Lubricating oil on-line monitoring system suitable for power plant unit non-stop scene

Abstract

This invention relates to the field of lubricating oil monitoring technology, specifically to an online lubricating oil monitoring system suitable for power plant unit non-stop operation scenarios. The system includes a data acquisition module for collecting vibration and viscosity signals; a data analysis module for determining the instantaneous vibration influence coefficient based on the amplitude value and the difference between the vibration frequency and the natural frequency; determining the cumulative interference index at the current moment based on the instantaneous vibration influence coefficient at historical times; determining the instantaneous frequency risk weight based on the cumulative interference index, the vibration frequency at the current moment, and the natural frequency, and determining the resonance risk index through recursive fusion calculation; and determining the target process noise covariance based on the cumulative interference index and the resonance risk index. The output module substitutes the target process noise covariance into the Kalman filter prediction step to filter the viscosity signal, obtaining a viscosity value reflecting the true state of the lubricating oil, thereby reducing the probability of false alarms and missed alarms regarding oil condition.

Description

Technical Field

[0001] This invention relates to the field of lubricating oil monitoring technology, and more specifically to an online lubricating oil monitoring system suitable for power plant unit shutdown scenarios. Background Technology

[0002] As the core equipment of the power system, the long-term stable operation of power plant units directly determines the reliability and security of the power grid. Lubricating oil, as a critical protective medium for the rotating components of the unit, makes online monitoring of its quality a vital industry process. Vibratory viscometers are easy to install online and reflect viscosity by measuring the damping effect of the oil on the vibrating unit. Generally, abnormal increases or decreases in oil viscosity can lead to excessively thick or insufficient lubricating film, causing accelerated component wear, localized overheating, and even shaft jamming. In severe cases, shutdown for maintenance is necessary, resulting in significant economic losses and the risk of power outages.

[0003] Currently, in existing technologies, traditional Kalman filters are often used to smooth and denoise viscosity signals. However, this method usually uses fixed process noise and measurement noise covariance matrices. When faced with dynamic changes in operating conditions such as power plant unit load and speed, filters with fixed parameters show obvious limitations and cannot accurately distinguish between unit vibration interference and actual oil viscosity changes, resulting in poor filtering performance. Summary of the Invention

[0004] To address the technical problem of inaccurately distinguishing between unit vibration interference and actual oil viscosity changes due to the use of fixed process noise and measurement noise covariance matrices, this invention provides an online lubricating oil monitoring system suitable for power plant unit non-stop operation scenarios. The specific technical solution adopted is as follows:

[0005] This invention proposes an online lubricating oil quality monitoring system suitable for power plant unit non-stop operation scenarios. The system includes:

[0006] The data acquisition module is used to simultaneously acquire vibration signal data and lubricating oil viscosity signal data of power plant unit operation. The vibration signal includes at least amplitude and vibration frequency.

[0007] The data analysis module, communicating with the data acquisition module, is used to determine the instantaneous vibration influence coefficient based on the distribution characteristics of amplitude values ​​in amplitude data and the difference between vibration frequency and the natural frequency of the viscosity signal; based on the instantaneous vibration influence coefficient at historical moments and combined with the time-domain attenuation characteristics of vibration frequency at historical moments under disturbance, it determines the cumulative interference index characterizing the cumulative effect of historical disturbance at the current moment; based on the cumulative interference index and the Gaussian function relationship between the vibration frequency at the current moment and the natural frequency, it determines the instantaneous frequency risk weight, and determines the resonance risk index at the current moment through recursive fusion calculation; based on the cumulative interference index and the resonance risk index, it dynamically adjusts the preset baseline process noise covariance to determine the target process noise covariance.

[0008] The output module, which communicates with the data analysis module, is used to substitute the target process noise covariance into the prediction step of the Kalman filter to filter the viscosity signal and obtain a viscosity value that reflects the true state of the lubricating oil.

[0009] Furthermore, the synchronous acquisition of vibration signal data and lubricating oil viscosity signal data of the power plant unit operation includes:

[0010] A unified clock reference is provided for the acquisition of vibration signal data and viscosity signal data, and the vibration signal is sampled. The sampling frequency is configured to be sufficient to obtain the main frequency components of the unit vibration.

[0011] Furthermore, the process for determining the instantaneous vibration influence coefficient includes:

[0012] A set of continuous amplitude data is sorted in ascending order of amplitude value to generate an amplitude sequence. The subset of amplitudes greater than the upper quartile is extracted as strong amplitude data. The arithmetic mean of all amplitude values ​​in the strong amplitude data is calculated as the reference amplitude value.

[0013] For any given moment, the ratio of the amplitude value to the reference amplitude value is calculated and used as the amplitude influence factor.

[0014] For any given moment, calculate the absolute difference between the vibration frequency and the natural frequency, which is taken as the frequency difference; and calculate the ratio of the frequency difference to the natural frequency, which is taken as the frequency deviation index.

[0015] The instantaneous vibration influence coefficient is determined based on the amplitude influence factor, frequency deviation index, and preset attenuation coefficient.

