Bearing bush surface oil film thickness and pressure distribution measuring method based on ultrasonic guided waves

Abstract

The invention discloses a method for measuring the thickness of an oil film on the surface of a bearing bush and pressure distribution based on ultrasonic guided waves. The method comprises the following specific steps: step 1, embedding a piezoelectric plate array in the back of a to-be-measured bearing bush; step 2, acquiring a reference signal, changing oil film pressure and film thickness, and acquiring a guided wave signal; step 3, acquiring a noise signal, and performing trend noise fitting on the noise signal; 4, after trend noise fitting is completed, a segmented RMS index is established, and pressure and film thickness features are extracted; and 5, importing the extracted guided wave data into a convolutional neural network in deep learning to decouple coupling features, and imaging the surface of the bearing bush by using an RAPID imaging algorithm based on a probability damage imaging principle to complete measurement. According to the bearing bush surface oil film thickness and pressure distribution measuring method based on the ultrasonic guided waves, the thickness of the oil film can be accurately calculated, and the pressure of the oil film can be calculated.

Description

Technical Field

[0001] The invention belongs to the technical field of ultrasonic guided waves, and in particular relates to a method for measuring the oil film thickness and pressure distribution on the surface of a bearing bush based on ultrasonic guided waves. Background Art

[0002] Sliding bearings are key components in many mechanical systems, and their performance directly affects the efficiency and life of the machine. The oil film pressure of the sliding bearing affects the bearing's load-bearing capacity and operating stability, while the oil film thickness determines the lubrication effect and friction resistance. Therefore, the oil film pressure and oil film thickness of the sliding bearing are important indicators to measure whether the working state of the sliding bearing is normal. The existing ultrasonic oil film detection technology calculates the distance between the upper and lower surfaces of the oil film through the reflection time difference of the ultrasonic body wave. Its measurement accuracy has extremely high requirements on the ultrasonic emission and sampling rate, resulting in high costs and the accuracy being greatly affected by the monitoring environment. At the same time, its detection range is limited by the propagation distance of the ultrasonic body wave, and each measuring point can only detect local oil film thickness and pressure and other parameters. Therefore, it is difficult to promote in practical applications.

[0003] Ultrasonic guided wave detection technology uses the propagation characteristics of Lamb waves, such as low attenuation and long propagation distance, along the propagation path in plate-like media to detect changes in the waveguide properties on the propagation path. On the surface of the sliding bearing bush, the oil film thickness and pressure will affect the dispersion characteristics and energy leakage of the Lamb waves propagating therein, thereby affecting its propagation speed and the energy attenuation characteristics of each frequency component. Therefore, the thickness and pressure of the oil film can be inferred by analyzing the propagation time and waveform characteristics of the Lamb waves propagating on the surface of the bush. The ultrasonic guided wave detection technology has much lower operating frequency requirements for Lamb waves than the ultrasonic body wave detection method, and the emission and reception of ultrasonic waves can be achieved through the piezoelectric chip embedded at the bottom of the bush contact surface, which is inexpensive and can detect the oil film pressure and thickness of the entire bush surface through the array. Summary of the invention

[0004] The purpose of the present invention is to provide a method for measuring the oil film thickness and pressure distribution on the bearing surface based on ultrasonic guided waves, which can perform real-time monitoring under the operation state of the bearing, obtain the pressure distribution and oil film thickness at any point on the bearing surface, and avoid affecting the propagation speed of Lamb waves.

[0005] The technical solution adopted by the present invention is a method for measuring the oil film thickness and pressure distribution on the bearing surface based on ultrasonic guided waves, which is specifically implemented according to the following steps:

[0006] Step 1, embedding and arranging a piezoelectric sheet array on the back of the bearing to be tested;

[0007] Step 2, collecting the guided wave signal under the set working condition of the oil film as the reference signal. After the reference signal is collected, the pressure and film thickness of the oil film are changed to collect the guided wave signals under different pressures and different film thicknesses;

[0008] Step 3, obtaining the noise signal in the waveguide signal under different pressures and different film thicknesses in step 2, and performing trend noise fitting on the noise signal;

[0009] Step 4: After the trend noise fitting is completed, the segmented RMS index is established to extract the coupled RMS sensitive features of pressure and film thickness;

[0010] Step 5: Import the extracted waveguide data containing the coupled RMS sensitive features into the convolutional neural network in deep learning to decouple the coupling features, and use the RAPID imaging algorithm based on the probabilistic damage imaging principle to image the bearing surface to complete the measurement.

