Open cycle water pump vibration wear correlation diagnosis method and system

Abstract

The application discloses an open circulating water pump vibration and wear correlation diagnosis method and system, the method comprises the following steps: collecting multi-dimensional vibration signals through vibration sensors arranged at the bearing and pump shell of the circulating water pump, implementing frequency domain decoupling on the signals, and extracting high-frequency vibration features corresponding to fluid particle impact and low-frequency vibration features corresponding to the decrease of rotor radial constraint; combining the difference in change rate and the ratio trend of the two types of features within a preset time window, constructing a time sequence evolution correlation, and determining the wear evolution stage of the pump set. The application can effectively distinguish different failure mechanisms, overcome the defect of early warning lag in the traditional monitoring mode, realize dynamic tracking of the whole process of the wear state, improve the accuracy and timeliness of fault diagnosis, and provide reliable support for the safe operation and predictive maintenance of the open circulating water pump.

Description

Technical Field

[0001] This invention belongs to the field of rotating machinery condition monitoring and fault diagnosis technology, specifically relating to a method and system for the correlation diagnosis of vibration and wear in an open circulating water pump. Background Technology

[0002] In open-loop cooling water systems in industries such as power and chemical engineering, circulating water pumps often directly use natural water sources such as seawater and river water as cooling media. Because the water source contains silt, suspended solids, and impurities, the pump's flow components and supporting bearings operate under particulate-rich fluid conditions for extended periods, easily leading to abrasive wear and erosion. This results in increased component clearances, decreased rotor radial constraint, intensified vibration, and a positive feedback deterioration mechanism of wear-vibration-impact. Existing circulating water pump condition monitoring methods mostly rely on single-threshold alarms based on the effective value of vibration velocity or conventional spectrum analysis. These methods cannot adapt to the phased evolution characteristics of wear under open-loop conditions. Early-stage high-frequency vibration characteristics caused by particle impact are weak, mid-stage low-frequency oscillation characteristics caused by excessive clearance are delayed, and late-stage fault characteristics are significantly coupled. Traditional methods struggle to decouple high- and low-frequency vibration signals, cannot distinguish between the fault mechanisms of particle impact and clearance failure, and cannot accurately identify wear stages and evolution inflection points. They generally suffer from delayed early warnings, high false alarm and false negative rates, making predictive maintenance difficult and significantly increasing the risk of unplanned unit shutdowns. Summary of the Invention

[0003] The purpose of this invention is to provide a method and system for the correlation diagnosis of vibration and wear in open circulating water pumps, in order to solve the technical defects of existing monitoring methods, such as difficulty in decoupling high and low frequency vibrations, inability to distinguish wear mechanisms, delayed early warning, and high false alarm and missed alarm rates.

[0004] To achieve the above objectives, the present invention employs the following technical solutions.

[0005] In a first aspect, this application provides a method for diagnosing vibration and wear correlation in an open-loop circulating water pump, comprising: Vibration sensors installed on the bearings and casing of the circulating water pump are used to collect multi-dimensional vibration signals during the operation of the pump unit. The multidimensional vibration signal is decoupled in the frequency domain to extract high-frequency vibration features characterizing fluid particle impact and low-frequency vibration features characterizing the decrease in rotor radial constraint. Based on the difference in the rate of change and the ratio trend of the high-frequency vibration characteristics and the low-frequency vibration characteristics within a preset time window, a temporal evolution correlation is constructed. Based on the aforementioned temporal evolution correlation, the current wear evolution stage of the circulating water pump is determined.

[0006] Furthermore, the high-frequency vibration characteristics correspond to the inherent frequency response range of the circulating water pump bearing and pump casing components, and are used to characterize the abrasive impact of hard particles in the fluid on the pump structure. The low-frequency vibration characteristics correspond to the power frequency and sub-power frequency bands of the pump rotor system, and are used to characterize the operating state of the rotor due to decreased radial support stiffness and failure of radial constraint caused by bearing wear.

[0007] Furthermore, constructing the temporal evolution correlation specifically includes: Within a preset sliding time window, the first rate of change corresponding to the high-frequency vibration feature and the second rate of change corresponding to the low-frequency vibration feature are calculated respectively. Based on the relative difference between the first and second rates of change, and combined with the temporal variation trend of the ratio of high-frequency vibration characteristics to low-frequency vibration characteristics, a temporal evolution correlation is established to characterize the wear evolution law of the pump group.

[0008] Further, extracting the high-frequency vibration features includes: The multidimensional vibration signals collected from the bearing and casing of the circulating water pump are subjected to bandpass filtering to extract the high-frequency signal components of the corresponding component's natural frequency response segment. Perform a Hilbert transform on the high-frequency signal components to obtain the signal envelope; Perform a Fast Fourier Transform on the envelope signal to generate envelope spectrum data; The peak energy of the spectral frequency range within the envelope spectrum is extracted and used as a high-frequency vibration characteristic to characterize abrasive impact.

[0009] Further, extracting the high-frequency vibration features includes: High-pass filtering was performed on the multidimensional vibration signals collected from the bearings and casing of the circulating water pump to obtain high-frequency impact signals characterizing particle impact. Calculate the peak value to root mean square ratio of the high-frequency impulse signal within a preset sampling period; The signal exceeding the preset impact energy threshold is integrated to obtain the vibration peak energy, which is then used as the high-frequency vibration characteristic to characterize abrasive impact.

[0010] Further, extracting the low-frequency vibration features includes: The multidimensional vibration signals collected from the bearings and casing of the circulating water pump are subjected to low-pass filtering to obtain the low-frequency signal components of the rotor power frequency and sub-power frequency bands. The low-frequency signal components are sequentially resampled and windowed to suppress spectral leakage, and then subjected to fast Fourier transform to obtain a high-resolution refined spectrum. Based on the real-time speed of the pump unit, the rotor's first harmonic component and the preset order subharmonic component are identified. The low-frequency vibration characteristics that characterize the decrease in rotor radial constraint are obtained by weighting and summing the first harmonic amplitude and the subharmonic amplitude according to a preset weighting coefficient.

[0011] Furthermore, determining the wear evolution stage specifically includes: When the first rate of change of the high-frequency vibration characteristics is greater than the second rate of change of the low-frequency vibration characteristics, and the ratio of high-frequency to low-frequency characteristics continues to rise, the circulating water pump bearing and rotor system are determined to be in the early wear stage. When the second rate of change of the low-frequency vibration characteristics exceeds the preset acceleration threshold and the ratio of high-frequency to low-frequency characteristics reverses and decreases, it is determined to be in the middle wear stage. When both high-frequency and low-frequency vibration characteristics exceed the preset danger threshold and their rates of change diverge, the system is determined to be in the late wear stage.

[0012] Furthermore, the method also includes: Based on the determined wear evolution stage of the circulating water pump bearing and rotor system, a corresponding level of early warning instruction is generated; When the bearing is in the early wear stage, a first warning command is generated to improve the bearing lubrication condition. When the vehicle is in the mid-term wear and tear stage, a second early warning instruction is generated for planned maintenance. When the machine is in the late wear stage, a third warning command is generated for immediate shutdown and inspection. The warning instruction will then be pushed to the operation and maintenance terminal.

