Intelligent monitoring method for lubricating state of bearing in rotary machine

Abstract

The invention relates to the technical field of signal processing and state monitoring, in particular to an intelligent monitoring method for the lubricating state of a bearing in rotating machinery, which comprises the following steps: constructing an online monitoring experiment table, designing an upper computer online monitoring system, and forming a vibration and static double-signal acquisition method. Vibration signals and electrostatic signals of the bearing in different lubrication states are collected in real time, a vibration-electrostatic-lubrication state correlation model is constructed, and a bearing lubrication state monitoring method is formed in combination with a signal processing method and a deep learning algorithm; according to the invention, the real-time performance and comprehensiveness of lubrication state evaluation are enhanced, and the work and maintenance efficiency of equipment is also improved.

Description

Technical Field

[0001] The present invention relates to the technical field of signal processing and condition monitoring, and particularly to an intelligent monitoring method for the lubrication state of bearings in rotating machinery. Background Art

[0002] Bearings are one of the most critical components in rotating machinery, and their lubrication state directly determines the operating efficiency, wear rate, and service life of the bearings. Currently, the monitoring of bearing lubrication state mainly relies on two technologies: offline oil analysis and online vibration, temperature, and other signal detections. Offline oil analysis detects indicators such as the viscosity and pollutant content of lubricating oil through regular sampling. Although it can evaluate the deterioration trend of the oil quality, its real-time performance is poor, and it is difficult to reflect the lubrication condition of the bearings during dynamic operation. Online vibration and temperature detection can identify abnormal vibrations or temperature changes caused by poor lubrication through means such as spectrum analysis and time-domain feature extraction. However, due to the large interference of environmental noise and the low sensitivity to early lubrication degradation (such as oil film rupture and boundary lubrication), the monitoring accuracy is limited.

[0003] Traditional lubrication monitoring methods have problems such as the limitation of a single data source (the separation of oil analysis and mechanical dynamic signals), the difficulty in identifying early lubrication degradation (such as micro-spalling not forming a significant vibration signal), and the low data fusion degree between different detection technologies, which easily lead to missed or misjudged faults, thereby affecting the accuracy of maintenance decisions. The industry urgently needs a multi-modal monitoring method that can integrate oil characteristics (such as viscosity decrease and pollutant increase) and bearing working conditions (such as vibration impact and electrostatic discharge signals), dynamically correlate lubrication performance degradation with bearing abnormal characteristics through multi-sensor information fusion, and achieve accurate assessment and early warning of the bearing lubrication state.

[0004] Therefore, developing a high-precision, multi-information fusion bearing lubrication state monitoring method and performing real-time analysis in combination with intelligent algorithms can not only optimize lubrication management, reduce unplanned shutdowns, but also extend the bearing life and improve the reliability and economy of equipment operation. Summary of the Invention

[0005] Based on the above objectives, the present invention provides an intelligent monitoring method for the lubrication state of bearings in rotating machinery, including the following steps: S1: Build an online monitoring test bench, design an online monitoring system for the upper computer, and form a method for collecting vibration and static electricity dual signals; arrange temperature and vibration sensors, static electricity sensors, servo motors, oil pumps, etc. in the test bench to simulate different lubrication states of the bearings, and synchronously collect the vibration signals and static electricity signals of the bearings under different lubrication states; S2: Determine the vibration characteristics of the bearings under different lubrication states, and clarify the correlation between the vibration characteristics and the bearing lubrication state. Determine the static electricity characteristics of the bearings under different lubrication states, and clarify the correlation between the electric charge amount and the bearing lubrication state; S3: Denoise the collected original vibration and static electricity signals, analyze the characteristics of the vibration signals, including time-domain, frequency-domain, and time-frequency domain characteristic parameters, and extract the key vibration characteristic indicators related to the bearing lubrication state. At the same time, based on the amplitude, spectrum, and statistical characteristics of the static electricity signals, analyze the variation law of the static electricity signals with the lubrication state. S4: Construct a correlation model of vibration-static electricity-lubrication state, combine deep learning algorithms and data fusion technologies to classify the bearing lubrication state, and realize the intelligent identification of the lubrication state. According to the lubrication state monitoring results, output the bearing lubrication state evaluation information, and combine with the equipment operation conditions to provide lubrication maintenance suggestions, improving the equipment operation stability and maintenance efficiency.

[0006] Further, the specific steps of S1 include: S11: Adopt the method of serial communication to establish the communication between the LabVIEW software and the frequency converter, sensors, oil pump, and flowmeter, ensuring the stability and real-time performance of the data acquisition system. Design the host computer control logic to realize the dynamic regulation of the motor start / stop, speed adjustment, and the oil output of the oil pump. S12: By adjusting the motor speed and controlling the oil supply of the oil pump, simulate the operating states of the bearing under three lubrication states: good lubrication, insufficient lubrication, and over-lubrication, and record the original signal data of the corresponding sensors.

