Method and device for monitoring axial vibration signal of high-speed rotating mechanism

Abstract

The invention discloses a high-speed rotating mechanism axial vibration signal monitoring method and device, and the method comprises the following steps: creating feature library data which is used for the training of a recognition model of a mechanical fault of a high-speed rotating mechanism, and the recognition model comprises a feature comparison model, a prediction model and an evaluation model; acquiring an axial vibration signal of the high-speed rotating mechanism in real time, and preprocessing the axial vibration signal; taking a numerical value of a maximized interval sample as a threshold value, inputting the preprocessed axial vibration signal into the feature comparison model, comparing the preprocessed axial vibration signal with the threshold value, and if the preprocessed axial vibration signal is within the threshold value range, determining that the mechanical state of the high-speed rotating mechanism is normal; if not, the mechanical state of the high-speed rotating mechanism is abnormal; and inputting an axial vibration signal with an abnormal mechanical state into the prediction model and the evaluation model for prediction and evaluation, generating a prediction result and an evaluation result, and triggering an alarm if the prediction result and the evaluation result are abnormal. The method is mainly embodied in that the health state of the high-speed rotating mechanism is evaluated, faults or abnormities possibly occurring in the future are predicted, and the operation efficiency and safety of the high-speed rotating mechanism are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of mechanical vibration measurement, and in particular to a method and device for monitoring axial vibration signals of a high-speed rotating mechanism. Background Art

[0002] High-speed rotating mechanism usually refers to a mechanical device or component that can rotate at a high speed. This mechanism is widely used in many fields, such as motors, turbines, compressors, pumps, etc. The high-speed rotating mechanism axial vibration signal monitoring device is a device specially used to monitor the axial vibration signal of the high-speed rotating mechanism in real time. This device usually includes a rotor, a stator, a thrust plate and an axial vibration sensing component. Its working principle is to use the sensor to measure the axial vibration by radial installation, so as to realize the real-time monitoring of the rotor axial vibration.

[0003] When collecting data, a high-speed rotating mechanism axial vibration signal monitoring device disclosed in announcement number CN214200366U includes a rotor, two stators mounted on the rotor, and a thrust plate mounted on the rotor and located between the two stators. The thrust plate is provided with a detection surface, and a detachable axial vibration sensing component is installed between the two stators, which is radially aligned with the detection surface and used to monitor the axial displacement of the detection surface. However, the device may cause signal interruption during detection due to sensor failure or communication fault light, resulting in data loss and detection interruption, affecting the reliability of the monitoring device, and during the monitoring process, the vibration signal may be affected by environmental noise, interference, etc., resulting in a decrease in signal quality, affecting subsequent fault diagnosis. In addition, the axial vibration of the high-speed rotating mechanism may also have potential faults and abnormal conditions, but if they are not discovered and predicted in time, it may cause equipment accidents or shutdowns, affecting operating efficiency and safety. Summary of the invention

[0004] The purpose of the present invention is to overcome the deficiencies in the prior art and to provide a method and device for monitoring axial vibration signals of a high-speed rotating mechanism.

[0005] The purpose of the present invention is achieved through the following technical solutions:

[0006] A method for monitoring axial vibration signals of a high-speed rotating mechanism comprises the following steps:

[0007] S1. Create feature library data for training a recognition model for mechanical faults of high-speed rotating mechanisms, wherein the recognition model includes a feature comparison model, a prediction model, and an evaluation model;

[0008] The feature comparison model is trained by the SVM algorithm, specifically including:

[0009] S11, selecting a training data set from the feature library data,

[0010] {(\mathbf{x}_1, y_1, mathbf{x}_2, y_2, \ldots, mathbf{x}_n, y_n)}

[0011] Where mathbf{x}_i is the input feature vector; y_1 is the corresponding category label (y_i\in{-1,+1});

[0012] S12, use SVM classifier to find a hyperplane to separate samples of different categories; the hyperplane is expressed as

[0013] mathbf{w}\cdot\mathbf{x}+b=0, where mathbf{w} normal vector determines the direction of the hyperplane and b is the bias term;

