute a new feature space.

[0151] 5.4. Select the eigenvector with the largest eigenvalue of the preset number as the principal component matrix.

[0152] In the process of dimensionality reduction, we need to select principal components, namely The eigenvector with the largest eigenvalue. Usually, The value of is determined based on the cumulative variance explained. Specifically, principal components, so that the The variance explained by the eigenvalues accounts for 95% of the total variance. Cumulative variance explained The calculation formula is: .

[0153] 5.5. According to the standard feature matrix and principal component matrix, the final extracted features are obtained.

[0154] The reduced-dimensional features are obtained by multiplying the standard feature matrix with the selected principal component matrix, that is, the final extracted features. The expression of the final extracted features is as follows:

[0155]

[0156] Where, Represents the final extracted features; represents the standard feature matrix; Represents the principal component matrix composed of the eigenvectors with the first k largest eigenvalues, where k represents the preset number of principal components.

[0157] Final feature extraction The dimension is , where M is the number of samples, is the preset number of principal components.

[0158] Final feature extraction This is the representation of the original fused feature vector in the new principal component space. By projecting high-dimensional features onto a low-dimensional subspace, redundant information can be reduced, computational efficiency can be improved, and model performance can be enhanced. Therefore, the final extracted features are the result obtained using the feature extraction method for mechanical vibration defect signals of electrical equipment based on fractional Fourier transform and information entropy quantization.

[0159] Step 6: Input the final extracted features into the pre-trained equipment fault diagnosis model to obtain the equipment fault diagnosis results.

[0160] 6.1. Based on the final extracted features, perform standardization and time series slicing to obtain slice data.

[0161] Based on preset normalization parameters, the final extracted features are precisely calibrated. Each feature must be normalized according to the data distribution during training to ensure that the input data used during inference is consistent with the training data. To maintain the continuity and integrity of time series features, the data is sliced using the same sliding window strategy used during training. By setting the window size and sliding step size, the continuous signal is cut into a series of fixed-length segments. This step effectively captures dynamic changes during device operation.

[0162] 6.2. Based on the sliced data, perform tensor format conversion and virtual batch dimension embedding to obtain preprocessed data.

[0163] After normalization and time-series slicing, the sliced data is converted into a tensor format that meets the model's requirements, and a virtual batch dimension is embedded in the data dimension. This processing not only ensures that the data meets the model input shape requirements but also improves inference efficiency through batch processing.

[0164] 6.3. Input the preprocessed data into the pretrained equipment fault diagnosis model to obtain the equipment fault diagnosis results; the equipment fault diagnosis model adopts block loading technology based on memory mapping mechanism and parallel computing method based on multi-threaded reasoning.

[0165] Before fault diagnosis, a fixed random seed is used to ensure a consistent initial state each time the device fault diagnosis model is run, thereby ensuring the stability and reproducibility of prediction results. When dealing with large-scale data and complex models, memory-mapped files or block loading techniques are used to effectively alleviate memory pressure. This avoids resource bottlenecks caused by loading all data at once while ensuring continuous data transmission. The multi-threading and GPU acceleration interfaces provided by the deep learning framework enable parallel model inference. The system dynamically adjusts parallel computing parameters based on the current hardware load, ensuring high accuracy while minimizing inference response time and enabling real-time online diagnosis.

[0166] The output data after model inference is decoded by the post-processing module, and the abstract tensor results are converted into interpretable fault type and status information to obtain the equipment fault diagnosis results.

[0167] The equipment fault diagnosis model in this embodiment is further provided with a historical fault database, a credibility evaluation module, a diagnosis report module and an intelligent decision-making module.

[0168] The credibility assessment module first compares the equipment fault diagnosis results with the historical fault database to achieve comprehensive integration of the results. The credibility of the equipment fault diagnosis results is then assessed using an expert system and pre-set rules. By analyzing the confidence score of the predicted output and the degree of match with historical data, the actual risk and severity of the fault are determined.