[0013] Furthermore, after acquiring the multidimensional vibration signal, the process also includes: Adaptive filtering is performed on the multidimensional vibration signals collected from the bearings and casing of the circulating water pump to remove environmental noise and electromagnetic interference. Temperature drift compensation is performed on the vibration signal based on the real-time temperature data of the pump unit. By removing outliers from the compensation signal, a clean target vibration signal is obtained. A vibration health baseline model was established based on the historical normal operation data of the pump set.

[0014] A second aspect of this application provides a vibration and wear correlation diagnostic system for an open-type circulating water pump, comprising: The data acquisition module is configured to collect multi-dimensional vibration signals during the operation of the pump unit through vibration sensors installed on the bearings and casing of the circulating water pump. The data processing module is configured to perform frequency domain decoupling on the multidimensional vibration signal, extract high-frequency vibration features characterizing fluid particle impact and low-frequency vibration features characterizing the decrease in rotor radial constraint; construct a temporal evolution correlation based on the difference in the rate of change and the ratio trend of the high-frequency vibration features and low-frequency vibration features within a preset time window; and determine the current wear evolution stage of the circulating water pump based on the temporal evolution correlation.

[0015] Compared with the prior art, the present invention has the following beneficial effects: 1. By collecting multi-dimensional vibration signals and performing frequency domain decoupling, the high-frequency characteristics representing particle impact and the low-frequency characteristics representing the decrease in rotor radial constraint can be effectively separated. Then, based on the difference in rate of change and the trend of ratio, a time-series evolution correlation is constructed, which can determine the wear evolution stage of the circulating water pump. This fundamentally solves the problem that traditional single vibration index cannot distinguish fault mechanisms and has a lag in early warning. It enables dynamic tracking of the entire wear process, improves diagnostic accuracy and real-time performance, and provides a reliable basis for the safe operation and predictive maintenance of open circulating water pumps.

[0016] 2. By clearly defining the high-frequency vibration characteristics corresponding to the natural frequency range of the bearing and pump casing, and the low-frequency characteristics corresponding to the rotor's power frequency and sub-power frequency range, the extracted features have clear physical meanings. This enables the differentiation between two fault mechanisms: abrasive impact and radial constraint failure. It avoids the problems of feature confusion and high misjudgment rate in traditional spectrum analysis, significantly improving the pertinence and reliability of wear cause identification. This provides a clear mechanism and high signal-to-noise ratio feature basis for subsequent stage judgments, enhancing the stability and reliability of the diagnostic method.

[0017] 3. By calculating the rate of change of high and low frequency characteristics within a sliding time window and constructing a correlation with the time-series trend of the ratio, the wear evolution law can be dynamically reflected, effectively filtering out operating condition fluctuations and random noise interference. Compared with the fixed threshold judgment method, it can more sensitively capture the key inflection point of wear transformation from the early stage to the middle stage, improve the sensitivity and accuracy of stage judgment, realize the quantitative assessment of wear development trend, and provide a stable and reliable judgment basis for early warning and trend analysis.

[0018] 4. By employing bandpass filtering, Hilbert transform, and envelope spectrum analysis to extract high-frequency features, it can amplify weak abrasive impact signals, effectively suppress background noise interference, accurately capture the impact response caused by early wear, solve the problem that traditional methods are difficult to identify early minor wear, enable early wear hazards to be detected in time, and improve the early identification capability and fault sensitivity of diagnostic methods.

[0019] 5. High-frequency vibration features are extracted by high-pass filtering, peak-to-root-mean-square ratio calculation and impact energy integration. It has strong robustness to random mud and sand particle impacts and can effectively identify abrasive impact intensity without relying on a specific resonant frequency. It is suitable for open-cycle complex working conditions and can stably extract weak impact information. It avoids feature loss due to the non-periodic nature of impacts, further improving the applicability and reliability of high-frequency feature extraction and enhancing the ability to identify early wear.

[0020] 6. By extracting low-frequency features through low-pass filtering, resampling, windowing, and refined spectrum processing, the rotor's first harmonic and subharmonic components can be identified, effectively suppressing spectral leakage and improving low-frequency resolution. It can sensitively reflect the decrease in radial constraint caused by bearing wear, accurately capture shaft wobble and stiffness degradation information, solve the problem of traditional methods being insensitive to increased clearance, improve the accuracy of mid-term wear identification, and provide stable and reliable low-frequency feature support for wear stage determination.

[0021] 7. Quantify the judgment rules for early, middle and late wear stages to achieve automatic division of wear stages. It can clearly distinguish between three states: abrasive impact-dominated, gap failure-dominated and rapid deterioration. It can accurately capture the inflection point of wear evolution, avoid misjudgment and omission, make the diagnosis results more consistent with the actual deterioration process, greatly improve the timeliness and accuracy of early warning, provide maintenance personnel with clear and quantifiable status basis, and effectively reduce the risk of unplanned downtime.

[0022] 8. Generate graded early warning instructions based on different wear stages and push them to the operation and maintenance terminal to achieve direct matching of diagnostic results with operation and maintenance strategies.

[0023] 9. Through adaptive filtering, temperature drift compensation, outlier removal, and the establishment of a healthy baseline, the influence of environmental noise, temperature changes, and transient interference on vibration signals is effectively eliminated, and high-purity target vibration data is obtained. Attached Figure Description

[0024] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1 This is a schematic diagram of the vibration and wear correlation diagnosis method for open circulating water pumps provided by the present invention. Figure 2 This is a schematic diagram of the vibration-wear correlation evolution trend and stage determination logic in the vibration-wear correlation diagnosis method for open circulating water pumps provided by the present invention. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0027] To address the technical deficiencies mentioned in the background section, this embodiment provides a method and system for diagnosing vibration-wear correlation in open-loop circulating water pumps. The invention will be further described in detail below with reference to the accompanying drawings: In a first aspect, the present invention provides a method for diagnosing vibration and wear correlation in an open-loop circulating water pump, such as... Figure 1 As shown, it includes: S101. Vibration sensors installed on the bearings and casing of the circulating water pump are used to collect multi-dimensional vibration signals during the operation of the pump unit. For example, piezoelectric or MEMS accelerometer vibration sensors are deployed at key load-bearing and vibration transmission components of the circulating water pump, such as the guide bearing housing, motor bearing, and pump casing outer wall. These sensors simultaneously collect raw vibration waveform data in three directions: horizontal, vertical, and axial. The sampling frequency covers a wide frequency range from the rotor's power frequency to the component's natural frequency, ensuring complete preservation of vibration signal components corresponding to abrasive impact and rotor radial constraint reduction. After acquiring the multi-dimensional vibration signals, adaptive filtering is performed to remove environmental noise and electromagnetic interference. Temperature drift compensation is then performed based on the pump's real-time temperature data to eliminate outliers in the compensation signal, resulting in a clean target vibration signal. Finally, a vibration health baseline model is established based on the pump's historical normal operation data, providing high-quality, high-reliability data support for subsequent frequency domain decoupling and feature extraction.

