Hydraulic Motor Fault Prelocation Method Based on Sound Intensity Image and Feature Signal Extraction

Abstract

The present invention proposes a method for pre-locating hydraulic motor faults based on sound intensity images and feature signal extraction, including the following steps: Step S1: Obtain the vibration characteristic frequencies of each component inside the hydraulic motor; Step S2: Obtain the noise time-domain signal; Step S3: Obtain the time-domain signal with a large contribution amount through an algorithm and remove the interference signal, thereby obtaining an unstable time-domain signal composed of the noise signals of each component inside the motor; Step S4: Initially judge the main components generating vibration noise inside the motor; Step S5: Obtain the sound intensity visualization image of the hydraulic motor; Step S6: Reconstruct the sound intensity visualization image with high resolution to obtain the noise distribution on the outer surface of the motor and judge the main parts generating vibration noise; Step S7: Further obtain the main components precisely generating vibration noise to achieve pre-location of hydraulic motor faults; The present invention can solve the problem that traditional fault feature analysis methods cannot accurately judge fault components and achieve pre-location of motor faults.

Description

Technical Field

[0001] The present invention relates to the technical field of noise source localization and fault pre - localization of hydraulic motors, in particular to a fault pre - localization method for hydraulic motors based on sound intensity image and feature signal extraction. Background Art

[0002] As an engineering rotating mechanical device, a hydraulic motor acts as a key actuator in a hydraulic system and is widely used in various industrial fields. It is also one of the main noise sources in the hydraulic system. The noise of a hydraulic motor is divided into mechanical noise and fluid noise. Among them, mechanical noise mainly comes from the collision between the plunger and the plunger hole, bearing faults and improper installation, vibration caused by eccentricity and imbalance during the rotation of the rotating body, etc.; fluid noise is mainly caused by flow pulsation, flow distribution impact, etc. When a hydraulic motor fails, it usually shows non - stationary and non - linear vibration signals, and is mixed with strong noise. Extracting its fault characteristic signals is easily interfered and difficult to separate from the original information. Traditional fault characteristic analysis methods include time - domain analysis, frequency - domain analysis, and time - frequency domain analysis methods. Currently, fault diagnosis technologies are widely applied to single and simple parts such as gears and bearings. For a plunger motor with a complex internal structure and closely related components, the application of fault diagnosis is in its infancy. Due to the complex internal structure of the plunger motor and the existence of noise sources with similar characteristic frequencies, if traditional fault characteristic analysis methods are directly used, it will be impossible to accurately judge the faulty components.

[0003] To solve the above problems, the following methods can be adopted: design a signal extraction technology that can process non - linear and non - stationary vibration signals, and can effectively remove interference noise signals and separate the vibration characteristic frequencies of each component for a hydraulic motor composed of multiple closely related and coupled components; utilize the characteristic that compressive sensing technology can obtain high - resolution signals under under - sampling conditions to perform high - resolution reconstruction on the sound intensity image of the hydraulic motor, obtain the accurate external surface noise distribution, and compare and fit the two results to further obtain the main components generating noise. Summary of the Invention

[0004] The present invention proposes a fault pre - localization method for hydraulic motors based on sound intensity image and feature signal extraction. By combining the high - resolution reconstruction technology of the sound intensity image of the hydraulic motor with the feature signal extraction technology, it can solve the problem that traditional fault characteristic analysis methods cannot accurately judge faulty components and realize the pre - localization of motor faults.

[0005] The present invention adopts the following technical solutions.

[0006] A fault pre - localization method for hydraulic motors based on sound intensity image and feature signal extraction includes the following steps;

[0007] Step S1: Obtain the vibration characteristic frequencies of each component inside the hydraulic motor;

[0008] Step S2: Based on the sound pressure test experiment, obtain the noise time-domain signal when the hydraulic motor is working;

[0009] Step S3: Based on the vibration and noise time-domain signal of the hydraulic motor obtained in Step S2, use a signal denoising algorithm to obtain the time-domain signal with a large contribution amount and remove the interference signal, thereby obtaining an unstable time-domain signal composed of the noise signals of each component inside the motor;

[0010] Step S4: Based on the unstable time-domain signal obtained in Step S3, use a feature signal extraction algorithm to decompose it into multiple signal components, extract the time-domain signal with a high correlation degree with the original signal and convert it into a frequency-domain signal to obtain the main frequencies of the vibration and noise generated inside the hydraulic motor. By comparing and fitting with the vibration characteristic frequencies obtained in Step S1, obtain the components corresponding to the frequencies, and then preliminarily determine the main components generating vibration and noise inside the motor;

[0011] Step S5: Based on the sound intensity test experiment, obtain the sound intensity visualization image of the hydraulic motor;

[0012] Step S6: Based on the sound intensity visualization image obtained in Step S5, use compressive sensing technology to perform high-resolution reconstruction on it to obtain the noise distribution on the outer surface of the motor with accurate noise intensity points, and then judge the main parts where the hydraulic motor generates vibration and noise from the outside;

[0013] Step S7: Combine the noise distribution on the outside of the motor obtained in Step S6 with the components preliminarily judged to generate vibration and noise inside the motor in Step S4, screen out the components with the same vibration and noise frequencies but not at the noise intensity points, and further obtain the accurate main components generating vibration and noise, realizing the fault pre-positioning of the hydraulic motor.