[0016] Furthermore, determining the instantaneous vibration influence coefficient based on the amplitude influence factor, frequency deviation index, and preset attenuation coefficient includes:

[0017] Using the natural constant as the base, the frequency influence factor is obtained by taking the opposite of the product of the frequency deviation index and the preset attenuation coefficient and then performing an exponential operation.

[0018] Calculate the product of the amplitude influence factor and the frequency influence factor, and use it as the instantaneous vibration influence coefficient at any given moment.

[0019] Furthermore, the process for determining the cumulative interference index includes:

[0020] The time span required for the vibration disturbance state to go from the initial state to the steady state under the preset standard vibration frequency is obtained and recorded as the reference time constant; and the time length between the historical moment and the current moment is obtained and recorded as the time difference.

[0021] For each historical moment within a preset time window, extract the historical instantaneous vibration influence coefficient and historical vibration frequency of the historical moments preceding the current moment;

[0022] For each historical moment within a preset time window, the ratio of the preset standard vibration frequency to the historical vibration frequency is calculated as an indicator of frequency level.

[0023] Based on the frequency high / low index, the reference time constant and the corresponding time difference, the decay rate weight of historical moments is obtained through a preset decay weight calculation function.

[0024] The sum of the products of all historical instantaneous vibration influence coefficients and their corresponding decay rate weights within a preset time window is calculated as the cumulative disturbance index.

[0025] Furthermore, the step of obtaining the attenuation rate weight for historical moments based on the frequency high / low index, the reference time constant, and the corresponding time difference, through a preset attenuation weight calculation function, includes:

[0026] Using the frequency index as the base and the preset sensitivity constant as the exponent, a power operation is performed to obtain the attenuation rate coefficient; the product of the attenuation rate coefficient and the reference time constant is calculated as the attenuation time constant.

[0027] Calculate the product of the order of historical moments within the preset time window and the corresponding time difference, and use it as the first product;

[0028] Using the natural constant as the base, the inverse of the ratio of the first product to the corresponding decay time constant is exponentially calculated to obtain the decay rate weight.

[0029] Furthermore, the instantaneous frequency risk weight determination process includes:

[0030] Obtain the historical cumulative interference index for each historical moment within the preset time window; calculate the arithmetic mean of all historical cumulative interference indices as the reference cumulative interference index;

[0031] Calculate the ratio of the current cumulative interference index to the reference cumulative interference index, and use it as the interference intensity factor;

[0032] Based on the current vibration frequency and natural frequency, the instantaneous probability value for characterizing resonance caused by the vibration frequency is obtained through a Gaussian function.

[0033] Calculate the product of the interference intensity factor and the instantaneous probability value, and use it as the instantaneous frequency risk weight.

[0034] Furthermore, the recursive fusion calculation process is as follows:

[0035] Calculate the product of the resonance risk index of the previous time step and the preset risk forgetting factor, and use it as the second product;

[0036] Calculate the difference between 1 and the preset risk forgetting factor, and use it as the first difference;

[0037] Calculate the product of the instantaneous frequency risk weight at the current moment and the first difference, and use it as the resonance risk indicator at the current moment.

[0038] Furthermore, the target process noise covariance determination process includes:

[0039] Obtain the first preset weight coefficient and the second preset weight coefficient;

[0040] Calculate the product of the first preset weight coefficient and the resonance risk indicator as the resonance risk adjustment indicator; calculate the product of the second preset weight coefficient and the cumulative interference indicator as the interference adjustment indicator.

[0041] Calculate the sum of the positive integer 1, the interference adjustment index, and the resonance risk adjustment index, and use it as the overall adjustment factor;

[0042] The product of the preset baseline process noise covariance and the overall adjustment factor is calculated as the target process noise covariance.

[0043] Furthermore, the online lubricating oil monitoring system suitable for power plant unit non-stop operation also includes an oil evaluation module, which is connected to the output module to receive viscosity values, output oil deterioration scores through a pre-trained oil deterioration model, and generate an early warning signal when the oil deterioration score exceeds a preset threshold.

[0044] The present invention has the following beneficial effects:

[0045] This invention constructs a multi-level interference assessment process consisting of instantaneous vibration influence coefficient, cumulative interference index, and resonance risk index. This process accurately identifies and quantifies the dynamic interference mechanism of unit vibration on viscometer-measured viscosity data. Specifically, the process considers not only the instantaneous energy of the vibration but also the difference between the vibration frequency and the natural frequency, as well as the frequency-related time-domain attenuation characteristics, thus achieving a refined and comprehensive characterization of complex interference. By integrating the two core output cumulative interference indices (characterizing objective interference intensity) and resonance risk index (characterizing subjective inaccuracy risk) into the dynamic adjustment of the Kalman filter process noise covariance, the adaptive capability and robustness of the Kalman filter under dynamic operating conditions are improved. After adaptive filtering, the system can output a smoother and more accurate viscosity value reflecting the true state of the lubricating oil. This provides a high-quality data foundation for subsequent oil degradation analysis based on viscosity parameters, thereby significantly reducing the probability of false alarms and missed alarms regarding oil condition. Attached Figure Description

[0046] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0047] Figure 1 A schematic diagram of the structure of an online lubricating oil monitoring system suitable for power plant unit non-stop operation scenarios provided in an embodiment of the present invention;

[0048] Figure 2 An example diagram illustrating the process of determining cumulative interference indicators provided in one embodiment of the present invention;

[0049] Figure 3 This is an example diagram illustrating the instantaneous frequency risk weight determination process provided in one embodiment of the present invention. Detailed Implementation

[0050] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of an online lubricating oil monitoring system suitable for power plant unit non-stop operation scenarios proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0051] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0052] The following description, in conjunction with the accompanying drawings, details a specific solution for an online lubricating oil monitoring system suitable for power plant unit non-stop operation scenarios provided by the present invention.