[0014] The optimization method is:

[0015] min_{\mathbf{w},b}\frac{1}{2}|\mathbf{w}|^2

[0016] And satisfy the constraints:

[0017] y_i(\mathbf{w}\cdot\mathbf{x}_i+b)\geq1,\quad\text{for all}i=1,2,\dots,nwhere \cdot|^2 represents the L2 norm of the vector;

[0018] mathbf{w}\cdot\mathbf{x} represents the inner product of vectors;

[0019] S13, calculating the distances of samples of different categories from the optimal hyperplane found by the SVM algorithm, wherein the sample farthest from the hyperplane is the maximized margin sample;

[0020] The prediction model is trained by a linear regression algorithm.

[0021] hat{y}=w_0+w_1x_1+w_2x_2+\cdots+w_nx_n, where hat{y} is the predicted value, w_0, w_1, \cdots, W_n are model parameters, and x_1, x_2, \cdots, x_n are feature values;

[0022] The evaluation model is trained by variance, variance ∫[xE(x)]^2f(x)dx, where E(x) represents mathematical expectation;

[0023] S2, collecting the axial vibration signal of the high-speed rotating mechanism in real time and preprocessing it;

[0024] S3, using the value of the maximized interval sample as a threshold, inputting the preprocessed axial vibration signal into the feature comparison model, and comparing it with the threshold. If it is within the threshold range, the mechanical state of the high-speed rotating mechanism is normal; if it is not within the threshold range, the mechanical state of the high-speed rotating mechanism is abnormal;

[0025] S4. Inputting the axial vibration signal of abnormal mechanical state into the prediction model and the evaluation model for prediction and evaluation, generating prediction results and evaluation results, and triggering an alarm when the prediction results and the evaluation results are abnormal.

[0026] Preferably, the feature library data includes collecting vibration data of the high-speed rotating mechanism under different mechanical states, extracting frequency domain, time domain and statistical features from the vibration data, and arranging and summarizing them.

[0027] Preferably, in step S2, "real-time collection of axial vibration signals of the high-speed rotating mechanism" is specifically as follows: using a first collection unit to collect the axial vibration signals of the high-speed rotating mechanism; if the collected axial vibration signals are abnormal, switching to a second collection unit to re-collect the axial vibration signals of the high-speed rotating mechanism; the second collection unit and the first collection unit are redundant with each other.

[0028] Preferably, “the collected axial vibration signal is abnormal” specifically includes: using a vibration detection algorithm to monitor the axial vibration signal in real time, and the vibration monitoring algorithm is:

[0029] text{{if}\(\text{Peak Value}>T)\\text{then}\\text{Anomnaly}

[0030] Whether the peak value of the axial vibration signal collected by the first collection unit or the second collection unit exceeds a set threshold value, if it exceeds the set threshold value, it is judged as abnormal.

[0031] Preferably, "preprocessing it" in step S2 is specifically: converting the axial vibration signal into the frequency domain to identify the frequency component of the interference signal, analyzing the periodicity or burstiness of the interference signal in the time domain, and then using a filter to remove the interference signal in the axial vibration signal and perform noise reduction.

[0032] Preferably, “using a filter to remove interference signals in the axial vibration signal” specifically includes:

[0033] The filter output is y(n)=w(n)Tx(n), where y(n) is the output of the filter, w(n) is a vector containing the filter coefficients, and x(n) is the input signal vector;

[0034] Error calculation: e(n) = d(n) - y(n), where e(n) is the error signal and d(n) is the desired response signal;

[0035] Coefficient update: w(n+1)=w(n)+2μe(n)x(n), where μ is the step size parameter.

[0036] Preferably, “denoising” is to perform denoising using a SureShrink algorithm, wherein the steps of the SureShrink algorithm are as follows:

[0037] Wavelet decomposition: Perform wavelet decomposition on the axial vibration signal to obtain a series of wavelet coefficients, including approximate coefficients and detail coefficients;

[0038] Thresholding: Thresholding is applied to the wavelet coefficients at each level. In SureShrink, the selection of the threshold is based on the noise level of each wavelet subband. The threshold is set using the estimated noise level.