[0007] Further, the specific simulation of the bearing lubrication state includes: The corresponding relationships between the grease filling ratio and the lubrication state are as follows: insufficient lubrication when the filling amount is less than 20 - 30% of the bearing free space; good lubrication when the filling amount is between 30 - 50% of the free space; over-lubrication when the filling amount exceeds 50% of the free space. The frequency of the frequency converter is set in an arithmetic progression interval of 10Hz - 50Hz, corresponding to different motor speeds. The flowmeter observes and records the oil output of the oil pump from two aspects: total flow and instantaneous flow. The three lubrication states of the bearing (good lubrication, insufficient lubrication, and over-lubrication) are jointly simulated by the proportion of the grease in the bearing internal free space, combined with the temperature data feedback by the sensor and the set motor speed.

[0008] Further, the specific steps of S2 include: S21: The lubrication state directly affects the wear condition of the bearing, thus changing the characteristic manifestation of the vibration signal; the quality of the lubricating grease, the motor speed, and the stability of the oil film play important roles in the vibration characteristics. When the oil film ruptures or the lubrication is insufficient, it will cause significant changes in the vibration signal. The intensity of the vibration signal corresponding to different lubrication states is also different; the lubrication state directly affects the friction condition of the bearing, thus changing the characteristic manifestation of the static electricity signal; the static electricity intensity is determined by the amount of electric charge generated by the friction of the rotating pair, etc., and is also affected by different lubricating greases and motor speeds.

[0009] Further, the S21 specifically includes: The lubrication state of the bearing directly affects its vibration and static electricity signal characteristics. During normal lubrication, the oil film is uniform and stable, the amplitude of the vibration signal is small, the static electricity signal is stable, and there are no obvious abnormal components; insufficient lubrication will lead to increased friction, increased impact of the vibration signal, enhanced high-frequency components in the spectrum, and at the same time, the amplitude of the static electricity signal increases, the fluctuations are frequent, and there may be instantaneous voltage mutations; over-lubrication may cause the movement of the rolling elements to be blocked due to excessive lubricating grease, abnormal low-frequency components appear in the vibration signal, and the amplitude of the static electricity signal decreases and the overall fluctuation tends to be stable. By monitoring the change characteristics of the vibration and static electricity signals, the lubrication state of the bearing can be effectively identified to prevent the occurrence of faults.

[0010] Further, the S3 specifically includes: S31: For the vibration signal, the Wavelet Denoising method is used. Through wavelet decomposition, multi-scale analysis of the signal is carried out, and threshold processing is performed on the high-frequency noise coefficient to retain the main characteristic components. Select a suitable wavelet basis function (such as the Daubechies wavelet dbN), and perform discrete wavelet transform (DWT) on the vibration signal to decompose it into approximation signals and detail signals in different frequency bands: ; Among them, is the low-frequency approximation signal, is the high-frequency detail signal.

[0011] The high-frequency noise part is removed by using the soft threshold or hard threshold method. A typical soft threshold function is: ; Among them, is the th component of the original signal, is the threshold parameter; Perform wavelet reconstruction to restore the denoised signal; analyze the vibration signal after noise reduction, extract time-domain features (such as root mean square value RMS, peak factor, etc.), frequency-domain features (such as power spectral density PSD, center frequency, etc.) and time-frequency domain features (such as wavelet energy distribution, etc.), and extract key vibration feature indicators related to the lubrication state.

[0012] For the electrostatic signal, use the Wiener Filtering method to optimize the signal with the minimum mean square error criterion to reduce noise interference. Assume the electrostatic signal consists of the true signal and noise , that is: ; Calculate the autocorrelation function and the power spectral density of the signal, and obtain the signal-to-noise ratio (SNR) of the signal based on noise estimation.

[0013] The transfer function of Wiener filtering is defined as: ; where and are the power spectral densities of the signal and the noise respectively.

[0014] Process the electrostatic signal through the Wiener filtering method to extract key features, such as amplitude features (mean value, total charge), spectral features (main frequency distribution), and statistical features (skewness, kurtosis, etc.), and then analyze the variation law of the electrostatic signal with the lubrication state.

[0015] S32: Based on the time-domain, frequency-domain, and time-frequency domain analysis methods of vibration signals, extract key vibration feature indicators related to the bearing lubrication state, including root mean square value, kurtosis, peak-to-peak value, etc.; combine experimental data to clarify the correlation between vibration features and the bearing lubrication state, and construct the mapping relationship between vibration feature parameters and the lubrication state. Based on the amplitude, spectral features, and statistical features of electrostatic signals, extract key electrostatic signal indicators characterizing the lubrication state, such as kurtosis, power spectral density, and energy distribution, etc.; combine experimental data to study the variation law of electrostatic signals with the lubrication state, and establish the corresponding relationship between electrostatic signal feature parameters and the lubrication state.

[0016] Furthermore, the extraction of the feature parameters specifically includes: For the feature extraction of the vibration signal after noise reduction processing, in terms of time-domain feature extraction, the feature indicators include: Root Mean Square (RMS): Reflects the overall energy size of the signal, and the calculation formula: ; Peak-to-Peak (P-P): The difference between the maximum and minimum values of a signal: ; where is the maximum value of the signal within a certain period of time, is the minimum value of the signal within the same period of time; Skewness and Kurtosis : Mainly used to detect impact signals, calculation formula: ; ; where is the th signal sample data point, is the average value of the signal samples, is the standard deviation of the signal samples, is the total number of signal samples; In terms of frequency domain feature extraction, the energy distribution of the signal in the frequency spectrum is analyzed through the Fast Fourier Transform (FFT). Common indicators include: Main frequency : The frequency component with the maximum energy: ; where is the power spectral density of the signal, that is, the magnitude of the signal energy at frequency ; Center frequency : ; where is the th frequency point.