[0014] The specific method of Step S1 is: Based on the structure of the hydraulic motor and the vibration mechanism during operation, establish a motion mechanism model of the motor, clarify the motion modes and their laws of each component inside the motor, and then obtain the vibration characteristic frequencies of each component inside it;

[0015] The specific method of Step S2 is: Based on the sound pressure test standard, arrange sound pressure probes around the hydraulic motor by the hemisphere method in an anechoic chamber to obtain the noise time-domain signal of the hydraulic motor.

[0016] Step S2 includes the following steps;

[0017] Step S21: The sound pressure test experiment is carried out in an anechoic chamber, and the experimental instruments are B&K acoustic equipment and software;

[0018] Step S22: Based on the sound pressure test standard, sound pressure probes are respectively arranged at positions 1 m in front of, above, and on the side of the hydraulic motor through the hemispherical method to obtain the noise time-domain signal of the motor.

[0019] The said step S3 includes the following steps:

[0020] Step S31: Using a noise reduction algorithm, decompose the noise time-domain signal of the hydraulic motor obtained in step S2 into several frequency range nodes;

[0021] Step S32: Select the frequency range nodes with large contribution amounts among them, and filter out the redundant frequency ranges to filter out interference signals and achieve the noise reduction effect;

[0022] Step S33: Combine and reconstruct the time-domain signals of the selected frequency range nodes to obtain an unstable time-domain signal composed of the noise signals of each component inside the hydraulic motor;

[0023] The specific method of step S3 is: Use wavelet packet analysis to denoise the noise time-domain signal obtained in S2 and filter out interference signals: According to the noise time-domain signal obtained in S2, obtain its sampling frequency f s ; Set the wavelet packet decomposition layer number n according to the sampling frequency;

[0024] Obtain the frequency range of each node:

[0025] Calculate the energy distribution of each node frequency range, and select the frequency segments with larger energy contributions among them;

[0026] Superpose and reconstruct the obtained frequency segments to filter out interference signals, achieve the denoising effect, and further obtain an unstable time-domain signal composed of the noise signals of each component inside the motor.

[0027] The said step S4 includes the following steps:

[0028] Step S41: Separate the unstable time-domain signal obtained in step S3 through a feature signal extraction algorithm to obtain several signal components;

[0029] Step S42: Extract the signal components with large correlation with the original signal among them, and convert them into frequency-domain signals to obtain the main frequencies generating vibration noise in the hydraulic motor;

[0030] Step S43: Based on the characteristic frequencies of each component inside the motor obtained in step S1, compare and fit the obtained frequencies to obtain the components corresponding to the frequencies, and then preliminarily judge the main components generating vibration noise.

[0031] The specific method of step S4 is:

[0032] Step S41: First, use the Ensemble Empirical Mode Decomposition (EEMD) algorithm to separate and extract the denoised time-domain signal to obtain n Intrinsic Mode Functions (IMFs) and a residual component; then select the IMF components with a relatively high correlation with the original signal; perform Hilbert-Huang transform on the obtained IMF components to obtain a marginal spectrogram and acquire the main frequencies generating vibration noise in the hydraulic motor.

[0033] Step S42: Based on the vibration characteristic frequencies of each component inside the motor obtained in Step S1, fit the decomposed frequency signals to obtain the corresponding components of the frequencies, and then preliminarily determine the main components generating vibration noise.

[0034] The specific method of Step S5 is as follows: Based on the size of the hydraulic motor, arrange a measuring point grid around it, and collect the measuring point signals point by point through a sound intensity probe to obtain the sound intensity visualization images in different view directions, including the following steps;

[0035] Step S51: Based on the size of the hydraulic motor, arrange an m×n grid in its front view direction and top view direction respectively, and set the distance between adjacent measuring points according to the probe size;

[0036] Step S52: Set the corresponding grid in the test software, and collect the sound intensity signals point by point in the two grids respectively through the sound intensity probe, so as to generate the sound intensity visualization cloud maps in the front view and top view directions and obtain the approximate distribution of the motor noise.

[0037] Step S6 includes the following steps:

[0038] Step S61: Design a compressive sensing algorithm framework, which includes a sparse matrix, an observation matrix, and a reconstruction algorithm;

[0039] Step S62: Based on the designed compressive sensing algorithm, perform high-resolution reconstruction on the motor sound intensity visualization image obtained in Step S5 to obtain the noise distribution on the outer surface of the hydraulic motor with accurate noise intensity points.

