CN119023261A

CN119023261A - A method for early warning of eccentric wear of guide bearing bushes of hydropower units

Info

Publication number: CN119023261A
Application number: CN202411523465.5A
Authority: CN
Inventors: 张红伟; 司汉松; 刘钊; 陈中志; 李帅访; 张民威; 汪洋; 苏疆东
Original assignee: Jiangchuan Jinsha Hydropower Development Co ltd
Current assignee: Jiangchuan Jinsha Hydropower Development Co ltd
Priority date: 2024-10-30
Filing date: 2024-10-30
Publication date: 2024-11-26
Anticipated expiration: 2044-10-30
Also published as: CN119023261B

Abstract

The invention discloses an early warning method for eccentric wear of a guide bearing bush of a hydropower unit, and relates to the field of electronic technology. According to characteristic signals such as vibration swing, vibration, bush temperature, etc. installed on a hydropower generator unit, the invention adopts wavelet packet filtering and Hilbert transform method to perform envelope demodulation to identify impact characteristics caused by eccentric wear, combines with an SVR machine learning method to identify bush temperature anomalies, and realizes identification, early warning, and detection of eccentric wear of the guide bearing bush of the unit by comparing the impact orientation and the bush temperature change at the corresponding orientation.

Description

Method for early warning eccentric wear of guide bearing bush of hydroelectric generating set

Technical Field

The invention relates to the technical field of electronics, in particular to a method for pre-warning eccentric wear of a guide bearing bush of a hydroelectric generating set.

Background

The statements in this section merely provide background information related to the present disclosure and may not constitute prior art.

The eccentric wear is one of dynamic and static rubbing, and means rubbing occurring in a certain local orientation. The rub-impact refers to abnormal contact between the rotating component and the static component, and is a common fault in large-scale rotating machinery such as a hydro-turbo generator set and a turbo generator set. The main reason is that the dynamic and static gaps of the unit become smaller due to some reason. The friction causes the rotor to generate very complex vibration, which is an important reason for the instability of the rotor system, the light person causes the machine set to generate strong vibration, and the serious permanent bending of the rotating shaft and even the damage of the whole shaft system can be caused.

Rub/bias is a concomitant fault, secondary to other faults. The rubbing/bias contact may occur in radial, axial, or both directions, and is referred to as localized rubbing when the actual rubbing contact occurs at a fraction of the vibration cycle. When the contact occurs throughout the vibration cycle and has continuity, it is referred to as full cycle rub. For rub faults, localized rub is the most common rub manifestation, and localized rub involves three physical phenomena, collisions, friction, and changes in shafting stiffness.

For the hydroelectric generating set, the parts where the friction occurs mainly have the following conditions:

1) Eccentric wear between the large shaft and the guide bearing due to the misalignment of the bearing bush and the eccentricity of the large shaft.

The first aspect is that when the installation deviation of the unit bearing bush is uneven or eccentric, and the unit bearing bush gap is larger than the minimum gap, the bearing bush eccentric wear occurs; for a unit of three guide bearings, when the three guide bearings are not concentric, the eccentric wear of the bearing bush can also occur at the non-concentric part;

the second aspect is that for the thrust tile structure of the rigid support, when the horizontal adjustment deviation of the thrust tile is large, the phenomenon of eccentric wear of the bearing bush of the unit can be caused in the running process;

The third aspect is that during the running process of the machine set, the bearing bush is offset due to the offset of the rotation center of the guide bearing caused by the acting force in the fixed direction, and the force mainly comes from two parts, namely the imbalance of the magnetic pulling force in the fixed direction caused by the non-concentricity of the stator and the rotor of the machine set, and the imbalance of the water power caused by the uneven water flow along the circumference of the rotating wheel.

2) The unit runner blade collides with the runner chamber.