[0039] Hard threshold or soft threshold: Depending on the selected threshold, the wavelet coefficients are processed by hard threshold or soft threshold. The hard threshold sets the coefficients whose absolute value is less than the threshold to zero, while the soft threshold shrinks these coefficients to zero.

[0040] Inverse wavelet transform: Perform inverse wavelet transform on the processed wavelet coefficients to reconstruct the denoised signal;

[0041] Data fusion is performed using the denoised signals from multiple signal sources.

[0042] Preferably, the prediction results are specifically vibration displacement characteristics and vibration velocity characteristics, and the vibration displacement characteristics and vibration velocity characteristics are specifically vibration frequency, displacement and velocity characteristics including crack location of high-speed rotating mechanism, inner and outer ring failure of bearing, misalignment position of rotating shaft and foundation deformation.

[0043] Preferably, the evaluation results at least include energy distribution index, probability distribution curve index and frequency domain analysis index;

[0044] The energy distribution index is used to analyze the energy distribution of the axial vibration signal and identify and distinguish different fault types;

[0045] The probability distribution curve indicator is used to analyze the probability distribution curve of the axial vibration signal data and evaluate the state of the rolling bearing;

[0046] The frequency domain analysis index is used to identify specific fault frequency components.

[0047] A high-speed rotating mechanism axial vibration signal monitoring device, comprising:

[0048] Acquisition module: collects axial vibration signals of high-speed rotating mechanisms;

[0049] Signal processing module: monitors the quality and interference of the axial vibration signal in real time and adjusts processing parameters, converts the axial vibration signal into the frequency domain to identify the frequency components of the interference signal, analyzes the periodicity or burstiness of the interference signal in the time domain, and then uses a filter to remove the interference signal in the axial vibration signal and perform noise reduction;

[0050] Pattern recognition module: Use feature library data to train the pattern recognition algorithm and establish a recognition model for mechanical faults. The feature library data includes frequency domain, time domain and statistical features extracted from vibration data in normal and fault states. The recognition model recognizes different mechanical states based on the axial vibration signal after noise reduction processing.

[0051] Diagnostic module: matches the vibration data collected in real time with the features in the feature comparison model to diagnose and determine whether there is any abnormality or failure in the current mechanical state;

[0052] Prediction and evaluation module: Based on the identified mechanical state and the diagnosis results, the health status of the mechanism is evaluated and predicted, and an alarm is triggered when the identified mechanical state and the diagnosis results are abnormal;

[0053] A real-time monitoring module is electrically connected to the first acquisition unit and the second acquisition unit.

[0054] The beneficial effects of the present invention are mainly reflected in:

[0055] 1. Predict and evaluate the health status of high-speed rotating mechanisms, predict possible failures or abnormalities in the future, and keep abreast of the health status of high-speed rotating mechanisms. This will help to discover potential problems in a timely manner, prevent accidents, and improve the operating efficiency and safety of high-speed rotating mechanisms.

[0056] 2. The feature comparison model is trained through the SVM algorithm, which can quickly process nonlinear fault signals with less axial vibration data, avoiding the current deep learning methods' reliance on big data, thereby ensuring the accuracy of the axial vibration signal.

[0057] 3. The prediction model is trained through a linear regression algorithm. The model is simple and intuitive, easy to understand and implement. It can also be used to predict continuous values ​​such as the remaining service life of key components of the shaft system, greatly improving safety.

[0058] 4. The evaluation model measures the statistical difference between the predicted value and the true value through variance, conducts real-time training, continuously improves the generalization ability of the model, and has high system stability.

[0059] 5. The axial vibration signal is analyzed in the frequency domain and time domain, and then filtered and denoised. The specific processing parameters are adjusted by judging the quality and interference of the axial vibration signal. Overall, the comprehensive optimization and improvement of the axial vibration signal monitoring of the high-speed rotating mechanism is achieved, which not only improves the accuracy and stability of signal processing, but also provides a high-quality data basis for subsequent fault diagnosis, enhances the ability to extract fault features, improves the accuracy and efficiency of fault diagnosis, and supports early fault warning.