[0040] In Step S61, the specific method is as follows: Perform sparse representation on the sound intensity visualization image obtained in Step S5: Utilize the orthogonality of the Fourier transform to construct an FFT sparse basis matrix with good decorrelation. The specific steps are as follows:

[0041] Step A1: According to the signal dimension N of the obtained sound intensity visualization image, determine that the dimension of the sparse basis is N×N; construct an N×N identity matrix;

[0042] Step A2: Perform FFT transform on the identity matrix;

[0043] Step A3: Finally, invert the matrix to complete the construction of the FFT sparse basis matrix;

[0044] The method of step S62 is specifically as follows: According to the acoustic field characteristics of the motor, using the orthogonality of the Hadamard matrix, construct a pseudo-random partial Hadamard matrix as the observation matrix. The specific steps are as follows:

[0045] Step B1: According to the signal dimension N of the obtained sound intensity visualization image, construct a Hadamard matrix of dimension N×N. This matrix is distributed with -1 and 1 as follows:

[0046]

[0047] Step B2: Set the number of observations M, and randomly select M rows from the constructed Hadamard matrix to obtain a partial Hadamard matrix of M×N;

[0048] Step B3: According to the characteristic that one measurement point in the sound intensity image corresponds to one signal, assign the -1 part to 0;

[0049] Step B4: Randomly shuffle its rows and columns to complete the construction of the pseudo-random partial Hadamard matrix; the constructed observation matrix is as follows:

[0050]

[0051] Step S6 further includes step S63, that is: adopt the orthogonal matching pursuit algorithm (OMP) as the reconstruction algorithm, combine the sparse basis matrix and the observation matrix, and perform high-resolution reconstruction on the sound intensity visualization cloud map to obtain the accurate surface noise distribution of the hydraulic motor.

[0052] The specific method of step S7 is as follows: Aiming at the characteristics of the compact internal structure of the hydraulic motor, the close coupling of each component, and the problem that the vibration frequencies of its internal components have the same frequency or multiple frequency relationship, based on the high-resolution sound intensity strong point distribution image of the hydraulic motor obtained in step S6, screen out the components with the same vibration frequency but not in the noise strong point position among the components generating vibration noise obtained through step S4, so as to reduce the influence of the components with the same frequency or multiple frequency vibration frequencies, and further obtain the accurate main components generating vibration noise, realizing the fault pre-positioning of the hydraulic motor.

[0053] The present invention provides a method for pre - locating hydraulic motor faults based on high - resolution sound intensity images and feature signal extraction. First, by analyzing the vibration mechanism of the motor, the vibration characteristic frequencies of each component are obtained. Then, through the feature signal extraction method, the sound pressure signal is separated and extracted to preliminarily obtain the components generating vibration noise. Finally, based on the high - resolution sound intensity image obtained by the compressive sensing method, the components with the same frequency but not at the noise intensity peak positions are screened out, and the main components generating vibration noise are accurately located, thereby realizing the pre - location of hydraulic motor faults. Compared with the prior art, the present invention has the following beneficial effects:

[0054] The present invention can effectively distinguish components with similar frequencies, accurately locate the main components generating noise in the hydraulic motor, and effectively improve the localization of the excitation source for integrated components with complex internal structures and closely related and coupled components. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] The following further details the present invention in conjunction with the drawings and specific embodiments:

[0056] Att Figure 1 is a schematic flow chart of the method of the embodiment of the present invention;

[0057] Att Figure 2 is a schematic diagram of the motor structure and motion mechanism model of the embodiment of the present invention;

[0058] Att Figure 3 is a schematic diagram of the sound pressure test experiment of the hydraulic motor of the embodiment of the present invention;

[0059] Att Figure 4 is a schematic diagram of the time - domain signal of noise reduction using wavelet packet analysis of the embodiment of the present invention;

[0060] Att Figure 5 is a schematic diagram of the frequency spectrum extracted by EEMD - Hilbert of the embodiment of the present invention;

[0061] Att Figure 6 is a schematic diagram of the sound intensity test grid distribution of the hydraulic motor of the embodiment of the present invention;

[0062] Att Figure 7 is a schematic diagram of the high - resolution reconstructed image of the sound intensity of the hydraulic motor of the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0063] As shown in the figure, the method for pre - locating hydraulic motor faults based on sound intensity images and feature signal extraction includes the following steps;

[0064] Step S1: Obtain the vibration characteristic frequencies of each component inside the hydraulic motor;

[0065] Step S2: Based on the sound pressure test experiment, obtain the noise time - domain signal when the hydraulic motor is working;

[0066] Step S3: Based on the time-domain signal of the hydraulic motor vibration noise obtained in Step S2, use a signal denoising algorithm to obtain the time-domain signal with a large contribution amount and remove the interference signal, thereby obtaining an unstable time-domain signal composed of the noise signals of each component inside the motor;

[0067] Step S4: Based on the unstable time-domain signal obtained in Step S3, use a feature signal extraction algorithm to decompose it into multiple signal components, extract the time-domain signal with a high correlation with the original signal and convert it into a frequency-domain signal to obtain the main frequencies of the vibration noise generated inside the hydraulic motor. By comparing and fitting with the vibration characteristic frequencies obtained in Step S1, obtain the components corresponding to the frequencies, and then preliminarily judge the main components generating vibration noise inside the motor;