[0028] S102. Decouple the multidimensional vibration signal in the frequency domain and extract the high-frequency vibration characteristics that characterize the impact of fluid particles and the low-frequency vibration characteristics that characterize the decrease of rotor radial constraint. For example, the pre-processed pure target vibration signal is decoupled in the frequency domain by splitting it into frequency bands. Among them, the high-frequency vibration characteristics correspond to the natural frequency response range of the bearing and pump casing, which are used to reflect the abrasive impact caused by hard fluid particles. They can be extracted in two ways: First, the signal is band-pass filtered to extract the high-frequency components of the corresponding frequency band, and the signal envelope is obtained by Hilbert transform. Then, the envelope spectrum is generated by fast Fourier transform, and the peak energy of the spectrum within the preset impact frequency range is selected as the high-frequency feature. Second, the high-frequency impact signal is obtained by high-pass filtering, the ratio of the signal peak value to the root mean square is calculated, and the energy of the signal exceeding the threshold is integrated to obtain the vibration peak energy, which is used as the high-frequency feature. Low-frequency vibration characteristics correspond to the power frequency and sub-power frequency bands of the rotor system, reflecting the decrease in radial support stiffness and constraint failure caused by bearing wear. The low-frequency components are obtained by low-pass filtering of the signal, and after resampling and windowing to suppress spectral leakage, a fast Fourier transform is performed to obtain a refined spectrum. Combined with the pump speed, the rotor's first harmonic and subharmonic components are identified. The amplitudes of the two types of components are weighted and summed to obtain the low-frequency vibration characteristics, ultimately achieving accurate separation of different fault characteristics.

[0029] S103. Based on the difference in the rate of change and the ratio trend of the high-frequency vibration characteristics and the low-frequency vibration characteristics within a preset time window, construct a temporal evolution correlation relationship; For example, using a preset sliding time window as the statistical interval, the first rate of change corresponding to high-frequency vibration characteristics and the second rate of change corresponding to low-frequency vibration characteristics are calculated respectively, quantifying the dynamic change amplitude of the two types of characteristics. Combining the relative difference between the first and second rates of change, and simultaneously statistically analyzing the evolution of the ratio of high-frequency and low-frequency vibration characteristics over time, and integrating the temporal trends of both, a temporal evolution correlation that can quantify the wear evolution law of the pump group is established, fully reflecting the overall development process of circulating water pump wear.

[0030] S104. Based on the time-series evolution relationship, determine the current wear evolution stage of the circulating water pump.

[0031] For example, the wear stage is determined by the dynamic change of the rate of change and the ratio of features in the temporal evolution relationship: if the first rate of change of the high-frequency vibration feature is greater than the second rate of change of the low-frequency vibration feature, and the ratio of high and low frequency features continues to rise, the circulating water pump bearing and rotor system is determined to be in the early wear stage; if the second rate of change of the low-frequency vibration feature exceeds the preset acceleration threshold, and the ratio of high and low frequency features reverses and declines, it is determined to be in the middle wear stage; if both the high-frequency and low-frequency vibration features exceed the preset danger threshold, and the rates of change of the two are divergent, the equipment is determined to have entered the late wear stage.

[0032] After the phase determination is completed, corresponding early warning instructions are generated based on the actual wear level of the pump unit. The first early warning instruction to improve bearing lubrication is output during the early wear stage, the second early warning instruction to plan maintenance is output during the mid-wear stage, and the third early warning instruction to immediately stop the machine for inspection is output during the late wear stage. Finally, all kinds of early warning instructions are pushed to the operation and maintenance terminal.

[0033] When implementing the above methods and steps, in an open-loop cooling water system, the circulating water pump directly draws natural water containing impurities such as silt and suspended solids. Its operating environment is characterized by strong background noise and variable operating conditions. Therefore, the multidimensional vibration signal obtained here should not be understood as a single effective value of vibration velocity or a simple snapshot of the spectrum, but rather as a sequence of original vibration data containing rich time-domain waveform information and wideband response characteristics.

[0034] For example, acceleration sensors placed in key parts such as pump guide bearing housings and motor bearings can simultaneously collect the original vibration waveforms in the X, Y, and Z directions. The sampling frequency needs to cover a wide frequency range from the rotor rotation frequency to the component's natural frequency to ensure that the signal components characterizing different fault modes can be completely preserved. This step provides the necessary data foundation for subsequent feature decoupling and avoids the problem of losing weak wear characteristics due to oversimplification of front-end signal processing.

[0035] Extracting high-frequency vibration characteristics representing the impact of hard particles in the fluid and low-frequency vibration characteristics representing the decrease in rotor radial constraint from multidimensional vibration signals is the core concept that distinguishes this invention from traditional single vibration monitoring methods. Under open-cycle conditions, the wear process is not a linear deterioration in a single dimension, but rather involves two parallel and coupled physical mechanisms: First, the hard particles carried in the fluid cause microscopic cutting and impact on the surfaces of the flow components and bearings. Second, the cumulative wear leads to changes in macroscopic geometric clearance and degradation of support stiffness. If the judgment is based solely on the overall vibration level, it will often be missed in the early stage because the impact energy is drowned out by background noise, or misjudged in the middle stage due to vibration changes caused by load fluctuations.

[0036] Therefore, this embodiment adopts a dual-channel feature extraction architecture to decouple two types of heterogeneous information mixed in the same vibration signal. The high-frequency vibration feature is specifically used to capture the transient elastic wave response excited by hard particles impacting the metal surface, and it is highly sensitive to the particle activity in the early stage of wear. The low-frequency vibration feature focuses on the power frequency and its harmonic components that reflect the stability of the rotor's motion in the water-lubricated bearing. Its amplitude change directly maps the decrease in radial constraint capability caused by the increase in bearing clearance. Through this feature separation with clear physical orientation, the diagnostic system can simultaneously perceive the cause and effect of wear, thereby establishing a three-dimensional perception capability for the entire life cycle of wear.

[0037] Based on the temporal evolution correlation between high-frequency vibration characteristics and low-frequency vibration characteristics, the wear evolution stage of the circulating water pump is determined. After obtaining the above two types of characteristics, this invention does not directly trigger an alarm based on the absolute value of the characteristics at a certain moment, but instead constructs a dynamic correlation model between the two in the time dimension. This is because in a complex open circulation environment, the absolute amplitude of the vibration signal is easily affected by external operating conditions such as flow rate, head, and tidal water level, and the static threshold is difficult to adapt to this non-stationary characteristic.

[0038] In contrast, the relative evolution trend between high-frequency and low-frequency characteristics exhibits stronger intrinsic stability and physical interpretability. For example, in the early stages of wear, the growth rate of high-frequency impact characteristics may be significantly higher than that of low-frequency oscillation characteristics, indicating that the primary issue at this time is abrasive erosion. However, as wear progresses into the middle stage, when gap widening becomes the dominant factor, low-frequency characteristics may show an accelerated upward trend, even reversing the ratio between the two. By continuously tracking this temporal evolution correlation, the system can automatically identify the key turning point from quantitative accumulation to qualitative deterioration of wear, thereby accurately determining the specific wear evolution stage of the equipment. This method effectively filters out common-mode interference, elevating the diagnostic logic from simple over-limit alarms to the level of mechanism tracing, providing reliable condition criteria for predictive maintenance.

[0039] It should be understood that, although Figure 1 The linear execution flow is shown, but in actual engineering applications, the above steps can be executed periodically or triggered by events. At the same time, the specific extraction methods of high-frequency vibration features and low-frequency vibration features, the mathematical expression of the temporal evolution correlation, and the classification criteria of wear evolution stages can all be adaptively adjusted according to the specific pump structure, water quality conditions, and operation and maintenance requirements. As long as its core logic follows the concept of dual feature decoupling and dynamic correlation determination, it should be covered within the protection scope of this invention.

