Early anomaly detection method for rotating machinery based on Grubrum and gray matrix co-occurrence

Abstract

The invention discloses a rotating machine early-stage anomaly detection method based on Grubrum and gray matrix co-occurrence, and the method comprises the steps: converting a one-dimensional time sequence signal into a two-dimensional image signal through a Grubrum angle field, maintaining the time dependence of the signal, and enhancing the feature expression capability of the signal; and carrying out multi-dimensional texture feature extraction on the reconstructed image signal through gray-level co-occurrence matrix analysis so as to meticulously represent an abnormal mode in the signal. A dynamic threshold value calculation and continuous alarm mechanism integrated in the system performs adaptive adjustment based on a characteristic value extracted in real time, and when the characteristic value exceeds a calculation threshold value, an alarm is triggered immediately, so that high-sensitivity monitoring of the equipment state is realized. In addition, the system is equipped with a data preprocessing module and a feature fusion module which are respectively responsible for denoising filtering and principal component analysis dimensionality reduction of the vibration signals, so that the signal quality and the feature effectiveness are ensured. The method has excellent anomaly detection accuracy and reliability in a high-noise and heavy-load environment.

Description

Technical Field

[0001] The invention belongs to the field of mechanical equipment maintenance, and in particular relates to a method for detecting early abnormalities of rotating machinery based on the symbiosis of Gram and grayscale matrices. Background Art

[0002] As the complexity of industrial equipment continues to increase, early fault detection of heavy rotating machinery is particularly critical in industrial applications. Such equipment usually operates in high-load and high-intensity environments, such as power plants, petrochemicals, mining, and marine engineering. Once a failure occurs, it often brings serious economic losses and safety risks. Therefore, early fault detection of heavy rotating machinery has become a key research area in industrial maintenance and safety assurance. Current fault detection methods can be mainly divided into three categories: knowledge-driven, model-driven, and data-driven. However, these methods have certain limitations in the application of complex working conditions and changing environments.

[0003] Knowledge-driven methods rely on expert knowledge and a preset fault rule base, and judge the known failure modes of the equipment through expert experience. This type of method can achieve certain results under simple working conditions, but for the complex structure and diverse failure modes of heavy rotating machinery, the coverage and effectiveness of expert knowledge are limited. Especially on new equipment, experts do not have a sufficient understanding of its internal working principles and potential failure modes, which makes the knowledge-driven method face challenges in accuracy and reliability. In addition, the knowledge-driven method relies on manually defining rules and building a knowledge base. The updating and maintenance of rules is particularly difficult under complex working conditions, and it is difficult to adapt to the dynamic changes in equipment status.

[0004] Model-driven methods are usually based on physical models and attempt to detect faults by describing the normal and abnormal behaviors of equipment. However, for heavy rotating machinery, the establishment of physical models has obvious limitations. Since such equipment usually operates under unpredictable load fluctuations and environmental changes, a single physical model can hardly fully describe the normal and abnormal states of the equipment. In addition, complex model calculations make it difficult to achieve real-time detection in actual operation. Therefore, the adaptability and real-time performance of model-driven methods under complex working conditions are poor, which limits its widespread application in heavy rotating machinery.

[0005] Data-driven methods are based on the historical data of the equipment and perform fault detection through signal analysis and pattern recognition methods, without relying on expert knowledge or physical models. This type of method has certain potential in adapting to complex systems, but in practical applications it faces problems such as missing and incomplete data, especially the monitoring data of heavy rotating machinery is often uneven under multiple operating conditions. In addition, data-driven methods rely on a large amount of labeled data, while industrial sites often lack labeled data for new fault modes, which limits the generalization ability of the model and makes it difficult to deal with new faults.

[0006] The application of signal processing technology in heavy rotating machinery fault detection has gradually gained attention. Traditional time domain, frequency domain and time-frequency domain analysis methods, such as short-time Fourier transform (STFT), wavelet transform and envelope analysis, can extract fault features from vibration signals. However, with the increasing complexity of the equipment operating environment, traditional signal processing methods have shown shortcomings when facing variable working conditions, especially the difficulty in capturing hidden early fault features. Therefore, there is an urgent need for an early abnormality detection method for rotating machinery to achieve more accurate fault warning. Summary of the invention

[0007] Purpose of the invention: The purpose of the present invention is to provide a method for early abnormality detection of rotating machinery based on the symbiosis of Gram and grayscale matrices. It can adapt to complex working conditions and achieve efficient and accurate fault warning. Based on advanced signal processing technology, the present invention overcomes the limitations of existing methods through multi-step feature extraction, feature fusion, dynamic threshold calculation and adaptive alarm mechanism, and achieves high-sensitivity and high-reliability fault detection.