[0053] Please see Figure 1 The diagram illustrates a schematic of an online lubricating oil monitoring system for power plant unit shutdown scenarios, according to an embodiment of the present invention. The method includes:

[0054] The data acquisition module 10 is used to simultaneously acquire vibration signal data and lubricating oil viscosity signal data of the power plant unit operation, wherein the vibration signal includes at least amplitude and vibration frequency.

[0055] It is understood that the data acquisition module can collect vibration signals of the unit operation through one or more vibration sensors (such as piezoelectric accelerometers). These sensors are preferably installed in key locations that best reflect the overall vibration state of the unit, such as bearing housings, gearbox housings, or the unit base.

[0056] It is understandable that the data acquisition module can also acquire the viscosity signal of lubricating oil through a vibratory viscometer. Generally speaking, the viscometer vibrates in the oil through its internal vibrating element. The viscosity damping of the oil will change the vibration characteristics of the vibrating element. These changes are converted into electrical signal output, which is the original viscosity signal.

[0057] In this embodiment, a unified clock reference is provided for the acquisition of vibration signal data and viscosity signal data, and the vibration signal is sampled. The sampling frequency is configured to be sufficient to obtain the main frequency components of the unit vibration.

[0058] It should be noted that the specific value of the sampling frequency is determined according to actual needs, and this embodiment does not impose a specific limitation. For example, in order to perform high-frequency sampling of the vibration signal and ensure that the Nyquist sampling theorem is satisfied so as to restore the main frequency components of the unit vibration without distortion, the sampling frequency is not less than 10kHz.

[0059] The data analysis module 12, which communicates with the data acquisition module, is used to determine the instantaneous vibration influence coefficient based on the distribution characteristics of the amplitude value in the amplitude data and the difference between the vibration frequency and the natural frequency of the viscosity signal; based on the instantaneous vibration influence coefficient at historical moments and combined with the time-domain attenuation characteristics of the vibration frequency at historical moments when subjected to disturbance, it determines the cumulative interference index characterizing the cumulative effect of historical disturbance at the current moment; based on the cumulative interference index and the Gaussian function relationship between the vibration frequency at the current moment and the natural frequency, it determines the instantaneous frequency risk weight, and determines the resonance risk index at the current moment through recursive fusion calculation; based on the cumulative interference index and the resonance risk index, it dynamically adjusts the preset baseline process noise covariance to determine the target process noise covariance.

[0060] It is important to understand that the impact of unit vibration on the viscometer is not simply a matter of noise superposition, but rather it is achieved by altering the dynamic environment of its vibrating unit. Since resonance is highly likely to occur when the external vibration frequency is close to the viscometer's natural operating frequency, even a small external energy input can cause the vibrating unit to produce a huge amplitude response. The stronger the interference effect, the greater the impact. Furthermore, it is known that amplitude can characterize the intensity of vibration, and vibration frequency is the dominant frequency component of vibration. Therefore, by analyzing the energy level of vibration (determined by amplitude) and the risk of triggering resonance (determined by the difference between vibration frequency and natural frequency), the intensity of interference at different times can be determined.

[0061] In order to eliminate the dimensional differences in absolute vibration amplitude between different units or sensors, and given that resonance is usually caused by high-frequency components, it is necessary to establish a benchmark, i.e., a reference amplitude value, that represents the typical vibration level of the unit under healthy, high-load steady-state operation.

[0062] The inherent frequency of the viscosity signal, the mechanical resonance frequency of the internal viscometer when its internal vibrating element vibrates freely in a specific designed medium, is a core physical property of the viscometer itself, and the inherent frequency cannot be zero. For example, for ultra-high precision viscometers based on quartz crystals, the frequency can reach 5MHz.

[0063] In this embodiment, a set of continuous amplitude data is arranged in ascending order of amplitude value to generate an amplitude sequence. The subset of amplitude data with amplitude values ​​greater than the upper quartile is extracted as strong amplitude data. The arithmetic mean of all amplitude values ​​in the strong amplitude data is calculated as a reference amplitude value. For any amplitude value at any given time, the ratio of the amplitude value to the reference amplitude value is calculated as an amplitude influence factor. For any vibration frequency at any given time, the absolute difference between the vibration frequency and the natural frequency is calculated as a frequency difference. The ratio of the frequency difference to the natural frequency is also calculated as a frequency deviation index. Based on the amplitude influence factor, the frequency deviation index, and a preset attenuation coefficient, the instantaneous vibration influence coefficient is determined.