[0068] Step S5: Based on the sound intensity test experiment, obtain the sound intensity visualization image of the hydraulic motor;

[0069] Step S6: Based on the sound intensity visualization image obtained in Step S5, use compressive sensing technology to reconstruct it with high resolution to obtain the noise distribution on the outer surface of the motor with precise noise intensity points, and then judge the main parts generating vibration noise of the hydraulic motor from the outside;

[0070] Step S7: Combine the noise distribution on the outside of the motor obtained in Step S6 with the components generating vibration noise inside the motor preliminarily judged in Step S4, screen out the components with the same vibration noise frequency but not at the noise intensity points, and further obtain the main components generating vibration noise accurately, realizing the fault pre-positioning of the hydraulic motor.

[0071] The specific method of Step S1 is: Based on the structure of the hydraulic motor and the vibration mechanism during operation, establish a motion mechanism model of the motor, clarify the motion modes and their laws of each component inside the motor, and then obtain the vibration characteristic frequencies of each component inside it;

[0072] The specific method of Step S2 is: Based on the sound pressure test standard, arrange sound pressure probes around the hydraulic motor by the hemispherical method in an anechoic chamber to obtain the time-domain signal of the hydraulic motor noise.

[0073] Step S2 includes the following steps;

[0074] Step S21: The sound pressure test experiment is carried out in an anechoic chamber, and the experimental instruments are B&K acoustic equipment and software;

[0075] Step S22: Based on the sound pressure test standard, arrange sound pressure probes at positions 1 m in front of, above, and to the side of the hydraulic motor by the hemispherical method to obtain the time-domain signal of the noise of this motor.

[0076] Step S3 includes the following steps:

[0077] Step S31: Using a noise reduction algorithm, decompose the time-domain signal of the hydraulic motor noise obtained in step S2 into several frequency range nodes;

[0078] Step S32: Select the frequency range nodes with large contribution amounts, screen out the redundant frequency ranges, so as to filter out the interference signals and achieve the noise reduction effect;

[0079] Step S33: Combine and reconstruct the time-domain signals of the selected frequency range nodes, so as to obtain an unstable time-domain signal composed of the noise signals of each component inside the hydraulic motor;

[0080] The specific method of step S3 is: Use wavelet packet analysis to denoise the time-domain noise signal obtained in S2 and filter out the interference signals: According to the time-domain noise signal obtained in S2, obtain its sampling frequency f s ; Set the wavelet packet decomposition layer number n according to the sampling frequency;

[0081] Obtain the frequency ranges of each node:

[0082] Calculate the energy distribution of each node frequency range, and select the frequency segments with larger energy contributions;

[0083] Overlay and reconstruct the obtained frequency segments to filter out the interference signals, achieve the denoising effect, and further obtain an unstable time-domain signal composed of the noise signals of each component inside the motor.

[0084] Step S4 includes the following steps:

[0085] Step S41: Separate the unstable time-domain signal obtained in step S3 through a feature signal extraction algorithm to obtain several signal components;

[0086] Step S42: Extract the signal components with high correlation with the original signal and convert them into frequency-domain signals to obtain the main frequencies generating vibration noise in the hydraulic motor;

[0087] Step S43: Based on the characteristic frequencies of each component inside the motor obtained in step S1, compare and fit the obtained frequencies to obtain the corresponding components, and then preliminarily judge the main components generating vibration noise.

[0088] The specific method of step S4 is:

[0089] Step S41: First, use the Ensemble Empirical Mode Decomposition (EEMD) algorithm to separate and extract the denoised time-domain signal to obtain n Intrinsic Mode Functions (IMFs) and a residual component; then select the IMF components with a relatively high correlation with the original signal; perform Hilbert-Huang transform on the obtained IMF components to obtain a marginal spectrogram and acquire the main frequencies generating vibration noise in the hydraulic motor.

[0090] Step S42: Based on the vibration characteristic frequencies of each component inside the motor obtained in Step S1, fit the decomposed frequency signals to obtain the corresponding components of the frequencies, and then preliminarily determine the main components generating vibration noise.

[0091] The specific method of Step S5 is as follows: Based on the size of the hydraulic motor, arrange a measuring point grid around it, and collect the measuring point signals point by point through a sound intensity probe to obtain the sound intensity visualization images in different view directions, including the following steps;

[0092] Step S51: Based on the size of the hydraulic motor, arrange grids of m×n in its front view direction and top view direction respectively, and set the distance between adjacent measuring points according to the probe size;

[0093] Step S52: Set the corresponding grids in the test software, and collect the sound intensity signals point by point in the two grids through the sound intensity probe respectively, so as to generate the sound intensity visualization cloud maps in the front view and top view directions and obtain the approximate distribution of the motor noise.

[0094] Step S6 includes the following steps:

[0095] Step S61: Design a compressive sensing algorithm framework, which includes a sparse matrix, an observation matrix, and a reconstruction algorithm;

[0096] Step S62: Based on the designed compressive sensing algorithm, perform high-resolution reconstruction on the motor sound intensity visualization image obtained in Step S5 to obtain the noise distribution on the outer surface of the hydraulic motor with precise noise intensity points.

