A method and system for monitoring abnormal sound of mechanical equipment

Abstract

The present invention relates to a method and system for monitoring abnormal noise of mechanical equipment, comprising: S1: collecting sound signals and image signals of the mechanical equipment under test; S2: identifying whether the sound signal contains abnormal noise; S3: if so, performing sound source location processing on the abnormal noise; S4: performing sound field imaging processing according to the generation direction and noise intensity of the abnormal noise and the image signal; S5: performing data analysis and fault diagnosis on the sound signal and the information after the sound field imaging processing. By monitoring the abnormal noise of mechanical equipment, the present invention can not only determine the faulty parts of the equipment, but also effectively diagnose the cause of the fault, so that maintenance can be carried out purposefully to reduce unplanned shutdowns, thereby avoiding affecting the normal operation of nuclear power units as much as possible.

Description

Technical Field

[0001] The present invention relates to the technical field of equipment detection, and in particular to a method and system for monitoring abnormal noise of mechanical equipment. Background Art

[0002] In nuclear power plants, there are many types of mechanical equipment, and they work continuously for a long time, so it is inevitable that mechanical equipment will fail. Usually, when mechanical equipment fails, it will cause abnormal noise in the equipment, and this abnormal noise is usually used as the basis for judging the existence of the fault. For example, when the sound is dull and accompanied by a certain vibration phenomenon, it is easier to judge that the fault exists. When the abnormal noise emitted by the equipment is very slight or complex, the staff will not be able to effectively identify whether it has abnormal noise, which makes fault diagnosis difficult and cannot timely and accurately identify the early failure of the equipment. Abnormal noise problems occur from time to time, seriously affecting the safety, stability, reliability and load rate of nuclear power plant units. Summary of the invention

[0003] The technical problem to be solved by the present invention is to provide a method and system for monitoring abnormal noise of mechanical equipment in view of at least one defect in the prior art.

[0004] The technical solution adopted by the present invention to solve the technical problem is: constructing a method for monitoring abnormal noise of mechanical equipment, comprising:

[0005] S1: Collect the sound and image signals of the tested mechanical equipment;

[0006] S2: Identify whether the sound signal contains abnormal noise;

[0007] S3: If yes, perform sound source localization processing on the abnormal noise to confirm the generation direction and noise intensity of the abnormal noise;

[0008] S4: performing sound field imaging processing according to the generation direction and noise intensity of the abnormal noise and the image signal to realize imaging of the generation direction and noise intensity of the abnormal noise;

[0009] S5: Performing data analysis and fault diagnosis on the sound signal and the information processed by the sound field imaging to identify the component of the mechanical equipment where the mechanical failure occurs.

[0010] Preferably, in the abnormal sound monitoring method for mechanical equipment of the present invention, S1 further comprises:

[0011] S11: preprocessing the sound signal;

[0012] S12: performing image vectorization processing and pixel angle conversion processing on the image signal.

[0013] Preferably, in the abnormal sound monitoring method for mechanical equipment of the present invention, the S11 comprises:

[0014] The sound signal is subjected to one or more processing measures of removing direct current components, normalizing, filtering and pre-emphasis.

[0015] Preferably, in the abnormal sound monitoring method for mechanical equipment described in the present invention, after S11, it also includes: referring to the sound characteristics emitted by the components in the mechanical equipment when working normally or abnormally, performing feature extraction on the preprocessed sound signal to extract the characteristic components of the sound signal, thereby providing a detection basis for the subsequent identification of whether the preprocessed sound signal contains abnormal noise.

[0016] Preferably, in the abnormal sound monitoring method for mechanical equipment of the present invention, S2 comprises:

[0017] A deep neural network detection algorithm or an unsupervised abnormal sound detection algorithm is performed on the characteristic component to identify whether the preprocessed sound signal contains abnormal noise.