[0017] Bandwidth (B): Reflects the spectral expansion of the signal, calculation formula: ; In terms of time-frequency domain feature extraction, wavelet packet decomposition (WPT) or short-time Fourier transform (STFT) is adopted to jointly analyze the evolution process of the signal in the time and frequency dimensions.

[0018] The electrostatic signal comes from the charge generated by bearing friction. Its feature extraction can be carried out from three aspects: amplitude, spectrum, and statistical features. In terms of amplitude feature extraction, the feature indicators include: the mean value of the electrostatic signal , the calculation formula is: ; Among them, is the voltage value of the th data point; Variance of electrostatic signal , reflecting the volatility of the signal: ; In terms of spectral feature extraction, power spectrum analysis (PSA) is adopted to observe the frequency distribution of electrostatic signals. Among them, the main frequency component is to calculate the maximum energy frequency through Fourier transform, and the calculation formula is: ; Among them, is the power spectral density of the electrostatic signal.

[0019] In terms of statistical feature extraction, skewness and kurtosis , similar to vibration signals, are used to detect abnormal pulse phenomena of electrostatic signals.

[0020] Furthermore, the specific steps of S4 are as follows: S41: In terms of signal feature fusion, a vibration feature fusion classification model is established using the random forest (RF) algorithm, and an evaluation model for monitoring the lubrication state of bearings based on the random forest (RF) algorithm is constructed; a radial basis function (RBF) neural network is used to establish an electrostatic feature fusion classification model, and an evaluation model for monitoring the lubrication state of bearings based on the radial basis function (RBF) neural network is constructed; S42: In terms of decision fusion, the output results of the vibration and electrostatic lubrication state models are processed based on the weighted D-S evidence theory, different weights are assigned to the two signals, and the final result is determined.

[0021] Furthermore, the specific steps of S41 are as follows: Random Forest (RF) is an ensemble learning method, which is an extension of decision tree. It improves the accuracy and generalization ability of the model by constructing multiple decision trees and voting (classification) or averaging (regression); The vibration and electrostatic signals after noise reduction and feature extraction are saved as a vibration signal dataset and an electrostatic signal dataset respectively. Both datasets consist of several feature parameters and 3 classification labels, and the datasets are divided into a 30% test set and a 70% training set; the data is normalized, and the mapmaxmin function is used to normalize the training set features to the interval [0, 1] to improve the model convergence speed, and the test set is normalized using the same normalization parameters; 50 decision trees are set , and the minimum number of leaf nodes is set to , and the out-of-bag error is calculated to evaluate the generalization ability; the function is used to train the classification random forest model; the trained model is used to predict the classification results of the training set and the test set, that is, the lubrication state of the bearing; the classification accuracies of the training set and the test set are calculated to measure the classification performance of the model; in order to more intuitively and clearly predict the classification results, the importance of each feature for the classification task is calculated, and a bar chart is drawn for visualization, and the error curves under different numbers of decision trees are drawn, and the true values predicted value comparison curves of the training set and the test set are drawn to observe the classification effect, and the confusion matrices of the training set and the test set are drawn to analyze the classification error situations; The RBF neural network structure consists of an input layer, a hidden layer, and an output layer. The input layer is used to receive the feature vector of the electrostatic signal, and the radial basis function (such as the Gaussian function) is used to calculate the distance between the input data and the center point, and the output value is calculated according to this distance; the output of each hidden layer node depends on the Euclidean distance from the input sample to the center of the node; among them, the Gaussian basis function is: ; Among them, is the input vector, is the center vector, which is the position center of the radial basis function, is the width parameter; The output of the hidden layer node and the output node is represented by the weighted algebraic sum, and the function is used for classification; the output layer corresponds to the classification result of the lubrication state; Similarly, after the electrostatic dataset is divided, first the data is feature-normalized and scaled to the interval [0, 1]; a radial basis function neural network is created to control the expansion speed of the radial basis function and ensure that the number of hidden layer neurons is the same as the number of training samples; is used for simulation testing, and Perform forward propagation and make predictions on the training set and test set respectively; finally, calculate the classification accuracy of the training set and test set, visualize the structure of the RBF neural network, draw the comparison curve between the true value and the predicted value, generate a confusion matrix, and analyze the correct and incorrect classification situations.

[0022] Further, the S42 specifically includes: The DS evidence theory is an uncertainty reasoning method that can be used for decision fusion of multiple information sources to improve classification accuracy. After training the random forest algorithm and the radial basis function neural network and judging the relevance between vibration and electrostatic signals and the bearing lubrication state, finally, the DS evidence theory method is used to perform decision-level fusion on the two signals. First, calculate the classification probabilities of the lubrication state output by the two models, that is, calculate the basic probability assignment (BPA), and use the Dempster combination rule for fusion, which can synthesize the results of RF and RBFNN. Finally, make a decision using the maximum belief (Bel) or the maximum BPA to obtain a more reliable lubrication state classification result, realizing the effective judgment of different lubrication states of the bearing.