[0169] The diagnostic reporting module generates detailed fault diagnosis reports that not only list the fault type and risk assessment, but also provide specific maintenance recommendations and early warning measures. When necessary, the intelligent decision-making module automatically triggers subsequent processing, such as real-time alarms, equipment shutdowns, or dispatching maintenance personnel, thus achieving full-chain intelligent monitoring and response from data collection to decision execution.

[0170] This method uses wavelet packet decomposition to reduce noise in the original signal, followed by a fractional Fourier transform. This method can extract multi-scale time-frequency features from complex mechanical vibration signals. It also combines information entropy quantization to extract signal complexity information. Ultimately, it fuses these multi-dimensional features to improve the accuracy of defect identification and fault diagnosis, effectively enhancing the accuracy and reliability of fault diagnosis. This method is highly adaptable and can be widely used in health monitoring and fault diagnosis of electrical equipment.

[0171] Example 2

[0172] This embodiment 2 provides an equipment fault diagnosis system based on FrFT and information entropy quantization, which is applicable to the equipment fault diagnosis method based on FrFT and information entropy quantization in embodiment 1, including:

[0173] A data acquisition module is used to obtain real-time pre-processed mechanical vibration signals of electrical equipment;

[0174] A feature extraction module is used to perform fractional Fourier transform feature extraction based on the preprocessed mechanical vibration signal to obtain feature data;

[0175] The information entropy module is used to perform information entropy quantification processing based on the feature data to obtain the corresponding information entropy data;

[0176] The feature fusion module is used to perform feature fusion based on feature data and corresponding information entropy data through weighted average method to obtain fused features;

[0177] The feature dimensionality reduction module is used to perform dimensionality reduction based on the fusion features and the principal component analysis method to obtain the final extracted features;

[0178] The fault diagnosis module is used to input the final extracted features into the pre-trained equipment fault diagnosis model to obtain the equipment fault diagnosis results.

[0179] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0180] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0181] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0182] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0183] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. The equipment fault diagnosis method based on FrFT and information entropy quantification is characterized by: The following steps are involved: Obtain real-time pre-processed mechanical vibration signals of electrical equipment; Performing fractional Fourier transform feature extraction on the preprocessed mechanical vibration signal to obtain feature data; Performing information entropy quantization processing on the characteristic data to obtain corresponding information entropy data; According to the feature data and the corresponding information entropy data, feature fusion is performed by a weighted average method to obtain a fused feature; According to the fusion features, dimension reduction is performed based on principal component analysis to obtain the final extracted features; The final extracted features are input into a pre-trained equipment fault diagnosis model to obtain equipment fault diagnosis results.

2. The device fault diagnosis method based on FrFT and information entropy quantization according to claim 1 is characterized in that: The method of obtaining a real-time pre-processed mechanical vibration signal of an electrical device comprises the following steps: Obtain real-time original mechanical vibration signals of electrical equipment; According to the original mechanical vibration signal, recursively decompose it based on a preset number of decomposition layers by wavelet packet decomposition to obtain low-frequency data containing all low-frequency signals; Reconstructing the signal by inverse wavelet packet transform according to the low-frequency data to obtain a preprocessed mechanical vibration signal; Among them, the wavelet packet decomposition includes the following steps: during the first decomposition, the original mechanical vibration signal is decomposed by applying wavelet transform to obtain the low-frequency signal and high-frequency signal of the first decomposition; in addition to the first decomposition, the high-frequency signal obtained by the previous decomposition is low-pass filtered and high-pass filtered each time to obtain the low-frequency signal and high-frequency signal of the current decomposition.

3. The device fault diagnosis method based on FrFT and information entropy quantization according to claim 1 is characterized in that: The fractional Fourier transform feature extraction includes the following data: According to the preprocessed mechanical vibration signal, fractional Fourier transform is performed based on a plurality of preset angle parameters to obtain a plurality of signal features; Obtaining feature data according to multiple signal features; The expression of the fractional Fourier transform is as follows: ； Where, Indicates angle-based The signal characteristics at time t obtained by performing fractional Fourier transform; Represents the process of fractional Fourier transform; represents the preprocessed mechanical vibration signal at time t; Represents the time variable The differential of Indicates about time The kernel function of the fractional Fourier transform of .