[0018] Preferably, in the abnormal sound monitoring method for mechanical equipment of the present invention, the sound source localization processing in S3 includes:

[0019] The sound signal is collected through a sensor array, and the phases of the collected array signals are compensated and summed to obtain preliminary direction information; at the same time, according to the frequency domain characteristics of the abnormal noise, the beam peak corresponding to the abnormal noise is found in the sound signal; according to the preliminary direction information and the beam peak, corresponding phase compensation and weighted summation processing are performed to obtain a three-dimensional beam diagram for characterizing the generation direction of the abnormal noise and the sound field intensity in the corresponding direction.

[0020] Preferably, in the abnormal sound monitoring method for mechanical equipment of the present invention, the sound field imaging processing in S4 includes:

[0021] The three-dimensional beam map is converted into a two-dimensional color contour map; wherein the characterization points in the two-dimensional color contour map are used to display the direction in which abnormal noise occurs at the pixel points in the video image, and the intensity of the sound field is displayed using the depth of the color of the characterization points; the two-dimensional color contour map is made transparent, and the characterization points are superimposed on the pixel points to achieve imaging of the direction in which the abnormal noise occurs and the noise intensity.

[0022] Preferably, in the abnormal sound monitoring method for mechanical equipment of the present invention, the data analysis and fault diagnosis in S5 include:

[0023] A fault simulation test bench is set up to simulate the failure of various components of the mechanical equipment, and the sample data generated by different fault modes are calibrated and recorded as a fault sample model. The sound signal is subjected to time domain analysis, spectrum analysis and envelope analysis, and the analyzed sound signal and the data information after imaging processing are compared with the fault sample model to realize fault diagnosis of the mechanical equipment under test.

[0024] Preferably, in the abnormal sound monitoring method for mechanical equipment of the present invention, performing time domain analysis on the sound signal further comprises:

[0025] Diagnose whether a component has a mechanical failure by judging whether the time domain characteristic parameter after time domain analysis exceeds a preset value;

[0026] The time domain characteristic parameters include one or more combinations of mean, peak, root mean square, crest factor, peak-to-peak value and kurtosis.

[0027] The present invention also constructs a mechanical equipment abnormal sound monitoring system, the monitoring system comprising:

[0028] A collection unit, used for collecting sound signals and image signals of the mechanical equipment under test;

[0029] an abnormal noise identification unit, used to identify whether the sound signal contains abnormal noise;

[0030] A sound localization unit, used for performing sound source localization processing on the abnormal noise;

[0031] An imaging processing unit, used for performing sound field imaging processing according to the generation direction and noise intensity of the abnormal noise and the image signal;

[0032] The data analysis and fault diagnosis unit is used to perform data analysis and fault diagnosis on the sound signal and the information processed by the sound field imaging.

[0033] The implementation of the present invention has the following beneficial effects: the present invention monitors and diagnoses abnormal sounds of mechanical equipment, perceives abnormal sounds of mechanical equipment through sound source localization processing, determines the specific faulty parts of the mechanical equipment, and can effectively diagnose the cause of the fault, so that maintenance can be carried out purposefully to reduce unplanned shutdowns, thereby avoiding as much as possible affecting the normal operation of nuclear power units. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] The present invention will be further described below with reference to the accompanying drawings and embodiments, in which:

[0035] Figure 1 This is the first embodiment of the abnormal sound monitoring method for mechanical equipment provided by the present invention;

[0036] Figure 2 is a flow chart of a second embodiment of the abnormal sound monitoring method for mechanical equipment provided by the present invention;

[0037] Figure 3 It is a schematic diagram of pixel angle conversion processing in the abnormal sound monitoring method of mechanical equipment provided by the present invention;

[0038] Figure 4 It is a three-dimensional beam diagram in the abnormal sound monitoring method of mechanical equipment provided by the present invention;

[0039] Figure 5 It is a two-dimensional color contour map in the abnormal sound monitoring method of mechanical equipment provided by the present invention;

[0040] Figure 6 It is a schematic diagram after the sound field imaging processing in the abnormal sound monitoring method of mechanical equipment provided by the present invention;

[0041] Figure 7 It is a schematic diagram of superimposing the characterization points and the pixel points in the abnormal sound monitoring method of mechanical equipment provided by the present invention;

[0042] Figure 8 It is a comparison diagram of the visualization processing in the abnormal sound monitoring method of mechanical equipment provided by the present invention. DETAILED DESCRIPTION

[0043] In order to have a clearer understanding of the technical features, purposes and effects of the present invention, specific embodiments of the present invention are now described in detail with reference to the accompanying drawings.