[0097] In Step S61, the specific method is as follows: Perform sparse representation on the sound intensity visualization image obtained in Step S5: Utilize the orthogonality of the Fourier transform to construct an FFT sparse basis matrix with good incoherence, and its specific steps are as follows:

[0098] Step A1: According to the signal dimension N of the obtained sound intensity visualization image, determine that the dimension of the sparse basis is N×N; construct an N×N identity matrix;

[0099] Step A2: Perform FFT transform on the identity matrix;

[0100] Step A3. Finally, invert the matrix to complete the construction of the FFT sparse basis matrix;

[0101] The method of step S62 is as follows: According to the acoustic field characteristics of the motor, using the orthogonality of the Hadamard matrix, construct a pseudo-random partial Hadamard matrix as the observation matrix. The specific steps are as follows:

[0102] Step B1. According to the signal dimension N of the obtained sound intensity visualization image, construct a Hadamard matrix of dimension N×N. This matrix is distributed with -1 and 1, as shown below:

[0103]

[0104] Step B2. Set the number of observations M, and randomly select M rows from the constructed Hadamard matrix to obtain a partial Hadamard matrix of M×N;

[0105] Step B3. According to the characteristic that one measurement point in the sound intensity image corresponds to one signal, assign the -1 part to 0;

[0106] Step B4. Randomly shuffle its rows and columns to complete the construction of the pseudo-random partial Hadamard matrix; the constructed observation matrix is as shown below:

[0107]

[0108] Step S6 further includes step S63, that is: adopt the orthogonal matching pursuit algorithm (OMP) as the reconstruction algorithm, combine the sparse basis matrix and the observation matrix, and perform high-resolution reconstruction on the sound intensity visualization cloud map to obtain the accurate surface noise distribution of the hydraulic motor.

[0109] The specific method of step S7 is as follows: Aiming at the characteristics of the compact internal structure of the hydraulic motor, the close coupling of each component, and the problem that the vibration frequencies of its internal components have the relationship of the same frequency or multiple frequencies, based on the high-resolution sound intensity strong point distribution image of the hydraulic motor obtained in step S6, screen out the components with the same vibration frequency but not in the noise strong point position among the components generating vibration noise obtained through step S4, so as to reduce the influence of the components with the same frequency or multiple frequency vibration frequencies, and further obtain the accurate main components generating vibration noise, realizing the fault pre-positioning of the hydraulic motor.

[0110] Example:

[0111] As Figure 1 shown, this example is specifically divided into the following six steps:

[0112] Step S1: As Figure 2 shown is the bent-axis piston motor and its motion mechanism model studied in this example. According to its working principle, perform vibration mechanism analysis to obtain the vibration characteristic frequencies of each component inside the motor. The specific content is as follows:

[0113] In this embodiment, the working process of the axial piston motor is as follows: The high-pressure oil output by the hydraulic pump enters the cylinder block through the motor oil inlet and the distribution plate, pushing the piston outwards, causing the ball joint at the end of the piston to be articulated and extruded with the ball socket of the main shaft, making the main shaft rotate. At the same time, the piston collides and squeezes with the inner wall of the cylinder block, thereby driving the cylinder block to rotate. The hydraulic oil returns to the fuel tank through the distribution plate and the motor oil outlet. Therefore, the noise during the operation of the hydraulic motor mainly consists of mechanical noise caused by the relative movement and collision of various components inside the motor, and fluid noise caused by flow pulsation and distribution impact. The vibration generation methods and characteristic frequencies of various components inside it can be obtained as follows:

[0114] 1. Vibration caused by the eccentricity and imbalance of the main shaft. When the motor is operating, the main shaft collides with the connected components due to eccentricity and imbalance, thereby causing vibration. Its frequency is:

[0115] 2. Vibration caused by the bearing. Vibration occurs between the inner and outer rings of the bearing and the rollers. Its frequency is:

[0116] 3. Vibration caused by flow pulsation. Flow pulsation in the pipeline during the oil suction and discharge of the hydraulic motor causes vibration. Its frequency is:

[0117] 4. Vibration caused by piston collision and ball joint clearance. When the cylinder block rotates, it collides with the piston, causing vibration. There is a clearance between the ball joint and the ball socket of the main shaft, resulting in collision and vibration. Their frequencies are both:

[0118] Among them, n is the rotational speed of the main shaft, N is the number of bearing rollers, and Z is the number of pistons.

[0119] In this embodiment, the number of bearings N of the swashplate-type piston motor under study is 12, the number of pistons Z is 7, the set rotational speed n is 1000 rpm, and the first-order frequencies of vibration and noise at each part are as follows:

[0120]

[0121] Step S2: Conduct a sound pressure test experiment on the target hydraulic motor to obtain the time-domain signal of the motor vibration noise. The specific content is as follows:

[0122] As Figure 3 shown, based on the sound pressure measurement standard, sound pressure probes are arranged at positions 1 m away in the front, above, and side of the hydraulic motor respectively by the hemisphere method, so as to collect the time-domain signal of the noise during the operation of the motor. This experiment is carried out in an anechoic chamber, and the experimental instrument is B&K acoustic equipment.