[0011] The technical solution of the present invention is also characterized in that:

[0012] Step 1 is specifically as follows: a sensor array consisting of N pairs of piezoelectric sensors arranged symmetrically on the back of the bearing to be tested is embedded in the bearing back, and the excitation of the guided wave signal and data acquisition are controlled by the host computer.

[0013] Step 2 is specifically as follows: select an operating range of oil film pressure of 1-20MPa and oil film thickness of 10-200μm, use a pair of sensors with horizontal paths in the piezoelectric sensor array as the transmitting end and the receiving end, first, collect the guided wave signal within the operating range as the reference signal, and then change the oil film pressure and oil film thickness at intervals of 1MPa each time, and collect the guided wave signals under different pressures and different film thicknesses.

[0014] Step 3 is specifically as follows: a polynomial fitting function based on the least square method is used to fit the noise interference, the order of the polynomial fitting is 4 to 7, and the difference between the original signal and the electromagnetic interference is the actual signal.

[0015] Step 4 is as follows: the waveguide signal is divided into multiple windows of fixed length in the time domain, the entire signal is divided into k segments for processing, the root mean square (RMS) value is calculated in each time window, and the changes under different pressure and film thickness conditions are analyzed by comparing the RMS value of the entire signal with the local RMS value after segmentation. The RMS value in each time window is used as a feature to construct a feature matrix.

[0016] The formula for calculating the RMS value in each time window in step 4 is:

[0017]

[0018] Where k is the number of windows into which the waveguide signal is divided in the time domain.

[0019] Step 5 is specifically as follows: the feature matrix of the extracted coupling characteristic information about pressure and film thickness is used as the input feature of the neural network, and the convolutional neural network in deep learning is used for decoupling to obtain the relationship between the pressure distribution on the bearing surface and the change of the guided wave signal waveform, as well as the relationship between the oil film thickness on the bearing surface and the change of the guided wave signal waveform. Finally, the RAPID imaging algorithm based on the principle of probabilistic damage imaging is used to establish the oil film pressure distribution and oil film thickness distribution map on the bearing surface.

[0020] Step 5: The RAPID imaging algorithm based on the principle of probabilistic damage imaging is as follows:

[0021] S1, calculate the difference signal on the excitation-sensing path, and calculate the wavelet coefficient modulus curve of the difference signal to obtain the amplitude information time of the difference signal;

[0022] The calculation method of the wavelet coefficient modulus curve of the difference signal r(t) is as follows:

[0023] Φ(t)=abs(W r (t)) (2)

[0024]

[0025] Where Φ(t) is the modulus of the wavelet coefficient of the difference signal r(t), a is the scale parameter of the wavelet, and its value is related to the center frequency of the Lamb wave, and b is the translation parameter of the wavelet. is the mother wavelet function ψ a，b (t) Conjugation after extension and translation;

[0026] The expression is as follows:

[0027]

[0028] Where a is the scale parameter of the wavelet, and b is the translation parameter of the wavelet;

[0029] The mother wavelet function ψ(t) often selects the Morlet continuous complex wavelet as the wavelet function. Through the Morlet continuous complex wavelet transform, the envelope of the difference signal can be obtained to obtain the coupling characteristic information of the oil film pressure and oil film thickness at each time point;

[0030] S2, normalize the wavelet coefficient modulus curve of the difference signal r(t), the specific formula is as follows:

[0031]

[0032] Secondly, for any coordinate point (x, y) in the bearing to be tested, the surface pressure and film thickness are calculated from the exciter i (x i ,y i) to the receiving exciter j(x j ,y j ), the propagation time of Lamb wave in the path of excitation-any point on the bearing surface-receiving can be expressed as:

[0033]

[0034] Where t0 is the time offset of Lamb wave excitation, v g is the group velocity of Lamb wave;

[0035] Using the sensor array and probability superposition method, the pressure and oil film thickness change P(x, y) corresponding to any point (x, y) on the bearing surface can be calculated. The specific calculation method is as follows:

[0036]

[0037] Where N is the number of piezoelectric sensors, and the pressure coordinates are determined by superimposing the pressure and oil film thickness probability maps established by multiple excitation sensing paths;

[0038] Finally, the calculated surface field change P(x, y) is used as an input parameter and input into the convolutional neural network in deep learning. The coupling values ​​of the oil film pressure characteristics and the oil film thickness characteristics are decoupled, and the relationship between the pressure field on the bearing surface to be tested and the guided wave waveform, as well as the relationship between the film thickness distribution on the bearing surface and the guided wave waveform, are obtained. The pressure distribution and oil film thickness at any point on the bearing surface are obtained to complete the measurement.