[0044] Figure 1 This is the first embodiment of the abnormal sound monitoring method for mechanical equipment provided by the present invention. Figure 1 As shown, the present invention constructs a method for monitoring abnormal noise of mechanical equipment, including: step S1, step S2, step S3, step S4 and step S5.

[0045] S1: Collect the sound and image signals of the mechanical equipment under test.

[0046] Figure 2 This is a flow chart of a second embodiment of the abnormal sound monitoring method for mechanical equipment provided by the present invention. Compared with the first embodiment, the difference is that step S1 also includes: step S11 and step S12.

[0047] S11: Preprocessing the sound signal, including removing DC components, normalizing, filtering and pre-emphasis processing, or one or more processing measures to ensure the integrity of the sound signal and prevent distortion of the sound signal. After preprocessing the sound signal, the sound signal after preprocessing is also extracted with reference to the sound characteristics of the components in the mechanical equipment when they are working normally or abnormally, so as to extract the characteristic components of the sound signal, thereby providing a detection basis for the subsequent identification of whether the sound signal after preprocessing contains abnormal noise.

[0048] Feature extraction is a means of analyzing and processing the pre-processed sound signal. The principle of feature extraction is: since the sound generation mechanism of different devices or components is different, different devices or components have unique characteristic components when they are working normally or when abnormal failure occurs. Therefore, the characteristic components can be extracted as the basis for detecting whether the sound signal contains abnormal noise. The characteristic components extracted in this method include STFT, MFCC and HPSS.

[0049] S12: performing image vectorization processing and pixel angle conversion processing on the image signal to obtain a video image, wherein the image vectorization processing is performed on the image signal so as to convert the image signal into a vector matrix. Figure 3 Schematic diagram of pixel angle conversion processing in the abnormal sound monitoring method of mechanical equipment provided by the present invention, such as Figure 3 As shown, the pixel angle conversion processing of the image signal is to form a corresponding relationship between each pixel point in the image signal and its respective angle in the sound field, and map it to the vector matrix, and finally form a transformation matrix. After the image signal is processed by image vectorization and pixel angle conversion, a video image is obtained to prepare for the subsequent sound field imaging processing.

[0050] S2: Identify whether the sound signal contains abnormal noise, wherein the identification method includes: performing a deep neural network detection algorithm or an unsupervised abnormal sound detection algorithm on the feature component to identify whether the preprocessed sound signal contains abnormal noise.

[0051] Furthermore, this method can be divided into two detection methods according to whether signal detection requires a large number of positive and negative samples for preliminary learning, including a deep neural network detection algorithm based on CRNN and an unsupervised abnormal sound detection algorithm based on AutoEncoder.

[0052] The CRNN-based deep neural network detection algorithm is suitable for detection environments where the positive and negative samples of abnormal noise of the equipment are sufficient and the number of positive and negative samples is relatively balanced. The model can be trained by collecting the characteristic components of the sound signal of the mechanical equipment under test to identify and diagnose one or more abnormal noises, and to quickly warn in abnormal conditions. Since the algorithm requires a large number of positive and negative samples as the basis for deep learning, the final recognition accuracy of the algorithm is highly affected by the comprehensiveness and generalization of the positive and negative samples of the abnormal noise of the equipment. Therefore, the characteristics of the algorithm are as follows: it is suitable for situations where the number of positive and negative samples is sufficient; the detection is highly targeted, and the abnormal type can be directly diagnosed based on the characteristic components of the sound signal; the initial data accumulation time is long, and the sample requirements are high.