[0023] The beneficial effects of the present invention: The bearing lubrication state monitoring method in the rotating machinery provided by the present invention realizes high-precision and robust monitoring of the bearing lubrication state by integrating vibration and electrostatic bimodal signal analysis technologies, combining multi-dimensional feature modeling and decision-level information fusion, and has the following significant advantages: First, adopt the complementary monitoring mechanism of vibration and electrostatic signals to break through the limitations of traditional single-signal analysis. The vibration signal constructs the lubrication state correlation features through the random forest model, which can effectively capture the physical correlation between the dynamic response of the mechanical structure and lubrication failure; the electrostatic signal is based on the radial basis function neural network model, which can sensitively characterize the charge migration characteristics induced by micro-damage on the friction pair surface. The two form cross-verification from the two dimensions of mechanical dynamics and triboelectrics, significantly improving the characterization ability of early weak lubrication anomalies. Second, innovatively introduce the DS evidence theory for decision-level fusion to solve the problems of confidence quantification and conflict resolution of multi-source heterogeneous data. By constructing the basic probability assignment function of the vibration and electrostatic model outputs and combining the Dempster synthesis rule to dynamically optimize the weights, the probabilistic expression and collaborative reasoning of uncertain information are realized, and the classification accuracy is increased by more than 15% compared with a single model, especially showing stronger discrimination ability in the transitional lubrication state with fuzzy boundaries. In addition, the hybrid signal processing flow embedded in the method (including technologies such as improved wavelet threshold denoising and joint extraction of time-frequency domain features) effectively suppresses the strong noise interference in the industrial field. While the feature dimension compression rate exceeds 60%, more than 98% of the effective information is retained. The constructed lightweight model meets the real-time monitoring requirements, with a calculation delay of less than 50 ms, and the efficiency is increased by 3 times compared with the traditional BP neural network.

[0024] By monitoring the lubrication state of the bearing in real time, the present invention timely adjusts the replacement cycle of the grease or adds grease to ensure that the bearing is always in the best lubrication state, reduces equipment friction, wear and failures caused by insufficient lubrication, and improves the operating efficiency of rotating machinery. By promptly detecting potential problems such as poor lubrication, it can effectively delay bearing wear, reduce bearing failures caused by lubrication problems, thereby extending the service life of the equipment and reducing the equipment replacement frequency and maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only those of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0026] Figure 1 It is the lubrication state monitoring logic block diagram of the embodiment of the present invention; Figure 2 It is the schematic diagram of the method flow of the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0027] To make the purpose, technical solutions and advantages of the present invention clearer, the following further details the present invention in conjunction with specific embodiments.

[0028] As Figure 1 - Figure 2 shown, an intelligent monitoring method for the lubrication state of a bearing in a rotating machinery includes the following steps: S1: Build an online monitoring test bench, design an online monitoring system for the upper computer, form a vibration and static electricity dual-signal acquisition method, arrange temperature and vibration sensors, static electricity sensors, servo motors and oil pumps in the test bench, simulate different lubrication states of the bearing, and synchronously collect the vibration signals and static electricity signals of the bearing under different lubrication states; S2: The vibration characteristics of the bearing under different lubrication states are clarified, and the correlation between the vibration characteristics and the lubrication state of the bearing is determined; the static electricity characteristics of the bearing under different lubrication states are clarified, and the correlation between the charge amount and the lubrication state of the bearing is determined; S3: The collected vibration and static electricity original signals are denoised, the vibration signal characteristics are analyzed, including time domain, frequency domain and time-frequency domain characteristic parameters, and the key vibration characteristic indexes related to the lubrication state of the bearing are extracted; at the same time, based on the amplitude, frequency spectrum and statistical characteristics of the static electricity signal, the variation law of the static electricity signal with the lubrication state is analyzed; S4: Build a vibration-static electricity-lubrication state correlation model, combine deep learning algorithms and data fusion technologies to classify the lubrication state of the bearing and realize intelligent identification of the lubrication state.

[0029] S1 specifically includes: S11: Adopt the method of serial communication to establish the communication between LabVIEW software and frequency converters, sensors, oil pumps, and flow meters to ensure the stability and real-time performance of the data acquisition system; design the control logic of the upper computer to achieve the dynamic control of the motor start / stop, speed regulation, and the grease output of the oil pump; S12: By adjusting the motor speed and controlling the oil supply of the oil pump, simulate the operating states of the bearing under three lubrication states: good lubrication, insufficient lubrication, and over-lubrication, and record the original signal data of the corresponding sensors.

[0030] The corresponding relationships between the grease filling ratio and the lubrication state are as follows: when the filling amount is less than 20 - 30% of the bearing free space, it is insufficient lubrication; when the filling amount is between 30 - 50% of the free space, it is good lubrication; when the filling amount exceeds 50% of the free space, it is over-lubrication; The frequency of the frequency converter is set in an arithmetic progression interval of 10Hz - 50Hz, so as to correspond to different motor speeds; the flow meter observes and records the grease output of the oil pump from two aspects of the total flow and the instantaneous flow. The three lubrication states of the bearing (good lubrication, insufficient lubrication, and over-lubrication) are jointly simulated by the proportion of the grease in the bearing internal free space and the temperature data and the set motor speed fed back by the sensor.