The first aspect is that the displacement of the unit shaft causes the clearance between the blades and the rotating wheel chamber to be reduced, so that the rotating wheel chamber and the rotating wheel are rubbed. The friction is usually generated in a shutdown process or a vibration area or a non-cooperative area of the unit, and when the unit is impacted by water hammer or unstable water flow, the clearance between the blades and the rotating wheel chamber is reduced to generate friction when the unit generates obvious axial displacement;

the second aspect is that the rotor chamber is deformed (oval or other irregular shape) due to long-term uneven water flow or other reasons, on one hand, radial hydraulic imbalance is caused, radial displacement is caused to the rotation center of the rotor, and the gaps between the blades and the rotor chamber are locally reduced, so that friction is caused;

the third aspect is that due to the deviation of the installation angle of the guide vanes of the unit or the inconsistent operation of the guide vane operating mechanism, such as blocking of the guide vanes or blockage of foreign matters, the unit generates more obvious radial water thrust in the load adjustment process, so that the gap between the blades and the rotating wheel chamber is locally reduced, and friction is caused.

3) The friction between the crown and the labyrinth ring of the mixed flow unit runner.

For a mixed-flow water turbine, the gap between the runner crown and the labyrinth ring is smaller, usually only 2-4mm, and under the conditions of long-term water flow impact and sediment abrasion, the labyrinth ring can loosen or even fall off, so that friction between the runner crown and the labyrinth ring can be caused.

Friction results in a non-linear change in force simultaneously with the dynamic stiffness, and thus the dynamic conditions of the friction rotor become complex. When rubbing occurs, the contact force suddenly appears and disappears. When the rotating member contacts the stationary member, the stationary member pushes the rotor while the rotating member pushes the stationary member with equal but opposite forces. Such contact forces can be broken down into radial and tangential (friction) components. The direction of the radial force is directed towards the centre of the rotor and thus the rotating part is accelerated significantly away from the contact point. The radial force varies during the residence time of the contact cycle and is proportional to the instantaneous radial acceleration a (f=ma). After a short time, the rotating part deviates due to inertia, the average radial speed is reduced, and after the radial speed reaches zero, the rotating part bounces and the rubbing contact is finished.

Tangential friction forces occur during contact, the magnitude of which is equal to the instantaneous magnitude of the radial force multiplied by the friction factor of the contact surface. The tangential frictional force is directed against the surface velocity of the rotating member, creating a torque on the rotating member while attempting to accelerate the rotating member against the direction of movement. This is why in rub-impact, an inverse component is usually generated in the whole spectrum. A side effect of tangential friction is that it acts as a medium, transmitting rotational energy to radial vibrations. For localized rubbing, if a slight frictional contact is produced, the system stress and the overall stiffness of the system are not significantly changed, but as the degree of rubbing increases, the change in stress and stiffness is more significant, and the radial and tangential friction forces caused by rubbing are essentially impact forces characterized by sudden onset, high forces and then abrupt disappearance, similar to striking a rotor shaft with a hammer, which can produce an impact response when the rubbing is intense to some extent, including multiple free radial vibration modes in the rotating member that vibrate in free vibration form at one or more natural frequencies when the rotating member bounces off the contact, which is also a major cause of resonance in rubbing.

The eccentric wear of the guide bearing is one condition of local friction, and at the moment of eccentric wear, radial and tangential impact force is caused, and impact response is generated, so that instantaneous local resonance signals are caused. Generally, such rub-impact belongs to weak rub-impact with transient periodicity, and has transient characteristics. Such localized deformation and rebound motion of the jounce due to impact forces causes the jounce vibration signal to contain a large number of unsynchronized harmonic components, usually accompanied by the occurrence of a resonance signal. Friction and impact give localized rubbing a strong nonlinearity. The main basis for diagnosing the rub-impact fault is a characteristic frequency spectrum for a long time, but the characteristic frequency spectrum is similar to the spectrum characteristics of vibration faults caused by other reasons, and meanwhile, the transient impact characteristics expressed by the time domain where rub-impact is located are desalted after Fourier transformation, which is equivalent to the average of local characteristic information in the whole analysis frequency domain, so that the analysis result has larger error, and therefore, the accurate diagnosis of the rub-impact fault by the frequency spectrum alone has larger difficulty.