[0053] The unsupervised abnormal sound detection algorithm based on AutoEncoder uses the AutoEncoder autoencoder to encode and decode the characteristic components of the sound signal in normal state, and forms an encoder that only detects positive samples through deep neural network training. When the system collects abnormal noise in the abnormal working state of the tested mechanical equipment, the output result of the encoder will produce a significant deviation, so the output result of the encoder can be used as the main basis for whether the tested mechanical equipment has abnormal noise. Since the algorithm only requires a small number of positive samples to realize the construction of the neural network of the autoencoder, and does not rely on the support of a large number of negative samples. Therefore, the algorithm is easy to use. The core point that affects the accuracy of the algorithm is the generalization and complexity of the positive samples. For equipment with stable working noise state, the algorithm has excellent detection performance. Therefore, the characteristics of the algorithm are as follows: it is suitable for sound detection in an unsupervised state with a small number of samples; it does not rely on various negative samples, and the algorithm deployment cycle is fast; it can only warn of changes in the noise state, and cannot accurately diagnose the actual fault type.

[0054] S3: If it exists, the abnormal noise is processed for sound source localization to confirm the direction and noise intensity of the abnormal noise. The sound localization process includes collecting sound signals through the sensor array, compensating and summing the phases of the collected array signals to obtain preliminary direction information; at the same time, according to the frequency domain characteristics of the abnormal noise, searching for the beam peak corresponding to the abnormal noise in the sound signal; according to the preliminary direction information and the beam peak, performing corresponding phase compensation and weighted summation processing to obtain a three-dimensional beam diagram used to characterize the direction of the abnormal noise and the sound field intensity in the corresponding direction.

[0055] S4: Performing sound field imaging processing according to the direction and noise intensity of the abnormal noise and the image signal to realize imaging of the direction and noise intensity of the abnormal noise. Figure 4 , Figure 5 , Figure 6 and Figure 7As shown, the sound field imaging processing includes: converting the three-dimensional beam map into a two-dimensional color contour map; wherein the representation points in the two-dimensional color contour map are used to display the direction of abnormal noise generated by the pixel points in the video image, and the depth of the color of the representation points is used to display the sound field intensity; the two-dimensional color contour map is transparently processed, and the representation points are superimposed on the pixel points to realize the imaging of the direction of abnormal noise and the noise intensity.

[0056] Furthermore, the sound field imaging process also includes visualization processing; the visualization processing is used to intercept the imaging of the portion with higher color contour values ​​in the image after the sound field imaging process. Figure 8 1 is a comparison chart of the visualization processing in the abnormal sound monitoring method of mechanical equipment provided by the present invention. Figure 8 As shown, the left side is the image of the mechanical equipment under test when it is working normally, and the right side is the image when a failure occurs. By comparison, it can be seen that after visualization processing, the position of the abnormal noise on the image can be highlighted.

[0057] S5: Perform data analysis and fault diagnosis on the sound signal and the information processed by the sound field imaging to identify the parts of the mechanical equipment that have mechanical failures. The data analysis and fault diagnosis include: setting up a fault simulation test bench for simulating the failure of various parts of the mechanical equipment, calibrating and recording the sample data generated by different fault modes as a fault sample model, performing time domain analysis, spectrum analysis and envelope analysis on the sound signal, and comparing the analyzed sound signal and the data information after imaging processing with the fault sample model to achieve fault diagnosis of the tested mechanical equipment.

[0058] The fault simulation test bench includes a test bench base, a small drive motor, a motor mounting frame, a motor, a shaft, a coupling, a bearing load kit, a shaft friction kit, and a shaft vibration measurement kit. The fault modes of the fault simulation test bench include: misalignment fault simulation, simulating the working state of various components when the shafts are misaligned, and also simulating the working state when couplings with different stiffness affect the power and vibration of the rotor; shaft friction simulation of rotating mechanical equipment, by using a shaft friction kit to simulate the working state of shaft or rotor friction; resonance simulation, by installing different numbers of rotors at different positions on the shaft to excite the resonance frequency, to simulate the working state of the rotor and shaft resonance phenomenon, and by adjusting the position of the rotor and support to study the influence of mass and stiffness on the resonance frequency and modal vibration type; bearing fault simulation, using a rolling bearing kit to simulate the working state of the inner ring, outer ring, rolling element and cage of the rolling bearing; shaft eccentricity fault simulation, using a shaft radial vibration kit to simulate the working state of the shaft eccentricity fault.