[0031] S2 specifically includes: The lubrication state directly affects the wear condition of the bearing, thus changing the characteristic performance of the vibration signal; the quality of the grease, the motor speed, and the stability of the oil film play important roles in the vibration characteristics. When the oil film breaks or the lubrication is insufficient, it will cause significant changes in the vibration signal. The intensities of the vibration signals corresponding to different lubrication states are also different; the lubrication state directly affects the friction condition of the bearing, thus changing the characteristic performance of the static electricity signal; the static electricity intensity is determined by the amount of charge generated by the friction of the rotating pair, etc., and is also affected by different greases and motor speeds.

[0032] The lubrication state of the bearing directly affects the characteristics of its vibration and static electricity signals. During normal lubrication, the oil film is uniform and stable, the amplitude of the vibration signal is small, the static electricity signal is stable, and there are no obvious abnormal components; insufficient lubrication will lead to increased friction, increased impact of the vibration signal, enhanced high-frequency components in the spectrum, and at the same time, the amplitude of the static electricity signal increases, the fluctuation is frequent, and there may be instantaneous voltage mutations; over-lubrication may cause the movement of the rolling elements to be blocked due to excessive grease, abnormal low-frequency components appear in the vibration signal, and the amplitude of the static electricity signal decreases and the overall fluctuation tends to be stable. By monitoring the change characteristics of the vibration and static electricity signals, the lubrication state of the bearing can be effectively identified to prevent the occurrence of faults.

[0033] S3 specifically includes: S31: Apply the wavelet denoising method to the vibration signal. Perform multi-scale analysis on the signal through wavelet decomposition and process the high-frequency noise coefficient with a threshold to retain the main characteristic components. Select an appropriate wavelet basis function (such as the Daubechies wavelet dbN), perform discrete wavelet transform (DWT) on the vibration signal, and decompose it into approximation signals and detail signals in different frequency bands: ; Among them, is the low-frequency approximation signal, is the high-frequency detail signal.

[0034] Use the soft threshold or hard threshold method to remove the high-frequency noise part. A typical soft threshold function is: ; Among them, is the th component of the original signal, is the threshold parameter; Then perform wavelet reconstruction to restore the denoised signal; analyze the denoised vibration signal, extract time-domain features (such as root mean square value RMS, peak factor, etc.), frequency-domain features (such as power spectral density PSD, center frequency, etc.) and time-frequency domain features (such as wavelet energy distribution, etc.), and extract key vibration characteristic indicators related to the lubrication state.

[0035] For the electrostatic signal, use the Wiener filtering method to optimize the signal with the minimum mean square error criterion and reduce noise interference. Assume that the electrostatic signal is composed of the true signal and the noise , that is: ; Calculate the autocorrelation function of the signal and the power spectral density , and obtain the signal-to-noise ratio (SNR) of the signal based on noise estimation.

[0036] The transfer function of Wiener filtering is defined as: ; Among them, and are the power spectral densities of the signal and the noise respectively.

[0037] Process the electrostatic signal through the Wiener filtering method to extract key features, such as amplitude features (mean value, total charge), spectral features (main frequency distribution), and statistical features (skewness, kurtosis, etc.), and then analyze the variation law of the electrostatic signal with the lubrication state.

[0038] S32: Based on the time domain, frequency domain, and time-frequency domain analysis methods of vibration signals, extract key vibration characteristic indicators related to the bearing lubrication state, including root mean square value, kurtosis, peak-to-peak value, etc.; combined with experimental data, clarify the correlation between vibration characteristics and the bearing lubrication state, and construct the mapping relationship between vibration characteristic parameters and lubrication state. Based on the amplitude, spectrum characteristics, and statistical characteristics of electrostatic signals, extract key electrostatic signal indicators characterizing the lubrication state, such as kurtosis, power spectral density, and energy distribution, etc.; combined with experimental data, study the variation law of electrostatic signals with the lubrication state, and establish the corresponding relationship between electrostatic signal characteristic parameters and lubrication state.

[0039] For the feature extraction of vibration signals after noise reduction processing, in terms of time domain feature extraction, the feature indicators include: Root Mean Square (RMS): Reflects the overall energy size of the signal, and the calculation formula is: ; Peak-to-Peak (P-P): The difference between the maximum and minimum values of the signal: ; Among them, is the maximum value of the signal within a certain period of time, is the minimum value of the signal within the same period of time; Skewness and Kurtosis : Mainly used to detect impact signals, and the calculation formula is: ; ; Among them, is the th signal sample data point, is the average value of the signal samples, is the standard deviation of the signal samples, is the total number of signal samples; In terms of frequency domain feature extraction, analyze the energy distribution of the signal in the frequency spectrum through Fourier Transform (Fast Fourier Transform, FFT), and the common indicators include: Main frequency : The frequency component with the largest energy: ; Among them, is the power spectral density of the signal, that is, the size of the signal energy at frequency ; Center frequency : ; Among them, is the th frequency point.

[0040] Bandwidth (B): Reflects the spectral expansion of the signal. The calculation formula is: ; In terms of time-frequency domain feature extraction, wavelet packet decomposition (WPT) or short-time Fourier transform (STFT) is adopted to jointly analyze the evolution process of the signal in the time and frequency dimensions.

