Operation monitoring system and method for a vertical pipe pump

Abstract

The application relates to the technical field of industrial pump state monitoring, in particular to a running monitoring system and method of a vertical pipeline pump. The system comprises: a signal acquisition module configured to acquire a mechanical vibration signal of a bearing assembly at a first preset sampling rate and to acquire a sound signal in the running process of the bearing assembly at a second preset sampling rate; an edge computing node configured to input the sound signal in a first time window into a pre-trained model to obtain a classification result; if at least two of the first judgment conditions are met, a quantitative lubrication instruction and an oil discharge instruction are generated in the case that the classification result is poor lubrication. The first judgment conditions include that a time domain feature is greater than a first time domain threshold, an acoustic feature is greater than a first acoustic threshold, and a bearing feature evaluation index is less than a first index threshold and a frequency domain feature is greater than a first frequency domain threshold; it can be seen that the system provides reliable support for intelligent running monitoring of the vertical pipeline pump, and the monitoring accuracy is high.

Description

Technical Field

[0001] This application relates to the field of industrial pump condition monitoring technology, and in particular to an operation monitoring system and method for a vertical pipeline pump. Background Technology

[0002] Vertical pipeline pumps, as core equipment for fluid transportation, are widely used in municipal water supply, industrial circulation, and other scenarios. The operating condition of their bearing assemblies directly determines the reliability of the pump body. Existing monitoring technologies for vertical pipeline pumps mostly rely on single vibration signal detection, which is susceptible to fluid disturbances and environmental noise interference. Furthermore, they are difficult to distinguish subtle differences in conditions such as poor lubrication and early wear, often resulting in misjudgments or missed detections. Summary of the Invention

[0003] This application provides an operation monitoring system and method for a vertical pipeline pump to address the problem of insufficient monitoring methods during the operation of the vertical pipeline pump.

[0004] In a first aspect, embodiments of this application provide an operation monitoring system for a vertical pipeline pump, applied to a vertical pipeline pump. The vertical pipeline pump includes at least a pump shaft and a bearing assembly. The bearing assembly includes a bearing housing, a first bearing and a second bearing disposed inside the bearing housing. The first bearing and the second bearing are spaced apart along the axial direction. An upper bearing retainer is disposed on the first bearing closer to the second bearing, and a lower bearing retainer is disposed on the second bearing farther from the first bearing. The bearing housing is provided with a lubrication hole and a first oil drain hole in parallel. The lubrication hole is located between the first bearing and the second bearing, and the first oil drain hole is located on the side of the lower bearing retainer. The lower bearing retainer is provided with a second oil drain hole, and the first oil drain hole and the second oil drain hole are interconnected.

[0005] The system includes: a signal acquisition module configured to: acquire mechanical vibration signals of a bearing assembly at a first preset sampling rate, and acquire sound signals of the bearing assembly during operation at a second preset sampling rate; an edge computing node configured to: extract features from the mechanical vibration signals within a first time window according to a first preset frequency, obtaining time-domain features, frequency-domain features, and envelope demodulation spectrum features; the first time window has a period length equal to the first preset frequency; calculate a bearing feature evaluation index using the envelope demodulation spectrum features, the bearing feature evaluation index being used to characterize the degree of deterioration of the mechanical state of the bearing assembly; and the edge computing node is further configured to: extract features from the sound signals within the first time window according to the first preset frequency, obtaining acoustic features; and extract sound signals within the first time window according to the first preset frequency. The audio signal is input into the pre-trained model to obtain the classification results of the bearing assembly's operating status. The operating status categories include at least normal, poor lubrication, early wear, and bearing failure. The edge computing node is also configured to generate a quantitative lubrication command and an oil discharge command if at least two of the first judgment conditions are met when the classification result is poor lubrication. The first judgment conditions include that the time domain feature is greater than a first time domain threshold, the acoustic feature is greater than a first acoustic threshold, and the bearing feature evaluation index is less than a first index threshold and the frequency domain feature is greater than a first frequency domain threshold. The quantitative lubrication command is used to drive the power assembly to inject a quantitative amount of lubricating grease into the first and second bearings through the filling hole, and the oil discharge command is used to drive the power assembly to discharge waste lubricating grease through the first and second oil discharge holes.

[0006] In one possible implementation, the edge computing node is further configured to: acquire acoustic features corresponding to X consecutive first time windows in real time, and construct a first vector; calculate the first similarity between the first vector and historical feature vectors; wherein the historical feature vectors are constructed from the acoustic features corresponding to the 1st to Xth first time windows after the vertical pipeline pump is started; if the first similarity calculated for Y consecutive times shows a monotonically decreasing trend and the absolute value of the decreasing slope is greater than a first slope threshold, calculate the second similarity between the first vector and a preset feature vector; wherein the preset feature vectors are constructed based on the acoustic features corresponding to X consecutive first time windows when the bearing assembly is in a poor lubrication state; and generate a quantitative lubrication command and an oil discharge command when the first similarity is less than the first similarity threshold and the second similarity is greater than the second similarity threshold.

[0007] In one possible implementation, the envelope demodulation spectrum features include the theoretical fault frequency of the bearing outer ring and the amplitude at each harmonic, the total energy of the target characteristic frequency band, and the total energy of the reference frequency band.

[0008] The formula for calculating the bearing characteristic evaluation index is as follows:

[0009] ;

[0010] in, I Indicates the bearing characteristic evaluation index, V V represents the total order of the target characteristic frequencies, where n is the order of the target characteristic frequencies, 0 < n ≤ V, and the nth order target characteristic frequency. f n It is a harmonic frequency determined based on the theoretical fault frequency of the bearing outer ring; A ( f n ) represents the amplitude corresponding to the characteristic frequency of the nth target; E f This represents the total energy of the target's characteristic frequency band; E b M represents the total energy of the reference frequency band; M is the total frequency order of the entire spectrum of the mechanical vibration signal.

[0011] In one possible implementation, the edge computing node is further configured to: generate a warning signal, a quantitative lubrication command, and an oil discharge command if at least two of the second judgment conditions are met when the classification result is early wear; wherein the second judgment conditions include: the time domain feature is greater than a second time domain threshold; the acoustic feature is greater than a second acoustic threshold; and the bearing feature evaluation index is greater than or equal to a first index threshold and the frequency domain feature is greater than a second frequency domain threshold; the second time domain threshold is greater than the first time domain threshold, the second acoustic threshold is greater than the first acoustic threshold, and the second frequency domain threshold is greater than the first frequency domain threshold.

[0012] In one possible implementation, the edge computing node is further configured to generate an alarm signal if at least two of the third judgment conditions are met when the classification result is a bearing fault; wherein the third judgment conditions include: the time-domain feature is greater than a third time-domain threshold; the acoustic feature is greater than a third acoustic threshold; and the bearing feature evaluation index is greater than a second exponential threshold and the frequency-domain feature is greater than a third frequency-domain threshold; wherein the third time-domain threshold is greater than the second time-domain threshold, the third acoustic threshold is greater than the second acoustic threshold, the second exponential threshold is greater than the first exponential threshold, and the third frequency-domain threshold is greater than the second frequency-domain threshold.

[0013] In one possible implementation, the time-domain feature is the effective value of vibration velocity or the kurtosis index; the frequency-domain feature is the energy proportion of the target high-frequency band; and the acoustic feature is the energy proportion of the target frequency band or the acoustic kurtosis.

[0014] In one possible implementation, the system further includes a model training module configured to: collect sound signals of the vertical pipeline pump when it is delivering various different fluids under multiple combined operating conditions to obtain first training data; and collect sound signals of the vertical pipeline pump when it is not delivering fluid under multiple combined operating conditions to obtain second training data; and use the training dataset to perform supervised training on the initial deep neural network model to obtain a pre-trained model; the training dataset includes at least the first training data and the second training data.

[0015] In one possible implementation, the combined operating conditions are the Cartesian product of the first bearing operating condition set and the second bearing operating condition set; wherein, the first bearing operating condition set consists of the Cartesian product of m lubrication levels and n mechanical states of the first bearing, and the second bearing operating condition set consists of the Cartesian product of m lubrication levels and n mechanical states of the second bearing; different lubrication levels correspond to different percentage levels of lubricating grease filling, and different mechanical states correspond to the degradation levels of the first or second bearing from a normal state to a fault state; the combined operating conditions include a total of (m×n). 2 kind.

[0016] In one possible implementation, the vertical inline pump also includes a skeleton oil seal located at the junction of the pump shaft and the bearing assembly.

[0017] Secondly, this application provides an operation monitoring method for a vertical pipeline pump, which is applied to a vertical pipeline pump. The vertical pipeline pump includes at least a pump shaft and a bearing assembly. The bearing assembly includes a bearing housing, a first bearing and a second bearing disposed inside the bearing housing. The first bearing and the second bearing are spaced apart along the axial direction. An upper bearing retainer is disposed on the first bearing closer to the second bearing, and a lower bearing retainer is disposed on the second bearing farther from the first bearing. The bearing housing is provided with a lubrication hole and a first oil drain hole in parallel. The lubrication hole is located between the first bearing and the second bearing, and the first oil drain hole is located on the side of the lower bearing retainer. The lower bearing retainer is provided with a second oil drain hole, and the first oil drain hole and the second oil drain hole are interconnected.

[0018] The method includes: acquiring mechanical vibration signals of a bearing assembly at a first preset sampling rate, and acquiring sound signals of the bearing assembly during operation at a second preset sampling rate; extracting features from the mechanical vibration signals within a first time window according to a first preset frequency to obtain time-domain features, frequency-domain features, and envelope demodulation spectrum features; the period length corresponding to the first time window and the first preset frequency is equal; calculating a bearing feature evaluation index using the envelope demodulation spectrum features, the bearing feature evaluation index being used to characterize the degree of deterioration of the mechanical state of the bearing assembly; extracting features from the sound signals within the first time window according to the first preset frequency to obtain acoustic features; and inputting the sound signals within the first time window into a pre-trained model according to the first preset frequency. The pre-trained model is used to classify the operating state of the bearing assembly. The operating state categories include at least normal, poor lubrication, early wear, and bearing failure. If the classification result is poor lubrication, and at least two of the first judgment conditions are met, a quantitative lubrication command and an oil discharge command are generated. The first judgment conditions include that the time domain feature is greater than a first time domain threshold, the acoustic feature is greater than a first acoustic threshold, and the bearing feature evaluation index is less than a first index threshold and the frequency domain feature is greater than a first frequency domain threshold. The quantitative lubrication command is used to drive the power assembly to inject a quantitative amount of lubricating grease into the first and second bearings through the filling hole, and the oil discharge command is used to drive the power assembly to discharge waste lubricating grease through the first and second oil discharge holes.