4. The device fault diagnosis method based on FrFT and information entropy quantization according to claim 3 is characterized in that: About the moment The kernel function of the fractional Fourier transform of The expression is as follows: ； Where, Indicates the preset angle for fractional Fourier transform; represents the imaginary part; Indicates the center frequency of the preprocessed mechanical vibration signal; Indicates the time of the pre-processed mechanical vibration signal.

5. The device fault diagnosis method based on FrFT and information entropy quantization according to claim 1 is characterized in that: The information entropy quantization process comprises the following steps: For each signal feature in the feature data, calculating the corresponding information entropy value and quantizing it to obtain a quantized information entropy value; Obtaining information entropy data according to the quantized information entropy values corresponding to all signal features in the feature data; The calculation expression of the information entropy value is as follows: ； Where, Indicates that the nth angle parameter is used The signal characteristics at time t obtained by performing fractional Fourier transform; Indicates signal characteristics The corresponding information entropy value; Indicates signal characteristics Probability density on the mth segment; Indicates signal characteristics characteristic length.

6. The device fault diagnosis method based on FrFT and information entropy quantization according to claim 1 is characterized in that: The step of obtaining the fusion features comprises the following steps: For each signal feature in the feature data, assign a weight according to the corresponding quantized information entropy value; According to all signal features and corresponding weights in the feature data, feature fusion is performed by weighted average method to obtain a high-dimensional fusion feature; The expression of the fusion feature is as follows: ； Where, represents fusion features; Shows the nth angle parameter The quantitative information entropy value of Indicates that the nth angle parameter is used The signal characteristics at time t are obtained by performing fractional Fourier transform.

7. The device fault diagnosis method based on FrFT and information entropy quantization according to claim 1 is characterized in that: The final extraction feature is obtained by: According to the fusion features, data standardization is performed to obtain a standard feature matrix; According to the standard characteristic matrix, a covariance matrix is obtained; Performing eigenvalue decomposition on the covariance matrix to obtain eigenvalues and eigenvectors corresponding to the covariance matrix; wherein an eigenvalue represents the importance of a principal component, and the corresponding eigenvector represents the direction of the corresponding principal component; Select the eigenvectors with the largest eigenvalues in the preset number of eigenvalues as the principal component matrix; According to the standard feature matrix and the principal component matrix, the final extracted features are obtained.

8. The device fault diagnosis method based on FrFT and information entropy quantization according to claim 7 is characterized in that: The expression of the final extracted features is as follows: ； Where, Represents the final extracted features; represents the standard feature matrix; Represents the principal component matrix composed of the eigenvectors with the largest first k eigenvalues, where k represents the preset number of principal components.

9. The device fault diagnosis method based on FrFT and information entropy quantization according to claim 1 is characterized in that: Obtaining the equipment fault diagnosis result includes the following steps: Performing standardization and time series slicing processing according to the final extracted features to obtain slice data; According to the slice data, tensor format conversion and virtual batch dimension embedding are performed to obtain preprocessed data; The preprocessed data is input into a pre-trained equipment fault diagnosis model to obtain an equipment fault diagnosis result; wherein, the equipment fault diagnosis model adopts a block loading technology based on a memory mapping mechanism and a parallel computing method based on multi-threaded reasoning.

10. An equipment fault diagnosis system based on FrFT and information entropy quantization, applicable to the equipment fault diagnosis method based on FrFT and information entropy quantization according to any one of claims 1 to 9, characterized in that: include: A data acquisition module is used to obtain real-time pre-processed mechanical vibration signals of electrical equipment; A feature extraction module, configured to perform fractional Fourier transform feature extraction based on the preprocessed mechanical vibration signal to obtain feature data; An information entropy module is used to perform information entropy quantization processing based on the feature data to obtain corresponding information entropy data; A feature fusion module is used to perform feature fusion based on the feature data and the corresponding information entropy data by a weighted average method to obtain a fused feature; A feature dimensionality reduction module is used to perform dimensionality reduction based on the fusion features and the principal component analysis method to obtain the final extracted features; The fault diagnosis module is used to input the final extracted features into a pre-trained equipment fault diagnosis model to obtain equipment fault diagnosis results.

Images