[0064] It should be noted that the upper quartile is the 75th percentile. Determining the upper quartile is an existing technical method, and will not be elaborated on in this embodiment.

[0065] The reference amplitude value characterizes the typical vibration intensity of the unit during stable high-load operation.

[0066] It should be noted that a power plant unit is a huge and complex rotating mechanical system. During its operation, the rotor cannot achieve absolute dynamic balance. Even in a new unit, there are slight deformations and gaps. Therefore, under any operating condition, including the most stable steady-state condition, there must be a basic level of non-zero vibration, that is, the reference amplitude value cannot be zero.

[0067] Since the amplitude of the vibration directly determines the amount of kinetic energy applied to the viscometer's vibration unit, a larger amplitude means more energy is transferred, which may force the viscometer's vibration unit to produce undesired resonance or amplitude distortion, thus causing the viscosity reading to be distorted. Therefore, a larger amplitude influence factor indicates that the current vibration is stronger than typical, and the risk of interference is high.

[0068] It is important to understand that the interference effect decays exponentially over time. However, "how fast it decays" is a problem that needs to be precisely controlled. Different unit-viscometer systems have different sensitivities to frequency proximity. Therefore, it is necessary to further analyze the interference intensity experienced by the unit-viscometer system at different times by combining the different decay conditions of the interference effect.

[0069] In this embodiment, the frequency influence factor is obtained by taking the opposite of the product of the frequency deviation index and the preset attenuation coefficient with the natural constant as the base, and then performing an exponential operation. The product of the amplitude influence factor and the frequency influence factor is calculated as the instantaneous vibration influence coefficient at any time.

[0070] It should be noted that the preset attenuation coefficient quantifies the system's sensitivity to frequency deviation and is an adjustable parameter greater than 0. The larger the preset attenuation coefficient, the more sensitive the system is to frequency deviation; that is, even a slight frequency deviation causes the factor value to decay rapidly. Therefore, the specific value of the preset attenuation coefficient is determined based on the actual situation, and this embodiment does not impose a specific limitation. For example, different models of viscometers have different quality factors and damping characteristics of their vibrating elements, resulting in different sharpness of the resonance peaks. A system with a high quality factor has a very sharp resonance peak, corresponding to a larger preset attenuation coefficient (e.g., 3); a system with a low quality factor has a wider resonance peak, corresponding to a smaller preset attenuation coefficient (e.g., 0.5).

[0071] According to mechanical vibration theory, resonance occurs when the vibration frequency of the unit approaches the natural frequency of the viscometer. In this case, even a small external energy input can cause the viscometer's vibration unit to produce a huge amplitude response, leading to severe measurement inaccuracies. If the vibration frequency is the same as the natural frequency, the frequency deviation is zero, the frequency influence factor is at its maximum, and the risk of resonance is highest. Conversely, if the frequency difference increases, the frequency influence factor decreases rapidly (tending towards 0), indicating a sharp reduction in the risk of resonance. Therefore, assuming t represents the current moment, the instantaneous vibration influence coefficient can be expressed by the following formula:

[0072]

[0073] in, This represents the instantaneous vibration influence coefficient at the current moment; Indicates the amplitude influence factor; The frequency of vibration at the current moment is represented by F; F represents the natural frequency. Indicates the preset attenuation coefficient; This represents an exponential function with the natural constant as its base.

[0074] It is important to understand that the interference of unit vibration on the viscometer is not an instantaneous phenomenon. The noise interference will persist for a period of time through two paths: mechanical transmission and oil dynamic transmission. Furthermore, the attenuation characteristics of vibrations at different frequencies are different (high frequencies attenuate quickly, low frequencies attenuate slowly), and load changes will also cause the dominant vibration frequency to shift, thereby changing the attenuation behavior of the interference. The interference of unit vibration on the viscometer is not an instantaneous phenomenon, but a dynamic process with continuous and frequency-dependent attenuation. Therefore, the degree of noise interference at the current moment can be assessed by combining the changes in the interference effect at historical moments.

[0075] The process of determining the cumulative interference index is as follows: Figure 2 As shown, it includes:

[0076] S101-1: Obtain the time span required for the vibration disturbance state to go from the initial state to the steady state under the preset standard vibration frequency, and record it as the reference time constant; and obtain the time length between the historical moment and the current moment, and record it as the time difference value.

[0077] The preset standard vibration frequency is a pre-set constant frequency with a clear physical meaning, so that the vibration attenuation characteristics of different frequencies can be measured and compared under the same benchmark. The specific value of the preset standard vibration frequency is determined according to the actual situation, and this embodiment does not make a specific limitation. For example, the preset standard vibration frequency can be the rated rotation frequency of the generator set. For example, for a generator set with a rated speed of 3000 RPM, the rotation frequency of the generator set is 50 Hz, then the preset standard vibration frequency should be set to 50 Hz.