[0123] Step S3: Denoise the noise time-domain signal obtained in Step S2 to remove the interference noise signal, obtaining an unstable time-domain signal composed of the noise signals of each component inside the hydraulic motor. Then, use the feature signal extraction algorithm to separate and extract this signal to obtain the main frequencies generating vibration noise. By comparing and fitting with the vibration characteristic frequencies obtained in Step S1, the components conforming to the noise frequencies are obtained, and the main components generating vibration noise are preliminarily judged. The specific content is as follows:

[0124] S31: Use wavelet packet analysis to denoise the noise time-domain signal measured in S2:

[0125] It can be seen from the time-domain signal obtained in S2 that its sampling frequency is 65600 Hz;

[0126] Set the number of wavelet packet decomposition layers to 8 layers, and finally there are 256 nodes;

[0127] The frequency range of each node is

[0128] After decomposing the signal, superimpose and reconstruct the signals in the two node ranges with larger contribution amounts, namely 128.125 Hz - 256.25 Hz and 256.25 Hz - 384.275 Hz, to obtain the time-domain signal after removing the interference noise, as Figure 4 shown is the denoising process of the time-domain signal using wavelet packet analysis.

[0129] S32: Use EEMD-Hilbert transform to separate and extract the denoised signal to obtain the spectrogram of the noise signal:

[0130] For the non-stationary signal of the hydraulic motor, in this example, ensemble empirical mode decomposition (EEMD) is used to decompose the denoised time-domain signal, obtaining n intrinsic mode functions (IMFs) with different frequencies and a residual component;

[0131] Obtain the IMF components with a relatively large correlation with the original signal;

[0132] Perform Hilbert-Huang transform on the extracted IMF components to obtain the spectrograms of each component signal, as Figure 5 shown is the spectrogram extracted by EEMD-Hilbert, and the main frequencies generating vibration noise are 205 Hz and 369 Hz.

[0133] In this example, in Embodiment S32; "relatively large correlation" means that the amplitude of the extracted signal component accounts for a relatively large proportion in the original signal and has the most significant and important information of the original signal.

[0134] S33: Based on the vibration frequencies of the internal components of the motor obtained in step S1, it is obtained that the frequencies in the spectrogram respectively correspond to the first-order frequency of bearing vibration, the triple frequency of plunger collision and ball joint clearance collision, and the triple frequency of the flow distribution impact at the inlet and outlet ports. That is, the components that conform to the vibration frequency are bearing vibration, plunger collision, ball joint clearance collision, and the flow distribution impact at the inlet and outlet ports.

[0135] Step S4: Conduct an acoustic intensity test experiment on the target motor to obtain the acoustic intensity visualization image of the motor. The specific content is as follows:

[0136] S41: As Figure 6 shown, according to the size of the hydraulic motor 165mm×252mm×266mm, arrange a 7×10 grid in the front view direction and the top view direction respectively, and set the adjacent measurement points to be 25mm apart.

[0137] S42: Set the corresponding grid in the test software, and use the acoustic intensity probe to collect the acoustic intensity signals point by point in the two grids respectively, so as to generate the acoustic intensity visualization cloud map in the front view and top view directions, and obtain the noise distribution image when the hydraulic motor is working.

[0138] Step S5: Use the compressive sensing framework to perform high-resolution reconstruction on the acoustic intensity visualization image obtained in step S4 to obtain the accurate noise distribution of the motor. The specific content is as follows:

[0139] S51: Perform sparse representation on the obtained acoustic intensity visualization image:

[0140] In this example, using the orthogonality of the Fourier transform (FFT), construct an FFT sparse basis matrix with good non-correlation;

[0141] According to the data dimension of the motor acoustic intensity visualization image, determine the number of rows and columns of the sparse basis matrix. The number of data points of the image obtained through the acoustic intensity test experiment is 315. From this, it is determined that the dimension of the constructed FFT sparse basis matrix is 315×315.

[0142] First, construct a 315×315 identity matrix, then perform an FFT transform on this identity matrix, and finally invert the matrix to obtain the FFT sparse basis matrix.

[0143] S52: Design an observation matrix according to the acoustic field characteristics:

[0144] In this example, using the orthogonality of the Hadamard matrix, construct a pseudo-random partial Hadamard matrix as the observation matrix. The steps are as follows:

[0145] First, according to the data dimension of the motor acoustic intensity visualization image being 315, construct a 315×315 Hadamard matrix. This matrix is distributed with 1 and -1, as shown below:

[0146]

[0147] Next, set the observed value to 100, randomly select 100 rows from the Hadamard matrix to obtain a partial Hadamard matrix of 100×315. According to the characteristic that one measurement point in the sound intensity image corresponds to one signal, assign 0 to the part where -1 is located. Finally, shuffle its rows and columns to complete the construction of the pseudo-random partial Hadamard matrix, and its matrix is shown as follows:

[0148]

[0149] S53: In this example, the orthogonal matching pursuit algorithm (OMP) is used as the reconstruction algorithm. Combining the sparse basis matrix and the observation matrix, the sound intensity visualization cloud map is reconstructed with high resolution to obtain the accurate surface noise distribution of the hydraulic motor, as Figure 7 shown as the high-resolution reconstructed image of the hydraulic motor sound intensity.