[0008] Technical solution: A method for detecting early abnormalities of rotating machinery based on the symbiosis of Gram and grayscale matrices of the present invention comprises the following steps:

[0009] Step S1, collecting the original vibration signal of the target machine, and performing segmented aggregation approximation processing on it to obtain a one-dimensional time series vibration signal, and reconstructing the one-dimensional time series vibration signal using Gram angular field (GAF) to convert the one-dimensional signal into a two-dimensional image signal to obtain a GAF image;

[0010] Step S2, using gray level co-occurrence matrix to perform texture analysis on the GAF image, extracting texture features, including contrast, angular second-order moment, entropy, inverse moment and correlation features, to capture subtle texture changes in the image and achieve multi-dimensional feature representation, sorting the texture features by monotony using Spearman correlation coefficient, and using PCA for dimensionality reduction to obtain processed texture features;

[0011] Step S3: Calculate the dynamic alarm threshold and design the alarm trigger mechanism. Based on the real-time texture features extracted in step S2, use statistical characteristics to calculate the dynamic threshold of abnormality detection, improve the real-time performance of the alarm response, and achieve early warning of target mechanical failures.

[0012] Furthermore, in step S1, the segmented aggregation approximation processing is specifically: segmenting the original signal into segments and calculating the average value of each segment through segmented aggregation approximation processing, removing signal redundancy and improving the calculation efficiency of subsequent feature extraction. The calculation formula of the segmented aggregation approximation processing is as follows:

[0013]

[0014] Among them, w is the segment length, N is the time series length, x i is the original signal value, and X' is the one-dimensional time series signal obtained after segmented aggregation approximation processing.

[0015] Furthermore, in step S1, the Gram angular field GAF ​​is used to reconstruct the one-dimensional spatial sequence vibration signal, specifically: for the one-dimensional time series signal X'={x1,x2,...,x i}, the calculation formula of Gram angle field GAF ​​is as follows:

[0016] G i,j = cos(φ i +φ j )

[0017] Among them, φ i and φ j are the results of normalizing the data points at time i and j in the time series and mapping them to polar coordinate angles. The calculation formula is as follows:

[0018]

[0019] The generated two-dimensional image signal retains the temporal dependency of the original signal and enhances its feature expression capability.

[0020] Further, in step S2, the texture analysis of the GAF image using the gray level co-occurrence matrix and the extraction of texture features are specifically as follows: the texture analysis of the GAF image using the gray level co-occurrence matrix is ​​performed to extract key features of contrast, angular second-order moment, entropy, inverse moment and correlation to capture subtle texture changes in the image and achieve multi-dimensional feature characterization. The specific calculation formula is as follows:

[0021] Contrast ratio is defined as follows:

[0022]

[0023] The angular second moment is defined as follows:

[0024]

[0025] Entropy is defined as follows:

[0026]

[0027] The inverse moment difference is defined as follows:

[0028]

[0029] The correlation is defined as follows:

[0030]

[0031] Among them, P(i,j) is the element value of the gray level co-occurrence matrix, μ i and σ i are the mean and standard deviation of the corresponding rows, respectively. Texture features describe the local symmetry, detail, and smoothness information of the image and identify abnormal patterns in mechanical vibration signals.

[0032] Furthermore, the calculation of the gray-level co-occurrence matrix involves the analysis of multi-directional texture features, specifically including the extraction of primary information in the 135° direction and secondary information in the 0° or 90° direction, so as to achieve a comprehensive analysis of multi-directional texture features, thereby improving the reliability of abnormal pattern recognition.

[0033] Furthermore, in step S3, the formula for calculating the dynamic alarm threshold is as follows:

[0034] Ψ=μ=k·σ

[0035] Wherein, μ and σ represent the mean and standard deviation of the real-time feature respectively, k represents the adjustment coefficient, and the alarm mechanism triggers an alarm signal when it detects that the feature value exceeds the dynamic threshold Ψ, thereby realizing early warning of mechanical failure.