[0060] 6. The real-time monitoring module can be used to ensure the accuracy of the collected axial vibration signal. By switching the first acquisition unit and the second acquisition unit, it can be ensured that the data collection work will not be interrupted due to the failure of the first acquisition unit, thereby improving the reliability of the monitoring system and ensuring that the monitoring of the axial vibration signal of the high-speed rotating mechanism can be carried out continuously. DETAILED DESCRIPTION

[0061] The following specific embodiments describe the present invention in detail, but these embodiments are not limited to the present invention, and any results, methods, or functional changes made by a person skilled in the art based on these embodiments are all within the protection scope of the present invention.

[0062] The present invention discloses a method for monitoring axial vibration signals of a high-speed rotating mechanism, comprising the following steps:

[0063] S1. Create feature library data for training the recognition model of mechanical faults of high-speed rotating mechanisms, wherein the recognition model includes a feature comparison model, a prediction model, and an evaluation model. In the above, the feature library data includes collecting vibration data of the high-speed rotating mechanism under different mechanical states, extracting frequency domain, time domain and statistical features from the vibration data, and arranging and summarizing them. In addition, in this preferred embodiment, the different mechanical states refer to the states of the high-speed rotating mechanism under normal and fault conditions.

[0064] In this preferred embodiment, the feature contrast model is trained by the SVM algorithm, specifically including:

[0065] S11, selecting a training data set from the feature library data,

[0066] {(\mathbf{x}_1, y_1, mathbf{x}_2, y_2, \ldots, mathbf(x}_n, y_n)}

[0067] Where mathbf{x}_i is the input feature vector; y_1 is the corresponding category label (y_i\in{-1,+1});

[0068] S12, use SVM classifier to find a hyperplane to separate samples of different categories; the hyperplane is expressed as

[0069] mathbf{w}\cdot\mathbf{x}+b=0, where mathbf{w} normal vector determines the direction of the hyperplane and b is the bias term;

[0070] The optimization method is:

[0071] min_{\mathbf{w},b}\frac{1}{2}|\mathbf{w}|^2

[0072] And satisfy the constraints:

[0073] y_i(\mathbf{w}\cdot\mathbf{x}_i+b)\geq1,\quad\text{for all}i=1,2,\ldots,nwhere |\cdot|^2 represents the L2 norm of the vector;

[0074] mathbf{w}\cdot\mathbf{x} represents the inner product of vectors;

[0075] S13. Calculate the distances between samples of different categories and the optimal hyperplane found by the SVM algorithm, where the sample farthest from the hyperplane is the maximized margin sample.

[0076] The above-mentioned feature comparison model is trained through the SVM algorithm, which can quickly process nonlinear fault signals with less axial vibration data, avoiding the reliance of various current deep learning methods on big data, thereby ensuring the accuracy of the axial vibration signal.

[0077] The prediction model is trained by a linear regression algorithm.

[0078] hat{y}=w_0+w_1x_1+w_2x_2+\cdots+w_nx_n, where hat{y} is the predicted value, w_0, w_1, \cdots, w_n are model parameters, and x_1, x_2, \cdots, x_n are feature values.

[0079] The evaluation model measures the statistical difference between the predicted value and the true value through variance, performs real-time training, continuously improves the generalization ability of the model, and has high system stability.

[0080] The evaluation model is trained by variance, variance ∫[xE(x)]^2f(x)dx, where E(x) represents the mathematical expectation.

[0081] The evaluation results include failure risk scores and health status registration. Through verification and optimization, the performance of the evaluation model is improved. The prediction model is trained by a linear regression algorithm. The model is simple and intuitive, easy to understand and implement, and can be used to predict continuous values ​​such as the remaining service life of key shafting components, greatly improving safety.

[0082] S2. Collect the axial vibration signal of the high-speed rotating mechanism in real time and pre-process it.