[0078] The reference time constant quantifies the duration for which vibration interference "survives" in the system at a preset standard vibration frequency. For example, it can be used to identify the duration of vibration interference from its initial state (e.g., the moment of load change) to the point where the vibration amplitude decays to a steady state (e.g., the steady state refers to the vibration amplitude decaying to its initial value) in historical viscosity signal data. It is obtained by fitting the time span of the 10% fluctuation band, for example, 10 seconds.

[0079] S101-2: For each historical moment within a preset time window, extract the historical instantaneous vibration influence coefficient and historical vibration frequency of the historical moments before the current moment.

[0080] It should be noted that the preset time window size can be adjusted. A larger window will take into account more overall trends and be more robust to noise interference, but it will be less sensitive to local data trend changes. On the other hand, a smaller window will be too sensitive to local data trends and be easily affected by noise interference. Therefore, the specific value of the preset time window is determined according to actual needs. This embodiment does not impose a specific limitation. For example, if the preset time window value is 10 and the sampling interval is 0.1 seconds, then the length of the time window is 10 × 0.1 seconds = 1 second. That is, when calculating the total interference at the current moment, the data in the past second will be considered.

[0081] S101-3: For each historical moment within a preset time window, calculate the ratio of the preset standard vibration frequency to the historical vibration frequency, which serves as an indicator of frequency level.

[0082] When the historical vibration frequency is greater than the preset standard vibration frequency, the frequency high / low index is less than 1, indicating that it is a high frequency; when the historical vibration frequency is less than the preset standard vibration frequency, the frequency high / low index is greater than 1, indicating that it is a low frequency.

[0083] It should be noted that, under actual operating conditions, the rotor rotates as long as the unit is running. According to Newton's laws, any unbalanced rotating mass will generate periodic centrifugal force, thereby causing vibration. Even under the most stable operating conditions of the unit, the measured vibration signal contains background noise from various sources. This noise energy is also distributed in the frequency band greater than zero, so the historical vibration frequency cannot be zero.

[0084] S101-4: Based on the frequency high / low index, the reference time constant and the corresponding time difference, the attenuation rate weight of historical moments is obtained through a preset attenuation weight calculation function.

[0085] It's important to understand that vibration disturbances don't disappear instantly. Generally, high-frequency vibrations are more easily dissipated by damping at structural connections, decaying rapidly. Low-frequency vibrations, on the other hand, are more likely to excite the overall modes of large structures, are more persistent, and decay more slowly. Therefore, without frequency-based weighting, a recently occurring high-frequency strong disturbance and a low-frequency moderate disturbance that has been present for some time might be incorrectly assessed as having the same residual effect because they decay at the same rate. In reality, however, the high-frequency disturbance may have almost disappeared, while the low-frequency disturbance continues to cause problems. Therefore, it's necessary to introduce a decay rate weight that considers both the passage of time and frequency characteristics to quantify how much residual effect a disturbance from a specific past moment has on the current measurement results.

[0086] To accurately obtain the attenuation rate weight, as an example, the calculation process of the preset attenuation weight calculation function is as follows: using the frequency index as the base and the preset sensitivity constant as the exponent, perform a power operation to obtain the attenuation rate coefficient; calculate the product of the attenuation rate coefficient and the reference time constant as the attenuation time constant; calculate the product of the order of the historical time within the preset time window and the corresponding time difference as the first product; using the natural constant as the base, perform a power operation on the negative of the ratio of the first product to the corresponding attenuation time constant to obtain the attenuation rate weight.

[0087] The decay rate coefficient quantifies the proportion of decay rate corresponding to the frequency of vibration at a given historical moment. If the frequency index is less than 1, the decay rate coefficient becomes smaller after exponentiation. This means high-frequency vibrations correspond to a smaller coefficient, indicating faster decay. Conversely, if the frequency index is greater than 1, the decay rate coefficient becomes larger after exponentiation. This means low-frequency vibrations correspond to a larger coefficient, indicating slower decay.

[0088] It should be noted that the preset sensitivity constant is used to amplify this difference in attenuation rate, and its specific value is not limited in this embodiment. The preset sensitivity constant is an adjustable parameter greater than 0. The larger the preset sensitivity constant, the smaller the coefficient calculated for high-frequency vibration (faster attenuation) and the larger the coefficient calculated for low-frequency vibration (slower attenuation), and the higher the system's sensitivity to frequency.

[0089] The decay time constant characterizes the specific decay time constant corresponding to the vibration at a historical moment, reflecting how quickly the historical disturbance will decay in the system. A small decay time constant indicates fast decay; a large decay time constant indicates slow decay.

[0090] It's important to understand that the longer the time (the larger the first product), the smaller the weight of the decay rate. However, for low-frequency vibrations (large decay time constant), the denominator is large, the exponential term decays slowly, and even over a long period, its decay rate weight is relatively large; for high-frequency vibrations (small decay time constant), the denominator is small, the exponential term decays quickly, and its decay time constant decreases rapidly.

[0091] S101-5: Calculate the sum of the products of all historical instantaneous vibration influence coefficients and their corresponding decay rate weights within the preset time window, and use this as the cumulative disturbance index.