[0059] In this embodiment, in order to ensure the accuracy of the sound source localization processing, a fault simulation test bench can be used to simulate the failure modes of various components to verify the accuracy of the sound source localization processing in this method. For example, when using a fault simulation test bench to simulate a bearing outer ring failure, a bearing outer ring failure kit is installed on a vibration test bench to simulate a bearing outer ring failure, and the test results of this method at different test points are compared. If the test results at different test points all show that the problem bearing part is abnormal, it proves that the function of the sound source localization processing is normal.

[0060] Furthermore, the fault simulation test bench can be used to simulate the failure modes of various components to verify and debug the test accuracy of the method. By using the fault simulation test bench to simulate various failure modes, steps S1-S5 in the method are performed in sequence, and finally the test results of the method are compared with the failure components of the fault simulation test bench to verify whether the method is accurate. In the test process of various failure modes, representative data of the test process can also be calibrated and stored as a fault sample model for comparison.

[0061] Furthermore, time domain analysis of the sound signal also includes: diagnosing whether a component has a mechanical failure by determining whether the time domain characteristic parameters after the time domain analysis exceed a preset value; wherein the time domain characteristic parameters include one or more combinations of mean, peak, root mean square, crest factor, peak-to-peak value and kurtosis.

[0062] The mean is used to reflect the static part of the sound signal. It has no effect on diagnosis, but has a great influence on the calculation of other parameters. Therefore, when calculating, the mean is first taken out from the sound signal, leaving the dynamic part that is useful for diagnosis.

[0063] The peak value is used to characterize the maximum value of the noise intensity within a certain period of time. It varies greatly over time and can therefore be used to detect impact signals.

[0064] The root mean square is the effective value that represents the sound intensity. It is the most important indicator in the monitoring process and can be used as an important indicator for judging wear-related faults.

[0065] The crest factor is the ratio of the peak value to the root mean square. The cause of the failure of the tested mechanical equipment can be predicted based on the change and size of the crest factor. For example, when the crest factor increases, it may be a pitting fault, and when the crest factor decreases, it may be a wear fault. The normal crest factor value of a bearing is about 4 to 5. When the crest factor is greater than 10, it means that the impact is caused by a local defect in a certain component; when the crest factor becomes smaller, it indicates abnormal conditions such as poor bearing lubrication and wear, which makes the noise larger.

[0066] The peak-to-peak value is used to characterize the amplitude of vibration, so the peak-to-peak value can be used to describe the displacement value of the component to determine whether the component has a peeling failure.

[0067] Kurtosis is used to characterize the peak degree of waveforms and is particularly sensitive to suddenly changing sound signals. Under normal circumstances, the kurtosis index should be below 3. When the index is greater than 3.5, it generally indicates the occurrence of a fault. Therefore, kurtosis can be used to diagnose early faults, especially for surface damage faults. For example, when an early fault occurs in a bearing, the kurtosis value will increase immediately. As the bearing fault develops, the kurtosis value will remain unchanged or gradually decrease.

[0068] Spectrum analysis is one of the basic methods for studying fault characteristics. For the monitoring object of this method, relatively comprehensive and accurate fault information can be obtained in the spectrum.

[0069] Envelope analysis is used to extract the periodic impact signal emitted when a bearing fails, so that component failure can be diagnosed.

[0070] Furthermore, after completing the fault diagnosis of the mechanical equipment under test, the method also includes trend analysis processing. The purpose of trend analysis processing is to arrange maintenance work according to the actual situation, arrange maintenance at the right time, make scientific decisions, and minimize economic losses. Fault diagnosis is the basis of trend analysis processing. The time history curve of the mechanical equipment under test is drawn according to the characteristic components of the collected sound signals to analyze the short-term and long-term trend status of the mechanical equipment under test.