[0150] Step S6: Based on the motor noise distribution obtained in step S5, screen out the components that generate noise initially judged in step S3, and further accurately locate the main components that generate vibration noise to achieve the fault pre-positioning of the hydraulic motor. The specific content is as follows:

[0151] First, combine the hydraulic motor structure analyzed in S1 and its vibration mechanism with the components that match the vibration signal frequency obtained in step S3 to obtain the internal vibration distribution. Then, based on the surface noise distribution of the hydraulic motor obtained in S5, exclude the components with the same vibration frequency but not at the sound intensity strong point positions, and accurately locate that the main components generating vibration noise inside the motor are the bearing vibration of the hydraulic motor and the flow distribution impact at the oil inlet and outlet, so as to achieve the fault pre-judgment of the hydraulic motor.

Claims

1. A hydraulic motor fault pre - positioning method based on sound intensity image and feature signal extraction, characterized in that: including the following steps; Step S1: Obtain the vibration characteristic frequencies of each component inside the hydraulic motor; Step S2: Based on the sound pressure test experiment, obtain the noise time-domain signal when the hydraulic motor is working; Step S3: Based on the vibration and noise time-domain signal of the hydraulic motor obtained in Step S2, use the signal denoising algorithm to obtain the time-domain signal with a large contribution amount and remove the interference signal, thereby obtaining the unstable time-domain signal composed of the noise signals of each component inside the motor; Step S4: Based on the unstable time-domain signal obtained in Step S3, use the feature signal extraction algorithm to decompose it into multiple signal components, extract the time-domain signal with a large correlation degree with the original signal and convert it into a frequency-domain signal to obtain the main frequencies generating vibration and noise inside the hydraulic motor. By comparing and fitting with the vibration characteristic frequencies obtained in Step S1, obtain the components corresponding to the frequencies, and then preliminarily judge the main components generating vibration and noise inside the motor; Step S5: Based on the sound intensity test experiment, obtain the sound intensity visualization image of the hydraulic motor; Step S6: Based on the sound intensity visualization image obtained in Step S5, perform high-resolution reconstruction on it through compressive sensing technology to obtain the noise distribution on the outer surface of the motor with accurate noise intensity points, and then judge the main parts generating vibration and noise of the hydraulic motor from the outside; Step S7: Combine the noise distribution on the outside of the motor obtained in Step S6 with the components preliminarily judged to generate vibration and noise inside the motor in Step S4, screen out the components with the same vibration and noise frequencies but not at the noise intensity points, and further obtain the accurate main components generating vibration and noise to realize the fault pre-positioning of the hydraulic motor.

2. The hydraulic motor fault pre - positioning method based on sound intensity image and feature signal extraction according to claim 1, characterized in that: The specific method of Step S1 is: Based on the structure of the hydraulic motor and the vibration mechanism during operation, establish a motor motion mechanism model, clarify the motion modes and their laws of each component inside the motor, and then obtain the vibration characteristic frequencies of each component inside it; The specific method of Step S2 is: Based on the sound pressure test standard, arrange sound pressure probes around the hydraulic motor by the hemispherical method in an anechoic chamber to obtain the noise time-domain signal of the hydraulic motor.

3. The hydraulic motor fault pre-positioning method based on sound intensity image and feature signal extraction according to claim 2, characterized in that: Step S2 includes the following steps; Step S21: The sound pressure test experiment is carried out in an anechoic chamber, and the experimental instruments are B&K acoustic equipment and software; Step S22: Based on the sound pressure test standard, arrange sound pressure probes at positions 1 m in front of, above, and on the side of the hydraulic motor by the hemispherical method to obtain the noise time-domain signal of the motor.

4. The hydraulic motor fault pre - positioning method based on sound intensity image and feature signal extraction according to claim 1, wherein: The specific method of step S3 is as follows: perform denoising processing on the noise time-domain signal obtained in S2 by using wavelet packet analysis to filter out interference signals: based on the noise time-domain signal obtained in S2, obtain its sampling frequency f s ; Set the wavelet packet decomposition layer number n according to the sampling frequency; Obtain the frequency range of each node: Calculate the energy distribution of each node frequency range and select the frequency band with a larger energy contribution; Stack and reconstruct the obtained frequency bands to filter the interference signal, achieve the effect of denoising, and then obtain the unstable time-domain signal composed of the noise signals of each component inside the motor.