[0019] As can be seen from the above, this application provides an operation monitoring system and method for a vertical pipeline pump. This system can be applied to the aforementioned vertical pipeline pump and specifically includes: a signal acquisition module configured to: acquire mechanical vibration signals of a bearing assembly at a first preset sampling rate, and acquire sound signals of the bearing assembly during operation at a second preset sampling rate; an edge computing node configured to: extract features from the mechanical vibration signals within a first time window according to a first preset frequency, obtaining time-domain features, frequency-domain features, and envelope demodulation spectrum features; the first time window has the same period length as the first preset frequency; calculate a bearing feature evaluation index using the envelope demodulation spectrum features, the bearing feature evaluation index being used to characterize the degree of mechanical degradation of the bearing assembly; and the edge computing node is further configured to: extract features from the sound signals within the first time window according to the first preset frequency. Acoustic features are obtained; according to a first preset frequency, the sound signal within the first time window is input into the pre-trained model to obtain the classification result of the pre-trained model on the operating state of the bearing assembly; the operating state categories include at least normal, poor lubrication, early wear, and bearing failure; and the edge computing node is also configured to: in the case of poor lubrication, if at least two of the first judgment conditions are met, generate a quantitative lubrication command and an oil discharge command; wherein, the first judgment conditions include the time domain feature being greater than a first time domain threshold, the acoustic feature being greater than a first acoustic threshold, and the bearing feature evaluation index being less than a first index threshold and the frequency domain feature being greater than a first frequency domain threshold; the quantitative lubrication command is used to drive the power assembly to inject a quantitative amount of lubricating grease into the first bearing and the second bearing through the oil filling hole, and the oil discharge command is used to drive the power assembly to discharge waste lubricating grease through the first oil discharge hole and the second oil discharge hole. As can be seen, the vertical pipeline pump operation monitoring system provided in this application, through the collaborative work of the signal acquisition module and the edge computing node, realizes multi-dimensional, high-frequency acquisition and feature extraction of mechanical vibration signals and sound signals of bearing components. Combined with the intelligent classification and multi-condition threshold verification logic of the pre-trained model, it can not only accurately identify different operating states of the bearing, such as normal, poor lubrication, early wear, and faults, but also output targeted quantitative lubrication and oil discharge commands for poor lubrication states. This effectively avoids misjudgments caused by single signal features or model classification, and significantly improves the accuracy of condition monitoring and the pertinence of maintenance decisions. At the same time, the localized processing mode of the edge computing node ensures the real-time nature of monitoring and command generation, enabling rapid response to abnormal bearing states. By intervening in poor lubrication problems in advance, it can delay the bearing deterioration process, reduce the risk of downtime due to failure, and extend the service life of the equipment. Furthermore, the high-dimensional feature learning capability of the pre-trained model, based on refined joint operating condition training, can adapt to the monitoring needs of different fluid types and operating conditions, providing reliable support for the intelligent and routine operation monitoring of vertical pipeline pumps, reducing operation and maintenance costs, and improving overall operating efficiency. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the structure of a vertical pipeline pump provided in an embodiment of this application;

[0021] Figure 2 A schematic diagram of the bearing housing, upper bearing retainer, and lower bearing retainer provided in an embodiment of this application;

[0022] Figure 3 A schematic diagram of the operation monitoring system for a vertical pipeline pump provided in an embodiment of this application;

[0023] Figure 4 A flowchart illustrating the operation monitoring method for a vertical pipeline pump provided in an embodiment of this application;

[0024] Figure 5 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application;

[0025] Among them, 100-pump shaft; 200-bearing assembly; 201-bearing housing; 202-first bearing; 203-second bearing; 204-upper bearing retainer; 205-lower bearing retainer; 206-oil filling hole; 207-first oil drain hole; 208-second oil drain hole; 301-flushing pipe; 302-pump cover; 303-impeller; 304-impeller nut; 305-pump body; 306-bearing gland; 307-motor frame; 308-mechanical seal; 309-skeletal oil seal; 1001-signal acquisition module; 1002-edge computing node; 1003-model training module. Detailed Implementation

[0026] To enable those skilled in the art to better understand the technical solutions in this application, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of this application.

[0027] Figure 1 This is a schematic diagram of the structure of a vertical pipeline pump provided in an embodiment of this application.

[0028] like Figure 1 As shown, this application embodiment provides a vertical pipeline pump, which includes at least a pump shaft 100 and a bearing assembly 200. The bearing assembly 200 can be sleeved on the outside of the pump shaft 100 to provide stable radial constraint and axial positioning for the high-speed rotation of the pump shaft 100.

[0029] Figure 2This is a schematic diagram of the structure of the bearing housing, upper bearing retainer and lower bearing retainer provided in the embodiments of this application.

[0030] Furthermore, Figure 2 Figure (a) shows the structure of the bearing housing 201, as follows: Figure 1 and Figure 2 As shown in (a), the bearing assembly 200 includes a bearing housing 201, a first bearing 202 and a second bearing 203 disposed inside the bearing housing 201, the first bearing 202 and the second bearing 203 being spaced apart along the axial direction. Further, Figure 2 Figure (b) shows the structure of the upper bearing retainer 204, as follows: Figure 1 and Figure 2 As shown in (b), an upper bearing retainer 204 is provided on the first bearing 202 in the direction close to the second bearing 203. Figure 2 Image (c) shows the structure of the lower bearing retainer 205, as follows: Figure 1 and Figure 2 As shown in (c), the second bearing 203 is provided with a lower bearing retainer 205 in the direction away from the first bearing 202.

[0031] It is worth noting that, because the lubricating grease inside the bearing can fall into the lower part of the bearing as the pump operates, this can lead to poor lubrication, increased temperature, and shortened bearing life. The first bearing 202 provided in this embodiment is equipped with an upper bearing retainer 204, which prevents the lubricating grease from slipping out of the first bearing 202, thus extending its service life. Similarly, the second bearing 203 is equipped with a lower bearing retainer 205, which also prevents the lubricating grease from slipping out of the second bearing 203, extending its service life. Therefore, the vertical pipeline pump provided in this embodiment can avoid frequent lubrication, saving manpower and resources.

[0032] Further details can be found by referring to [link / reference]. Figure 1 and Figure 2 The bearing housing 201 has a lubrication hole 206 and a first oil drain hole 207 arranged side by side. The lubrication hole 206 is located between the first bearing 202 and the second bearing 203, and the first oil drain hole 207 is located on the side of the lower bearing retainer 205. The lower bearing retainer 205 has a second oil drain hole 208, and the first oil drain hole 207 and the second oil drain hole 208 are interconnected. Therefore, when the lubricating grease in the first bearing 202 or the second bearing 203 is insufficient, lubricating grease can be added through the lubrication hole 206. Furthermore, if there is too much lubricating grease in the lower part, the waste lubricating grease can also be discharged through the first oil drain hole 207 and the second oil drain hole 208 in this embodiment.

[0033] See also Figure 1The vertical pipeline pump provided in this application embodiment may further include: a flushing pipe 301, a pump cover 302, an impeller 303, an impeller nut 304, a pump body 305, a bearing cap 306, a motor frame 307, and a mechanical seal 308; wherein, the impeller 303 is sleeved on the lower end of the pump shaft 100 and axially locked by the impeller nut 304 to ensure that the pump shaft 100 rotates synchronously with the impeller 303 to achieve fluid transportation; the pump cover 302 and the pump body 305 are sealed together to form the pump's flow cavity, and the impeller 303 is housed inside the flow cavity; the mechanical seal 308 is disposed between the pump shaft 100 and the pump cover 302. At the junction of 2, it is used to block the fluid in the flow cavity from leaking outward along the pump shaft 100; one end of the flushing pipe 301 is connected to the high-pressure fluid area of ​​the pump body 305, and the other end is connected to the sealing cavity of the mechanical seal 308. By introducing high-pressure fluid, the mechanical seal 308 is flushed and cooled to improve the sealing reliability; the bearing cover 306 is closed on the end of the bearing seat 201 to restrict the axial movement of the first bearing 202, and at the same time to provide dustproof sealing for the inside of the bearing seat 201; the motor frame 307 is fixedly connected to the upper end of the bearing seat 201 to support the motor assembly that drives the pump shaft 100 to rotate, so as to realize the coaxial assembly of the motor and the pump body.

[0034] Further details can be found by referring to [link / reference]. Figure 1 The vertical pipeline pump provided in this application embodiment also includes a skeleton oil seal 309, which is disposed at the junction of the pump shaft 100 and the bearing assembly 200, specifically at the through-hole position where the pump shaft 100 passes through the bearing housing 201. This skeleton oil seal 309 replaces the dust cover structure, eliminating the potential for friction between rotating and stationary parts caused by loose set screws leading to the dust cover sinking and fitting against the bearing cap 306, thus improving operational safety.

[0035] Furthermore, to achieve precise monitoring of the operation of vertical pipeline pumps and avoid misjudgments or omissions, this application provides a vertical pipeline pump operation monitoring system and method. This system and method can be widely applied to vertical pipeline pump equipment in petrochemical, municipal water supply, nuclear power plant cooling loops, and large data center cooling pumps. Specifically, it can be deployed locally on the pump unit or remotely in the cloud. By real-time acquisition of vibration-acoustic multimodal signals and combining them with a pre-trained model, it progressively judges four states: normal, poor lubrication, early wear, and bearing failure, achieving precise monitoring of the vertical pipeline pump. This effectively avoids misjudgments and omissions, significantly reduces the probability of unplanned equipment downtime, extends bearing life, and saves on maintenance manpower and material costs.

[0036] Figure 3 This is a schematic diagram of the operation monitoring system for a vertical pipeline pump provided in an embodiment of this application.

[0037] like Figure 3As shown, the operation monitoring system for the vertical pipeline pump provided in this application embodiment may include:

[0038] The signal acquisition module 1001 is configured to perform the following steps S100.

[0039] S100: Acquire mechanical vibration signals of bearing assembly 200 at a first preset sampling rate, and acquire sound signals of bearing assembly 200 during operation at a second preset sampling rate.

[0040] In this embodiment, a vibration sensor can be installed inside or outside the vertical pipeline pump to collect mechanical vibration signals from the bearing assembly 200. The vibration sensor can be a vibration velocity sensor or a piezoelectric accelerometer; this embodiment does not specifically limit the type. Furthermore, the vibration sensor can be specifically installed on the inner or outer wall of the bearing housing 201. The first preset sampling rate can be 50kHz to 100kHz to ensure complete capture of high-frequency vibration signals generated by abnormal bearing impacts.

[0041] Furthermore, in this embodiment, an acoustic sensor can be placed near the bearing assembly 200 to collect sound signals during the operation of the bearing assembly 200. Specifically, the acoustic sensor can be a piezoelectric acoustic sensor, a capacitive measuring microphone, or an acoustic array sensor; this embodiment does not impose specific limitations on this. The second preset sampling rate can be 44.1kHz or 48kHz, which can accurately collect sound signals caused by friction and impact during bearing operation.