[0040] Based on the acquisition of multidimensional vibration signals during the operation of the circulating water pump, and considering the complexity of the open circulation environment, this embodiment further provides a signal preprocessing and health baseline establishment method for this specific operating condition.

[0041] Specifically, adaptive filtering is performed on the multidimensional vibration signal to remove environmental noise and electromagnetic interference signals. In an open-loop pump house, the background noise is not stationary white noise, but exhibits significant time-varying non-stationary characteristics due to changes in tidal water level, unit load adjustment, and the start-up and shutdown of surrounding equipment.

[0042] Using bandpass or lowpass filters with fixed parameters often fails to meet the noise reduction requirements under different operating conditions, and may even filter out useful weak wear signals. Therefore, this embodiment preferably uses an adaptive filtering algorithm based on minimum mean square error (LMS) or recursive least squares (RLS). This algorithm can adjust the filter weight coefficients in real time according to the statistical characteristics of the input signal. When the ambient noise spectrum drifts, the filter can automatically track and suppress new interference components, while preserving vibration information that characterizes the mechanical state to the greatest extent.

[0043] For example, when the pump group causes the frequency band of fluid turbulence noise to shift due to flow regulation, the adaptive filter can converge within several sampling periods, raising the signal-to-noise ratio to a level that meets the requirements of subsequent feature extraction. This step effectively solves the problem of unstable feature extraction caused by dynamic changes in background noise under open-loop operating conditions.

[0044] The system acquires real-time temperature data of the circulating water pump and performs temperature drift compensation on the filtered multidimensional vibration signal based on the real-time temperature data. Since the piezoelectric ceramic or MEMS sensitive element inside the industrial-grade vibration sensor has inherent temperature sensitivity, and the open circulating pump room is usually in a semi-open environment with significant day-night and seasonal temperature differences, the sensor output is very prone to zero-point drift or sensitivity changes. If such false trend changes are not corrected, they will be misjudged as equipment deterioration.

[0045] This embodiment synchronously collects the temperature at the measuring point, pre-calibrates the temperature-sensitive characteristic curve of the sensor, and constructs a polynomial fitting compensation model for temperature-vibration amplitude. During actual operation, the system looks up the current deviation amount in a table or calculates it based on the real-time temperature, and subtracts the deviation amount from the vibration signal.

[0046] It should be understood that, in addition to polynomial fitting, piecewise linear interpolation or neural network regression can also be used to establish compensation models, as long as they can reflect the mapping relationship between temperature and sensor output. By compensating for temperature drift, the influence of ambient temperature fluctuations on the vibration amplitude benchmark is eliminated, ensuring the longitudinal comparability of long-term trend analysis data.

[0047] During data transmission and acquisition, it is inevitable that the signal will be affected by factors such as transient electromagnetic pulses, poor contact, or A / D conversion errors, resulting in spike noise with amplitudes far exceeding the normal range. However, in open-cycle wear diagnosis, the actual hard particle impact signal also manifests as transient spikes. If a fixed threshold is simply used for truncation, it is very easy to mistakenly delete the true early wear characteristics.

[0048] Therefore, this embodiment employs anomaly detection criteria based on statistical distribution, such as the 3σ principle or the interquartile range (IQR) rule, or uses unsupervised learning algorithms such as Isolation Forest. These methods can identify outliers that significantly deviate from the overall data distribution in terms of statistical characteristics, while retaining transient waveforms that, although having high amplitudes, conform to the attenuation law of impact signals.

[0049] The target vibration signal obtained after the above cleaning process not only eliminates dirty data with non-physical significance, but also fully retains the effective components containing the wear mechanism, providing a high-confidence data source for subsequent steps.

[0050] Traditional health baselines are often set as a single absolute threshold. However, under open-cycle operating conditions, even if the equipment is in a healthy state, its vibration level will fluctuate within a certain range depending on external conditions such as tide level and water temperature. Therefore, the health baseline established in this embodiment is not a static value, but a probability distribution model constructed based on historical normal data. Specifically, target vibration signals that have been inspected and accepted and have been running stably under various typical operating conditions for a period of time are selected as the sample set. Their time-domain and frequency-domain statistical features are extracted to construct a multidimensional Gaussian mixture model or a nonparametric kernel density estimation model.

[0051] This model defines the probability density boundary of the normal state in the feature space, rather than a simple upper or lower bound. When subsequent real-time monitoring data falls within this probability boundary, it is judged as normal; otherwise, it is considered as deviating from the baseline. This dynamic baseline mechanism can effectively accommodate the inherent variability of open-cycle operating conditions and significantly reduce the false alarm rate caused by operating condition switching.

[0052] After completing the aforementioned preprocessing and baseline establishment, the extraction of high-frequency vibration features characterizing the impact of hard particles in the fluid, and low-frequency vibration features characterizing the decrease in rotor radial constraint from the multidimensional vibration signal in subsequent embodiments, specifically refers to extracting the high-frequency vibration features and the low-frequency vibration features from the target vibration signal. This means that all feature extraction operations are performed on clean data after adaptive filtering, temperature drift compensation, and outlier removal, thereby fundamentally ensuring the accuracy of dual-channel feature decoupling and the reliability of wear stage determination.

[0053] Building upon the extraction of high-frequency and low-frequency vibration characteristics, this embodiment further clarifies the physical orientation of these two types of characteristics and elaborates on the specific construction dimensions of the temporal evolution correlation. Specifically, the high-frequency vibration characteristics point to the inherent frequency response range of the circulating water pump components, used to characterize abrasive impact; the low-frequency vibration characteristics point to the rotor power frequency and sub-power frequency band of the circulating water pump, used to characterize the decrease in radial constraint. This limitation is not an arbitrary frequency band division, but rather a deep decoupling based on the wear mechanism of the open circulating water pump.

[0054] like Figure 2 As shown, under the lubrication conditions of fluids containing impurities, the impact of hard particles on the surface of the guide bearing or impeller is a transient broadband excitation. Its energy is widely distributed across the spectrum and has a weak amplitude, making it easily submerged by fluid turbulence noise. However, when the impact frequency coincides with the natural frequency of pump components (such as bearing housings, pump casings, or shaft systems), it will excite structural resonance, significantly amplifying the vibration response within that frequency band.

[0055] Therefore, anchoring high-frequency vibration characteristics to the component's natural frequency response range essentially utilizes the mechanical structure's own transfer function as a natural amplifier to extract a high signal-to-noise ratio signal characterizing the activity of abrasive impact from strong background noise.

[0056] It should be understood that the inherent frequency response range of the component here can be pre-calibrated through experimental modal analysis or hammer test under operating conditions. Its specific frequency range depends on the pump body structure and material properties, rather than a fixed high-frequency value.

[0057] In contrast, the reason why low-frequency vibration characteristics point to the rotor's mains frequency and sub-mains frequency bands is that as wear leads to increased bearing clearance, the load-bearing capacity of the water lubrication film decreases. Under hydrodynamic action, the rotor will generate oil film eddies or nonlinear friction. These fault mechanisms will manifest in the frequency spectrum as significant subharmonic components such as the 0.5th and 0.38th harmonics, as well as an abnormal increase in the mains frequency amplitude. If only the mains frequency is monitored, it is easy to confuse it with vibration caused by rotor mass imbalance; however, by introducing sub-mains frequency characteristics, the specific physical state of decreased radial constraint caused by increased clearance can be accurately captured, thereby achieving a specific characterization of the wear-causing factors.