In addition, from the point of view of signal analysis, the resonance signal usually caused by instantaneous rub-and-bump appears in a high frequency band, and this signal is often modulated by a periodic 1X vibration signal, and the signal amplitude is not very strong, so this signal tends to be submerged in other periodic new vibration signals and other noise signals.

Therefore, the collision fault between the parts can not be identified by using the characteristic parameters of the time domain such as peak value, effective value and the like, and the impact signal is difficult to separate and identify from the frequency domain by using the conventional FFT change. Fig. 1 is a waveform of a swing degree signal (a rub feature appears in a signal) after a partial weak rub occurs on a hydraulic guide bearing of a certain hydraulic turbine of a certain power station, in a time domain waveform, the swing degree change caused by the weak rub is very small, and the swing degree signal is submerged in a 1X signal of the swing degree, and has no influence on a through-frequency peak value of the signal. It is therefore difficult to identify the localized weak rub fault using conventional peaks and peaks.

In addition, a feature that is significantly different from other weak rubs is that the occurrence of eccentric wear of the guide bearing shoes is accompanied by an increase in the bearing shoe temperature, which is also an important feature of eccentric wear and other weak rubs.

In summary, the bearing pad eccentric wear failure is characterized in the signal as follows:

(1) The bias wear causes short-time high-frequency shock resonance signals;

(2) The stable eccentric wear occurs periodically, so that the caused impact resonance signal occurs periodically, and the eccentric wear occurs once or more times when the unit rotates for one circle;

(3) The bias wear causes the resonance signal of short-time high frequency to be modulated by the rotating speed frequency;

(4) The resonance signal can be detected by the wobble, radial vibration measurement point, but is often masked in other vibration signals and noise.

(5) A bias wear failure is typically accompanied by a rising deviation of the corresponding shaft shoe temperature.

The hydraulic generator set guide bearing (such as an upper guide bearing, a lower guide bearing and a water guide bearing) is a supporting component of a hydraulic generator set rotating component (comprising a rotor, a large shaft, a rotating wheel and the like) and is used for supporting and positioning the rotating part of the water turbine, and meanwhile, the hydraulic generator set guide bearing has good bearing capacity and enough strength. However, due to the defects of tortuous axis, improper bearing clearance, non-concentricity of the guide bearing and the like of the unit, the eccentric wear problem is generated between a large shaft and a bearing bush surface of the unit, and the long-term running ‌ can not only lead to abrasion, damage, heat generation, even bush burning and the like of the bearing, but also cause adverse effects including vibration increase, stability reduction and the like on the running of the unit. Therefore, the eccentric wear of the guide bearing bush of the unit is detected on line, and the health state of the bearing bush is identified, so that the method is particularly important for safe operation of the unit.

Disclosure of Invention

The invention aims at: aiming at the problems existing in the prior art, the invention provides a hydroelectric generating set guide bearing bush eccentric wear early warning method, which adopts wavelet packet filtering and Hilbert transform methods to carry out envelope demodulation and identify impact characteristics caused by eccentric wear according to characteristic signals such as vibration swing degree, vibration and bearing bush temperature collected on a hydroelectric generating set, and combines SVR machine learning methods to identify bearing bush temperature abnormality, and realizes the identification early warning detection of the set guide bearing bush eccentric wear by comparing impact azimuth and corresponding bearing bush temperature changes, thereby solving the problems.

The technical scheme of the invention is as follows:

A hydroelectric generating set guide bearing bush eccentric wear early warning method comprises the following steps:

Step S1: training according to the historical health data sample to obtain a prediction model of the temperature change of each guide bearing bush and the maximum allowable bush temperature deviation;

Step S2: based on wavelet belting pass filtering, extracting impact characteristic indexes of guide bearing swing degree signals and vibration signals acquired in real time through an impact envelope demodulation algorithm;

Step S3: based on wavelet band pass filtering, through a shock envelope demodulation algorithm, selecting a plurality of historical health data to calculate an envelope average peak value basic value in the historical sample data in an accumulated mode The upper guide X-direction swing degree signal, the upper guide Y-direction swing degree signal, the upper frame X-direction vibration signal and the upper frame X-direction vibration signal are respectively corresponding;

Step S4: and carrying out unit guide bearing bush eccentric wear early warning detection.