[0083] In the above, real-time collection of the axial vibration signal of the high-speed rotating mechanism is specifically as follows: using a first collection unit to collect the axial vibration signal of the high-speed rotating mechanism, if the collected axial vibration signal is abnormal, switching to the second collection unit to re-collect the axial vibration signal of the high-speed rotating mechanism, the second collection unit and the first collection unit are redundant with each other.

[0084] In the above, “the collected axial vibration signal is abnormal” specifically means: using a vibration detection algorithm to monitor the axial vibration signal in real time, and the vibration monitoring algorithm is:

[0085] text{if}\(\text{Peak Value}>T)\\text{then}\\text{Anomaly}

[0086] Whether the peak value of the axial vibration signal collected by the first collection unit or the second collection unit exceeds a set threshold value, if it exceeds the set threshold value, it is judged as abnormal.

[0087] After the axial vibration signal of the high-speed rotating mechanism is collected, it is preprocessed, specifically: the axial vibration signal is converted into the frequency domain to identify the frequency component of the interference signal, the periodicity or burstiness of the interference signal is analyzed in the time domain, and then a filter is used to remove the interference signal in the axial vibration signal and perform noise reduction.

[0088] In addition, in this preferred embodiment, the filter is used to remove the interference signal in the axial vibration signal:

[0089] The filter output is y(n)=w(n)Tx(n), where y(n) is the output of the filter, w(n) is a vector containing the filter coefficients, and x(n) is the input signal vector;

[0090] Error calculation: e(n) = d(n) - y(n), where e(n) is the error signal and d(n) is the desired response signal;

[0091] Coefficient update: w(n+1)=w(n)+2μe(n)x(n), where μ is the step size parameter.

[0092] The method for monitoring axial vibration signals of high-speed rotating mechanisms according to claim 6 is characterized in that: "performing noise reduction" is to perform noise reduction processing using a SureShrink algorithm, and the steps of the SureShrink algorithm are as follows:

[0093] Wavelet decomposition: Perform wavelet decomposition on the axial vibration signal to obtain a series of wavelet coefficients, including approximate coefficients and detail coefficients;

[0094] Thresholding: Thresholding is applied to the wavelet coefficients at each level. In SureShrink, the selection of the threshold is based on the noise level of each wavelet subband. The threshold is set using the estimated noise level.

[0095] Hard threshold or soft threshold: Depending on the selected threshold, the wavelet coefficients are processed by hard threshold or soft threshold. The hard threshold sets the coefficients whose absolute value is less than the threshold to zero, while the soft threshold shrinks these coefficients to zero.

[0096] Inverse wavelet transform: Perform inverse wavelet transform on the processed wavelet coefficients to reconstruct the denoised signal;

[0097] Data fusion is performed using the denoised signals from multiple signal sources.

[0098] The above-mentioned frequency domain and time domain analysis of the axial vibration signal, followed by filtering and denoising, and the adjustment of specific processing parameters by judging the quality and interference of the axial vibration signal, have achieved overall optimization and improvement of the axial vibration signal monitoring of the high-speed rotating mechanism, which not only improves the accuracy and stability of signal processing, but also provides a high-quality data basis for subsequent fault diagnosis, enhances the ability to extract fault features, improves the accuracy and efficiency of fault diagnosis, and supports early fault warning.

[0099] S3. Using the value of the maximized interval sample as a threshold, the preprocessed axial vibration signal is input into the feature comparison model and compared with the threshold. If it is within the threshold range, the mechanical state of the high-speed rotating mechanism is normal; if it is not within the threshold range, the mechanical state of the high-speed rotating mechanism is abnormal.

[0100] S4. Inputting the axial vibration signal of abnormal mechanical state into the prediction model and the evaluation model for prediction and evaluation, generating prediction results and evaluation results, and triggering an alarm when the prediction results and the evaluation results are abnormal.

[0101] The prediction results described above are specifically vibration displacement characteristics and vibration velocity characteristics, and the vibration displacement characteristics and vibration velocity characteristics are specifically vibration frequency, displacement and velocity characteristics including crack location of high-speed rotating mechanism, inner and outer ring failure of bearing, misalignment position of rotating shaft and foundation deformation.