[0058] Having clarified the physical connotations of the features, this embodiment further concretizes the abstract correlation into a computable mathematical dimension. The determination of the wear evolution stage of the circulating water pump based on the temporal evolution correlation between the high-frequency vibration features and the low-frequency vibration features includes: Within a preset time window, a first rate of change of the high-frequency vibration characteristic and a second rate of change of the low-frequency vibration characteristic are calculated. Based on the difference between the first and second rates of change, and / or the trend of the ratio of the high-frequency vibration characteristic to the low-frequency vibration characteristic, the temporal evolution correlation is constructed. According to the temporal evolution correlation, the current wear evolution stage of the circulating water pump is determined. The core value of this step lies in eliminating the impact of operating condition fluctuations through dynamic trend comparison. Specifically, the preset time window can be a fixed-length sliding window (e.g., the past 7 days or 30 days), or an adaptive window based on equipment operating hours or start-stop cycles.

[0059] Within the window, the first and second rates of change can be calculated using the first-order difference method, which is the difference between the feature value at the current moment and the feature value at the beginning of the window divided by the time span. Preferably, to suppress random fluctuations, the least squares method can also be used to linearly fit the feature sequence within the window, and the slope of the fitted line can be used as the rate of change. This approach can reflect the overall evolution of the feature over a period of time, rather than instantaneous jumps.

[0060] More importantly, the ratio trend (i.e., the trajectory of the ratio of high-frequency vibration characteristics to low-frequency vibration characteristics over time) constitutes the core criterion for determining the shift in the dominant wear mechanism. In the early stages of wear, the impact of hard particles is the main problem, the high-frequency characteristics increase rapidly, while the gap has not yet widened significantly, and the low-frequency characteristics are relatively stable. At this time, the ratio trend shows a continuous upward trend, indicating that the equipment is in a stage of active impact but still under reasonable constraint.

[0061] As wear accumulates, when the gap expands beyond the critical threshold, the decrease in rotor radial constraint becomes the dominant factor, and the low-frequency characteristics begin to rise rapidly, even exceeding the growth rate of the high-frequency characteristics, causing the ratio trend to reverse and show a downward state. This non-monotonic change law of the ratio essentially reflects the qualitative change process of wear failure mode from surface damage to system instability.

[0062] It should be understood that, in addition to directly using the ratio of the original features, in some embodiments, the normalized feature difference trend or weighted combination trend can also be used instead, as long as it can characterize the dynamic evolution of the relative strength relationship between the two types of features.

[0063] Through the construction of the above multi-dimensional temporal correlation, this invention transforms discrete vibration data into state evolution trajectories with clear physical semantics, providing solid mathematical and mechanistic support for the subsequent accurate division of wear stages.

[0064] Based on the clear understanding of the physical orientation and correlation dimensions of high-frequency vibration features and low-frequency vibration features, this embodiment further provides specific extraction algorithm implementation methods for the above two types of features. It should be understood that the algorithm paths described below are only illustrative examples, intended to demonstrate how to transform abstract physical representations into computable digital signal processing flows, and not to limit the scope of protection of this invention.

[0065] Specifically, extracting high-frequency vibration features characterizing the impact of hard particles in the fluid from the multidimensional vibration signal includes the following two parallel implementation paths.

[0066] The first implementation path is applicable to scenarios where the impact signal has obvious periodicity, such as when the guide bearing or rolling bearing suffers discrete damage such as local peeling or cracks; this path specifically includes: performing bandpass filtering on the multidimensional vibration signal to extract the high-frequency signal component that contains the inherent frequency response segment of the component.

[0067] The center frequency of the bandpass filter here should be determined in advance through modal testing or spectrum peak search under operating conditions, and the bandwidth should cover the half-power point width of the structural resonance peak to ensure complete preservation of modulation information.

[0068] Subsequently, a Hilbert transform is performed on the high-frequency signal component to obtain its analytic signal, and the magnitude of the analytic signal is calculated to obtain the envelope signal. The Hilbert transform can convert a real signal into a complex analytic signal, and its magnitude is the instantaneous amplitude envelope of the signal. This step essentially completes the demodulation of the high-frequency carrier, extracting the low-frequency impulse pulse sequence hidden in the resonant response.

[0069] Next, a fast Fourier transform is performed on the envelope signal to generate envelope spectrum data; finally, the spectral peak energy located within the preset impact frequency range in the envelope spectrum data is extracted, and the spectral peak energy is used as the high-frequency vibration feature.

[0070] The preset impact frequency range here usually corresponds to the bearing fault characteristic frequency or its harmonics, or the impeller blade passing frequency and its sideband. By calculating the area or sum of squares of amplitude under the spectral peak in this frequency band, the intensity of periodic impact can be quantified. The advantage of this path based on resonance demodulation is that it can effectively amplify weak early fault signals. However, when facing a large number of non-periodic, randomly distributed mud and sand abrasive particles impacting under open cyclic conditions, the envelope spectrum may become blurred due to the lack of a stable modulation carrier.

[0071] The second implementation path is specifically designed to address the highly random and transient wideband characteristics of hard particle impacts under open-cycle conditions. This path specifically includes: high-pass filtering the multidimensional vibration signal to obtain a high-frequency impact signal. The cutoff frequency of the high-pass filter is generally set at a frequency band much higher than the rotor power frequency and its lower harmonics (e.g., above 2kHz) to thoroughly filter out low-frequency background interference from hydrodynamic noise and mechanical vibration. Subsequently, the ratio of the peak value to the root mean square value of the high-frequency impact signal within a preset sampling period is calculated, and this ratio is compared with a preset impact energy threshold. This ratio is a variant of the kurtosis index, which is extremely sensitive to transient spikes in the signal but insensitive to stable background noise.

[0072] The preset impact energy threshold can be dynamically adjusted based on the statistical distribution of healthy baseline data (e.g., mean plus three standard deviations). Then, the integrated value of the signal energy that meets the preset impact energy threshold is taken as the vibration peak energy, and this vibration peak energy is taken as the high-frequency vibration characteristic. This method, based on time-domain statistical filtering followed by integration, is essentially an adaptive impact event counter and energy accumulator. It does not depend on a specific modulation frequency, and therefore has better robustness to non-periodic cutting and impacts caused by random sediment particles carried by water flow. In practical applications, automatic switching or weighted fusion between the two paths can be performed based on the current operating status of the pump unit or historical diagnostic experience to adapt to the complex process of wear mechanism evolution from discrete damage to random abrasive wear.

[0073] On the other hand, extracting low-frequency vibration features characterizing the decrease in rotor radial constraint from the multidimensional vibration signal includes the following steps: First, the multidimensional vibration signal is subjected to low-pass filtering to extract the low-frequency signal components containing the rotor power frequency and sub-power frequency bands. The cutoff frequency of the low-pass filter should be slightly higher than 2 to 3 times the rotor power frequency to retain necessary subharmonic information while removing high-frequency interference.