[0042] Furthermore, the vertical pipeline pump operation monitoring system provided in this application embodiment may also include an edge computing node 1002. The edge computing node 1002 is a dedicated computing device deployed near the device, specifically an embedded artificial intelligence (AI) computing board, a micro industrial personal computer (IPC), etc.

[0043] In some implementations, embodiments of this application may also employ a cloud-edge collaborative architecture, utilizing edge computing nodes 1002 on the edge side to achieve real-time processing, and utilizing cloud servers on the cloud side to achieve model training, data management, etc.

[0044] Edge computing node 1002 can be configured to perform the following steps S201-S202.

[0045] S201: According to the first preset frequency, the mechanical vibration signal within the first time window is subjected to feature extraction to obtain time domain features, frequency domain features and envelope demodulation spectrum features; the period length corresponding to the first time window and the first preset frequency is equal.

[0046] The first preset frequency can be 30 seconds / time, 1 minute / time, 3 minutes / time, 5 minutes / time, or 10 minutes / time, and can be designed based on actual needs. This application embodiment does not specifically limit this. It can be understood that when the first preset frequency is 30 seconds / time, the corresponding period length is 30 seconds, the length of the first time window is 30 seconds, and so on, which will not be elaborated here.

[0047] Furthermore, the time-domain characteristics of mechanical vibration signals can reflect the overall level and transient impact characteristics of the vibration, specifically the effective value of vibration velocity or the kurtosis index. The effective value of vibration velocity is a statistical measure obtained by taking the square root of the squared average of the mechanical vibration velocity signals within the first time window. The calculation formula is as follows:

[0048] ;

[0049] in, V rms Indicates the effective value of vibration velocity. T Indicates the length of the first time window. v ( t () indicates instantaneous vibration velocity.

[0050] It should be noted that when the vibration sensor is a vibration velocity sensor, it can capture the velocity changes of the bearing vibration and output the instantaneous vibration velocity value at each moment in real time (instantaneous vibration velocity). When the vibration velocity sensor is an acceleration sensor, it can collect the instantaneous data of vibration acceleration, and then obtain the instantaneous vibration velocity value (instantaneous vibration velocity) through integration calculation (converting the acceleration data into velocity data).

[0051] The magnitude of the effective value of vibration velocity directly characterizes the overall intensity of vibration of the bearing assembly 200. The larger the effective value of vibration velocity, the worse the smoothness of bearing operation, and there may be problems such as wear, loosening or poor lubrication.

[0052] The kurtosis index is a dimensionless parameter describing the steepness of the probability density distribution of a vibration signal. The calculation formula is:

[0053] ;

[0054] in, K α Indicates kurtosis index, x w These are vibration acceleration sample values. μ x The average of the sampled values. s Standard deviation W The number of sampling points, 0 < w ≤W .

[0055] Under normal operating conditions, the kurtosis value of the mechanical vibration signal is close to 3. When the bearing has local defects such as pitting or cracks, the kurtosis value will increase significantly (usually > 5), which can realize the accurate identification of early bearing failures.

[0056] Furthermore, embodiments of this application can combine preset time-domain thresholds to measure time-domain characteristics. Specifically, a first time-domain threshold, a second time-domain threshold, and a third time-domain threshold can be set, where the second time-domain threshold is greater than the first time-domain threshold, and the third time-domain threshold is greater than the second time-domain threshold.

[0057] For example, when the time-domain characteristic is the effective value of vibration velocity, the first time-domain threshold can be equal to 1.8 mm / s, the second time-domain threshold can be equal to 2.8 mm / s, and the third time-domain threshold can be equal to 4.5 mm / s. The corresponding state judgment logic is as follows: when the effective value of vibration velocity is < 1.8 mm / s, it indicates that the bearing is running smoothly without obvious abnormalities; when 1.8 mm / s ≤ effective value of vibration velocity < 2.8 mm / s, it indicates that the bearing has slight vibration abnormalities (such as poor lubrication or early wear); when 2.8 mm / s ≤ effective value of vibration velocity < 4.5 mm / s, it indicates that the bearing vibration abnormalities are aggravated (such as wear development or component loosening); when the effective value of vibration velocity is ≥ 4.5 mm / s, it indicates that the bearing has serious faults (such as rolling element jamming or outer ring cracks).

[0058] When the time-domain feature is a kurtosis index, the first time-domain threshold can be equal to 3, the second time-domain threshold can be equal to 5, and the third time-domain threshold can be equal to 8. The corresponding state judgment logic is as follows: when the kurtosis index < 3, it indicates that the bearing vibration signal distribution is stable and in a normal state; when 3 ≤ kurtosis index < 5, it indicates that the bearing has slight local defects (such as early pitting or minor wear); when 5 ≤ kurtosis index < 8, it indicates that the bearing defects have further developed (such as pitting expansion or rolling element damage); when the kurtosis index ≥ 8, it indicates that the bearing has serious local faults.

[0059] The specific values ​​of the first time domain threshold, the second time domain threshold, and the third time domain threshold can be adjusted based on the actual situation, and the embodiments of this application do not impose specific limitations on this.

[0060] Furthermore, frequency domain characteristics are the frequency domain features obtained after performing a Fast Fourier Transform (FFT) on the mechanical vibration signal. Specifically, they can include characteristic frequency amplitude, target high-frequency band energy proportion, and harmonic distortion rate: characteristic frequency amplitude refers to the spectral peak value corresponding to the theoretical fault frequency of key components such as the outer ring, inner ring, and rolling elements of the bearing, which is directly related to the deterioration state of specific components; target high-frequency band energy proportion is the ratio of the energy of the preset fault-sensitive frequency band (such as 2000-5000Hz) to the energy of the entire frequency band, which can highlight the contribution of the fault characteristic frequency; harmonic distortion rate is used to measure the energy proportion of the harmonic components of the characteristic frequency, reflecting the severity of the fault development.

[0061] Furthermore, embodiments of this application can combine preset frequency domain thresholds to measure frequency domain characteristics. Specifically, a first frequency domain threshold, a second frequency domain threshold, and a third frequency domain threshold can be set, where the second frequency domain threshold is greater than the first frequency domain threshold, and the third frequency domain threshold is greater than the second frequency domain threshold.

[0062] For example, when the frequency domain characteristic is the proportion of energy in the target high-frequency band, the first frequency domain threshold can be equal to 10%, the second frequency domain threshold can be equal to 20%, and the third frequency domain threshold can be equal to 30%. The corresponding state judgment logic is as follows: if the proportion of energy in the target high-frequency band is less than the first frequency domain threshold (10%), it indicates that the bearing vibration energy is mainly concentrated in the low-frequency band, the operating state is stable, and there is no obvious abnormality; if the first frequency domain threshold (10%) is less than the proportion of energy in the target high-frequency band and less than the second frequency domain threshold (20%), it indicates that the bearing has a slight abnormality (such as lubricating grease consumption, slight wear of rolling elements), and the high-frequency vibration energy begins to rise; if the second frequency domain threshold (20%) is less than the proportion of energy in the target high-frequency band and less than the third frequency domain threshold (30%), it indicates that the abnormal state of the bearing is aggravated (such as wear expansion, pitting development of the outer ring), and the high-frequency characteristic frequency energy increases significantly; if the proportion of energy in the target high-frequency band is greater than or equal to the third frequency domain threshold (30%), it indicates that the bearing has a serious fault (such as rolling element jamming, inner ring crack), the high-frequency impact signal dominates, and an emergency shutdown is required for investigation.

[0063] Furthermore, the envelope demodulation spectrum characteristics are obtained by performing a Fourier transform on the mechanical vibration signal after high-pass filtering and envelope detection. Specifically, they include the theoretical fault frequency of the bearing outer ring and the amplitude of its harmonics at each order (such as 1st to 3rd order), the total energy of the target characteristic frequency band, and the total energy of the reference frequency band.

[0064] The theoretical failure frequency of the bearing outer ring is calculated from the bearing structural parameters (number of rolling elements, pitch circle diameter, rolling element diameter, contact angle) and the pump shaft speed. Its harmonic amplitude can reflect the impact intensity of the outer ring defect. In practical applications, the embodiments of this application can calculate the theoretical failure frequencies of the bearing outer ring of the first bearing 202 and the second bearing 203 respectively.

[0065] For example, the formula for calculating the theoretical failure frequency of the bearing outer ring is:

[0066] ;

[0067] in, f oi This indicates the theoretical failure frequency of the bearing outer ring. i =1 indicates that the first bearing is 202. i =2 indicates the second bearing 203; f r Indicates the bearing rotational speed frequency. Z i Indicates the first i The number of rolling elements in a bearing. d i Indicates the first i The diameter of the rolling elements of the bearing, D pi Indicates the first i The pitch circle diameter of the bearing. α i Indicates the first i The contact angle of the bearing.

[0068] S202: Calculate the bearing characteristic evaluation index using envelope demodulation spectrum features. The bearing characteristic evaluation index is used to characterize the degree of deterioration of the mechanical condition of the bearing assembly.

[0069] The formula for calculating the bearing characteristic evaluation index is as follows:

[0070] ;

[0071] in, I Indicates the bearing characteristic evaluation index, V V represents the total order of the target characteristic frequencies, where n is the order of the target characteristic frequencies, 0 < n ≤ V, and the nth order target characteristic frequency. f n It is a harmonic frequency determined based on the theoretical fault frequency of the bearing outer ring; A ( f n ) represents the amplitude corresponding to the characteristic frequency of the nth target; E f This represents the total energy of the target's characteristic frequency band; E b M represents the total energy of the reference frequency band; M is the total frequency order of the entire spectrum of the mechanical vibration signal.

[0072] The following section provides a detailed explanation of the calculation process for the bearing characteristic evaluation index, using the case where the first bearing 202 and the second bearing 203 are of the same model and have a total harmonic order of 3 as an example.

[0073] Since the first bearing 202 and the second bearing 203 are of the same model, and their number of rolling elements, rolling element diameter, pitch circle diameter, and contact angle are completely identical, and they share the same pump shaft speed, their theoretical outer ring failure frequencies are equal. f o1 = f o2 = f o When the total harmonic order is 3, the set of target characteristic frequencies can be determined as { f o ,2 f o ,3 f o}, respectively corresponding to the first harmonic ( f o ), 2nd harmonic (2 f o ), 3rd harmonic (3 f o ).

[0074] The specific calculation steps for the bearing characteristic evaluation index are as follows:

[0075] First, extract the amplitude of the target feature frequencies at each order. A ( f n )|.

[0076] The mechanical vibration signals acquired within the first time window are sequentially subjected to high-pass filtering (to remove low-frequency interference below 500Hz), envelope detection (to extract the signal envelope of abnormal bearing impact), and Fourier transform (to convert the time-domain envelope signal into a frequency-domain signal) to obtain a smoothed envelope demodulation spectrum. The target characteristic frequencies corresponding to n=1, n=2, and n=3 are then located within this spectrum.