[0092] Among them, the larger the historical instantaneous vibration influence coefficient, the more residual interference will follow; if the decay rate weight is smaller, it means that it is very likely a high-frequency vibration that decays quickly, so there will be less residual interference.

[0093] It's important to understand that a high-risk operating condition requires the simultaneous fulfillment of two conditions: a significant level of interference, meaning the total interference level of the current system is far higher than its typical state; and a significant possibility of resonance, meaning the current vibration frequency is very close to the viscometer's natural frequency, making resonance highly likely. Therefore, it can be understood that if there is only strong interference but frequency mismatch, the risk is manageable; if the frequency matches but the interference is weak, the risk is also limited. Only when both conditions are present simultaneously does it constitute the highest level of instantaneous risk. Therefore, the possibility of resonance risk needs to be analyzed from these two aspects.

[0094] The process of determining the instantaneous frequency risk weight is as follows: Figure 3 As shown, it includes:

[0095] S102-1: Obtain the historical cumulative interference index for each historical moment within the preset time window; calculate the arithmetic mean of all historical cumulative interference indices as a reference cumulative interference index.

[0096] The reference cumulative interference index serves as a benchmark for "interference level," allowing for comparison of interference levels across different units and operating conditions. Therefore, the reference cumulative interference index cannot be zero.

[0097] S102-2: Calculate the ratio of the current cumulative interference index to the reference cumulative interference index, and use it as the interference intensity factor.

[0098] It is important to understand that a larger interference intensity factor indicates that the current interference level is higher than the historical average, posing a risk of energy overload; conversely, a smaller interference intensity factor indicates that the current interference level is lower than the historical average, making it relatively safe.

[0099] S102-3: Based on the current vibration frequency and natural frequency, obtain the instantaneous probability value used to characterize the resonance caused by the vibration frequency through a Gaussian function.

[0100] The instantaneous probability value quantifies the probability of resonance occurring at the current moment based solely on frequency relationships.

[0101] Since the closer the vibration frequency is to the natural frequency at the current moment, the greater the probability of resonance, the instantaneous probability value can be expressed using the existing Gaussian function as follows:

[0102]

[0103] in, This represents the instantaneous probability value at the current moment; The frequency of vibration at the current moment is represented by F; F represents the natural frequency. The standard deviation of the Gaussian function is represented by... The specific determination method is based on existing technical means, and will not be described in detail in this embodiment; This represents an exponential function with the natural constant as its base.

[0104] S102-4: Calculate the product of the interference intensity factor and the instantaneous probability value as the instantaneous frequency risk weight.

[0105] It is important to understand that the output instantaneous frequency risk weight will only be high when both the interference intensity factor and the instantaneous probability value are high. If either of them is zero or very low, the final risk weight will be significantly reduced.

[0106] In this embodiment, the recursive fusion calculation process is as follows: calculate the product of the resonance risk index of the previous time and the preset risk forgetting factor as the second product; calculate the difference between 1 and the preset risk forgetting factor as the first difference; calculate the product of the instantaneous frequency risk weight of the current time and the first difference as the resonance risk index of the current time.

[0107] It's important to understand that in engineering practice, risks don't appear and disappear instantly. For example, after a strong resonance event, even if the external vibration frequency temporarily deviates from the danger zone, the resulting mechanical stress, thermal effects, or system instability will persist for some time; the risk isn't immediately eliminated. Therefore, recursive calculations are needed to simulate this risk memory effect, ensuring the system maintains reasonable vigilance and preventing drastic changes in warning signals.

[0108] It should be noted that the preset risk forgetting factor is used to control the decay rate of historical risks. The specific value of the preset risk forgetting factor is determined according to the actual situation, and this embodiment does not impose a specific limitation. The preset risk forgetting factor ranges from 0 to 1. The larger the preset risk forgetting factor (closer to 1), the better the system's "memory" is, the greater the weight of the historical resonance risk index in the resonance risk index at the current moment, the slower the risk state decays, and the more "cautious" the system is. The smaller the preset risk forgetting factor (closer to 0), the greater the system's "forgetfulness" is, the faster the historical resonance risk index is forgotten, and the system focuses more on the present.

[0109] The first difference is the difference between the total weight of 1 and the weight of historical risk as a preset risk forgetting factor. Since the total weight is 1, the weight of the current instantaneous risk value is naturally the difference between 1 and the preset risk forgetting factor.

[0110] Since a larger second product indicates a longer duration of historical resonance risk up to the previous moment, and a more significant residual impact of historical risk status on the current moment; and a larger instantaneous frequency risk weight indicates a higher instantaneous probability of resonance at the current frequency, the resonance risk index at the current moment can be expressed by the following formula:

[0111]

[0112] in, This indicates a pre-set risk forgetting factor; The resonance risk indicator of the previous time step (i.e., t-1) at the current moment; This indicates the resonance risk indicator at the current moment; This represents the instantaneous frequency risk weight at the current moment.

[0113] Output module 13 is connected to the data analysis module and is used to substitute the target process noise covariance into the prediction step of Kalman filtering to filter the viscosity signal and obtain a viscosity value that reflects the true state of the lubricating oil.