[0074] Secondly, resampling and windowing are performed on the low-frequency signal components to improve the frequency resolution of the low-frequency signal components in the target low-frequency band and suppress spectral leakage. Since low-frequency characteristics (especially subharmonics such as 0.5X and 0.38X) are often close to the power frequency and have weak amplitude, if FFT is directly performed on the original long sequence, it may not be effectively separated due to the limited frequency resolution; while if only the short sequence is truncated, it will lead to serious spectral leakage.

[0075] Therefore, this embodiment employs resampling technology to downsample the signal to a sampling rate suitable for low-frequency analysis, and then applies window functions with good sidelobe suppression characteristics, such as the Hanning window or Flat Top window, thereby improving the detection signal-to-noise ratio of weak signals while ensuring frequency accuracy. Then, a Fast Fourier Transform is performed on the windowed low-frequency signal components to generate refined spectral data. Here, spectral refinement algorithms such as Zoom-FFT or Chirp-Z transform can be used to achieve local high-resolution analysis of the target low-frequency band without increasing the overall computational load.

[0076] After obtaining the high-resolution refined spectrum, the corresponding rotor first harmonic component and the subharmonic component of the preset order are identified in the refined spectrum data based on the real-time rotational speed of the circulating water pump. The real-time rotational speed can be obtained by tracking the power frequency peak of the key phase signal or the vibration signal itself, and this is used as a frequency reference to dynamically locate each characteristic component, so as to avoid frequency offset error caused by rotational speed fluctuation.

[0077] Finally, the amplitudes of the rotor's first harmonic component and the subharmonic component are extracted. Based on a preset weighting coefficient, the amplitudes of the rotor's first harmonic component and the subharmonic component are weighted and summed. The weighted summation result is used as the low-frequency vibration characteristic. The physical significance of this weighted summation step lies in comprehensively characterizing the combined effect of radial constraint reduction. The first harmonic amplitude primarily reflects the amplification of the unbalanced response caused by increased clearance, while the subharmonic amplitude directly indicates instability phenomena such as oil film eddy or nonlinear friction. For example, in a preferred embodiment, the weighted summation formula can be expressed as LF = α × A_1X + β × A_sub, where A_1X is the first harmonic amplitude and A_sub is the subharmonic amplitude (which can be the sum of 0.5X and 0.38X amplitudes or the maximum value). The weighting coefficients α and β can be calibrated according to the equipment type and wear sensitivity. For long-shaft vertical circulating water pumps, which are more sensitive to instability, α = 0.4 and β = 0.6 can be set to strengthen the weight of the subharmonic component in diagnosis; while for horizontal pumps with stronger rigid support, α = 0.7 and β = 0.3 can be set. Through this configurable weighted fusion mechanism, the low-frequency vibration characteristics can be flexibly adapted to the wear evolution characteristics of circulating water pumps with different structural forms, avoiding the one-sidedness of a single frequency index under complex working conditions.

[0078] Based on the established temporal evolution correlation, this embodiment further translates this correlation into specific quantitative judgment rules and hierarchical early warning mechanisms for wear evolution stages.

[0079] Specifically, based on the temporal evolution correlation, the current wear evolution stage of the circulating water pump is determined, including the following three determination scenarios.

[0080] In the first scenario, when the first rate of change is greater than the second rate of change, and the ratio of the high-frequency vibration feature to the low-frequency vibration feature shows a continuous upward trend, the circulating water pump is determined to be in the early wear stage. The physical essence of this determination condition lies in identifying the impact-dominated characteristics in the early stage of wear.

[0081] Under open-cycle conditions, when silt particles first begin to intrude into the bearing clearance, the radial constraint stiffness of the rotor remains relatively high because the bearing clearance is still within the design tolerance range. Therefore, the low-frequency vibration characteristics characterizing the shaft oscillation increase relatively slowly. However, the microscopic cutting and impact of hard particles on the metal surface are very active, leading to a rapid increase in the high-frequency vibration characteristics characterizing the impact.

[0082] At this time, the rate of change of high-frequency features is significantly higher than that of low-frequency features, and the ratio between the two continues to increase, indicating that the equipment is in a recoverable window period where abrasive erosion is active but the macroscopic geometry has not yet failed. By capturing this specific temporal correlation pattern, the present invention can keenly identify early abrasive wear hazards before the traditional total vibration value exceeds the standard.

[0083] In the second scenario, when the second rate of change exceeds a preset acceleration threshold, and the ratio trend reverses and shows a downward trend, the circulating water pump is determined to be in the mid-term wear stage. The preset acceleration threshold here is an empirical value derived from historical equipment degradation data, used to distinguish between normal operating fluctuations and abnormal accelerated degradation. This determination condition signifies a qualitative change in the wear mechanism, that is, from surface damage to system instability.

[0084] As wear accumulates, the bearing clearance gradually expands and exceeds the critical value. The load-bearing capacity of the water lubrication film decreases sharply, and the failure of the rotor radial constraint triggers oil film eddy or nonlinear friction, resulting in an explosive increase in low-frequency vibration characteristics. Its growth rate surpasses that of high-frequency characteristics, causing the originally rising ratio curve to turn into an inflection point and then decline. This non-monotonic reversal of the ratio is the most significant fingerprint feature of mid-term wear. It effectively eliminates the overall vibration increase caused by simply increased load or fluid turbulence, and accurately pinpoints the structural degradation caused by excessive clearance.

[0085] In the third scenario, when both the high-frequency vibration characteristic and the low-frequency vibration characteristic exceed their respective preset danger thresholds, and both the first rate of change and the second rate of change exhibit divergent behavior, the circulating water pump is determined to be in the late wear stage.

[0086] At this stage, wear has entered an irreversible and rapid deterioration phase. Not only do the absolute amplitudes of both high-frequency and low-frequency characteristics exceed safety limits, but more importantly, their rates of change no longer follow a stable correlation, instead exhibiting a random and violent divergence. This typically indicates severe peeling of the bearing bushing, scratches on the journal, or even intermittent rubbing between moving and stationary parts, pushing the system's dynamic behavior to the brink of chaos. At this point, the diagnostic focus shifts from trend prediction to emergency status confirmation, aiming to prevent catastrophic accidents.

[0087] It should be understood that the above three judgment scenarios do not necessarily have to be triggered in strict chronological order in actual implementation. Under certain extreme conditions, such as a sudden influx of a large amount of mud and sand or an interruption of cooling water, the equipment may skip the early stage and directly enter the middle or even late stage of wear.

[0088] Therefore, the determination logic of the present invention preferably adopts a parallel monitoring or multi-state machine mechanism to evaluate the confidence level of the current data in meeting the conditions of each stage in real time, and takes the stage with the highest confidence level as the final output, thereby ensuring the real-time performance and robustness of the diagnosis.

[0089] After accurately determining the wear evolution stage, in order to form a complete operation and maintenance closed loop, the method further includes: Based on the determined wear evolution stage, a corresponding level of early warning instruction is generated. Unlike existing technologies that trigger a single alarm solely based on vibration amplitude, the early warning instruction generated in this embodiment includes handling suggestions that are deeply matched to the physical stage of wear, achieving a leap from status notification to decision support.

[0090] Specifically, when the early wear stage is determined, a first warning instruction containing suggestions for improving lubrication is generated. Since early wear is mainly caused by abrasive impact and the clearance is still recoverable, the optimal strategy at this time is not shutdown for maintenance, but rather to alleviate wear by adjusting operating parameters. For example, the first warning instruction may specifically suggest increasing the cooling water flow rate to enhance the flushing and sand removal effect, switching to the standby water filter to reduce the sand content entering the pump, or appropriately reducing the pump speed to reduce the kinetic energy of particle impact. These measures can effectively extend the duration of the early wear stage without affecting production, and postpone the time when the equipment enters the irreversible deterioration stage.