[0077] When n=1, the target characteristic frequency is f o Read the spectral line height corresponding to that frequency on the envelope demodulated spectrum, which is | A ( f 1)|, which reflects the impact strength of the outer ring foundation failure of the first bearing 202 and the second bearing 203;

[0078] When n=2, the target characteristic frequency is 2. f o Read the spectral line height corresponding to that frequency on the envelope demodulated spectrum, which is | A ( f 2)|, which reflects the intensity of the second harmonic generated by the fault impact. The larger the value, the more obvious the fault development.

[0079] When n=3, the target characteristic frequency is 3.f o Read the spectral line height corresponding to that frequency on the envelope demodulated spectrum, which is | A ( f 3)|, which reflects the intensity of the third harmonic of the fault impact, will only show a significant amplitude when the bearing has severe wear and pitting.

[0080] For example, | A ( f 1) |=0.8m / s 2 、| A ( f 2) |=0.5m / s 2 、| A ( f 2) |=0.3m / s 2 .

[0081] The second step is to calculate the sum of the target characteristic frequency amplitudes. .

[0082] The amplitudes of the third-order target characteristic frequencies are summed sequentially, i.e.:

[0083] =| A ( f 1)|+| A ( f 2)|+| A ( f 3) |=0.8+0.5+0.3=1.6m / s 2 .

[0084] The result represents the sum of the combined impact strengths of the defects in the outer rings of the first bearing 202 and the second bearing 203. The value directly reflects the overall basic level of deterioration of the outer rings of the two bearings.

[0085] The third step is to calculate the logarithm of the band power ratio. .

[0086] The total energy of the target characteristic frequency band is the sum of the envelope spectrum energy within a preset fault-sensitive frequency band (e.g., 2000-5000Hz), which is the sum of the squares of the amplitudes of all spectral lines within the 2000-5000Hz band. The total energy of the reference frequency band is the sum of the envelope spectrum energy within a stable frequency band (e.g., 500-1000Hz), which is the sum of the squares of the amplitudes of all spectral lines within the 500-1000Hz band. The ratio of these two values ​​can eliminate environmental noise interference and improve the stability of the characteristic.

[0087] For example, E f =120 (m / s) 2 ) 2 , E b=15 (m / s) 2 ) 2 ,but =120÷15=8. Then perform a logarithmic transformation, log... 10 8 ≈0.903.

[0088] The fourth step is to determine the average order proportion coefficient. .

[0089] The total order M of the entire spectrum is determined by the sampling rate of the mechanical vibration signal and the duration of the first time window: assuming the first preset sampling rate is 50kHz and the first time window duration is 1 minute (60s), then the total order M of the entire spectrum = 50000Hz × 60s = 3 × 10 6 (That is, the entire spectrum contains a total of 3×10) 6 (Spectral lines).

[0090] Therefore, the average order proportion coefficient is calculated as follows: ≈6.67×10 -7 .

[0091] It is worth noting that the average order proportion coefficient can be used as a weight to reflect the contribution of high-order characteristics. The higher the total order and the more prominent the proportion of high-order characteristics, the larger the coefficient, which can accurately distinguish between "poor lubrication" (mainly low-order) and "bearing mechanical failure" (prominent high-order).

[0092] Step 5: Calculate the bearing characteristic evaluation index I .

[0093] Substituting the above calculation results into the formula, we get:

[0094] =1.6×0.903×6.67×10 -7 ≈9.63×10 -7 .

[0095] Furthermore, in this embodiment, the bearing characteristic evaluation index can be measured in conjunction with a preset index threshold. Specifically, a first index threshold and a second index threshold can be set, with the second index threshold being greater than the first index threshold.

[0096] For example, the first exponential threshold can be equal to 2 × 10 -6 The second exponential threshold can be equal to 1 × 10. -5 The specific values ​​of the first and second exponential thresholds can be adjusted based on actual circumstances, and this application embodiment does not impose specific limitations on them.

[0097] Furthermore, if the bearing characteristic evaluation index I < First exponential threshold (2×10) -6This can determine whether the bearing assembly 200 is in a normal state or a low-deterioration state corresponding to poor lubrication. If the first index threshold ≤ I <Second exponential threshold (1×10) -5 This indicates that bearing assembly 200 is in an early wear state (such as initial pitting on the outer ring or slight wear of the rolling elements). If I >The second index threshold can determine that the bearing assembly 200 is in a serious fault state (such as outer ring cracks or rolling element jamming).

[0098] As can be seen, the embodiments of this application can utilize bearing characteristic evaluation indices. I Characterizing the degree of mechanical degradation of bearing assembly 200, bearing characteristic evaluation index. I The larger the size, the greater the degree of degradation.

[0099] In addition, edge computing node 1002 is also configured to perform the following steps S203-S204.

[0100] S203: Extract features from the sound signal within the first time window according to the first preset frequency to obtain acoustic features.

[0101] Among them, acoustic characteristics are the target frequency band energy ratio or acoustic kurtosis.

[0102] Specifically, the target frequency band energy ratio refers to the ratio of the energy of the sound signal within the preset target frequency band to the energy of the full frequency band, used to highlight the characteristic sound signals when the bearing is poorly lubricated or worn. The target frequency band can be 1000-3000Hz. This frequency band is the concentrated area of ​​high-frequency noise caused by friction between the bearing rolling elements and raceways, and lack of lubricating grease, and is less affected by pump flow noise and environmental interference. The specific target frequency band can be determined based on actual conditions, and this application embodiment does not specifically limit it. The full frequency band is the complete frequency range of the sound signal, which can be determined by the second preset sampling rate. For example, if the second preset sampling rate is 44.1kHz, then the full frequency band is 0-22.05kHz.

[0103] The formula for calculating the target frequency band energy percentage is:

[0104] η = E l / E total ;

[0105] in, η This indicates the energy percentage of the target frequency band. E l The signal energy within the target frequency band can be obtained by integrating the square of the instantaneous value of the sound signal within the target frequency band. Etotal It represents the signal energy across the entire frequency band.

[0106] Understandably, when bearing lubrication is poor, the dry friction between the rolling elements and the raceway intensifies, significantly increasing the noise energy within the target frequency band, thus affecting the energy proportion of the target frequency band. η Enlarged; under normal conditions, η The value remains at a low level (e.g., ≤10%).

[0107] Furthermore, acoustic kurtosis is a dimensionless parameter describing the steepness of the probability density distribution of a sound signal, used to capture instantaneous impact noise caused by defects such as early wear and pitting in bearings. Its calculation formula is:

[0108] ;

[0109] in, K β Indicates acoustic kurtosis, s z The first sound signal within the first time window z Each sample value, μ s The average of the sampled values. σ s The standard deviation of the sampled values. Z This represents the number of sampling points for the sound signal within the first time window, where 0 < z ≤ Z .

[0110] During normal operation, the bearing's sound signal is relatively stable, with an acoustic kurtosis value close to 3. When the bearing exhibits defects such as early wear or rolling element pitting, intermittent impact noise is generated, causing the signal probability density distribution to become steeper and the acoustic kurtosis value to increase significantly (e.g., ≥5). The more severe the defect, the higher the kurtosis value. K β The higher the value, the more accurate the identification of early faults.

[0111] Furthermore, embodiments of this application can combine preset acoustic thresholds to measure acoustic characteristics. Specifically, a first acoustic threshold, a second acoustic threshold, and a third acoustic threshold can be set, where the second acoustic threshold is greater than the first acoustic threshold, and the third acoustic threshold is greater than the second acoustic threshold.

[0112] For example, when the acoustic characteristic is the energy percentage of the target frequency band, the first acoustic threshold can be equal to 10%, the second acoustic threshold can be equal to 20%, and the third acoustic threshold can be equal to 30%. The corresponding state judgment logic is as follows: when the energy percentage of the target frequency band is <10%, it indicates that the sound signal generated by the bearing operation is stable, with no abnormal friction or impact noise; when the energy percentage of the target frequency band is 10% ≤ the energy percentage of the target frequency band is <20%, it indicates that the bearing has slight lubrication problems, and the friction noise between the rolling elements and the raceway is enhanced; when the energy percentage of the target frequency band is 20% ≤ the energy percentage of the target frequency band is <30%, it indicates that the early wear of the bearing is aggravated, and the high-frequency components of the impact noise are significantly increased; when the energy percentage of the target frequency band is ≥30%, it indicates that the bearing has a serious fault, and intermittent impact noise dominates the sound signal.

[0113] For example, when the acoustic feature is acoustic kurtosis, the first acoustic threshold can be equal to 3, the second acoustic threshold can be equal to 5, and the third acoustic threshold can be equal to 8. The corresponding state judgment logic is as follows: when acoustic kurtosis < 3, it indicates that the probability distribution of the bearing sound signal is stable and there is no impact noise caused by local defects; when 3 ≤ acoustic kurtosis < 5, it indicates that the bearing has early pitting or slight wear, and intermittent impacts cause a slight increase in acoustic kurtosis; when 5 ≤ acoustic kurtosis < 8, it indicates that the bearing defects have further developed, and the frequency and intensity of impact noise have increased significantly; when acoustic kurtosis ≥ 8, it indicates that the bearing has serious local faults (such as rolling element breakage or outer ring cracks), and impact noise becomes the main component of the sound signal.

[0114] It is understood that the specific values ​​of the first acoustic threshold, the second acoustic threshold, and the third acoustic threshold can be adjusted based on the actual situation, and the embodiments of this application do not impose specific limitations on this.

[0115] In some implementations, embodiments of this application may preprocess the sound signal before extracting acoustic features: filter out fixed frequency noise (such as the fundamental frequency of motor operation and its harmonics) generated by the operation of the pump body through a band-stop filter, and then perform DC component removal and normalization processing to avoid interfering with the accuracy of the features; the preprocessed signal is used for feature calculation at a first preset frequency (consistent with the frequency of vibration signal feature extraction) to ensure that the feature extraction timing of the two types of signals is synchronized, providing a consistent data basis for subsequent joint judgment.

[0116] S204: Input the sound signal within the first time window into the pre-trained model according to the first preset frequency to obtain the classification result of the pre-trained model on the operating status of the bearing assembly; the operating status categories include at least normal, poor lubrication, early wear and bearing failure.

[0117] Understandably, the model's classification results can include four categories: normal, poor lubrication, early wear, and bearing failure.

[0118] The pre-trained model can be a hybrid model of convolutional neural network (CNN) and long short-term memory network (LSTM) (CNN-LSTM), or a deep residual network (ResNet) or gradient boosting decision tree (GBDT) model. This application does not specifically limit the specific model training process. The model pre-training process will be described in detail below and will not be repeated here.

[0119] For example, after the pre-trained model completes training, this embodiment of the application can compress the size of the pre-trained model through model quantization (such as INT8 quantization), and then deploy the pre-trained model to the edge computing node 1002; further, according to the first preset frequency, the edge computing node 1002 can input the pre-processed sound signal into the pre-trained model in real time, and the pre-trained model outputs the classification probability of various states (such as normal probability 0.05, poor lubrication probability 0.93, early wear probability 0.01, bearing failure probability 0.01), and then selects the category with the highest probability as the final classification result.