[0102] The evaluation results at least include energy distribution index, probability distribution curve index and frequency domain analysis index;

[0103] The energy distribution index is used to analyze the energy distribution of the axial vibration signal and identify and distinguish different fault types;

[0104] The probability distribution curve indicator is used to analyze the probability distribution curve of the axial vibration signal data and evaluate the state of the rolling bearing;

[0105] The frequency domain analysis index is used to identify specific fault frequency components.

[0106] Interpret and analyze the prediction and evaluation results, identify the causes and trends of failures, and make corresponding maintenance recommendations based on the evaluation and prediction results, including specifying maintenance plans, determining spare parts, etc.

[0107] The present invention can evaluate the health status of the high-speed rotating mechanism and predict possible failures or abnormalities in the future, so as to understand the health status of the high-speed rotating mechanism at any time, which helps to timely discover potential problems, prevent accidents, and improve the operating efficiency and safety of the high-speed rotating mechanism.

[0108] The present invention also discloses a high-speed rotating mechanism axial vibration signal monitoring device, comprising:

[0109] The acquisition module includes a first acquisition unit and a second acquisition unit. The first acquisition unit and the second acquisition unit are used to collect axial vibration data in real time. The second acquisition unit collects data after an abnormality occurs in the first acquisition unit. Similarly, after an abnormality occurs in the second acquisition unit, the first acquisition unit can be switched back to the first acquisition unit for collection. It should be pointed out that the number of specific acquisition units can be set according to the probability of failure. Generally speaking, two redundant units can be set. The first acquisition unit and the second acquisition unit both include a signal acquisition unit, an electromagnetic shielding and isolation unit, a sampling rate and resolution optimization unit, and an acquisition rule formulation unit. The signal acquisition unit monitors the axial vibration data of the ear of the high-speed rotating mechanism in real time. The electromagnetic shielding and isolation unit uses electromagnetic shielding materials to isolate the signal acquisition unit. The sampling rate and resolution optimization unit increases the details and accuracy of the data by improving the sampling rate and resolution. The acquisition rule formulation unit formulates reasonable acquisition rules to avoid data redundancy and errors.

[0110] Signal processing module: including real-time monitoring and feedback control unit, frequency domain analysis unit, time domain analysis unit, filter application unit, noise reduction algorithm unit and data fusion unit. It is used to monitor the quality and interference of the axial vibration signal in real time and adjust the processing parameters, convert the axial vibration signal to the frequency domain to identify the frequency components of the interference signal, analyze the periodicity or burst of the interference signal in the time domain, and then use a filter to remove the interference signal in the axial vibration signal and perform noise reduction.

[0111] Pattern recognition module: Use feature library data to train the pattern recognition algorithm and establish a recognition model for mechanical faults. The feature library data includes frequency domain, time domain and statistical features extracted from vibration data under normal and fault conditions. The recognition model identifies different mechanical states based on the axial vibration signal after noise reduction processing.

[0112] Diagnostic module: matches the vibration data collected in real time with the features in the feature comparison model to diagnose whether there is any abnormality or fault in the current mechanical state and obtain the diagnostic result.

[0113] Prediction and evaluation module: Based on the identified mechanical state and the diagnosis results, the health status of the mechanism is evaluated and predicted, and an alarm is triggered when the identified mechanical state and the diagnosis results are abnormal.

[0114] In this preferred embodiment, the real-time monitoring module is electrically connected to the first acquisition unit and the second acquisition unit. The real-time monitoring module can be used to ensure the accuracy of the collected axial vibration signal. By switching the first acquisition unit and the second acquisition unit, it can be ensured that the data collection work will not be interrupted due to the failure of the first acquisition unit, thereby improving the reliability of the monitoring system and ensuring that the monitoring of the axial vibration signal of the high-speed rotating mechanism can be carried out continuously.

[0115] It should be understood that although this specification is described according to implementation modes, not every implementation mode contains only one independent technical solution. This description of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each implementation mode may also be appropriately combined to form other implementation modes that can be understood by those skilled in the art.