[0091] When the system is determined to be in the intermediate wear stage, a second early warning instruction containing a planned maintenance recommendation is generated. Given that intermediate wear represents a substantial decrease in radial constraint, simply adjusting operating parameters cannot reverse the deterioration trend; physical intervention is necessary. The second early warning instruction should clearly indicate the location of the fault (e.g., "lower guide bearing clearance exceeds standard") and the rate of deterioration, and recommend a maintenance window accordingly. For example, the system can predict the remaining service life based on the current low-frequency characteristic second rate of change and generate a specific work order recommending the replacement of the guide bearing bushing during low-load periods within 30 days. This helps the maintenance department transform unplanned outages into controllable planned maintenance, minimizing the impact on power generation plans.

[0092] When the device is determined to be in the late wear stage, a third warning instruction is generated, which includes a recommendation for immediate shutdown and inspection. At this point, the equipment is on the verge of failure, and any delay could lead to serious consequences such as shaft breakage or motor burnout. The third warning instruction has the highest priority and, in addition to being pushed to the maintenance terminal, can also be directly linked to the DCS system to trigger audible and visual alarms or automatic trip protection.

[0093] The instructions should emphasize immediate emergency shutdown and list the essential inspection items after shutdown, such as checking the wear depth of the journal and measuring the concentricity of the bearing housing, to provide timely guidance for subsequent emergency repairs. By accurately mapping the content of the warning instructions to the wear evolution stage, this invention not only solves the problem of when to repair, but also answers the questions of how to repair and what to repair, significantly improving the intelligence level and economic benefits of open circulating water pump operation and maintenance management.

[0094] A second aspect of the present invention provides a vibration and wear correlation diagnostic system for an open-loop circulating water pump, comprising: The data acquisition module is configured to collect multi-dimensional vibration signals during the operation of the pump unit through vibration sensors installed on the bearings and casing of the circulating water pump. The data processing module is configured to perform frequency domain decoupling on the multidimensional vibration signal, extract high-frequency vibration features characterizing fluid particle impact and low-frequency vibration features characterizing the decrease in rotor radial constraint; construct a temporal evolution correlation based on the difference in the rate of change and the ratio trend of the high-frequency vibration features and low-frequency vibration features within a preset time window; and determine the current wear evolution stage of the circulating water pump based on the temporal evolution correlation.

[0095] To further verify the effectiveness and engineering practical value of the present invention, this embodiment uses a vertical circulating water pump in an open circulating cooling water system of a coastal power plant as the application object for actual scenario verification. It should be understood that this application scenario is only an illustrative example, intended to demonstrate the implementation effect of the technical solution of the present invention under typical lubrication conditions of fluids containing impurities, and is not intended to limit the scope of protection of the present invention.

[0096] In other embodiments, the present invention is also applicable to rotating mechanical equipment with similar wear mechanisms, such as river water intake pumping stations and open cooling tower circulating pumps in chemical plants.

[0097] In this verification scenario, the target pump unit is a long-shaft vertical structure, directly cooled by seawater. The sediment content in the water fluctuates with tides and seasons, and the guide bearing is water-lubricated, which conforms to the typical characteristics of open-circulation operation.

[0098] The system continuously collected and analyzed the pump set's operating data for 24 months according to the method described in the aforementioned embodiment. In order to intuitively demonstrate the wear evolution process, high-frequency vibration characteristics (specifically, the envelope spectrum energy index, denoted as HF) characterizing the impact of hard particles and low-frequency vibration characteristics (specifically, the 0.5 harmonic amplitude index, denoted as LF) characterizing the decrease in rotor radial constraint were selected as core observation variables, and the root mean square value of traditional vibration velocity (RMS) was recorded simultaneously as a comparison benchmark.

[0099] During the initial 0 to 6 months of normal operation after the equipment was put into use, monitoring data showed that both the HF and LF characteristics remained at low and stable levels, and the ratio between the two remained within the preset health baseline range. The system did not trigger any warning commands, which was consistent with the good equipment condition recorded in the on-site inspection.

[0100] After six months, the HF characteristic began to show a slow but continuous upward trend, while the LF characteristic remained relatively stable. In the ninth month, the system calculated that the first rate of change of the HF characteristic was significantly greater than the second rate of change of the LF characteristic, and the HF / LF ratio showed a continuous upward trend, satisfying the criteria for determining the early wear stage in Example 5. At this time, the traditional RMS index was only 1.2 mm / s, far below the attention zone threshold (2.8 mm / s) specified in the ISO 10816 standard. Relying on traditional methods would fail to identify potential problems at this stage.

[0101] Based on this, the system automatically generated a first warning instruction that suggested switching to the backup water filter and appropriately increasing the cooling water flow to enhance the sand removal effect. After the maintenance personnel adopted the suggestion, the upward slope of the HF characteristic slowed down significantly, successfully extending the healthy operation cycle of the equipment and verifying the intervention value of the invention in the early wear stage.

[0102] As the operating time progressed to around the 15th month, despite the implementation of improved lubrication measures, the guide bearing clearance gradually widened due to the long-term cumulative effect. The monitoring curve showed that the LF characteristic began to show signs of accelerated increase, and its second rate of change exceeded the preset acceleration threshold. At the same time, the previously continuously rising HF / LF ratio curve reached a significant inflection point in the 16th month and turned to a downward trend. This combination of characteristics accurately triggered the judgment logic for the mid-term wear period.

[0103] In contrast, the traditional RMS index has just risen to 3.5 mm / s at this time. Although it has entered the attention zone, it has not yet reached the alarm value. It is very easy to be misjudged as a general imbalance or load fluctuation. However, the present invention, based on the ratio reversal judgment mechanism, keenly captures the critical point of the transformation of the fault-dominant mechanism from surface impact to structural instability.

[0104] The system then generated a second warning instruction, which included a suspected excessive clearance in the lower guide bearing and suggested scheduled maintenance within 30 days during a low-load window. Subsequent disassembly and inspection confirmed that the wear of the guide bearing bushing had reached 1.8 times the design clearance, which was highly consistent with the diagnosis and avoided the risk of sudden bearing seizure caused by operating the equipment with defects.

[0105] During the late wear stage from month 21 to month 24, due to the failure to repair in a timely and thorough manner, both the HF and LF characteristics exceeded their respective preset danger thresholds, and the rate of change of both showed an irregular divergent oscillation state. Based on this, the system determined that the equipment had entered the late wear stage and immediately pushed out the third early warning command, and the linkage control system executed an emergency shutdown protection.

[0106] Subsequent inspection revealed minor scratches on the journal, but timely shutdown prevented catastrophic consequences. Throughout the verification period, compared to the traditional RMS threshold alarm method, this invention achieved an early wear detection lead time of approximately 9 months and an intermediate structural degradation detection lead time of approximately 5 months, and no false alarms caused by fluctuations in operating conditions occurred during the entire process.