[0114] In this embodiment, a first preset weight coefficient and a second preset weight coefficient are obtained; the product of the first preset weight coefficient and the resonance risk index is calculated as the resonance risk adjustment index; the product of the second preset weight coefficient and the cumulative interference index is calculated as the interference adjustment index; the sum of the positive integer 1, the interference adjustment index, and the resonance risk adjustment index is calculated as the overall adjustment factor; and the product of the preset baseline process noise covariance and the overall adjustment factor is calculated as the target process noise covariance.

[0115] The first preset weighting coefficient and the second preset weighting coefficient are adjustable gains. The second preset weighting coefficient controls the response intensity to objective physical interference energy, while the first preset weighting coefficient controls the response intensity to subjective resonance risk level. The specific values ​​of the first preset weighting coefficient and the second preset weighting coefficient are determined according to the actual situation. For example, if the interference intensity and resonance risk are given the same level of importance, then both the first preset weighting coefficient and the second preset weighting coefficient are set to 0.5.

[0116] The preset baseline process noise covariance quantitatively describes the inherent and unavoidable inaccuracy of the mathematical model used to describe the viscosity change of lubricating oil when the unit is running smoothly and the influence of vibration on the viscometer is negligible. The method of obtaining the preset baseline process noise covariance is a well-known technique in the art, and will not be described in detail in this embodiment.

[0117] It's important to understand that the process noise covariance matrix of a Kalman filter represents the uncertainty of the system model. A larger process noise covariance matrix value indicates a less reliable model, meaning the filter will rely more heavily on current measurements. Therefore, a high resonance risk index indicates the system is in or has just emerged from a high resonance risk state, making the measurement system extremely fragile and prone to inaccuracies. Even if the current cumulative interference index is not large, this factor will proactively increase the process noise covariance, improving the filter's sensitivity to measurements and enabling it to respond and correct immediately when distortion occurs, acting as a "risk warning." Conversely, a larger cumulative interference index indicates the presence of strong physical interference that may directly impact the measurement system, causing reading distortion. Increasing this factor will immediately increase the process noise covariance value, allowing the filter to more quickly adapt to changes caused by energy impacts.

[0118] In this embodiment, the online lubricating oil monitoring system suitable for power plant unit non-stop operation scenarios also includes an oil evaluation module, which is communicatively connected to the output module. It is used to receive viscosity values, output oil deterioration scores through a pre-trained oil deterioration model, and generate an early warning signal when the oil deterioration score exceeds a preset threshold.

[0119] It should be noted that the pre-trained oil deterioration model uses the received viscosity value as the core input and outputs an oil deterioration score. The training process of the oil deterioration model is a well-known technique in the art and will not be described in detail in this embodiment. For example, a machine learning model (such as random forest, support vector machine or neural network) can be trained using historical data (including viscosity value, other oil parameters and corresponding laboratory analysis results). This machine learning model can learn complex nonlinear relationships, integrate multiple indicators, and give a comprehensive oil deterioration score. The trained machine learning model is the pre-trained oil deterioration model.

[0120] It should be noted that the oil deterioration score can be compared with one or more preset thresholds. When the score exceeds a preset threshold, an early warning signal is automatically generated. The specific preset thresholds are determined based on actual conditions, and this embodiment does not impose specific limitations. For example, assuming the oil deterioration score ranges from 0 to 1, multiple threshold levels can be set to correspond to different levels of early warning. For example, a "Note" level (low threshold): when the oil deterioration score exceeds the first threshold (e.g., 0.3), a "Note" prompt is issued, reminding maintenance personnel to pay attention to the oil deterioration trend; a "Warning" level (medium threshold): when the oil deterioration score exceeds the second higher threshold (e.g., 0.6), a "Warning" signal is issued, indicating that oil testing or replacement needs to be arranged; and a "High" level (high threshold): when the oil deterioration score exceeds the highest threshold (e.g., 0.8), an "Alarm" signal is issued, indicating that the oil has severely deteriorated and immediate measures need to be taken to prevent equipment damage.