[0071] The present invention also constructs a mechanical equipment abnormal sound monitoring system, which includes: a collection unit, an abnormal noise identification unit, a sound positioning unit, an imaging processing unit, and a data analysis and fault diagnosis unit.

[0072] The acquisition unit is used to acquire the sound signal and image signal of the mechanical equipment under test.

[0073] In some embodiments, the system further includes a preprocessing unit for preprocessing the sound signal and performing image vectorization processing and pixel angle conversion processing on the image signal.

[0074] The preprocessing unit preprocesses the sound signal, including removing the DC component, normalizing, filtering and pre-emphasis processing, or one or more processing measures to ensure the integrity of the sound signal and prevent the sound signal from being distorted. After the sound signal is preprocessed, the sound signal after preprocessing is also extracted with reference to the sound characteristics of the components in the mechanical equipment when they are working normally or abnormally, so as to extract the characteristic components of the sound signal.

[0075] The preprocessing unit performs image vectorization processing on the image signal to convert the image signal into a vector matrix, and performs pixel angle conversion processing on the image signal to form a corresponding relationship between each pixel point in the image signal and its respective angle in the sound field, and maps them to the vector matrix, and finally forms a transformation matrix. After the image signal is subjected to image vectorization processing and pixel angle conversion processing, a video image is obtained.

[0076] The abnormal noise identification unit is used to identify whether the sound signal contains abnormal noise. The identification method includes: performing a deep neural network detection algorithm or an unsupervised abnormal sound detection algorithm on the characteristic component to identify whether the preprocessed sound signal contains abnormal noise.

[0077] The sound localization unit is used to localize the sound source of the abnormal noise to confirm the direction and noise intensity of the abnormal noise. The sound localization unit also includes a sensor array for collecting sound signals and a positioning analysis processor. The positioning analysis processor compensates and sums the phases of the collected array signals to obtain preliminary direction information; at the same time, according to the frequency domain characteristics of the abnormal noise, it searches for the beam peak corresponding to the abnormal noise in the sound signal; according to the preliminary direction information and the beam peak, it performs corresponding phase compensation and weighted summation processing to obtain a three-dimensional beam diagram for characterizing the direction of the abnormal noise and the sound field intensity in the corresponding direction.

[0078] The imaging processing unit is used to perform sound field imaging processing according to the direction and noise intensity of the abnormal noise and the image signal, so as to realize the imaging of the direction and noise intensity of the abnormal noise. The imaging processing unit converts the three-dimensional beam map into a two-dimensional color contour map; wherein the representation points in the two-dimensional color contour map are used to display the direction of the abnormal noise generated by the pixel points in the video image, and the sound field intensity is displayed by using the depth of the color of the representation points; the two-dimensional color contour map is transparently processed, and the representation points are superimposed on the pixel points to realize the imaging of the direction and noise intensity of the abnormal noise.

[0079] Furthermore, the imaging processing unit is also used to visualize the data after the sound field imaging processing.

[0080] The data analysis and fault diagnosis unit is used to perform data analysis and fault diagnosis on the sound signal and the information processed by the sound field imaging, so as to identify the parts of the mechanical equipment that have mechanical failures. The data analysis and fault diagnosis unit includes a fault simulation test bench for simulating the failure of various parts of the mechanical equipment, and a host computer for data analysis and fault diagnosis. The host computer calibrates and records the sample data generated by different fault modes, and performs time domain analysis, spectrum analysis and envelope analysis on the sound signal as a fault sample model, and compares the analyzed sound signal and the data information after imaging processing with the fault sample model to realize the fault diagnosis of the mechanical equipment under test.

[0081] The implementation of the present invention has the following beneficial effects: the present invention monitors and diagnoses abnormal sounds of mechanical equipment, perceives abnormal sounds of mechanical equipment through sound source localization processing, determines the specific faulty parts of the mechanical equipment, and can effectively diagnose the cause of the fault, so that maintenance can be carried out purposefully to reduce unplanned shutdowns, thereby avoiding as much as possible affecting the normal operation of nuclear power units.