[0039] Compared with the prior art, the present invention has the following beneficial effects:

[0040] The present invention provides a method for measuring the thickness and pressure distribution of the oil film on the bearing surface based on ultrasonic guided waves, which infers the thickness and pressure of the oil film by analyzing the propagation time and waveform characteristics of the ultrasonic wave. The measurement of the thickness of the oil film usually depends on the reflection time of the ultrasonic wave in the oil film. The ultrasonic wave is sent by the transmitter, and the reflected wave is received. The measured time difference can accurately calculate the thickness of the oil film, and the pressure measurement is achieved by analyzing the speed change of the ultrasonic wave when it propagates in the oil film. The pressure change in the oil film will affect the propagation speed of the ultrasonic wave, so the oil film pressure is inferred by establishing a corresponding model. This method can not only perform real-time monitoring under the operating state of the bearing, obtain the pressure distribution and oil film thickness at any point on the bearing surface, avoid affecting the Lamb wave propagation speed and the energy attenuation characteristics of each frequency component and interference with the bearing system, but the application of ultrasonic guided wave technology can significantly improve the accuracy and reliability of the measurement, which is helpful for the health monitoring and maintenance of sliding bearings. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1It is a schematic diagram of the process of measuring the oil film thickness and pressure distribution on the surface of the sliding bearing bush based on ultrasonic guided waves of the present invention.

[0042] Figure 2 It is a schematic diagram of installing a piezoelectric sensor in the gap between a sliding bearing bush and a bearing seat of the present invention;

[0043] Figure 3 This is an installation layout diagram of the piezoelectric sensor array of the present invention embedded in the back of the sliding bearing bush;

[0044] Figure 4 It is a schematic diagram comparing the actual signal and the reference signal received after the ultrasonic guided wave excitation bearing in Example 1 of the present invention.

[0045] In the figure, 1. bearing, 2. piezoelectric sensor, 3. bearing seat. DETAILED DESCRIPTION

[0046] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0047] The present invention provides a method for measuring the oil film thickness and pressure distribution on the bearing surface based on ultrasonic guided waves, such as Figure 1-3 As shown, the specific implementation steps are as follows:

[0048] Step 1, embedding and arranging a piezoelectric sheet array on the back of the bearing to be tested;

[0049] Step 1 is specifically as follows: a sensor array consisting of N pairs of piezoelectric sensors 2 arranged symmetrically on the left and right is embedded on the back of the bearing 1 to be tested, and the excitation and data acquisition of the waveguide signal are controlled by the host computer. The terminal board converts the signal into an electrical signal. The piezoelectric sensor 2 is used to excite and receive the waveguide signal, while the signal amplifier and charge amplifier are used to amplify the waveguide signal.

[0050] Step 2, collecting the guided wave signal under the set working condition of the oil film as the reference signal. After the reference signal is collected, the pressure and film thickness of the oil film are changed to collect the guided wave signals under different pressures and different film thicknesses;

[0051] Step 2 is specifically as follows: select an operating range of an oil film pressure of 1-20 MPa and an oil film thickness of 10-200 μm, use a pair of sensors with horizontal paths in the piezoelectric sensor 2 array as the transmitting end and the receiving end, firstly, collect the guided wave signal within the operating range as the reference signal, and secondly, change the oil film pressure and the oil film thickness at intervals of 1 MPa each time, and collect the guided wave signals under different pressures and different film thicknesses.

[0052] Step 3, obtaining the noise signal in the waveguide signal under different pressures and different film thicknesses in step 2, and performing trend noise fitting on the noise signal;

[0053] Step 3 is specifically as follows: a polynomial fitting function based on the least square method is used to fit the noise interference, the order of the polynomial fitting is 4 to 7, and the difference between the original signal and the electromagnetic interference is the actual signal.