5. The hydraulic motor fault pre - positioning method based on sound intensity image and feature signal extraction according to claim 1, characterized in that: The said Step S4 includes the following steps: Step S41: Separate the unstable time-domain signal obtained in Step S3 through the feature signal extraction algorithm to obtain several signal components; Step S42: Extract the signal component with a large correlation degree with the original signal and convert it into a frequency-domain signal to obtain the main frequencies generating vibration and noise in the hydraulic motor; Step S43: Based on the characteristic frequencies of each component inside the motor obtained in Step S1, compare and fit the obtained frequencies to obtain the components corresponding to the frequencies, and then preliminarily determine the main components generating vibration noise.

6. The hydraulic motor fault pre-positioning method based on sound intensity image and feature signal extraction according to claim 5, wherein: The specific method of Step S4 is as follows: Step S41: First, use the Ensemble Empirical Mode Decomposition (EEMD) algorithm to separate and extract the denoised time-domain signal to obtain n Intrinsic Mode Functions (IMFs) and a residual component; then select the IMF components with a relatively high correlation with the original signal; perform Hilbert-Huang transform on the obtained IMF components to obtain the marginal spectrogram and acquire the main frequencies generating vibration noise in the hydraulic motor. Step S42: Based on the vibration characteristic frequencies of each component inside the motor obtained in Step S1, fit the decomposed frequency signals to obtain the components corresponding to the frequencies, and then preliminarily determine the main components generating vibration noise.

7. The hydraulic motor fault pre - positioning method based on sound intensity image and feature signal extraction according to claim 1, wherein: The specific method of Step S5 is as follows: Based on the size of the hydraulic motor, arrange a measurement point grid around it, and collect the measurement point signals point by point through a sound intensity probe to obtain the sound intensity visualization images in different view directions, including the following steps; Step S51: Based on the size of the hydraulic motor, arrange an m×n grid in its front view direction and top view direction respectively, and set the distance between adjacent measurement points according to the probe size. Step S52: Set the corresponding grid in the test software, and collect the sound intensity signals point by point in the two grids through the sound intensity probe respectively, so as to generate the sound intensity visualization cloud maps in the front view and top view directions and obtain the approximate distribution of the motor noise.

8. The hydraulic motor fault pre - positioning method based on sound intensity image and feature signal extraction according to claim 1, wherein: Step S6 includes the following steps: Step S61: Design a compressive sensing algorithm framework, which includes a sparse matrix, an observation matrix, and a reconstruction algorithm; Step S62: Based on the designed compressive sensing algorithm, perform high-resolution reconstruction on the motor sound intensity visualization image obtained in Step S5 to obtain the noise distribution on the outer surface of the hydraulic motor with accurate noise intensity points.

9. The hydraulic motor fault pre - positioning method based on sound intensity image and feature signal extraction according to claim 8, wherein: In Step S61, the specific method is as follows: Perform sparse representation on the sound intensity visualization image obtained in Step S5: Utilize the orthogonality of the Fourier transform to construct an FFT sparse basis matrix with good incoherence. The specific steps are as follows: Step A1: According to the signal dimension N of the obtained sound intensity visualization image, determine the dimension of the sparse basis as N×N; Construct an N×N identity matrix; Step A2: Perform FFT transformation on this identity matrix; Step A3: Finally, invert this matrix to complete the construction of the FFT sparse basis matrix; The method of Step S62 is specifically as follows: According to the characteristics of the motor sound field, utilize the orthogonality of the Hadamard matrix to construct a pseudo-random partial Hadamard matrix as the observation matrix. The specific steps are as follows: Step B1: According to the signal dimension N of the obtained sound intensity visualization image, construct a Hadamard matrix of dimension N×N, and this matrix is distributed as -1, 1, as shown below: Step B2: Set the number of observations M, and randomly select M rows from the constructed Hadamard matrix to obtain an M×N partial Hadamard matrix; Step B3: According to the characteristic that one measurement point in the sound intensity image corresponds to one signal, assign the -1 part to 0; Step B4: Randomly shuffle its rows and columns to complete the construction of the pseudo-random part of the Hadamard matrix; the constructed observation matrix is as follows: Step S6 further includes Step S63, that is: using the Orthogonal Matching Pursuit algorithm OMP as the reconstruction algorithm, combining the sparse basis matrix and the observation matrix, performing high-resolution reconstruction on the sound intensity visualization cloud map to obtain the accurate surface noise distribution of the hydraulic motor.

10. The hydraulic motor fault pre-positioning method based on sound intensity image and feature signal extraction according to claim 1, wherein: The specific method of Step S7 is: aiming at the characteristics of the compact internal structure of the hydraulic motor and the close coupling of each component, as well as the problem that the vibration frequencies of its internal components have the relationship of the same frequency or multiple frequencies, based on the high-resolution sound intensity strong point distribution image of the hydraulic motor obtained in Step S6, screening out the components with the same vibration frequency but not in the noise strong point position among the components generating vibration noise obtained through Step S4, so as to reduce the influence of the components with the same frequency or multiple frequency vibration frequencies, and further obtain the accurate main components generating vibration noise, realizing the fault pre-positioning of the hydraulic motor.

Images