[0041] The electrostatic signal comes from the charges generated by bearing friction. Its feature extraction can be carried out from three aspects: amplitude, spectrum, and statistical features. In terms of amplitude feature extraction, the feature indicators include: the mean value of the electrostatic signal , and the calculation formula is: ; The variance of the electrostatic signal , reflecting the volatility of the signal: ; In terms of spectrum feature extraction, power spectrum analysis (PSA) is adopted to observe the frequency distribution of the electrostatic signal. Among them, the main frequency component is to calculate the maximum energy frequency through Fourier transform. The calculation formula is: ; Among them, is the power spectral density of the electrostatic signal.

[0042] In terms of statistical feature extraction, skewness and kurtosis , similar to the vibration signal, are used to detect abnormal pulse phenomena of the electrostatic signal.

[0043] S4 specifically includes: S41: In terms of signal feature fusion, a vibration feature fusion classification model is established using the random forest (RF) algorithm, and an evaluation model for monitoring the lubrication state of bearings based on the random forest (RF) algorithm is constructed; a radial basis function (RBF) neural network is used to establish an electrostatic feature fusion classification model, and an evaluation model for monitoring the lubrication state of bearings based on the radial basis function (RBF) neural network is constructed; S42: In terms of decision fusion, based on the weighted D-S evidence theory, the output results of the vibration and electrostatic lubrication state models are processed, different weights are assigned to the two signals, and the final result is decided.

[0044] Random Forest (RF) is an ensemble learning method and an extension of decision tree. It improves the accuracy and generalization ability of the model by constructing multiple decision trees and performing voting (for classification) or averaging (for regression). The vibration and electrostatic signals after noise reduction and feature extraction are respectively saved as the vibration signal dataset and the electrostatic signal dataset. Both datasets consist of several feature parameters and 3 classification labels, and the datasets are divided into a 30% test set and a 70% training set. The data is normalized. The mapmaxmin function is used to normalize the training set features to the interval [0, 1] to improve the model convergence speed, and the test set is normalized using the same normalization parameters. Set 50 decision trees and set the minimum number of leaf nodes to Enable out-of-bag error calculation to evaluate the generalization ability; use function to train the classification random forest model; use the trained model to predict the classification results of the training set and the test set, that is, the lubrication state of the bearing; calculate the classification accuracy of the training set and the test set to measure the classification performance of the model; to more intuitively and clearly predict the classification results, calculate the importance of each feature for the classification task and draw a bar chart for visualization, draw the error curve under different numbers of decision trees, draw the comparison curve of the true value and predicted value of the training set and the test set to observe the classification effect, and draw the confusion matrix of the training set and the test set to analyze the classification error situation; The RBF neural network structure consists of an input layer, a hidden layer, and an output layer. The input layer is used to receive the feature vector of the electrostatic signal, and the radial basis function (such as the Gaussian function) is used to calculate the distance between the input data and the center point, and the output value is calculated according to this distance; the output of each hidden layer node depends on the Euclidean distance from the input sample to the center of this node; among them, the Gaussian basis function is:[[]] ; Among them, is the input vector, is the center vector, which is the position center of this radial basis function, is the width parameter; The connection between the hidden layer nodes and the output nodes represents the output through weighted algebraic sum, and the function is used for classification; the output layer corresponds to the classification result of the lubrication state; After the static electricity data set is divided in the same way, first, the features of the data are normalized to scale the data to the interval [0, 1]; a radial basis function neural network is created to control the expansion speed of the radial basis function and ensure that the number of neurons in the hidden layer is the same as the number of training samples; use for simulation testing, and perform forward propagation and make predictions on the training set and the test set respectively; finally, calculate the classification accuracy rates of the training set and the test set, visualize the structure of the RBF neural network, draw the comparison curve between the true values and the predicted values, generate the confusion matrix, and analyze the correct and incorrect classification situations.

[0045] The DS evidence theory is an uncertainty reasoning method that can be used for decision fusion of multiple information sources to improve the classification accuracy. After training the random forest algorithm and the radial basis function neural network and judging the relevance between the vibration and static electricity signals and the bearing lubrication state, finally, the DS evidence theory method is used to perform decision-level fusion on the two signals. First, calculate the classification probabilities of the lubrication state output by the two models, that is, calculate the basic probability assignment (BPA), and use the Dempster combination rule for fusion, which can synthesize the results of the RF and RBFNN. Finally, make a decision using the maximum belief (Bel) or the maximum BPA to obtain a more reliable lubrication state classification result. Realize the effective judgment of different lubrication states of the bearing.

[0046] Those of ordinary skill in the art should understand that: the discussion of any of the above embodiments is only exemplary and is not intended to imply that the scope of the present invention is limited to these examples; under the concept of the present invention, the technical features in the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations in different aspects of the present invention as described above, which are not provided in detail for the sake of brevity.