Further, each of the guide bearing shafts includes: an upper guide bearing, a lower guide bearing and a water guide bearing.

Further, the step S1 includes:

Step S11: extracting feature index data which are in a steady state after grid connection from a stored historical data record;

Step S12: preprocessing the feature index data extracted in the step S11, removing abnormal points, and obtaining a new time sequence of the feature index data And a time sequence of predicted features; Wherein, ，，For the total number of feature indicators that all participate in the prediction,The length of the time sequence of each characteristic index;

Step S13: normalizing; mapping all time sequences of all characteristic indexes to [ -1,1] range to form new time sequences ；

Step S14: for new time seriesTime series of predicted featuresRandomly scrambling the sequence, and then extracting the first 70% of data as training set time sequenceAndThe remaining 30% of the data was used as test set time seriesAnd；

Step S15: invoking training set time seriesAndHandleAs an independent variable,As dependent variable input SVR predictive model for training, obtaining predictive model vector data after training; Wherein, the parameters of the SVR prediction model are set as follows: the penalty coefficient c=1, and the kernel function is selected as a gaussian kernel function;

Step S16: predictive model vector data obtained by training in step S15 Based on the time series of test setsAndCarrying out SVR prediction model, calculating prediction error and penalty coefficient;

Step S17: predictive model vector data obtained by training in step S15 Based on, willAndCarrying out point-by-point calculation on errors by taking the SVR prediction model;

step S18: repeating the steps S11-S17, traversing all the upper guide bearing bush temperature, the lower guide bearing bush temperature and the water guide bearing bush temperature to obtain SVR prediction model vector data and maximum allowable deviation of each bush temperature, and recording Vector data for a certain watt-temperature SVR predictive model,For the maximum allowable deviation value of the watt temperature,，The total temperature of the upper guide, the lower guide and the water guide shoe.

Further, the preprocessing in step S12 includes:

preprocessing the time sequence of each characteristic index by adopting Gaussian filtering with a 99.7% confidence interval;

the step S13 includes:

Wherein:

Is the first The maximum value of the time series of the individual characteristic indicators,Is the firstThe minimum value of the time series of the characteristic indexes.

Further, the step S16 includes:

In the above-mentioned method, the step of, For the actual measurements in the training set,As an average of the actual measured values in the training set,The predicted value is represented by a value of the prediction,To test the length in the time series of the set,A value closer to 1 indicates a better fitting effect with less model error, e.gIf the training set is smaller than 0.8, the training set and the SVR training parameters are required to be readjusted, and the step S11 is re-executed, otherwise, the next step is executed.

Further, the step S17 includes:

In the above-mentioned method, the step of, For actual measurements in the time series of the entire sample set,The predicted value is represented by a value of the prediction,Is an error value;

Then according to The distribution is 1.25 times as large as the maximum allowable temperature deviation, that is:

In the above-mentioned method, the step of, Representing from mIs calculated from the data of (a)Is recorded as the width range of (2)A deviation is allowed for this bushing temperature.

Further, the step S2 includes:

step S21: acquiring the acquired real-time swing and vibration original time domain waveform signals ，A key phase mark is recorded in the memory;

Step S22: morlet is chosen as the wavelet basis function, setting the decomposition layer number to 4, for Performing wavelet packet transformation to extract wavelet packet coefficients of each scale，；

Step S23: filtering low-frequency and high-frequency noise based on band-pass filtering of wavelet packets;

step S24: according to wavelet packet coefficients after filtering Performing wavelet packet inverse transformation ‌ to obtain reconstruction signal to form new vibration and swing degree signal；

Step S25: for a pair ofEnvelope waveform signal of impulse pulse signal is obtained by Hilbert transform digital envelope demodulation technology；

Step S26: and (5) impact feature extraction.