[0082] It can be understood that the above embodiments only express the preferred implementation modes of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the patent scope of the present invention. It should be pointed out that, for ordinary technicians in this field, the above technical features can be freely combined without departing from the concept of the present invention, and several deformations and improvements can be made, which all belong to the protection scope of the present invention. Therefore, all equivalent changes and modifications made to the scope of the claims of the present invention should belong to the coverage of the claims of the present invention.

Claims

1. A method for monitoring abnormal noise of mechanical equipment, characterized in that: include: S1: Collect the sound and image signals of the tested mechanical equipment; S2: Identify whether the sound signal contains abnormal noise; S3: If yes, perform sound source localization processing on the abnormal noise to confirm the generation direction and noise intensity of the abnormal noise; S4: performing sound field imaging processing according to the generation direction and noise intensity of the abnormal noise and the image signal to realize imaging of the generation direction and noise intensity of the abnormal noise; S5: performing data analysis and fault diagnosis on the sound signal and the information processed by the sound field imaging to identify the component of the mechanical equipment where the mechanical failure occurs; Before S2, it also includes: referring to the sound characteristics emitted by the components in the mechanical equipment when working normally or abnormally, performing feature extraction on the pre-processed sound signal to extract the characteristic components of the sound signal; correspondingly, S2 includes: performing a deep neural network detection algorithm on the characteristic components to identify whether the sound signal contains abnormal noise; The data analysis and fault diagnosis in S5 include: A fault simulation test bench is provided for simulating the failure of various components of the mechanical equipment, and the sample data generated by different fault modes are calibrated and recorded as a fault sample model. The sound signal is subjected to time domain analysis, spectrum analysis and envelope analysis, and the analyzed sound signal and the data information after imaging processing are compared with the fault sample model to realize the fault diagnosis of the mechanical equipment under test; the time domain analysis of the sound signal also includes: diagnosing whether a mechanical failure occurs in a component by judging whether the time domain characteristic parameters after the time domain analysis exceed a preset value, and the time domain characteristic parameters include mean value, peak value, root mean square, crest factor, peak-to-peak value and kurtosis; and determining the cause of the mechanical failure according to the time domain characteristic parameters, including: judging whether the mechanical equipment under test has a wear failure according to the root mean square, judging whether the mechanical equipment under test has a pitting failure, a wear failure, an impact caused by a local defect in a component and poor bearing lubrication according to the change of the crest factor, judging whether the mechanical equipment under test has a component peeling failure according to the peak-to-peak value, and judging whether the mechanical equipment has a component surface damage failure according to the kurtosis; The failure modes of the fault simulation test bench include misalignment fault simulation, shaft rubbing simulation, resonance simulation and shaft eccentricity fault simulation; wherein, the misalignment fault simulation includes forming a shaft misalignment structure of certain components in the fault simulation test bench to simulate the working state of shaft misalignment; the shaft rubbing simulation includes using a kit to simulate the working state of the shaft or rotor rubbing in the fault simulation test bench; the resonance simulation includes installing different numbers of rotors at different positions of a shaft of the fault simulation test bench to excite the resonance frequency to simulate the working state of the rotor and the shaft resonance phenomenon; the shaft eccentricity fault simulation includes using a kit to simulate the working state of the shaft eccentricity fault in the fault simulation test bench.

2. The method for monitoring abnormal noise of mechanical equipment according to claim 1, characterized in that: The S1 further comprises: S11: preprocessing the sound signal; S12: performing image vectorization processing and pixel angle conversion processing on the image signal to obtain a video image.

3. The method for monitoring abnormal noise of mechanical equipment according to claim 2, characterized in that: The S11 includes: The sound signal is subjected to one or more processing measures of removing direct current components, normalizing, filtering and pre-emphasis.