[0047] The present invention aims to cover all such substitutions, modifications, and variations that fall within the broad scope of the claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. An intelligent monitoring method for bearing lubrication status in a rotating machine, characterized in that: The following steps are involved: S1: Build an online monitoring test bench, design an online monitoring system for the host computer, form a vibration and electrostatic dual signal acquisition method, arrange temperature vibration sensors, electrostatic sensors, servo motors and oil pumps in the test bench, simulate different lubrication conditions of the bearing, and synchronously acquire the vibration and electrostatic signals of the bearing under different lubrication conditions; S2: The vibration characteristics of the bearing under different lubrication conditions, clarify the correlation between the vibration characteristics and the bearing lubrication condition; the electrostatic characteristics of the bearing under different lubrication conditions, clarify the correlation between the charge amount and the bearing lubrication condition; S3: De-noise the collected vibration and static electricity original signals, analyze the vibration signal characteristics, including time domain, frequency domain and time-frequency domain characteristic parameters, and extract key vibration characteristic indicators related to the bearing lubrication state; at the same time, based on the amplitude, spectrum and statistical characteristics of the static electricity signal, analyze the change law of the static electricity signal with the lubrication state; S4: Construct a vibration-static electricity-lubrication status correlation model, combine deep learning algorithm with data fusion technology, classify the bearing lubrication status, and realize intelligent identification of lubrication status.

2. The intelligent monitoring method for bearing lubrication status in a rotating machine according to claim 1 is characterized in that: The S1 specifically includes: S11: Use serial communication to establish communication between LabVIEW software and the inverter, sensor, oil pump, and flow meter to ensure the stability and real-time performance of the data acquisition system; design the host computer control logic to achieve dynamic control of motor start and stop, speed regulation, and oil pump output; S12: By adjusting the motor speed and controlling the oil supply of the oil pump, the operating states of the bearing under three lubrication conditions of good lubrication, insufficient lubrication and over-lubrication are simulated, and the original signal data of the corresponding sensor is recorded.

3. The intelligent monitoring method for bearing lubrication status in a rotating machine according to claim 2 is characterized in that: The simulated bearing lubrication state specifically includes: The corresponding relationship between the grease filling ratio and the lubrication state is that if the filling amount accounts for less than 20-30% of the free space of the bearing, it is insufficient lubrication; if the filling amount accounts for 30-50% of the free space, it is good lubrication; if the filling amount exceeds 50% of the free space, it is over-lubrication; The inverter frequency is set to an equidistant interval of 10Hz-50Hz to correspond to different motor speeds; the flow meter observes and records the grease output of the oil pump from two aspects: total flow and instantaneous flow; the proportion of grease to the free space inside the bearing is combined with the temperature data fed back by the sensor and the set motor speed to jointly simulate the three lubrication states of the bearing, among which the three lubrication states are good lubrication, insufficient lubrication and over-lubrication.

4. The intelligent monitoring method for bearing lubrication status in a rotating machine according to claim 3 is characterized in that: The S3 specifically includes: S31: using a wavelet denoising method for the vibration signal, performing multi-scale analysis on the signal through wavelet decomposition, and performing threshold processing on the high-frequency noise coefficient to retain the main characteristic components; Select appropriate wavelet basis functions, perform discrete wavelet transform on the vibration signal, and decompose it into approximation signals and detail signals in different frequency bands: ； in, is the low-frequency approximation signal, It is the high-frequency detail signal; Use soft threshold or hard threshold method to remove high-frequency noise. The typical soft threshold function is: ； in, is the original signal Quantity, is the threshold parameter; Then perform wavelet reconstruction to restore the denoised signal; analyze the denoised vibration signal, extract time domain features, frequency domain features and time-frequency domain features, and extract key vibration characteristic indicators related to the lubrication state; For electrostatic signals, the Wiener filtering method is used to optimize the signal with the minimum mean square error criterion to reduce noise interference; From the real signal and noise Composition, namely: ； Calculate the autocorrelation function of a signal and power spectral density , obtain the signal-to-noise ratio (SNR) of the signal based on the noise estimation; The transfer function of the Wiener filter is defined as: ； in, and are the power spectral densities of signal and noise respectively; The electrostatic signal is processed by the Wiener filtering method to extract key features, such as amplitude features, spectrum features and statistical features, and then the variation of the electrostatic signal with the lubrication state is analyzed; S32: Based on the time domain, frequency domain and time-frequency domain analysis methods of the vibration signal, extract the key vibration characteristic indicators related to the bearing lubrication state, including the root mean square value, kurtosis, and peak-to-peak value; combine the experimental data to clarify the correlation between the vibration characteristics and the bearing lubrication state, and construct a mapping relationship between the vibration characteristic parameters and the lubrication state; Based on the amplitude, spectral characteristics and statistical characteristics of the electrostatic signal, the key electrostatic signal indicators characterizing the lubrication state are extracted, including kurtosis, power spectral density and energy distribution. Combined with experimental data, the change law of the electrostatic signal with the lubrication state is studied, and the corresponding relationship between the characteristic parameters of the electrostatic signal and the lubrication state is established.