[0107] In summary, through application verification in actual industrial scenarios, it has been fully demonstrated that the diagnostic method proposed in this invention, based on the temporal evolution correlation between high-frequency impact characteristics and low-frequency constraint characteristics, can effectively overcome the technical challenges of strong background noise and high coupling degree of fault characteristics under open-cycle operating conditions. This method can not only accurately distinguish different physical stages of wear, but also transform the diagnostic results into executable operation and maintenance decisions through graded early warning instructions, realizing a substantial leap from passive post-event alarms to proactive predictive maintenance, and has significant engineering application value and promotion prospects.

[0108] It should be noted that the specific time points, characteristic values ​​and threshold parameters involved in the above verification process are all adaptive results for the specific pump set and water quality conditions. In practical applications, those skilled in the art can calibrate and adjust the above parameters according to the structural characteristics of the specific equipment, historical operating data and operation and maintenance strategies, referring to the principles described in the foregoing embodiments. Such adaptive modifications should all be covered within the protection scope of this invention.

[0109] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit its scope of protection. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that after reading the present invention, they can still make various changes, modifications or equivalent substitutions to the specific implementation of the invention, but these changes, modifications or equivalent substitutions are all within the scope of protection of the pending claims of the invention.

Claims

1. A method for diagnosing vibration and wear correlation in open-type circulating water pumps, characterized in that, include: Vibration sensors installed on the bearings and casing of the circulating water pump are used to collect multi-dimensional vibration signals during the operation of the pump unit. The multidimensional vibration signal is decoupled in the frequency domain to extract high-frequency vibration features characterizing fluid particle impact and low-frequency vibration features characterizing the decrease in rotor radial constraint. Based on the difference in the rate of change and the ratio trend of the high-frequency vibration characteristics and the low-frequency vibration characteristics within a preset time window, a temporal evolution correlation is constructed. Based on the aforementioned temporal evolution correlation, the current wear evolution stage of the circulating water pump is determined.

2. The method for diagnosing vibration and wear correlation in open-type circulating water pumps according to claim 1, characterized in that, The high-frequency vibration characteristics correspond to the natural frequency response range of the circulating water pump bearing and pump casing components, and are used to characterize the abrasive impact of hard particles in the fluid on the pump body structure. The low-frequency vibration characteristics correspond to the power frequency and sub-power frequency bands of the pump rotor system, and are used to characterize the operating state of the rotor due to decreased radial support stiffness and failure of radial constraint caused by bearing wear.

3. The vibration and wear correlation diagnosis method for open circulating water pumps according to claim 1, characterized in that, Constructing the aforementioned temporal evolution correlation specifically includes: Within a preset sliding time window, the first rate of change corresponding to the high-frequency vibration feature and the second rate of change corresponding to the low-frequency vibration feature are calculated respectively. Based on the relative difference between the first and second rates of change, and combined with the temporal variation trend of the ratio of high-frequency vibration characteristics to low-frequency vibration characteristics, a temporal evolution correlation is established to characterize the wear evolution law of the pump group.

4. The vibration and wear correlation diagnosis method for open circulating water pumps according to claim 1, characterized in that, Extracting the high-frequency vibration features includes: The multidimensional vibration signals collected from the bearing and casing of the circulating water pump are subjected to bandpass filtering to extract the high-frequency signal components of the corresponding component's natural frequency response segment. Perform a Hilbert transform on the high-frequency signal components to obtain the signal envelope; Perform a Fast Fourier Transform on the envelope signal to generate envelope spectrum data; The peak energy of the spectral frequency range within the envelope spectrum is extracted and used as a high-frequency vibration characteristic to characterize abrasive impact.

5. The method for diagnosing vibration and wear correlation in open-type circulating water pumps according to claim 1, characterized in that, Extracting the high-frequency vibration features includes: High-pass filtering was performed on the multidimensional vibration signals collected from the bearings and casing of the circulating water pump to obtain high-frequency impact signals characterizing particle impact. Calculate the peak value to root mean square ratio of the high-frequency impulse signal within a preset sampling period; The signal exceeding the preset impact energy threshold is integrated to obtain the vibration peak energy, which is then used as the high-frequency vibration characteristic to characterize abrasive impact.

6. The method for diagnosing vibration and wear correlation in open-type circulating water pumps according to claim 1, characterized in that, Extracting the low-frequency vibration features includes: The multidimensional vibration signals collected from the bearings and casing of the circulating water pump are subjected to low-pass filtering to obtain the low-frequency signal components of the rotor power frequency and sub-power frequency bands. The low-frequency signal components are sequentially resampled and windowed to suppress spectral leakage, and then subjected to fast Fourier transform to obtain a high-resolution refined spectrum. Based on the real-time speed of the pump unit, the rotor's first harmonic component and the preset order subharmonic component are identified. The low-frequency vibration characteristics that characterize the decrease in rotor radial constraint are obtained by weighting and summing the first harmonic amplitude and the subharmonic amplitude according to a preset weighting coefficient.

7. The method for diagnosing vibration and wear correlation in open-type circulating water pumps according to claim 1, characterized in that, The specific steps for determining the wear evolution stage include: When the first rate of change of the high-frequency vibration characteristics is greater than the second rate of change of the low-frequency vibration characteristics, and the ratio of high-frequency to low-frequency characteristics continues to rise, the circulating water pump bearing and rotor system are determined to be in the early wear stage. When the second rate of change of the low-frequency vibration characteristics exceeds the preset acceleration threshold and the ratio of high-frequency to low-frequency characteristics reverses and decreases, it is determined to be in the middle wear stage. When both high-frequency and low-frequency vibration characteristics exceed the preset danger threshold and their rates of change diverge, the system is determined to be in the late wear stage.

8. The method for diagnosing vibration and wear correlation in open-type circulating water pumps according to claim 1, characterized in that, The method further includes: Based on the determined wear evolution stage of the circulating water pump bearing and rotor system, a corresponding level of early warning instruction is generated; When the bearing is in the early wear stage, a first warning command is generated to improve the bearing lubrication condition. When the vehicle is in the mid-term wear and tear stage, a second early warning instruction is generated for planned maintenance. When the machine is in the late wear stage, a third warning command is generated for immediate shutdown and inspection. The warning instruction will then be pushed to the operation and maintenance terminal.

9. The method for diagnosing vibration and wear correlation in open-type circulating water pumps according to claim 1, characterized in that, After acquiring the multidimensional vibration signal, the following steps are also included: Adaptive filtering is performed on the multidimensional vibration signals collected from the bearings and casing of the circulating water pump to remove environmental noise and electromagnetic interference. Temperature drift compensation is performed on the vibration signal based on the real-time temperature data of the pump unit. By removing outliers from the compensation signal, a clean target vibration signal is obtained. A vibration health baseline model was established based on the historical normal operation data of the pump set.

10. A vibration and wear correlation diagnostic system for an open-type circulating water pump, characterized in that, include: The data acquisition module is configured to collect multi-dimensional vibration signals during the operation of the pump unit through vibration sensors installed on the bearings and casing of the circulating water pump. The data processing module is configured to perform frequency domain decoupling on the multidimensional vibration signal, extract high-frequency vibration features characterizing fluid particle impact and low-frequency vibration features characterizing the decrease in rotor radial constraint; construct a temporal evolution correlation based on the difference in the rate of change and the ratio trend of the high-frequency vibration features and low-frequency vibration features within a preset time window; and determine the current wear evolution stage of the circulating water pump based on the temporal evolution correlation.

Images