[0116] The series of detailed descriptions listed above are only specific descriptions of feasible implementation methods of the present invention. They are not intended to limit the scope of protection of the present invention. Any equivalent implementation methods or changes that do not deviate from the technical spirit of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for monitoring axial vibration signals of high-speed rotating mechanisms, characterized in that: The steps include: S1. Creating feature library data for training a recognition model for mechanical faults of high-speed rotating mechanisms, wherein the recognition model includes a feature comparison model, a prediction model, and an evaluation model; The feature comparison model is trained by the SVM algorithm, specifically including: S11, selecting a training data set from the feature library data, {(\mathbf{x}_1, y_1, mathbf{x}_2, y_2, \ldots, mathbf{x}__n, y_n)} Among them, mathbf{x}_i is the input feature vector; y_1 is the corresponding category label (y_i\in{-1,+1}), S12, use SVM classifier to find a hyperplane to separate samples of different categories; the hyperplane is expressed as mathbf{w}\cdot\mathbf{x}+b=0, where mathbF{w} normal vector determines the direction of the hyperplane and b is the bias term; The optimization method is: min_{\mathbf{w},b}\frac{1}{2}|\mnathbf{w}|^2 And satisfy the constraints: y_i(\mathbf{w}\cdot\mathbf{x}_i+b)\geq1,\quad\text{for all}i=1,2,\ldots,nwhere |\cdot|^2 represents the L2 norm of the vector; mathbf{w}\cdot\mathbf{x} represents the inner product of vectors; S13, calculating the distances of samples of different categories from the optimal hyperplane found by the SVM algorithm, wherein the sample farthest from the hyperplane is the maximized margin sample; The prediction model is trained by a linear regression algorithm. hat{y}=w_0+w_1x_1+w_2x_2+\cdots+w_nx_n, where hat{y} is the predicted value, w_0, w_1, \cdots, w_n are model parameters, and x_1, x_2, \cdots, x_n are feature values; The evaluation model is trained by variance, variance ∫[xE(x)]^2f(x)dx, where E(x) represents mathematical expectation; S2, collecting the axial vibration signal of the high-speed rotating mechanism in real time and preprocessing it; S3, using the value of the maximized interval sample as a threshold, inputting the preprocessed axial vibration signal into the feature comparison model, and comparing it with the threshold. If it is within the threshold range, the mechanical state of the high-speed rotating mechanism is normal; if it is not within the threshold range, the mechanical state of the high-speed rotating mechanism is abnormal; S4. Inputting the axial vibration signal of abnormal mechanical state into the prediction model and the evaluation model for prediction and evaluation, generating prediction results and evaluation results, and triggering an alarm when the prediction results and the evaluation results are abnormal.

2. The method for monitoring axial vibration signals of a high-speed rotating mechanism according to claim 1, characterized in that: The feature library data includes collecting vibration data of the high-speed rotating mechanism under different mechanical states, extracting frequency domain, time domain and statistical features from the vibration data, and arranging and summarizing the features.

3. The method for monitoring axial vibration signals of a high-speed rotating mechanism according to claim 1, characterized in that: In step S2, "real-time collection of axial vibration signals of high-speed rotating mechanisms" is specifically as follows: using a first collection unit to collect the axial vibration signals of the high-speed rotating mechanism; if the collected axial vibration signals are abnormal, switching to a second collection unit to re-collect the axial vibration signals of the high-speed rotating mechanism; the second collection unit and the first collection unit are redundant with each other.

4. The method for monitoring axial vibration signals of a high-speed rotating mechanism according to claim 3 is characterized in that: "The collected axial vibration signal is abnormal" specifically means: using a vibration detection algorithm to monitor the axial vibration signal in real time, and the vibration monitoring algorithm is: text{if}\(\text{Peak Value}>T)\\text{then}\\text{Anomaly} Whether the peak value of the axial vibration signal collected by the first collection unit or the second collection unit exceeds a set threshold value, if it exceeds the set threshold value, it is judged as abnormal.