[0036] Furthermore, in step S3, the alarm triggering mechanism design includes the following steps:

[0037] Step S3.1, smooth the extracted eigenvalues ​​through multiple rounds of first-order exponential smoothing to reduce the impact of signal noise on the eigenvalues ​​and improve the stability of feature extraction. The calculation formula is:

[0038] S t =αx t +(1-α)S t-1 ,0<α<1

[0039] Among them, S t is the current smoothing value, x t is the observed value, α is the smoothing coefficient;

[0040] Step S3.2: Perform Z-score normalization on the smoothed features. The normalization formula is as follows:

[0041]

[0042] Among them, x is the eigenvalue, μ is the mean of the eigenvalue, σ is the standard deviation of the feature, and Z-score normalization ensures the comparability between different features;

[0043] Step S3.3: When the normalized fusion feature value exceeds the preset warning threshold, an abnormal alarm signal is triggered, indicating a potential risk of equipment failure.

[0044] The present invention also discloses a system for realizing an early abnormality detection method for a rotating machine, comprising a signal reconstruction module, a texture analysis module, a feature fusion module and an alarm module;

[0045] The signal reconstruction module is used to receive the mechanical vibration signal and convert the time series signal into a two-dimensional image through GAF to form an image signal that can represent the vibration characteristics;

[0046] The texture analysis module is used to perform gray-level co-occurrence matrix analysis to extract texture features of image contrast, angular second-order moment, entropy, inverse moment and correlation, so as to identify abnormal signal features;

[0047] The feature fusion module performs monotony sorting on texture features by using the Spearman correlation coefficient and uses PCA for dimensionality reduction to enhance the representativeness of features and reduce redundant information;

[0048] The alarm module is used to calculate the real-time vibration threshold based on the fused features and generate an alarm signal based on the normalized vibration sequence. The alarm module has the function of analyzing the main and secondary features to ensure the real-time triggering of the alarm signal. The alarm module has the ability to adjust the threshold adaptively and can automatically adjust the alarm threshold according to different working conditions of the equipment to improve the system's abnormality detection sensitivity and accuracy, and reduce the probability of false alarms and missed alarms through the adaptive threshold mechanism.

[0049] Furthermore, the system further comprises a data preprocessing module, which is used to filter and denoise the input vibration signal to ensure the quality of the signal and provide optimized input for feature extraction.

[0050] Furthermore, the system is applied to early abnormality detection of high-speed and heavy-load mechanical equipment, and is particularly suitable for fault warning of complex mechanical components such as rolling bearings and gearboxes. The system continuously maintains detection accuracy under high noise and high load conditions to prevent potential mechanical failures.

[0051] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages:

[0052] 1. Enhanced feature expression capability: Through GAF reconstruction, the one-dimensional vibration signal is converted into a two-dimensional image, which effectively retains the time dependency of the signal, making the early fault features more prominent in the image and improving the expression capability of the fault features.

[0053] 2. Multi-dimensional texture feature extraction: Based on the gray-level co-occurrence matrix, a variety of texture features are extracted, including contrast, angular second-order moment, entropy, inverse moment and correlation, which can comprehensively characterize the tiny anomalies in the vibration signal and significantly improve the sensitivity and accuracy of detection.

[0054] 3. Efficient feature fusion and dimensionality reduction: The features are screened and reduced in dimension through Spearman correlation coefficient and principal component analysis (PCA), which effectively removes redundant information, simplifies calculations, improves detection efficiency, and ensures the retention of major fault features.

[0055] 4. Adaptive dynamic threshold calculation: The adaptive dynamic threshold calculation mechanism designed in the present invention can adjust the threshold according to the changes in real-time characteristics, thereby adapting to variable working conditions and improving the accuracy and real-time performance of fault detection in complex environments.

[0056] 5. Noise suppression and signal stability improvement: Through data preprocessing and first-order exponential smoothing, the noise interference in the signal is effectively reduced, and the comparability between different features is ensured through Z-score standardization, further improving the stability of fault detection.

[0057] In general, the present invention can achieve accurate detection and rapid response to early faults of heavy rotating machinery in complex industrial environments with high noise and changeable working conditions, significantly improving the reliability and applicability of the system, and providing strong technical support for the intelligent maintenance of heavy rotating machinery. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 It is a test bench for testing heavy rotary machinery-large roller press in this embodiment;

[0059] Figure 2 It is a curve diagram of operating condition parameter changes verified by experiments in the embodiment of the present invention;

[0060] Figure 3 It is a work flow chart of the present invention;

[0061] Figure 4 It is a conceptual diagram of the workflow of the present invention;

[0062] Figure 5 Reconstructing fault images for GAF of heavy-duty rotating machinery in the present invention;

[0063] Figure 6 Pseudo code for the continuous warning mechanism of the present invention;

[0064] Figure 7 This is a result diagram of early abnormality detection of a large roller press in the embodiment. DETAILED DESCRIPTION

[0065] The technical solution of the present invention is further described below in conjunction with the accompanying drawings.