[0054] Step 4: After the trend noise fitting is completed, the segmented RMS index is established to extract the coupled RMS sensitive features of pressure and film thickness;

[0055] Step 4 is as follows: the waveguide signal is divided into multiple windows of fixed length in the time domain, the entire signal is divided into k segments for processing, the RMS value is calculated in each time window, and the changes under different pressure and film thickness conditions are analyzed by comparing the RMS value of the entire signal with the local RMS value after segmentation. The RMS value in each time window is used as a feature to construct a feature matrix.

[0056] The formula for calculating the RMS value in each time window in step 4 is:

[0057]

[0058] Where k is the number of windows into which the waveguide signal is divided in the time domain.

[0059] Step 5: Import the extracted waveguide data containing the coupled RMS sensitive features into the convolutional neural network in deep learning to decouple the coupling features, and use the RAPID imaging algorithm based on the probabilistic damage imaging principle to image the bearing surface to complete the measurement.

[0060] Step 5 is specifically as follows: the feature matrix of the extracted coupling characteristic information about pressure and film thickness is used as the input feature of the neural network, and the convolutional neural network in deep learning is used for decoupling to obtain the relationship between the pressure distribution on the bearing surface and the change of the guided wave signal waveform, as well as the relationship between the oil film thickness on the bearing surface and the change of the guided wave signal waveform. Finally, the RAPID imaging algorithm based on the principle of probabilistic damage imaging is used to establish the oil film pressure distribution and oil film thickness distribution map on the bearing surface.

[0061] Step 5: The RAPID imaging algorithm based on the principle of probabilistic damage imaging is as follows:

[0062] S1, calculate the difference signal on the excitation-sensing path, and calculate the wavelet coefficient modulus curve of the difference signal to obtain the amplitude information time of the difference signal;

[0063] The calculation method of the wavelet coefficient modulus curve of the difference signal r(t) is as follows:

[0064] Φ(t)=abs(W r (t)) (2)

[0065]

[0066] Where Φ(t) is the modulus of the wavelet coefficient of the difference signal r(t), a is the scale parameter of the wavelet, and its value is related to the center frequency of the Lamb wave, and b is the translation parameter of the wavelet. is the mother wavelet function ψ a，b (t) Conjugation after extension and translation;

[0067] The expression is as follows:

[0068]

[0069] Where a is the scale parameter of the wavelet, and b is the translation parameter of the wavelet;

[0070] The mother wavelet function ψ(t) often selects the Morlet continuous complex wavelet as the wavelet function. Through the Morlet continuous complex wavelet transform, the envelope of the difference signal can be obtained to obtain the coupling characteristic information of the oil film pressure and oil film thickness at each time point;

[0071] S2, normalize the wavelet coefficient modulus curve of the difference signal r(t), the specific formula is as follows:

[0072]

[0073] Secondly, for any coordinate point (x, y) in the bearing to be tested, the surface pressure and film thickness are calculated from the exciter i (x i ,y i ) to the receiving exciter j(x j ,y j ), the propagation time of Lamb wave in the path of excitation-any point on the bearing surface-receiving can be expressed as:

[0074]

[0075] Where t0 is the time offset of Lamb wave excitation, v g is the group velocity of Lamb wave;

[0076] Using the sensor array and probability superposition method, the pressure and oil film thickness change P(x, y) corresponding to any point (x, y) on the bearing surface can be calculated. The specific calculation method is as follows:

[0077]

[0078] Where N is the number of piezoelectric sensors, and the pressure coordinates are determined by superimposing the pressure and oil film thickness probability maps established by multiple excitation sensing paths;

[0079] Finally, the calculated surface field change P(x, y) is used as an input parameter and input into the convolutional neural network in deep learning. The coupling values ​​of the oil film pressure characteristics and the oil film thickness characteristics are decoupled, and the relationship between the pressure field on the bearing surface to be tested and the guided wave waveform, as well as the relationship between the film thickness distribution on the bearing surface and the guided wave waveform, are obtained. The pressure distribution and oil film thickness at any point on the bearing surface are obtained to complete the measurement.

[0080] Example 1

[0081] This embodiment provides a method for measuring the oil film thickness and pressure distribution on the bearing surface based on ultrasonic guided waves. Figure 1-4 As shown, the specific implementation steps are as follows:

[0082] Step 1, embedding and arranging a piezoelectric sheet array on the back of the bearing to be tested;

[0083] Step 1 is specifically as follows: a sensor array consisting of 6 pairs of piezoelectric sensors 2 arranged symmetrically on the left and right is embedded in the back of the bearing 1 to be tested, and the excitation and data acquisition of the waveguide signal are controlled by the host computer. The terminal board converts the signal into an electrical signal. The piezoelectric sensor 2 is used to excite and receive the waveguide signal, while the signal amplifier and charge amplifier are used to amplify the waveguide signal.