[0120] In addition, edge computing node 1002 is also configured to perform the following steps S205-S207.

[0121] S205: If the classification result is poor lubrication, and at least two of the first judgment conditions are met, a quantitative lubrication command and an oil discharge command are generated; wherein, the first judgment conditions include that the time domain feature is greater than the first time domain threshold, the acoustic feature is greater than the first acoustic threshold, and the bearing feature evaluation index is less than the first index threshold and the frequency domain feature is greater than the first frequency domain threshold; the quantitative lubrication command is used to drive the power assembly to inject a quantitative amount of lubricating grease into the first bearing 202 and the second bearing 203 through the oil filling hole 206, and the oil discharge command is used to drive the power assembly to discharge the waste lubricating grease through the first oil discharge hole 207 and the second oil discharge hole 208.

[0122] It is understood that the embodiments of this application provide three first judgment conditions: ① the time domain feature is greater than the first time domain threshold, ② the acoustic feature is greater than the first acoustic threshold, and ③ the bearing feature evaluation index is less than the first index threshold and the frequency domain feature is greater than the first frequency domain threshold.

[0123] For example, for condition ①, when the time domain characteristic is greater than the first time domain threshold: when the effective value of the vibration velocity is greater than 1.8 mm / s or the kurtosis index is greater than 3, it indicates that the bearing impact energy has exceeded the normal fluctuation range, and there may be metal contact caused by oil film rupture.

[0124] For example, for condition ②, the acoustic characteristics are greater than the first acoustic threshold: when the target frequency band energy ratio is greater than 10% or the acoustic kurtosis is greater than 3, it indicates that the high-frequency components of friction noise or impact noise during bearing operation have increased significantly, the friction reduction effect of lubricating grease has failed, and the sound signal generated by direct contact between the metal surface has exceeded the normal threshold, which is highly matched with the acoustic characteristics of poor lubrication.

[0125] For example, for condition ③, the bearing characteristic evaluation index is less than the first index threshold and the frequency domain characteristic is greater than the first frequency domain threshold: when the bearing characteristic evaluation index is less than 2 × 10 -6 When the target high-frequency band energy accounts for more than 10%, it indicates that the bearing vibration energy is concentrated in the fault-sensitive frequency band, but the overall deterioration is low. Based on the classification results, it can be determined that the abnormal local characteristics are caused by poor lubrication, rather than a bearing body failure, and it needs to be restored to normal through lubrication maintenance.

[0126] It should be noted here that, based on the calculation formula of the bearing characteristic evaluation index and the bearing operating principle, when the bearing is poorly lubricated, the lubricating oil film between the rolling elements and the raceway breaks down, and the contact state degenerates from hydrodynamic lubrication to boundary lubrication. Microscopically, this manifests as continuous sliding friction rather than localized geometric impact. Since there is no mechanical damage such as spalling pits or corrosion on the raceway surface, and no periodic impact source, the envelope spectrum at the target characteristic frequency... f o There are no impact peaks at the harmonics, and at this time the bearing characteristic evaluation index calculation formula " "Approaching zero, the overall calculation results indicate that the bearing is in a healthy state with only a slight background impact, and the value of the bearing characteristic evaluation index is lower than the first index threshold."

[0127] Furthermore, when bearing lubrication is poor, dry friction or boundary friction between the rolling elements and raceways generates a high-frequency screeching sound. When the frequency domain characteristic is the target high-frequency energy percentage, the target high-frequency energy percentage increases, exceeding the first frequency domain threshold. At this point, combining the bearing characteristic evaluation index with the target high-frequency energy percentage allows it to be determined that the bearing is in a poor lubrication state, and there is no inherent fault in the bearing itself.

[0128] Based on the model classification results and the first judgment condition mentioned above, the embodiments of this application can output quantitative lubrication commands and oil discharge commands to solve the problem of poor lubrication.

[0129] The quantitative lubrication command is used to drive the power components (such as electric lubrication pumps and quantitative grease injection valve groups) to inject a quantitative amount of lubricating grease into the rolling elements of the first bearing 202 and the inner and outer raceways, and the rolling elements of the second bearing 203 and the inner and outer raceways, according to preset grease injection parameters and through the grease injection holes 206 on the bearing housing 201.

[0130] In some implementations, the grease injection parameters can be determined based on the specific values ​​of the target frequency band energy percentage or the target high-frequency band energy percentage. For example, if the target frequency band energy percentage is within the preset range of [10%, 15%], the amount of grease injected into a single bearing can be 1 / 3 of its lubrication cavity volume (e.g., when the lubrication cavity volume of the first bearing 202 is 10mL, the amount of grease injected is 3.3mL); if the target frequency band energy percentage is within the preset range of [15%, 20%], the amount of grease injected into a single bearing can be 1 / 2 of its lubrication cavity volume (e.g., when the lubrication cavity volume of the first bearing 202 is 10mL, the amount of grease injected is 5mL).

[0131] The oil drain command is used to drive the power components (such as electric oil drain pumps and electromagnetic oil drain valves) to open the first oil drain hole 207 and the second oil drain hole 208 at the bottom of the bearing housing 201, so as to completely drain the waste lubricating grease inside the bearing that has aged and deteriorated due to long-term operation and mixed with impurities (such as metal shavings and dust).

[0132] In practical applications, after the oil draining is completed, a quantitative lubrication command can be initiated to perform grease injection, forming a closed-loop lubrication maintenance process of "draining first and then injecting". This ensures that the newly injected lubricating grease can fully fill the bearing lubrication gap, effectively reduce the frictional loss between the rolling elements and the raceway, and restore the lubrication protection effect of the bearing.

[0133] S206: If the classification result is early wear, and at least two of the second judgment conditions are met, a warning signal, a quantitative lubrication command, and an oil discharge command are generated; wherein, the second judgment conditions include: the time domain feature is greater than the second time domain threshold; the acoustic feature is greater than the second acoustic threshold; and the bearing feature evaluation index is greater than or equal to the first index threshold and the frequency domain feature is greater than the second frequency domain threshold; the second time domain threshold is greater than the first time domain threshold, the second acoustic threshold is greater than the first acoustic threshold, and the second frequency domain threshold is greater than the first frequency domain threshold.

[0134] It is understood that the embodiments of this application provide three second judgment conditions: ① the time domain feature is greater than the second time domain threshold, ② the acoustic feature is greater than the second acoustic threshold, and ③ the bearing feature evaluation index is greater than or equal to the first exponential threshold and the frequency domain feature is greater than the second frequency domain threshold.

[0135] For example, for condition ①, the time domain characteristic is greater than the second time domain threshold: when the effective value of vibration velocity is greater than 2.8 mm / s or the kurtosis index is greater than 5, it indicates that the bearing impact energy has increased significantly, the raceway surface has developed micro-pitting or wear pits, and the rolling elements generate periodic impacts when passing through the defects. The fluctuation amplitude of the time domain signal far exceeds the stage of poor lubrication, which is consistent with the vibration characteristics of early wear.

[0136] For example, for condition ②, the acoustic characteristics are greater than the second acoustic threshold: when the target frequency band energy ratio is greater than 20% or the acoustic kurtosis is greater than 5, it indicates that the high-frequency components of the impact noise during bearing operation are further increased, the metal contact impact generated by early wear will excite more high-frequency harmonics, and the steepness of the sound signal is significantly improved, which is highly matched with the acoustic characteristics of early wear.

[0137] For example, for condition ③, the bearing characteristic evaluation index is greater than or equal to the first index threshold and the frequency domain characteristic is greater than the second frequency domain threshold: when the bearing characteristic evaluation index is greater than or equal to 2 × 10 -6 When the target high-frequency energy accounts for more than 20%, it indicates that the bearing not only has abnormal high-frequency vibration energy, but the overall deterioration has exceeded the threshold range of poor lubrication. Structural damage to the raceway or rolling elements has become the dominant factor. Combined with the classification results, it can be determined as early wear. Lubrication maintenance is needed to slow down the wear development and trigger an early warning to pay attention to changes in the bearing condition.

[0138] It should be noted here that, based on the calculation formula of the bearing characteristic evaluation index and the bearing operating principle, when early wear occurs in a bearing, pitting and wear pits on the raceway surface will form periodic impact sources, and the envelope spectrum at the target characteristic frequency... f o A significant impact peak appears at its harmonics, at which point the bearing characteristic evaluation index calculation formula " "Significantly increased, the overall calculation result exceeded the first exponential threshold (2×10) " -6 Meanwhile, the continuous impact caused by wear will further increase the energy proportion of the target high-frequency band (2000-5000Hz) to more than 20%. The combination of the two can accurately distinguish between early wear and poor lubrication, and avoid misjudging structural damage as a simple lubrication problem.

[0139] S207: If the classification result is a bearing failure, an alarm signal is generated if at least two of the third judgment conditions are met; wherein, the third judgment conditions include: the time domain feature is greater than the third time domain threshold; the acoustic feature is greater than the third acoustic threshold; and the bearing feature evaluation index is greater than the second exponential threshold and the frequency domain feature is greater than the third frequency domain threshold; wherein, the third time domain threshold is greater than the second time domain threshold, the third acoustic threshold is greater than the second acoustic threshold, the second exponential threshold is greater than the first exponential threshold, and the third frequency domain threshold is greater than the second frequency domain threshold.

[0140] It is understood that the embodiments of this application provide three third judgment conditions: ① the time domain feature is greater than the third time domain threshold, ② the acoustic feature is greater than the third acoustic threshold, and ③ the bearing feature evaluation index is greater than the second index threshold and the frequency domain feature is greater than the third frequency domain threshold.

[0141] For example, for condition ①, the time domain characteristic is greater than the third time domain threshold: when the effective value of vibration velocity is greater than 4.5 mm / s or the kurtosis index is greater than 8, it indicates that the bearing impact energy has reached a dangerous level. Serious faults such as raceway cracks and rolling element breakage will cause violent periodic impacts. The peak value and fluctuation amplitude of the time domain signal far exceed the early wear stage, which has threatened the safe operation of the equipment.

[0142] For example, for condition ②, the acoustic characteristics are greater than the third acoustic threshold: when the target frequency band energy ratio is greater than 30% or the acoustic kurtosis is greater than 8, it indicates that strong impact noise occurs during the operation of the bearing. The metal collision caused by the serious fault will excite the high frequency signal of the wide frequency band, and the abnormality of the acoustic characteristics reaches the peak, which is highly consistent with the acoustic performance of the bearing fault.