[0066] The test data used in this example comes from the full life cycle test of the bearing completed by Nanjing University of Aeronautics and Astronautics and Nanjing High Speed ​​Gear Co., Ltd. on a large roller press test bench. Figure 1 The site layout of the heavy rotating machinery-large roller press test bench used in the test is shown in the figure, and some parameters are marked in detail. Figure 2 The experimental parameter variation curves are shown in Figure 2.

[0067] like Figure 3 As shown in FIG. 1 , a real-time detection method for early abnormality of heavy rotating machinery based on the symbiosis of Gram and grayscale matrices of the present invention is shown in FIG. Figure 4 As shown. Specifically including the following steps:

[0068] Step S1, obtaining the vibration of heavy rotating machinery and completing signal reconstruction;

[0069] First, the collected mechanical vibration one-dimensional time series signal X = {x1, x2, ..., x i} to ensure that the data is within the interval [-1,1]. The standardization formula is as follows:

[0070]

[0071] Then, the normalized signal x' is mapped to the polar coordinate angle φ i , which is used to generate a two-dimensional image signal through the Gram angle field, the angle φ i The calculation formula is as follows:

[0072] φ i =arccos(x' i )

[0073] The two-dimensional image signal generated by the above steps retains the time dependency of the original signal and enhances its feature expression ability, laying the foundation for subsequent texture feature analysis. The reconstructed GAF ​​fault image is shown in Figure 5 shown.

[0074] The time series signal is processed by segmented aggregation approximation before reconstruction. The original signal is segmented into segments and the average value of each segment is calculated to effectively remove signal redundancy and improve the calculation efficiency of subsequent feature extraction. The calculation formula of the segmented aggregation approximation processing is as follows:

[0075]

[0076] Among them, w is the segment length, N is the time series length, x i is the original signal value, and X' is the new signal obtained after segmented aggregation approximation processing.

[0077] Step S2: Use the gray level co-occurrence matrix to perform texture analysis on the GAF image, extract key features such as contrast, angular second moment, entropy, inverse moment and correlation, so as to capture subtle texture changes in the image and achieve multi-dimensional feature representation. The specific calculation formula is as follows:

[0078] Contrast ratio is defined as follows:

[0079]

[0080] The angular second moment is defined as follows:

[0081]

[0082] Entropy is defined as follows:

[0083]

[0084] The inverse moment difference is defined as follows:

[0085]

[0086] The correlation is defined as follows:

[0087]

[0088] Among them, P(i,j) is the element value of the gray level co-occurrence matrix, μ i and σ i are the mean and standard deviation of the corresponding rows respectively. The texture feature describes the local symmetry, detail, smoothness and other information of the image, and can effectively identify abnormal patterns in mechanical vibration signals. The calculation of the gray level co-occurrence matrix involves the analysis of multi-directional texture features, specifically including the extraction of the main information in the 135° direction and the secondary information in the 0° or 90° direction, so as to achieve a comprehensive analysis of multi-directional texture features, thereby improving the reliability of abnormal pattern recognition.

[0089] Step S3: Calculate the dynamic alarm threshold and design the alarm trigger mechanism. Based on the real-time texture features extracted in step S2, use statistical characteristics to calculate the dynamic threshold of anomaly detection to improve the real-time performance of the alarm response. The dynamic threshold calculation formula is as follows:

[0090] Ψ=μ=k·σ

[0091] Wherein, μ and σ represent the mean and standard deviation of the real-time feature, respectively, and k represents the adjustment coefficient. The alarm mechanism triggers an alarm signal when it detects that the feature value exceeds the dynamic threshold Ψ, so as to achieve early warning of mechanical failure. The algorithm pseudo code of the alarm mechanism is as follows: Figure 6 shown.

[0092] Step S4: Smooth the extracted eigenvalues ​​through multiple rounds of first-order exponential smoothing to reduce the impact of signal noise on the eigenvalues ​​and improve the stability of feature extraction. The calculation formula is:

[0093] S t =αx t +(1-α)S t-1 ,0<α<1

[0094] Among them, S t is the current smoothing value, x t is the observed value, and α is the smoothing coefficient.