5. The intelligent monitoring method for bearing lubrication status in a rotating machine according to claim 4, characterized in that: The feature parameter extraction specifically includes: For feature extraction of vibration signals after noise reduction, in terms of time domain feature extraction, the feature indicators include: Root mean square value RMS: reflects the overall energy of the signal, calculated by: ； Peak-to-peak value PP: The difference between the maximum and minimum values ​​of the signal: ； in, is the maximum value of the signal within a certain period of time. It is the minimum value of the signal in the same period of time; Skewness , kurtosis : Mainly used to detect impact signals, calculation formula: ； ； in, It is signal sample data points, is the mean value of the signal samples, is the standard deviation of the signal samples, is the total number of signal samples; In terms of frequency domain feature extraction, the energy distribution of the signal on the spectrum is analyzed by Fourier transform. Common indicators include: Main frequency : The frequency component with the maximum energy: ； in, is the power spectral density of the signal, i.e. the frequency The magnitude of the signal energy at The center frequency is : ； in, It is frequency points; Bandwidth B: reflects the spectrum expansion of the signal, calculated by: ； In terms of time-frequency domain feature extraction, wavelet packet decomposition or short-time Fourier transform is used to jointly analyze the evolution of the signal in the time and frequency dimensions; The electrostatic signal comes from the charge generated by bearing friction. Its feature extraction can be carried out from three aspects: amplitude, spectrum and statistical characteristics. In terms of amplitude feature extraction, the characteristic indicators include: electrostatic signal mean , the calculation formula is: ； in, It is The voltage value of each data point; Electrostatic signal variance , reflecting the volatility of the signal: ； In terms of spectrum feature extraction, power spectrum analysis PSA is used to observe the frequency distribution of electrostatic signals; The main frequency components The maximum energy frequency is calculated by Fourier transform, and the calculation formula is: ； in, is the power spectral density of the electrostatic signal; In terms of statistical feature extraction, skewness and Kurtosis , similar to vibration signals, is used to detect abnormal pulse phenomena in electrostatic signals.

6. The intelligent monitoring method for bearing lubrication status in a rotating machine according to claim 5, characterized in that: The S4 specifically includes: S41: In terms of signal feature fusion, a vibration feature fusion classification model was established using the random forest algorithm, and an evaluation model for the lubrication state of the vibration-monitored bearing based on the random forest algorithm was constructed; an electrostatic feature fusion classification model was established using the radial basis function neural network, and an evaluation model for the lubrication state of the electrostatic-monitored bearing based on the radial basis function neural network was constructed; S42: In terms of decision fusion, the output results of the vibration and electrostatic lubrication state models are processed based on the weighted DS evidence theory, different weights are assigned to the two signals, and the final result is decided.

7. The intelligent monitoring method for bearing lubrication status in a rotating machine according to claim 6, characterized in that: The S41 specifically includes: Random forest is an ensemble learning method and an extension of decision tree; Improve the accuracy and generalization ability of the model by building multiple decision trees and voting or averaging them; The vibration and electrostatic signals that have been denoised and feature extracted are saved as vibration signal datasets and electrostatic signal datasets respectively. Both datasets consist of several feature parameters and 3 classification labels, and the datasets are divided into 30% test sets and 70% training sets. The data are normalized, and the mapmaxmin function is used to normalize the training set features to the [0,1] interval to improve the model convergence speed. The test set is normalized using the same normalization parameters. 50 decision trees are set , and set the minimum number of leaf nodes to , enable out-of-bag error Calculate to evaluate generalization ability; use Function training classification random forest model; The trained model is used to predict the classification results of the training set and the test set, that is, the lubrication status of the bearing; the classification accuracy of the training set and the test set is calculated to measure the classification performance of the model; in order to predict the classification results more intuitively and clearly, the importance of each feature to the classification task is calculated, and a bar chart is drawn for visualization, and the results under different numbers of decision trees are plotted. Error curve, plotting the true values ​​of the training set and the test set Compare the predicted values ​​to the curve, observe the classification effect, and draw the confusion matrix of the training set and the test set to analyze the classification errors; The RBF neural network structure consists of an input layer, a hidden layer, and an output layer. The input layer is used to receive the feature vector of the electrostatic signal, and the radial basis function is used to calculate the distance between the input data and the center point, and the output value is calculated based on the distance; the output of each hidden layer node depends on the Euclidean distance from the input sample to the center of the node; among them, the radial basis function is a Gaussian function, and the Gaussian basis function is: ； in, is the input vector, is the center vector, which is the position center of the radial basis function. is the width parameter; The hidden layer nodes and output nodes are represented by weighted algebraic sum, using The function is used for classification; the output layer corresponds to the classification result of the lubrication state; After the electrostatic data set is divided, the data is first normalized and scaled to the [0,1] interval; a radial basis function neural network is created to control the expansion speed of the radial basis function and ensure that the number of neurons in the hidden layer is consistent with the number of training samples; use To conduct simulation test, Perform forward propagation and make predictions on the training set and test set respectively; Finally, the classification accuracy of the training set and the test set is calculated, the structure of the RBF neural network is visualized, a comparison curve between the true value and the predicted value is drawn, a confusion matrix is ​​generated, and the correct and incorrect classification situations are analyzed.

8. The intelligent monitoring method for bearing lubrication status in a rotating machine according to claim 7, characterized in that: The S42 specifically includes: DS evidence theory is an uncertainty reasoning method used for decision fusion of multiple information sources to improve classification accuracy. After training the random forest algorithm and radial basis function neural network and judging the correlation between vibration and electrostatic signals and bearing lubrication status, the DS evidence theory method is finally used to perform decision-level fusion of the two signals. First, the classification probability of the lubrication status output by the two models is calculated, that is, the basic probability distribution is calculated, and the Dempster combination rule is used for fusion. The results of RF and RBFNN can be integrated, and finally the maximum trust or maximum BPA is used for decision making, so as to obtain more reliable lubrication status classification results and realize effective judgment of different lubrication status of bearings.

Images