[0143] For example, for condition ③, the bearing characteristic evaluation index is greater than the second index threshold and the frequency domain characteristic is greater than the third frequency domain threshold: when the bearing characteristic evaluation index is greater than 1×10 -5 When the target high-frequency band energy accounts for more than 30%, it indicates that the structural damage of the bearing has entered a serious stage. The fault impact dominates the vibration and sound signals, and the overall deterioration exceeds the safety threshold. Based on the classification results, it can be determined as a bearing fault. An alarm signal should be triggered immediately and shutdown and maintenance measures should be taken to prevent the fault from expanding and causing equipment damage.

[0144] It should be noted here that, based on the calculation formula of the bearing characteristic evaluation index and the bearing operating principle, when a bearing experiences a serious failure, defects such as cracks and fragmentation will generate a strong and continuous impact, and the envelope spectrum will be affected at the target characteristic frequency. f o The height of the impact peak at its harmonics increases significantly. "Reaching its maximum value, the bearing characteristic evaluation index exceeds the second index threshold (1×10⁻⁶)." -5 Meanwhile, the impact of a fault can cause the energy ratio of the target high-frequency band to surge to over 30%. The combination of these two factors can accurately identify severe fault conditions and provide a reliable basis for emergency alarms and shutdown decisions.

[0145] It should also be noted that in this embodiment, all features (time-domain features, acoustic features, frequency-domain features, and bearing feature evaluation index) are collected and calculated based on the same first time window. Furthermore, this embodiment can assign a unique identifier (Window_ID) to each window to ensure that each step is executed only for feature data under the same Window_ID, thus avoiding data confusion across windows.

[0146] In summary, the embodiments of this application can combine model classification results with multi-dimensional feature threshold determination to construct a closed-loop monitoring logic of "classification-verification-response". This not only enables rapid identification of bearing status through pre-trained models, but also avoids misjudgment through joint verification of time-domain, acoustic, and frequency-domain features. At the same time, it outputs differentiated instructions for different states, which greatly improves the accuracy and timeliness of bearing maintenance, effectively extends equipment life, and reduces the risk of downtime due to failure.

[0147] Furthermore, in the vertical pipeline pump operation monitoring system provided in this application embodiment, the edge computing node 1002 can also be configured to perform the following steps S301-S304.

[0148] S301: Real-time acquisition of acoustic features corresponding to X consecutive first time windows to construct the first vector.

[0149] This application embodiment can construct the acoustic feature vector of the current state through a sliding window mechanism. Specifically, it acquires the acoustic features corresponding to X consecutive first time windows before the current moment in real time. For each window, it extracts two indicators: the target frequency band energy ratio and acoustic kurtosis, and constructs a first vector with a dimension of 2X. For example, when X equals 3, the first vector is [η a (t-2), K a (t-2), η a (t-1), K a (t-1), η a (t), K a [(t)], a total of 6 elements.

[0150] S302: Calculate the first similarity between the first vector and the historical feature vector; wherein, the historical feature vector is constructed from the acoustic features corresponding to the first to the Xth first time windows after the vertical pipeline pump is started.

[0151] In this embodiment, the historical feature vector can characterize the state of the bearing assembly 200 after the vertical pipeline pump is started. By calculating the similarity between the first vector and the historical feature vector, the lubrication state and / or deterioration state of the bearing assembly 200 can be determined. Further, this embodiment specifically calculates the cosine similarity between the first vector and the historical feature vector as the first similarity.

[0152] It is understandable that the first vector has the same structure as the historical feature vectors.

[0153] S303: If the first similarity calculated Y times consecutively shows a monotonically decreasing trend and the absolute value of the decreasing slope is greater than the first slope threshold, calculate the second similarity between the first vector and the preset feature vector; wherein, the preset feature vector is constructed based on the acoustic features corresponding to X consecutive first time windows when the bearing assembly 200 is in a poor lubrication state.

[0154] For example, the value of Y can be equal to 3, and the first slope threshold can be 0.02 or 0.05. This application embodiment does not specifically limit this.

[0155] If the first similarity calculated Y times consecutively shows a monotonically decreasing trend and the absolute value of the decreasing slope is greater than the first slope threshold, it indicates that the deviation between the current acoustic features and the baseline state continues to expand and the expansion rate is significant. Further verification of the cause of the deviation is required. At this time, the second similarity between the first vector and the preset feature vector is calculated.

[0156] Furthermore, the preset feature vector is a reference vector constructed from the acoustic features corresponding to X consecutive first time windows of the bearing assembly 200 under typical poor lubrication conditions, collected through a large number of experiments on poor lubrication conditions (its structure is consistent with the first vector). The second similarity can also be achieved using the cosine similarity algorithm to determine the degree of matching between the current acoustic feature and the poor lubrication feature.

[0157] Furthermore, the calculation steps for the absolute value of the decreasing slope are as follows: First, fit a trend line using the first similarity data from Y consecutive times, calculate the slope of the line (representing the rate of decrease), and then take its absolute value to obtain the absolute value of the decreasing slope. It can be understood that if the decreasing slope is negative, it indicates that the first similarity shows a decreasing trend.

[0158] It should also be noted that the steps for fitting the trend line can be as follows: For the first similarity calculated Y times consecutively, with the "number of calculations" as the x-axis (x=1,2,...,Y) and the "first similarity" as the y-axis, fit the trend line by fitting a linear equation using the least squares method.

[0159] S304: If the first similarity is less than the first similarity threshold and the second similarity is greater than the second similarity threshold, generate a quantitative lubrication command and an oil discharge command.

[0160] For example, the first similarity threshold can be equal to 0.7. If the first similarity is less than the first similarity threshold, it indicates that the deviation between the current acoustic feature and the baseline state has reached a significant level. The second similarity threshold can be equal to 0.8. If the second similarity is greater than the second similarity threshold, it indicates that the current acoustic feature is highly matched with the lubrication failure feature. At this time, it can be determined that the acoustic abnormality of the bearing assembly 200 is caused by lubrication failure, and then a quantitative lubrication command and an oil drain command can be generated.

[0161] The specific values ​​of the first similarity threshold and the second similarity threshold can be adjusted based on the actual situation, and the embodiments of this application do not impose specific limitations on this.

[0162] Furthermore, the vertical pipeline pump operation monitoring system provided in this application embodiment may also include a model training module 1003, which is configured to perform the following steps S401-S402.

[0163] S401: Collect the sound signals of the vertical pipeline pump when it is delivering various different fluids under various combined operating conditions to obtain the first training data; and collect the sound signals of the vertical pipeline pump when it is not delivering fluid under various combined operating conditions to obtain the second training data.

[0164] The different fluids include typical transport media such as clean water, sandy water, industrial wastewater, and viscous media (such as lubricating oil and mud). Each fluid corresponds to different viscosity and solids content parameters to cover the differences in media characteristics in actual applications. At the same time, the sound signals of the vertical pipeline pump when it is not transporting fluid (no-load state) under various combined operating conditions are collected to obtain the second training data. High-precision acoustic sensors can be used during the collection process, and no less than 10 sets of samples can be continuously collected under each combined operating condition. The duration of each set of samples can be 60 seconds to ensure the sufficiency and representativeness of the training data.

[0165] S402: Supervised training of the initial deep neural network model is performed using the training dataset to obtain a pre-trained model; the training dataset includes at least the first training data and the second training data.

[0166] In practical applications, the embodiments of this application can annotate all training data, including the current joint operating condition type, fluid type, bearing status, etc. The initial deep neural model can be a hybrid model of convolutional neural network (CNN) and long short-term memory network (LSTM) (CNN-LSTM), or a deep residual network (ResNet) or gradient boosting decision tree (GBDT) model. The embodiments of this application do not specifically limit this.

[0167] The training process can use cross-validation to optimize model parameters with the goal of minimizing the loss function (such as cross-entropy loss). Training is stopped when the model’s state recognition accuracy on the validation set is ≥95%, ensuring that the pre-trained model has stable generalization ability.

[0168] Furthermore, the combined operating conditions are the Cartesian product of the first bearing operating condition set and the second bearing operating condition set; wherein, the first bearing operating condition set consists of the Cartesian product of m lubrication levels and n mechanical states of the first bearing 202, and the second bearing operating condition set consists of the Cartesian product of m lubrication levels and n mechanical states of the second bearing 203; different lubrication levels correspond to different percentage levels of lubricating grease filling, and different mechanical states correspond to the degradation levels of the first bearing 202 or the second bearing 203 from normal state to fault state; the combined operating conditions include a total of (m×n). 2 kind.

[0169] Specifically, different lubrication grades correspond to different percentage levels of lubricant filling. For example, with m=3, the lubrication grades include: 100% filling (normal lubrication), 50% filling (slightly low lubrication), and 20% filling (severe low lubrication).

[0170] Different mechanical states correspond to the degradation levels of the first bearing 202 or the second bearing 203 from normal state to failure state. For example, n=3, the mechanical states include: normal state, early wear state (raceway pitting, minor damage to rolling elements), and failure state (raceway cracks, rolling element breakage).

[0171] Therefore, the operating condition set of a single bearing contains m×n=3×3=9 operating conditions (normal lubrication + normal condition, normal lubrication + early wear, normal lubrication + failure, slight oil shortage + normal condition, ..., severe oil shortage + failure); the Cartesian product of the first bearing operating condition set and the second bearing operating condition set constitutes a joint operating condition, with a total number of (m×n). 2 =9×9=81 types, comprehensively covering the operating scenarios under different combinations of the two bearings.

[0172] It's worth noting that the Cartesian product is essentially a pairwise combination of all elements in two sets. That is, the joint operating condition is a pairwise combination of all elements in the first bearing operating condition set and the second bearing operating condition set. Furthermore, the first bearing operating condition set includes pairwise combinations of m lubrication levels and n mechanical states of the first bearing 202; that is, each element in the first bearing operating condition set is a combination of one lubrication level and one mechanical state of the first bearing 202. Correspondingly, the second bearing operating condition set includes pairwise combinations of m lubrication levels and n mechanical states of the second bearing 203; that is, each element in the second bearing operating condition set is a combination of one lubrication level and one mechanical state of the second bearing 203.

[0173] Taking the case of m=3 and n=3 as an example, the mapping relationship of some Cartesian products is introduced as follows:

[0174] Table 1. Schematic diagram of combined operating conditions

[0175] First bearing condition / Second bearing condition B1 (Normal lubrication + normal no wear) B2 (Normal lubrication + early wear) B3 (Normal lubrication + malfunction) …… B9 (Severe low oil level + malfunction) A1 (Normal lubrication + normal no wear) Combined operating condition 1 Combined operating condition 2 Combined operating condition 3 …… Combined operating condition 9 A2 (Normal lubrication + early wear) Combined operating condition 10 Combined operating condition 11 Combined operating condition 12 …… Combined operating condition 18 A3 (Normal lubrication + malfunction) Combined operating condition 19 Combined operating condition 20 Combined operating condition 21 …… Combined operating condition 27 …… …… …… …… …… …… A9 (Severe oil shortage + malfunction) Combined operating condition 73 Combined operating condition 74 Combined operating condition 75 …… Combined operating condition 81

[0176] The horizontal dimension represents the nine operating conditions of the second bearing 203, namely B1, B2, ..., B9, covering all lubrication levels and mechanical state combinations. The vertical dimension represents the nine operating conditions of the first bearing 202, namely A1, A2, ..., A9, covering all lubrication levels and mechanical state combinations. Each intersection of the horizontal and vertical dimensions corresponds to a unique combined operating condition, representing the combined operation scenario of the first bearing 202 and the second bearing 203 under specific conditions.