[0084] Step 2, collecting the guided wave signal under the set working condition of the oil film as the reference signal. After the reference signal is collected, the pressure and film thickness of the oil film are changed to collect the guided wave signals under different pressures and different film thicknesses;

[0085] Step 2 is specifically as follows: select an operating range where the oil film pressure is 20 MPa and the oil film thickness is 100 μm, and use a pair of sensors with horizontal paths in the piezoelectric sensor 2 array as the transmitting end and the receiving end. First, collect the guided wave signal within the operating range as the reference signal. Secondly, change the oil film pressure and the oil film thickness at intervals of 1 MPa each time, and collect the guided wave signals under different pressures and different film thicknesses.

[0086] Step 3, obtaining the noise signal in the waveguide signal under different pressures and different film thicknesses in step 2, and performing trend noise fitting on the noise signal;

[0087] Step 3 is specifically as follows: a polynomial fitting function based on the least square method is used to fit the noise interference, the order of the polynomial fitting is 4 to 7, and the difference between the original signal and the electromagnetic interference is the actual signal.

[0088] Step 4: After the trend noise fitting is completed, the segmented RMS index is established to extract the coupled RMS sensitive features of pressure and film thickness;

[0089] Step 4 is as follows: the waveguide signal is divided into multiple windows of fixed length in the time domain, the entire signal is divided into k segments for processing, the RMS value is calculated in each time window, and the changes under different pressure and film thickness conditions are analyzed by comparing the RMS value of the entire signal with the local RMS value after segmentation. The RMS value in each time window is used as a feature to construct a feature matrix.

[0090] The formula for calculating the RMS value in each time window in step 4 is:

[0091]

[0092] Where k is the number of windows into which the waveguide signal is divided in the time domain.

[0093] Step 5: Import the extracted waveguide data containing the coupled RMS sensitive features into the convolutional neural network in deep learning to decouple the coupling features, and use the RAPID imaging algorithm based on the probabilistic damage imaging principle to image the bearing surface to complete the measurement.

[0094] Step 5 is specifically as follows: the feature matrix of the extracted coupling characteristic information about pressure and film thickness is used as the input feature of the neural network, and the convolutional neural network in deep learning is used for decoupling to obtain the relationship between the pressure distribution on the bearing surface and the change of the guided wave signal waveform, as well as the relationship between the oil film thickness on the bearing surface and the change of the guided wave signal waveform. Finally, the RAPID imaging algorithm based on the principle of probabilistic damage imaging is used to establish the oil film pressure distribution and oil film thickness distribution map on the bearing surface.

[0095] Step 5: The RAPID imaging algorithm based on the principle of probabilistic damage imaging is as follows:

[0096] S1, calculate the difference signal on the excitation-sensing path, and calculate the wavelet coefficient modulus curve of the difference signal to obtain the amplitude information time of the difference signal;

[0097] The calculation method of the wavelet coefficient modulus curve of the difference signal r(t) is as follows:

[0098] Φ(t)=abs(W r (t)) (2)

[0099]

[0100] Where Φ(t) is the modulus of the wavelet coefficient of the difference signal r(t), a is the scale parameter of the wavelet, and its value is related to the center frequency of the Lamb wave, and b is the translation parameter of the wavelet. is the mother wavelet function ψ a，b (t) Conjugation after extension and translation;

[0101] The expression is as follows:

[0102]

[0103] Where a is the scale parameter of the wavelet, and b is the translation parameter of the wavelet;

[0104] The mother wavelet function ψ(t) often selects the Morlet continuous complex wavelet as the wavelet function. Through the Morlet continuous complex wavelet transform, the envelope of the difference signal can be obtained to obtain the coupling characteristic information of the oil film pressure and oil film thickness at each time point;

[0105] S2, normalize the wavelet coefficient modulus curve of the difference signal r(t), the specific formula is as follows:

[0106]

[0107] Secondly, for any coordinate point (x, y) in the bearing to be tested, the surface pressure and film thickness are calculated from the exciter i (x i ,y i ) to the receiving exciter j(x j ,y j ), the propagation time of Lamb wave in the path of excitation-any point on the bearing surface-receiving can be expressed as:

[0108]

[0109] Where t0 is the time offset of Lamb wave excitation, v g is the group velocity of Lamb wave;

[0110] Using the sensor array and probability superposition method, the pressure and oil film thickness change P(x, y) corresponding to any point (x, y) on the bearing surface can be calculated. The specific calculation method is as follows:

[0111]

[0112] Where N is the number of piezoelectric sensors, and the pressure coordinates are determined by superimposing the pressure and oil film thickness probability maps established by multiple excitation sensing paths;

[0113] Finally, the calculated surface field change P(x, y) is used as an input parameter and input into the convolutional neural network in deep learning. The coupling values ​​of the oil film pressure characteristics and the oil film thickness characteristics are decoupled, and the relationship between the pressure field on the bearing surface to be tested and the guided wave waveform, as well as the relationship between the film thickness distribution on the bearing surface and the guided wave waveform, are obtained. The pressure distribution and oil film thickness at any point on the bearing surface are obtained to complete the measurement.

[0114] Example 2

[0115] This embodiment provides a method for measuring the oil film thickness and pressure distribution on the bearing surface based on ultrasonic guided waves. Figure 1-4 As shown, the specific implementation steps are as follows:

[0116] Step 1, embedding and arranging a piezoelectric sheet array on the back of the bearing to be tested;

[0117] Step 2, collecting the guided wave signal under the set working condition of the oil film as the reference signal. After the reference signal is collected, the pressure and film thickness of the oil film are changed to collect the guided wave signals under different pressures and different film thicknesses;

[0118] Step 3, obtaining the noise signal in the waveguide signal under different pressures and different film thicknesses in step 2, and performing trend noise fitting on the noise signal;

[0119] Step 4: After the trend noise fitting is completed, the segmented RMS index is established to extract the coupled RMS sensitive features of pressure and film thickness;

[0120] Step 5: Import the extracted waveguide data containing the coupled RMS sensitive features into the convolutional neural network in deep learning to decouple the coupling features, and use the RAPID imaging algorithm based on the probabilistic damage imaging principle to image the bearing surface to complete the measurement.

[0121] Example 3

[0122] On the basis of Example 2, Figure 1-4 As shown, step 1 is specifically as follows: a sensor array consisting of N pairs of piezoelectric sensors arranged symmetrically on the left and right is embedded in the back of the bearing to be tested, and the excitation of the guided wave signal and data acquisition are controlled by the host computer.

[0123] Step 2 is specifically as follows: select an operating range of oil film pressure of 1-20MPa and oil film thickness of 10-200μm, use a pair of sensors with horizontal paths in the piezoelectric sensor array as the transmitting end and the receiving end, first, collect the guided wave signal within the operating range as the reference signal, and then change the oil film pressure and oil film thickness at intervals of 1MPa each time, and collect the guided wave signals under different pressures and different film thicknesses.

[0124] Step 3 is specifically as follows: a polynomial fitting function based on the least square method is used to fit the noise interference, the order of the polynomial fitting is 4 to 7, and the difference between the original signal and the electromagnetic interference is the actual signal.

[0125] Step 4 is as follows: the waveguide signal is divided into multiple windows of fixed length in the time domain, the entire signal is divided into k segments for processing, the RMS value is calculated in each time window, and the changes under different pressure and film thickness conditions are analyzed by comparing the RMS value of the entire signal with the local RMS value after segmentation. The RMS value in each time window is used as a feature to construct a feature matrix.

[0126] The formula for calculating the RMS value in each time window in step 4 is:

[0127]

[0128] Where k is the number of windows into which the waveguide signal is divided in the time domain.

[0129] Step 5 is specifically as follows: the feature matrix of the extracted coupling characteristic information about pressure and film thickness is used as the input feature of the neural network, and the convolutional neural network in deep learning is used for decoupling to obtain the relationship between the pressure distribution on the bearing surface and the change of the guided wave signal waveform, as well as the relationship between the oil film thickness on the bearing surface and the change of the guided wave signal waveform. Finally, the RAPID imaging algorithm based on the principle of probabilistic damage imaging is used to establish the oil film pressure distribution and oil film thickness distribution map on the bearing surface.