4. The method for monitoring abnormal noise of mechanical equipment according to claim 3, characterized in that: The sound source localization processing in S3 includes: The sound signal is collected through a sensor array, and the phases of the collected array signals are compensated and summed to obtain preliminary direction information; at the same time, according to the frequency domain characteristics of the abnormal noise, the beam peak corresponding to the abnormal noise is found in the sound signal; according to the preliminary direction information and the beam peak, corresponding phase compensation and weighted summation processing are performed to obtain a three-dimensional beam diagram for characterizing the generation direction of the abnormal noise and the sound field intensity in the corresponding direction.

5. The method for monitoring abnormal noise of mechanical equipment according to claim 4, characterized in that: The sound field imaging processing in S4 includes: The three-dimensional beam map is converted into a two-dimensional color contour map; wherein the characterization points in the two-dimensional color contour map are used to display the direction in which abnormal noise occurs at the pixel points in the video image, and the intensity of the sound field is displayed using the depth of the color of the characterization points; the two-dimensional color contour map is made transparent, and the characterization points are superimposed on the pixel points to achieve imaging of the direction in which the abnormal noise occurs and the noise intensity.

6. A mechanical equipment abnormal sound monitoring system, characterized in that: The monitoring system comprises: A collection unit, used to collect sound signals and image signals of the mechanical equipment under test; an abnormal noise identification unit, used to identify whether the sound signal contains abnormal noise; A sound localization unit, used for performing sound source localization processing on the abnormal noise; An imaging processing unit, used for performing sound field imaging processing according to the generation direction and noise intensity of the abnormal noise and the image signal; A data analysis and fault diagnosis unit, used for performing data analysis and fault diagnosis on the sound signal and the information processed by the sound field imaging; The monitoring system further includes: a preprocessing unit, which is used to extract features of the sound signal by referring to the sound characteristics of the components in the mechanical equipment under test when they are working normally or abnormally, so as to extract the characteristic components of the sound signal; correspondingly, the method for the abnormal noise identification unit to identify whether the sound signal contains abnormal noise includes: performing a deep neural network detection algorithm on the characteristic components to determine whether the sound signal contains abnormal noise; The data analysis and fault diagnosis unit includes a fault simulation test bench for simulating the failure of various components of mechanical equipment, and a host computer for data analysis and fault diagnosis; the host computer calibrates and records the sample data generated by different fault modes as a fault sample model, performs time domain analysis, spectrum analysis and envelope analysis on the sound signal, and compares the analyzed sound signal and the data information after imaging processing with the fault sample model to realize fault diagnosis of the mechanical equipment under test; the time domain analysis of the sound signal also includes: diagnosing whether the component is faulty by judging whether the time domain characteristic parameter after the time domain analysis exceeds the preset value whether a mechanical failure occurs, the time domain characteristic parameters include mean, peak, root mean square, crest factor, peak-to-peak value and kurtosis; the cause of the mechanical failure is also determined according to the time domain characteristic parameters, including: judging whether the measured mechanical equipment has a wear failure according to the root mean square, judging whether the measured mechanical equipment has a pitting failure, a wear failure, an impact caused by a local defect in a component and poor bearing lubrication according to the change of the crest factor, judging whether the measured mechanical equipment has a component peeling failure according to the peak-to-peak value, and judging whether the mechanical equipment has a component surface damage failure according to the kurtosis; Among them, the failure modes of the fault simulation test bench include misalignment fault simulation, shaft rubbing simulation, resonance simulation and shaft eccentricity fault simulation; wherein the misalignment fault simulation includes forming a shaft misalignment structure of certain components in the fault simulation test bench to simulate the working state of shaft misalignment; the shaft rubbing simulation includes using a kit to simulate the working state of the shaft or rotor rubbing in the fault simulation test bench; the resonance simulation includes installing different numbers of rotors at different positions of a shaft of the fault simulation test bench to excite the resonance frequency to simulate the working state of the rotor and the shaft resonance phenomenon; the shaft eccentricity fault simulation includes using a kit to simulate the working state of the shaft eccentricity fault in the fault simulation test bench.

Images