[0177] It is worth noting that this includes (m×n) 2 After training on the training dataset of various refined joint operating conditions, the pre-trained model does not directly output one of the 81 operating conditions. Instead, it learns a high-dimensional feature representation that can accurately characterize the essence of the complex state of the bearing. Through decision boundary optimization and state space compression mechanisms, the 81 fine-grained operating conditions are mapped to a 4-dimensional operation and maintenance decision space, and output four classification results: normal, poor lubrication, early wear, and bearing failure.

[0178] As can be seen from the above, this application provides an operation monitoring system for a vertical pipeline pump. This system can be applied to the aforementioned vertical pipeline pump and specifically includes: a signal acquisition module 1001 configured to: acquire mechanical vibration signals of a bearing assembly at a first preset sampling rate, and acquire sound signals of the bearing assembly during operation at a second preset sampling rate; an edge computing node 1002 configured to: extract features from the mechanical vibration signals within a first time window according to a first preset frequency, obtaining time-domain features, frequency-domain features, and envelope demodulation spectrum features; the first time window has the same period length as the first preset frequency; calculate a bearing feature evaluation index using the envelope demodulation spectrum features, the bearing feature evaluation index being used to characterize the degree of deterioration of the mechanical state of the bearing assembly; and the edge computing node 1002 is further configured to: extract features from the sound signals within the first time window according to the first preset frequency, obtaining acoustic features... The system performs the following steps: First, according to a first preset frequency, input the sound signal within the first time window into the pre-trained model to obtain the classification result of the bearing assembly's operating state by the pre-trained model; the operating state categories include at least normal, poor lubrication, early wear, and bearing failure; and, the edge computing node 1002 is further configured to: generate a quantitative lubrication command and an oil discharge command if at least two of the first judgment conditions are met when the classification result is poor lubrication; wherein, the first judgment conditions include a time domain feature greater than a first time domain threshold, an acoustic feature greater than a first acoustic threshold, and a bearing feature evaluation index less than a first index threshold and a frequency domain feature greater than a first frequency domain threshold; the quantitative lubrication command is used to drive the power assembly to inject a quantitative amount of lubricating grease into the first bearing 202 and the second bearing 203 through the oil filling hole 206, and the oil discharge command is used to drive the power assembly to discharge waste lubricating grease through the first oil discharge hole 207 and the second oil discharge hole 208.As can be seen, the vertical pipeline pump operation monitoring system provided in this application, through the collaborative work of the signal acquisition module 1001 and the edge computing node 1002, realizes multi-dimensional, high-frequency acquisition and feature extraction of mechanical vibration signals and sound signals of the bearing assembly 200. Combined with the intelligent classification and multi-condition threshold verification logic of the pre-trained model, it can not only accurately identify different operating states of the bearing, such as normal, poor lubrication, early wear and failure, but also output targeted quantitative lubrication and oil discharge commands for poor lubrication states, effectively avoiding misjudgments caused by single signal features or model classification, and greatly improving the accuracy of state monitoring and the pertinence of maintenance decisions. At the same time, the localized processing mode of the edge computing node 1002 ensures the real-time nature of monitoring and command generation, and can quickly respond to abnormal bearing states. By intervening in poor lubrication problems in advance, it can delay the bearing deterioration process, reduce the risk of failure and downtime, and extend the service life of the equipment. The high-dimensional feature learning capability formed by the pre-trained model based on refined joint working condition training can adapt to the monitoring needs of different fluid types and operating conditions, providing reliable support for the intelligent and normalized operation monitoring of vertical pipeline pumps, reducing operation and maintenance costs and improving overall operating efficiency.

[0179] Figure 4 This is a flowchart illustrating the operation monitoring method for a vertical pipeline pump provided in an embodiment of this application.

[0180] like Figure 4 As shown, this application provides an operation monitoring method for a vertical pipeline pump, which is applied to a vertical pipeline pump. The vertical pipeline pump includes at least a pump shaft and a bearing assembly 200. The bearing assembly 200 includes a bearing housing 201, a first bearing 202 and a second bearing 203 disposed inside the bearing housing 201. The first bearing 202 and the second bearing 203 are spaced apart along the axial direction. An upper bearing retainer 204 is provided on the first bearing 202 near the second bearing 203, and a lower bearing retainer 205 is provided on the second bearing 203 away from the first bearing 202. The bearing housing 201 has a lubrication hole 206 and a first oil drain hole 207 arranged side by side. The lubrication hole 206 is located between the first bearing 202 and the second bearing 203, and the first oil drain hole 207 is located on the side of the lower bearing retainer 205. The lower bearing retainer 205 has a second oil drain hole 208, and the first oil drain hole 207 and the second oil drain hole 208 are interconnected.

[0181] The method includes the following steps S501-S506.

[0182] S501: Acquire mechanical vibration signals of the bearing assembly 200 at a first preset sampling rate, and acquire sound signals of the bearing assembly 200 during operation at a second preset sampling rate;

[0183] S502: According to the first preset frequency, perform feature extraction on the mechanical vibration signal within the first time window to obtain time-domain features, frequency-domain features, and envelope demodulation spectrum features; the period length corresponding to the first time window and the first preset frequency is equal;

[0184] S503: Calculate the bearing characteristic evaluation index using envelope demodulation spectrum characteristics. The bearing characteristic evaluation index is used to characterize the degree of deterioration of the mechanical condition of bearing assembly 200.

[0185] S504: Extract features from the sound signal within the first time window according to the first preset frequency to obtain acoustic features;

[0186] S505: Input the sound signal within the first time window into the pre-trained model according to the first preset frequency to obtain the classification result of the pre-trained model on the operating state of the bearing assembly 200; the operating state categories include at least normal, poor lubrication, early wear and bearing failure;

[0187] S506: If the classification result is poor lubrication, and at least two of the first judgment conditions are met, a quantitative lubrication command and an oil discharge command are generated; wherein, the first judgment conditions include that the time domain feature is greater than the first time domain threshold, the acoustic feature is greater than the first acoustic threshold, and the bearing feature evaluation index is less than the first index threshold and the frequency domain feature is greater than the first frequency domain threshold; the quantitative lubrication command is used to drive the power assembly to inject a quantitative amount of lubricating grease into the first bearing 202 and the second bearing 203 through the oil filling hole 206, and the oil discharge command is used to drive the power assembly to discharge the waste lubricating grease through the first oil discharge hole 207 and the second oil discharge hole 208.

[0188] In one possible implementation, the method provided in this application embodiment further includes the following steps S601-S604.

[0189] S601: Real-time acquisition of acoustic features corresponding to X consecutive first time windows, and construction of the first vector;

[0190] S602: Calculate the first similarity between the first vector and the historical feature vector; wherein, the historical feature vector is constructed from the acoustic features corresponding to the first to the Xth first time windows after the vertical pipeline pump is started this time;

[0191] S603: If the first similarity calculated Y times consecutively shows a monotonically decreasing trend and the absolute value of the decreasing slope is greater than the first slope threshold, calculate the second similarity between the first vector and the preset feature vector; wherein, the preset feature vector is constructed based on the acoustic features corresponding to X consecutive first time windows when the bearing assembly 200 is in a poor lubrication state;

[0192] S604: If the first similarity is less than the first similarity threshold and the second similarity is greater than the second similarity threshold, generate a quantitative lubrication command and an oil discharge command.

[0193] In one possible implementation, the envelope demodulation spectrum features include the theoretical fault frequency of the bearing outer ring and the amplitude at each harmonic, the total energy of the target characteristic frequency band, and the total energy of the reference frequency band.

[0194] The formula for calculating the bearing characteristic evaluation index is as follows:

[0195] ;

[0196] in, I Indicates the bearing characteristic evaluation index, V V represents the total order of the target characteristic frequencies, where n is the order of the target characteristic frequencies, 0 < n ≤ V, and the nth order target characteristic frequency. f n It is a harmonic frequency determined based on the theoretical fault frequency of the bearing outer ring; A ( f n ) represents the amplitude corresponding to the characteristic frequency of the nth target; E f This represents the total energy of the target's characteristic frequency band; E b M represents the total energy of the reference frequency band; M is the total frequency order of the entire spectrum of the mechanical vibration signal.

[0197] In one possible implementation, step S505 is followed by step S507: if the classification result is early wear, and at least two of the second judgment conditions are met, a warning signal, a quantitative lubrication command, and an oil discharge command are generated; wherein the second judgment conditions include: the time domain feature is greater than a second time domain threshold; the acoustic feature is greater than a second acoustic threshold; and the bearing feature evaluation index is greater than or equal to a first index threshold and the frequency domain feature is greater than a second frequency domain threshold; the second time domain threshold is greater than the first time domain threshold, the second acoustic threshold is greater than the first acoustic threshold, and the second frequency domain threshold is greater than the first frequency domain threshold.

[0198] In one possible implementation, step S505 is followed by step S508: if the classification result is a bearing fault, an alarm signal is generated if at least two of the third judgment conditions are met; wherein the third judgment conditions include: the time domain feature is greater than the third time domain threshold; the acoustic feature is greater than the third acoustic threshold; and the bearing feature evaluation index is greater than the second exponential threshold and the frequency domain feature is greater than the third frequency domain threshold; wherein the third time domain threshold is greater than the second time domain threshold, the third acoustic threshold is greater than the second acoustic threshold, the second exponential threshold is greater than the first exponential threshold, and the third frequency domain threshold is greater than the second frequency domain threshold.

[0199] In one possible implementation, the time-domain feature is the effective value of vibration velocity or the kurtosis index; the frequency-domain feature is the energy proportion of the target high-frequency band; and the acoustic feature is the energy proportion of the target frequency band or the acoustic kurtosis.

[0200] In one possible implementation, the method provided in this application embodiment further includes the following steps S701-S702.

[0201] S701: Collect the sound signals of the vertical pipeline pump when it is delivering various different fluids under various combined operating conditions to obtain the first training data; and collect the sound signals of the vertical pipeline pump when it is not delivering fluid under various combined operating conditions to obtain the second training data.

[0202] S702: Supervised training of the initial deep neural network model is performed using the training dataset to obtain a pre-trained model; the training dataset includes at least the first training data and the second training data.