[0130] Step 5: The RAPID imaging algorithm based on the principle of probabilistic damage imaging is as follows:

[0131] S1, calculate the difference signal on the excitation-sensing path, and calculate the wavelet coefficient modulus curve of the difference signal to obtain the amplitude information time of the difference signal;

[0132] The calculation method of the wavelet coefficient modulus curve of the difference signal r(t) is as follows:

[0133] Φ(t)=abs(W r (t)) (2)

[0134]

[0135] Where Φ(t) is the modulus of the wavelet coefficient of the difference signal r(t), a is the scale parameter of the wavelet, and its value is related to the center frequency of the Lamb wave, and b is the translation parameter of the wavelet. is the mother wavelet function ψ a，b (t) Conjugation after extension and translation;

[0136] The expression is as follows:

[0137]

[0138] Where a is the scale parameter of the wavelet, and b is the translation parameter of the wavelet;

[0139] The mother wavelet function ψ(t) often selects the Morlet continuous complex wavelet as the wavelet function. Through the Morlet continuous complex wavelet transform, the envelope of the difference signal can be obtained to obtain the coupling characteristic information of the oil film pressure and oil film thickness at each time point;

[0140] S2, normalize the wavelet coefficient modulus curve of the difference signal r(t), the specific formula is as follows:

[0141]

[0142] Secondly, for any coordinate point (x, y) in the bearing to be tested, the surface pressure and film thickness are calculated from the exciter i (x i ,y i ) to the receiving exciter j(x j ,y j ), the propagation time of Lamb wave in the path of excitation-any point on the bearing surface-receiving can be expressed as:

[0143]

[0144] Where t0 is the time offset of Lamb wave excitation, v g is the group velocity of Lamb wave;

[0145] Using the sensor array and probability superposition method, the pressure and oil film thickness change P(x, y) corresponding to any point (x, y) on the bearing surface can be calculated. The specific calculation method is as follows:

[0146]

[0147] Where N is the number of piezoelectric sensors, and the pressure coordinates are determined by superimposing the pressure and oil film thickness probability maps established by multiple excitation sensing paths;

[0148] Finally, the calculated surface field change P(x, y) is used as an input parameter and input into the convolutional neural network in deep learning. The coupling values ​​of the oil film pressure characteristics and the oil film thickness characteristics are decoupled, and the relationship between the pressure field on the bearing surface to be tested and the guided wave waveform, as well as the relationship between the film thickness distribution on the bearing surface and the guided wave waveform, are obtained. The pressure distribution and oil film thickness at any point on the bearing surface are obtained to complete the measurement.

Claims

1. A method for measuring the oil film thickness and pressure distribution on the bearing surface based on ultrasonic guided waves, characterized in that: Follow the steps below to implement it: Step 1, embedding and arranging a piezoelectric sheet array on the back of the bearing to be tested; Step 2, collecting the guided wave signal under the set working condition of the oil film as the reference signal. After the reference signal is collected, the pressure and film thickness of the oil film are changed to collect the guided wave signals under different pressures and different film thicknesses; Step 3, obtaining the noise signal in the waveguide signal under different pressures and different film thicknesses in step 2, and performing trend noise fitting on the noise signal; Step 4: After the trend noise fitting is completed, the segmented RMS index is established to extract the coupled RMS sensitive features of pressure and film thickness; Step 5: Import the extracted waveguide data containing the coupled RMS sensitive features into the convolutional neural network in deep learning to decouple the coupling features, and use the RAPID imaging algorithm based on the probabilistic damage imaging principle to image the bearing surface to complete the measurement.

2. The method for measuring the oil film thickness and pressure distribution on the bearing surface based on ultrasonic guided waves according to claim 1 is characterized in that: The step 1 specifically comprises: embedding a sensor array consisting of N pairs of piezoelectric sensors (2) arranged symmetrically on the back of the bearing (1) to be tested, and controlling the excitation of the guided wave signal and data collection through a host computer.

3. The method for measuring the oil film thickness and pressure distribution on the bearing surface based on ultrasonic guided waves according to claim 1 is characterized in that: The step 2 specifically comprises: selecting an operating range in which the oil film pressure is 1-20 MPa and the oil film thickness is 10-200 μm, using a pair of sensors with horizontal paths in the piezoelectric sensor (2) array as the transmitting end and the receiving end, firstly collecting the guided wave signal within the operating range as the reference signal, and secondly changing the oil film pressure and the oil film thickness at intervals of 1 MPa each time, collecting the guided wave signals under different pressures and different film thicknesses.