5. The method for monitoring axial vibration signals of a high-speed rotating mechanism according to claim 4, characterized in that: The "preprocessing" in step S2 is specifically: converting the axial vibration signal into the frequency domain to identify the frequency component of the interference signal, analyzing the periodicity or burstiness of the interference signal in the time domain, and then using a filter to remove the interference signal in the axial vibration signal and perform noise reduction.

6. The method for monitoring axial vibration signals of a high-speed rotating mechanism according to claim 5, characterized in that: "Using a filter to remove the interference signal in the axial vibration signal" specifically means: The filter output is y(n)=w(n)Tx(n), where y(n) is the filter output, w(n) is a vector containing the filter coefficients, and x(n) is the input signal vector; Error calculation: e(n) = d(n) - y(n), where e(n) is the error signal and d(n) is the desired response signal; Coefficient update: w(n+1)=w(n)+2μe(n)x(n), where μ is the step size parameter.

7. The method for monitoring axial vibration signals of a high-speed rotating mechanism according to claim 6, characterized in that: "Performing noise reduction" is to perform noise reduction processing using the SureShrink algorithm, and the steps of the SureShrink algorithm are as follows: Wavelet decomposition: Perform wavelet decomposition on the axial vibration signal to obtain a series of wavelet coefficients, including approximate coefficients and detail coefficients; Thresholding: Thresholding is applied to the wavelet coefficients at each level. In SureShrink, the selection of the threshold is based on the noise level of each wavelet subband. The threshold is set using the estimated noise level. Hard threshold or soft threshold: Depending on the selected threshold, the wavelet coefficients are processed by hard threshold or soft threshold. The hard threshold sets the coefficients whose absolute value is less than the threshold to zero, while the soft threshold shrinks these coefficients to zero. Inverse wavelet transform: Perform inverse wavelet transform on the processed wavelet coefficients to reconstruct the denoised signal; Data fusion is performed using the denoised signals from multiple signal sources.

8. The method for monitoring axial vibration signals of a high-speed rotating mechanism according to claim 1, characterized in that: The prediction results are specifically vibration displacement characteristics and vibration velocity characteristics. The vibration displacement characteristics and vibration velocity indicators are specifically vibration frequency, displacement and velocity characteristics including crack positions of high-speed rotating mechanisms, inner and outer ring faults of bearings, misaligned positions of rotating shafts, and foundation deformation.

9. The method for monitoring axial vibration signals of a high-speed rotating mechanism according to claim 1, characterized in that: The evaluation results at least include energy distribution index, probability distribution curve index and frequency domain analysis index; The energy distribution index is used to analyze the energy distribution of the axial vibration signal and identify and distinguish different fault types; The probability distribution curve indicator is used to analyze the probability distribution curve of the axial vibration signal data and evaluate the state of the rolling bearing; The frequency domain analysis index is used to identify specific fault frequency components.

10. A high-speed rotating mechanism axial vibration signal monitoring device, characterized in that: include Acquisition module: collects axial vibration signals of high-speed rotating mechanisms; Signal processing module: monitors the quality and interference of the axial vibration signal in real time and adjusts processing parameters, converts the axial vibration signal into the frequency domain to identify the frequency components of the interference signal, analyzes the periodicity or burstiness of the interference signal in the time domain, and then uses a filter to remove the interference signal in the axial vibration signal and perform noise reduction; Pattern recognition module: Use feature library data to train the pattern recognition algorithm and establish a recognition model for mechanical faults. The feature library data includes frequency domain, time domain and statistical features extracted from vibration data in normal and fault states. The recognition model recognizes different mechanical states based on the axial vibration signal after noise reduction processing. Diagnostic module: matches the vibration data collected in real time with the features in the feature comparison model to diagnose and determine whether there is any abnormality or failure in the current mechanical state; Prediction and evaluation module: Based on the identified mechanical state and the diagnosis results, the health status of the mechanism is evaluated and predicted, and an alarm is triggered when the identified mechanical state and the diagnosis results are abnormal; A real-time monitoring module is electrically connected to the first acquisition unit and the second acquisition unit.