[0203] In one possible implementation, the combined operating conditions are the Cartesian product of the first bearing operating condition set and the second bearing operating condition set; wherein, the first bearing operating condition set consists of the Cartesian product of m lubrication levels and n mechanical states of the first bearing 202, and the second bearing operating condition set consists of the Cartesian product of m lubrication levels and n mechanical states of the second bearing 203; different lubrication levels correspond to different percentage levels of lubricating grease filling, and different mechanical states correspond to the degradation levels of the first bearing 202 or the second bearing 203 from a normal state to a fault state; the combined operating conditions include a total of (m×n). 2 kind.

[0204] In one possible implementation, the vertical pipeline pump also includes a skeleton oil seal 309, which is disposed at the junction of the pump shaft 100 and the bearing assembly 200.

[0205] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0206] like Figure 5 As shown, this application provides an electronic device including a processor 10 and a memory 20. The memory 20 stores a computer program, and the processor 10 is configured to run the computer program to execute the steps in any of the above-described embodiments of the vertical pipeline pump operation monitoring method.

[0207] In a specific implementation, the present invention also provides a computer storage medium, wherein the computer storage medium may store a program, which, when executed, may include some or all of the steps of the various embodiments of the vertical pipeline pump operation monitoring method provided by the present invention. The storage medium may be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0208] It is readily understood that, based on the several embodiments provided in this application, those skilled in the art can combine, split, or reorganize the embodiments of this application to obtain other embodiments, none of which exceed the protection scope of this application.

[0209] The above detailed embodiments further illustrate the purpose, technical solution, and beneficial effects of the embodiments of this application. It should be understood that the above are merely specific embodiments of the embodiments of this application and are not intended to limit the protection scope of the embodiments of this application. Any modifications, equivalent substitutions, improvements, etc., made on the basis of the technical solutions of the embodiments of this application should be included within the protection scope of the embodiments of this application.

Claims

1. A system for monitoring the operation of a vertical inline pump, characterized in that The application is applied to a vertical pipeline pump, the vertical pipeline pump at least includes a pump shaft and a bearing assembly, the bearing assembly includes a bearing seat, a first bearing and a second bearing arranged in the bearing seat, the first bearing and the second bearing are distributed along the axial direction, the first bearing is provided with an upper bearing stop pad in the direction close to the second bearing, the second bearing is provided with a lower bearing stop pad in the direction away from the first bearing, and the bearing seat is provided with an oiling hole and a first oil discharge hole side by side, the oiling hole is arranged corresponding to the first bearing and the second bearing, the first oil discharge hole is arranged corresponding to the side of the lower bearing stop pad, the lower bearing stop pad is provided with a second oil discharge hole, and the first oil discharge hole and the second oil discharge hole are communicated; The system comprises: A signal acquisition module is configured to acquire a mechanical vibration signal of the bearing assembly at a first preset sampling rate and to acquire a sound signal during operation of the bearing assembly at a second preset sampling rate; An edge computing node is configured to extract features of the mechanical vibration signal in a first time window at a first preset frequency to obtain time domain features, frequency domain features and envelope demodulation spectrum features; the first time window is equal in length to a period corresponding to the first preset frequency; A bearing feature evaluation index is calculated using the envelope demodulation spectrum features, and the bearing feature evaluation index is used to represent the degradation degree of the mechanical state of the bearing assembly; The edge computing node is further configured to extract features of the sound signal in the first time window at the first preset frequency to obtain acoustic features; The sound signal in the first time window is input into a pre-trained model at the first preset frequency to obtain a classification result of the pre-trained model on the operating state of the bearing assembly; the state categories of the operating state at least include normal, poor lubrication, early wear and bearing failure; If at least two of the first judgment conditions are met, a quantitative lubrication instruction and an oil discharge instruction are generated when the classification result is poor lubrication; wherein the first judgment conditions include that the time domain features are greater than a first time domain threshold, the acoustic features are greater than a first acoustic threshold, the bearing feature evaluation index is less than a first index threshold and the frequency domain features are greater than a first frequency domain threshold; the quantitative lubrication instruction is used to drive a power assembly to inject quantitative lubricating grease into the first bearing and the second bearing through the oiling hole, and the oil discharge instruction is used to drive the power assembly to discharge waste lubricating grease through the first oil discharge hole and the second oil discharge hole.

2. The operation monitoring system of a vertical pipe pump according to claim 1, characterized by, The edge computing node is further configured to: Real-time acquisition of the acoustic features corresponding to the first time window for X consecutive times to construct a first vector; Calculate the first similarity between the first vector and a historical feature vector; wherein the historical feature vector is constructed from the acoustic features corresponding to the first time window from 1 to X after the vertical pipeline pump is started this time. If the first similarity calculated for Y consecutive times presents a monotonically decreasing trend and the absolute value of the decreasing slope is greater than a first slope threshold, a second similarity between the first vector and a preset feature vector is calculated; wherein the preset feature vector is constructed based on the acoustic features corresponding to X consecutive first time windows when the bearing assembly is in a poor lubrication state; In a case where the first similarity is less than a first similarity threshold and the second similarity is greater than a second similarity threshold, the quantitative lubrication instruction and the oil discharge instruction are generated.

3. The operation monitoring system of a vertical pipe pump according to claim 1, characterized by, The envelope demodulation spectrum feature includes an amplitude at a bearing outer ring theoretical fault frequency and each order harmonic thereof, a target feature band total energy, and a reference band total energy; A calculation formula of the bearing feature evaluation index is: ； wherein, I represents the bearing characteristic evaluation index, V is the total order of the target characteristic frequency, n is the order of the target characteristic frequency, 0 < n < V, the nth order target characteristic frequency f n is a harmonic frequency determined based on the theoretical fault frequency of the bearing outer ring; A f n represents the amplitude corresponding to the nth order target characteristic frequency; E f represents the total energy of the target characteristic frequency band; E b represents the total energy of the reference frequency band; M is the total frequency order of the full spectrum of the mechanical vibration signal.​ 4. The operation monitoring system of a vertical pipe pump according to claim 1, characterized by, The edge computing node is further configured to, in a case where the classification result is early wear, if at least two of second judgment conditions are met, generate a warning signal, the quantitative lubrication instruction, and the oil discharge instruction; wherein the second judgment conditions include that the time domain feature is greater than a second time domain threshold, the acoustic feature is greater than a second acoustic threshold, and the bearing feature evaluation index is greater than or equal to the first index threshold and the frequency domain feature is greater than a second frequency domain threshold; the second time domain threshold is greater than the first time domain threshold, the second acoustic threshold is greater than the first acoustic threshold, and the second frequency domain threshold is greater than the first frequency domain threshold.

5. The operation monitoring system of a vertical pipe pump according to claim 4, characterized by The edge computing node is further configured to, in a case where the classification result is bearing failure, if at least two of third judgment conditions are met, generate an alarm signal; wherein the third judgment conditions include that the time domain feature is greater than a third time domain threshold, the acoustic feature is greater than a third acoustic threshold, and the bearing feature evaluation index is greater than a second index threshold and the frequency domain feature is greater than a third frequency domain threshold; wherein the third time domain threshold is greater than the second time domain threshold, the third acoustic threshold is greater than the second acoustic threshold, the second index threshold is greater than the first index threshold, and the third frequency domain threshold is greater than the second frequency domain threshold.

6. A system for monitoring the operation of a vertical pipeline pump according to any one of claims 1-5, characterized in that The time domain feature is a vibration velocity effective value or a kurtosis index; the frequency domain feature is a target high frequency band energy proportion; and the acoustic feature is a target frequency band energy proportion or an acoustic kurtosis.

7. The operation monitoring system of a vertical pipe pump according to claim 1, wherein The system further includes a model training module configured to: acquire sound signals of the vertical pipeline pump when conveying different fluids under various combined working conditions to obtain first training data, and acquire sound signals of the vertical pipeline pump when not conveying fluids under various combined working conditions to obtain second training data; perform supervised training on an initial deep neural network model using a training data set to obtain the pre-trained model; the training data set at least includes the first training data and the second training data.

8. The operation monitoring system of a vertical pipe pump according to claim 7, characterized by The combined working condition is a Cartesian product of a first bearing working condition set and a second bearing working condition set; wherein the first bearing working condition set is composed of a Cartesian product of m lubrication levels and n mechanical states of the first bearing, and the second bearing working condition set is composed of a Cartesian product of m lubrication levels and n mechanical states of the second bearing; different lubrication levels correspond to different percentage ranges of grease filling amount, and different mechanical states correspond to degradation levels of the first bearing or the second bearing from a normal state to a fault state; the combined working condition includes (m x n) 2 species in total.

9. The operation monitoring system of a vertical pipe pump according to claim 1, wherein The vertical pipeline pump further includes a skeleton oil seal, which is arranged at an intersection of the pump shaft and the bearing assembly.

10. A method of monitoring the operation of a vertical pipe pump, characterized by The application is applied to a vertical pipeline pump, which at least comprises a pump shaft and a bearing assembly, the bearing assembly comprises a bearing seat, a first bearing and a second bearing arranged inside the bearing seat, the first bearing and the second bearing are distributed along the axial direction, the first bearing is provided with an upper bearing stop pad in the direction close to the second bearing, and the second bearing is provided with a lower bearing stop pad in the direction away from the first bearing; and the bearing seat is provided with an oiling hole and a first oil discharge hole side by side, the oiling hole is arranged corresponding to the first bearing and the second bearing, the first oil discharge hole is arranged corresponding to the side of the lower bearing stop pad, the lower bearing stop pad is provided with a second oil discharge hole, and the first oil discharge hole and the second oil discharge hole are communicated; The method comprises: Collecting a mechanical vibration signal of the bearing assembly at a first preset sampling rate, and collecting a sound signal during operation of the bearing assembly at a second preset sampling rate; According to a first preset frequency, the mechanical vibration signal in a first time window is subjected to feature extraction to obtain time domain features, frequency domain features and envelope demodulation spectrum features; the first time window is equal in length to a period corresponding to the first preset frequency; A bearing feature evaluation index is calculated using the envelope demodulation spectrum features, and the bearing feature evaluation index is used to represent the degradation degree of the mechanical state of the bearing assembly; According to the first preset frequency, the sound signal in the first time window is subjected to feature extraction to obtain acoustic features; According to the first preset frequency, the sound signal in the first time window is input into a pre-trained model to obtain a classification result of the pre-trained model on the running state of the bearing assembly; the state categories of the running state at least include normal, poor lubrication, early wear and bearing failure; In the case that the classification result is poor lubrication, if at least two of the first judgment conditions are established, a quantitative lubrication instruction and an oil discharge instruction are generated; wherein the first judgment conditions include that the time domain features are greater than a first time domain threshold, the acoustic features are greater than a first acoustic threshold, and the bearing feature evaluation index is less than a first index threshold and the frequency domain features are greater than a first frequency domain threshold; the quantitative lubrication instruction is used to drive a power assembly to inject quantitative lubricating grease into the first bearing and the second bearing through the oiling hole, and the oil discharge instruction is used to drive the power assembly to discharge waste lubricating grease through the first oil discharge hole and the second oil discharge hole.

Images