[0095] Step S5: Perform Z-score normalization on the smoothed features. The normalization formula is as follows:

[0096]

[0097] Among them, x is the eigenvalue, μ is the mean of the eigenvalue, σ is the standard deviation of the feature, and Z-score normalization ensures the comparability between different features.

[0098] Step S6: When the normalized fusion feature value exceeds the preset warning threshold, an abnormal alarm signal is triggered, indicating a potential risk of equipment failure.

[0099] Step S7, signal reconstruction module establishment: used to receive the mechanical vibration signal, and convert the time series signal into a two-dimensional image through GAF to form an image signal that can represent the vibration characteristics.

[0100] Step S8, texture analysis module: used to perform gray-level co-occurrence matrix analysis to extract image texture features such as contrast, angular second-order moment, entropy, inverse moment and correlation, so as to identify abnormal signal features.

[0101] Step S9, to optimize the feature space, feature fusion module: monotonically sort the texture features by Spearman correlation coefficient, and use PCA for dimensionality reduction to enhance the representativeness of the features and reduce redundant information. The specific Spearman correlation coefficient calculation formula is as follows:

[0102]

[0103] Step S10, alarm module: used to calculate the real-time vibration threshold according to the fused features, and generate an alarm signal based on the normalized vibration sequence. The alarm module has the function of analyzing the main and secondary features to ensure the real-time triggering of the alarm signal.

[0104] The system further includes a data preprocessing module, which is used to filter and denoise the input vibration signal to ensure the quality of the signal and provide optimized input for feature extraction. The final result output is as follows: Figure 7 shown.

[0105] In summary, the present invention provides an early fault detection method and system suitable for heavy-duty rotating machinery, which realizes multi-dimensional analysis of vibration signals through signal reconstruction, texture feature extraction, feature fusion and dimensionality reduction, adaptive threshold calculation and alarm mechanism, and overcomes the limitations of traditional detection methods under complex working conditions. The present invention can effectively detect early faults in high-noise and variable environments, improves the accuracy and real-time performance of detection, and provides reliable technical support for the intelligent maintenance of heavy-duty rotating machinery.

Claims

1. A method for early abnormality detection of rotating machinery based on the symbiosis of Gram and grayscale matrices, characterized in that: The steps include: Step S1, collecting the original vibration signal of the target machine, and performing segmented aggregation approximation processing on it to obtain a one-dimensional time series vibration signal, and reconstructing the one-dimensional time series vibration signal using Gram angular field (GAF) to convert the one-dimensional signal into a two-dimensional image signal to obtain a GAF image; Step S2, using gray level co-occurrence matrix to perform texture analysis on the GAF image, extracting texture features, including contrast, angular second moment, entropy, inverse moment and correlation features, sorting the texture features by monotony using Spearman correlation coefficient, and using PCA to reduce the dimension to obtain the processed texture features; Step S3: Calculate the dynamic alarm threshold and design the alarm trigger mechanism. Based on the texture features processed in step S2, use statistical characteristics to calculate the dynamic threshold of anomaly detection, improve the real-time performance of the alarm response, and achieve early warning of target mechanical failures.

2. The method for early abnormality detection of rotating machinery based on the symbiosis of Gram and grayscale matrix according to claim 1 is characterized in that: In step S1, the segmented aggregation approximation processing is specifically: segmenting the original signal into segments and calculating the average value of each segment through segmented aggregation approximation processing, removing signal redundancy and improving the calculation efficiency of subsequent feature extraction. The calculation formula of the segmented aggregation approximation processing is as follows: Among them, w is the segment length, N is the time series length, x i is the original signal value, and X' is the one-dimensional time series signal obtained after segmented aggregation approximation processing.

3. The method for early abnormality detection of rotating machinery based on the symbiosis of Gram and gray matrix according to claim 1 is characterized in that: In step S1, the Gram angular field GAF ​​is used to reconstruct the one-dimensional spatial sequence vibration signal, specifically: for the one-dimensional time series signal X'={x1,x2,...,x i }, the calculation formula of Gram angle field GAF ​​is as follows: G i,j =cos(φ i +φ j ) Among them, φ i and φ j are the results of normalizing the data points at time i and j in the time series and mapping them to polar coordinate angles. The calculation formula is as follows: The generated two-dimensional image signal retains the temporal dependency of the original signal and enhances its feature expression capability.