4. The method for measuring the oil film thickness and pressure distribution on the bearing surface based on ultrasonic guided waves according to claim 1 is characterized in that: The step 3 is specifically as follows: a polynomial fitting function based on the least square method is used to fit the noise interference, the order of the polynomial fitting is 4 to 7, and the difference between the original signal and the electromagnetic interference is the actual signal.

5. The method for measuring the oil film thickness and pressure distribution on the bearing surface based on ultrasonic guided waves according to claim 1 is characterized in that: The step 4 is specifically as follows: the waveguide signal is divided into multiple windows of fixed length in the time domain, the entire signal is divided into k segments for processing, the RMS value is calculated in each time window, the changes under different pressure and film thickness conditions are analyzed by comparing the RMS value of the entire signal with the local RMS value after segmentation, and the RMS value in each time window is used as a feature to construct a feature matrix.

6. The method for measuring the oil film thickness and pressure distribution on the bearing surface based on ultrasonic guided waves according to claim 5 is characterized in that: The calculation formula for the RMS value in each time window described in step 4 is: Where k is the number of windows into which the waveguide signal is divided in the time domain.

7. The method for measuring the oil film thickness and pressure distribution on the bearing surface based on ultrasonic guided waves according to claim 1 is characterized in that: The step 5 is specifically as follows: the feature matrix of the extracted coupling feature information about pressure and film thickness is used as the input feature of the neural network, and the convolutional neural network in deep learning is used for decoupling to obtain the relationship between the pressure distribution on the bearing surface and the change of the guided wave signal waveform, as well as the relationship between the oil film thickness on the bearing surface and the change of the guided wave signal waveform, and finally, the RAPID imaging algorithm based on the probabilistic damage imaging principle is used to establish the oil film pressure distribution and oil film thickness distribution map on the bearing surface.

8. The method for measuring the oil film thickness and pressure distribution on the bearing surface based on ultrasonic guided waves according to claim 7 is characterized in that: The RAPID imaging algorithm based on the probabilistic damage imaging principle described in step 5 is specifically: S1, calculate the difference signal r(t) on the excitation-sensing path, and calculate the wavelet coefficient modulus curve of the difference signal r(t) to obtain the amplitude information time of the difference signal; The wavelet coefficient modulus curve calculation method of the difference signal r(t) is as follows: Φ(t)=abs(W r (t)) (2) Where Φ(t) is the modulus of the wavelet coefficient of the difference signal r(t), a is the scale parameter of the wavelet, and its value is related to the center frequency of the Lamb wave, and b is the translation parameter of the wavelet. is the mother wavelet function ψ a，b (t) Conjugation after extension and translation; Said The expression is as follows: Where a is the scale parameter of the wavelet, and b is the translation parameter of the wavelet; The mother wavelet function ψ(t) often selects the Morlet continuous complex wavelet as the wavelet function. Through the Morlet continuous complex wavelet transform, the envelope of the difference signal can be obtained to obtain the coupling characteristic information of the oil film pressure and oil film thickness at each time point; S2, normalizing the wavelet coefficient modulus curve of the difference signal r(t), the specific formula is as follows: Secondly, for any coordinate point (x, y) in the bearing to be tested, the surface pressure and film thickness are calculated from the exciter i (x i ,y i ) to the receiving exciter j(x j ,y j ), the propagation time of Lamb wave in the path of excitation-any point on the bearing surface-receiving can be expressed as: Where t0 is the time offset of Lamb wave excitation, v g is the group velocity of Lamb wave; Using the sensor array and probability superposition method, the pressure and oil film thickness change P(x, y) corresponding to any point (x, y) on the bearing surface can be calculated. The specific calculation method is as follows: Where N is the number of piezoelectric sensors, and the pressure coordinates are determined by superimposing the pressure and oil film thickness probability maps established by multiple excitation sensing paths; Finally, the calculated surface field change P(x, y) is used as an input parameter and input into the convolutional neural network in deep learning. The coupling values ​​of the oil film pressure characteristics and the oil film thickness characteristics are decoupled, and the relationship between the pressure field on the bearing surface to be tested and the guided wave waveform, as well as the relationship between the film thickness distribution on the bearing surface and the guided wave waveform, are obtained. The pressure distribution and oil film thickness at any point on the bearing surface are obtained to complete the measurement.

Images