[0121] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0122] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. An online lubricating oil quality monitoring system suitable for power plant unit non-stop operation scenarios, characterized in that, The system includes: The data acquisition module is used to simultaneously acquire vibration signal data and lubricating oil viscosity signal data of power plant unit operation. The vibration signal includes at least amplitude and vibration frequency. The data analysis module, which communicates with the data acquisition module, is used to determine the instantaneous vibration influence coefficient based on the distribution characteristics of the amplitude value in the amplitude data and the difference between the vibration frequency and the natural frequency of the viscosity signal. The process involves obtaining the time span required for the vibration disturbance state to transition from its initial state to a stable state at a preset standard vibration frequency, and recording this as the reference time constant; obtaining the time length between a historical moment and the current moment, and recording this as the time difference; extracting the historical instantaneous vibration influence coefficient and historical vibration frequency for each historical moment within the preset time window; calculating the ratio of the preset standard vibration frequency to the historical vibration frequency for each historical moment within the preset time window, as a frequency level indicator; obtaining the attenuation rate weight for each historical moment based on the frequency level indicator, the reference time constant, and the corresponding time difference, using a preset attenuation rate weight calculation function; and calculating the sum of the products of all historical instantaneous vibration influence coefficients and their corresponding attenuation rate weights within the preset time window, as the cumulative disturbance indicator. The system acquires historical cumulative interference indices for each historical moment within a preset time window; calculates the arithmetic mean of all historical cumulative interference indices as a reference cumulative interference indices; calculates the ratio of the current cumulative interference indices to the reference cumulative interference indices as the interference intensity factor; based on the current vibration frequency and natural frequency, it obtains the instantaneous probability value for characterizing resonance caused by vibration frequency through a Gaussian function; calculates the product of the interference intensity factor and the instantaneous probability value as the instantaneous frequency risk weight, and determines the resonance risk index for the current moment through recursive fusion calculation; and dynamically adjusts the preset baseline process noise covariance according to the cumulative interference indices and the resonance risk index to determine the target process noise covariance. The output module, which communicates with the data analysis module, is used to substitute the target process noise covariance into the prediction step of the Kalman filter to filter the viscosity signal and obtain a viscosity value that reflects the true state of the lubricating oil.

2. The online lubricating oil quality monitoring system for power plant unit non-stop operation as described in claim 1, characterized in that, The synchronous acquisition of vibration signal data and lubricating oil viscosity signal data from the power plant unit operation includes: A unified clock reference is provided for the acquisition of vibration signal data and viscosity signal data, and the vibration signal is sampled. The sampling frequency is configured to be sufficient to obtain the main frequency components of the unit vibration.

3. The online lubricating oil monitoring system for power plant unit shutdown scenarios according to claim 1, characterized in that, The process for determining the instantaneous vibration influence coefficient includes: A set of continuous amplitude data is sorted in ascending order of amplitude value to generate an amplitude sequence. The subset of amplitudes greater than the upper quartile is extracted as strong amplitude data. The arithmetic mean of all amplitude values ​​in the strong amplitude data is calculated as the reference amplitude value. For any given moment, the ratio of the amplitude value to the reference amplitude value is calculated and used as the amplitude influence factor. For any given moment, calculate the absolute difference between the vibration frequency and the natural frequency, which is taken as the frequency difference; and calculate the ratio of the frequency difference to the natural frequency, which is taken as the frequency deviation index. The instantaneous vibration influence coefficient is determined based on the amplitude influence factor, frequency deviation index, and preset attenuation coefficient.

4. The online lubricating oil monitoring system for power plant unit shutdown scenarios according to claim 3, characterized in that, The determination of the instantaneous vibration influence coefficient based on the amplitude influence factor, frequency deviation index, and preset attenuation coefficient includes: Using the natural constant as the base, the frequency influence factor is obtained by taking the opposite of the product of the frequency deviation index and the preset attenuation coefficient and then performing an exponential operation. Calculate the product of the amplitude influence factor and the frequency influence factor, and use it as the instantaneous vibration influence coefficient at any given moment.

5. The online lubricating oil quality monitoring system for power plant unit non-stop operation as described in claim 1, characterized in that, The attenuation rate weight for historical moments is obtained based on frequency level indicators, reference time constants, and corresponding time differences, using a preset attenuation weight calculation function. This includes: Using the frequency index as the base and the preset sensitivity constant as the exponent, a power operation is performed to obtain the attenuation rate coefficient; the product of the attenuation rate coefficient and the reference time constant is calculated as the attenuation time constant. Calculate the product of the order of historical moments within the preset time window and the corresponding time difference, and use it as the first product; Using the natural constant as the base, the inverse of the ratio of the first product to the corresponding decay time constant is exponentially calculated to obtain the decay rate weight.

6. The online lubricating oil quality monitoring system for power plant unit non-stop operation as described in claim 1, characterized in that, The recursive fusion calculation process is as follows: Calculate the product of the resonance risk index of the previous time step and the preset risk forgetting factor, and use it as the second product; Calculate the difference between 1 and the preset risk forgetting factor, and use it as the first difference; Calculate the product of the instantaneous frequency risk weight at the current moment and the first difference, and use it as the resonance risk indicator at the current moment.

7. The online lubricating oil quality monitoring system for power plant unit non-stop operation as described in claim 1, characterized in that, The target process noise covariance determination process includes: Obtain the first preset weight coefficient and the second preset weight coefficient; Calculate the product of the first preset weight coefficient and the resonance risk indicator as the resonance risk adjustment indicator; calculate the product of the second preset weight coefficient and the cumulative interference indicator as the interference adjustment indicator. Calculate the sum of the positive integer 1, the interference adjustment index, and the resonance risk adjustment index, and use it as the overall adjustment factor; The product of the preset baseline process noise covariance and the overall adjustment factor is calculated as the target process noise covariance.

8. The online lubricating oil quality monitoring system for power plant unit non-stop operation as described in claim 1, characterized in that, The system also includes an oil evaluation module, which is connected in communication with the output module. It receives viscosity values, outputs an oil deterioration score through a pre-trained oil deterioration model, and generates an early warning signal when the oil deterioration score exceeds a preset threshold.

Images