4. The method for early abnormality detection of rotating machinery based on the symbiosis of Gram and grayscale matrices according to claim 1 is characterized in that: In step S2, the texture analysis of the GAF image is performed using the gray level co-occurrence matrix to extract texture features. Specifically, the texture analysis of the GAF image is performed using the gray level co-occurrence matrix to extract key features of contrast, angular second moment, entropy, inverse moment and correlation to capture subtle texture changes in the image and achieve multi-dimensional feature representation. The specific calculation formula is as follows: Contrast ratio is defined as follows: The angular second moment is defined as follows: Entropy is defined as follows: The inverse moment difference is defined as follows: The correlation is defined as follows: Among them, P(i,j) is the element value of the gray level co-occurrence matrix, μ i and σ i are the mean and standard deviation of the corresponding rows, respectively. Texture features describe the local symmetry, detail, and smoothness information of the image and identify abnormal patterns in mechanical vibration signals.

5. A method for detecting early abnormality of rotating machinery based on the symbiosis of Gram and grayscale matrices according to claim 1 or 4, characterized in that: The calculation of the gray level co-occurrence matrix involves the analysis of multi-directional texture features, specifically including the extraction of primary information in the 135° direction and secondary information in the 0° or 90° direction.

6. The method for early abnormality detection of rotating machinery based on the symbiosis of Gram and grayscale matrices according to claim 1 is characterized in that: In step S3, the formula for calculating the dynamic alarm threshold is as follows: Ψ=μ=k·σ Wherein, μ and σ represent the mean and standard deviation of the real-time feature respectively, k represents the adjustment coefficient, and the alarm mechanism triggers an alarm signal when it detects that the feature value exceeds the dynamic threshold Ψ, thereby realizing early warning of mechanical failure.

7. The method for early abnormality detection of rotating machinery based on the symbiosis of Gram and grayscale matrices according to claim 1 is characterized in that: In step S3, the alarm triggering mechanism design includes the following steps: Step S3.1, smooth the extracted eigenvalues ​​through multiple rounds of first-order exponential smoothing to reduce the impact of signal noise on the eigenvalues ​​and improve the stability of feature extraction. The calculation formula is: S t =αx t +(1-α)S t-1 ,0<α<1 Among them, S t is the current smoothing value, x t is the observed value, α is the smoothing coefficient; Step S3.2: Perform Z-score normalization on the smoothed features. The normalization formula is as follows: Among them, x is the eigenvalue, μ is the mean of the eigenvalue, σ is the standard deviation of the feature, and Z-score normalization ensures the comparability between different features; Step S3.3: When the normalized fusion feature value exceeds the preset warning threshold, an abnormal alarm signal is triggered, indicating a potential risk of equipment failure.

8. A system for implementing the method for detecting early abnormality of a rotating machine as claimed in claim 1, characterized in that: It includes signal reconstruction module, texture analysis module, feature fusion module and alarm module; The signal reconstruction module is used to receive the mechanical vibration signal and convert the time series signal into a two-dimensional image through GAF to form an image signal that can represent the vibration characteristics; The texture analysis module is used to perform gray-level co-occurrence matrix analysis to extract texture features of image contrast, angular second-order moment, entropy, inverse moment and correlation, so as to identify abnormal signal features; The feature fusion module performs monotony sorting on texture features by using the Spearman correlation coefficient and uses PCA for dimensionality reduction to enhance the representativeness of features and reduce redundant information; The alarm module is used to calculate the real-time vibration threshold based on the fused features and generate an alarm signal based on the normalized vibration sequence. The alarm module has the function of analyzing the main and secondary features to ensure the real-time triggering of the alarm signal. The alarm module has the ability to adjust the threshold adaptively and can automatically adjust the alarm threshold according to different working conditions of the equipment to improve the system's abnormality detection sensitivity and accuracy, and reduce the probability of false alarms and missed alarms through the adaptive threshold mechanism.

9. The system according to claim 8, characterized in that The system further comprises a data preprocessing module, which is used to filter and denoise the input vibration signal to ensure the quality of the signal and provide optimized input for feature extraction.

10. The system according to claim 8, characterized in that The system is used for early abnormality detection of high-speed and heavy-load mechanical equipment, and is particularly suitable for early warning of failures of complex mechanical components such as rolling bearings and gearboxes. The system continuously maintains detection accuracy under high noise and high load conditions to prevent potential mechanical failures.

Images