Further, the step S23 includes:

Setting up For the filtered wavelet packet coefficientsObtained by calculation of the following formula:

In the above-mentioned method, the step of, Is thatA center frequency of the corresponding band pass frequency;

Then, in order to reduce the noise influence of the signals in the range of 20X-300 Hz, the following method is continuously adopted for noise filtering:

Is an adaptive threshold after the wavelet transform, Is noise level estimation, and the calculation method is as follows:

Wherein, To the number of data points in the data window used in performing the wavelet transform,The function represents the median value to be found,Is a soft threshold function:

Wherein, Is a threshold value parameter that is set to be,Is the coefficient after the transformation and is used to transform the coefficient,The input value of (2) is，The input value of (2) is；The function is a sign function, and the value of the function is as follows:

Representing the maximum value.

Further, the step S25 includes:

Setting up For the analysis signal of the signal envelope signal, there are:

Wherein, Is an imaginary unit of number and is,Representing the hilbert transform:

envelope spectrum recognition using fast fourier transform to extract envelope signal from impulse signal envelope signal waveform Wherein, the method comprises the steps of, wherein,Is the complex conjugate of the resolved signal:

the step S26 includes:

In the process of obtaining Then, according to key phase signals in the waveform, intercepting data of each period, and calculating envelope peak values and orientations of each period according to the data of each period;

From the slave Intercepting waveform signals between two adjacent key phases to generate the firstOf a cycle ofA signal;

For a pair of Sorting to remove the largest 50% data, and obtaining the maximum value from the rest 50% data as the noise baseline value in the signal；

Constructing a new time domain waveformFor a pair ofPulse peak value detection is carried out on the signals, pulse amplitude caused by impact is calculated, and phases corresponding to the pulse peak values are synchronously calculated according to key phase signals;

Thus repeatedly calculating The number of impact pulses, the pulse amplitude and the position of the pulse occurrence in all periods; recordingThe number of pulses is taken as the total number of pulses,For the amplitude of each pulse,For pulse azimuth, find:

For envelope waveform signals Performing full-period FFT to form an envelope spectrum, extracting the main frequency, and recording the maximum amplitude main frequency as。

Further, the step S4 includes:

Step S41: calculating average pulse amplitude and basic pulse amplitude growth coefficient of each guide bearing swing degree and vibration based on the extracted envelope peak value and azimuth of each period ；

Step S42: all the impact signals of the guide bearing swing degree and vibration are judged as follows:

Condition one: and (3) pendulum degree impact index judgment: And the main frequency Is thatThe frequency of the double-conversion, wherein,Is an integer and，The number of tiles for the corresponding guide bearing;

condition II: judging vibration impact indexes:

Vibration impact characteristic index for displacement type output: And the main frequency Is thatFrequency doubling;

vibration impact characteristic index for speed type output: And the main frequency Is thatFrequency doubling;

Vibration impact characteristic index for acceleration type output: And the main frequency Is thatFrequency doubling;

if all the conditions are not met, the non-guide bearing bush is subjected to eccentric wear, and the process returns, otherwise, the next step is executed;

step S43: detecting abnormal temperature of a bearing bush:

reading the phase of the corresponding impulse event from the vibration or yaw signal satisfying the above conditions According toCalculating the bearing bush number of the corresponding direction, and setting the bearing bush number as；

Bearing bush capable of being collected in real timeTemperature, settingIs a real-time bearing bushA measured value of temperature; collecting other input parameters used by a bearing bush temperature prediction model; predictive model vector data trained in step S1Based on the input parameters, the SVR prediction model is carried, the predicted temperature is calculated and recorded asRecordingBearing bush obtained for training according to step S1Allowable deviation of watt-temperature, then:

If it is Then the bearing shellThe eccentric wear early warning is carried out, and the bearing bush number is；

Otherwise, bearing bushThe bias is not established and step S43 is repeated until the impact pulse characteristics in all directions are detected.

Compared with the prior art, the invention has the beneficial effects that:

According to characteristic signals such as vibration swing degree, vibration, bearing bush temperature and the like which are installed on a hydroelectric generating set, a wavelet packet filtering and Hilbert transformation method is adopted to carry out envelope demodulation and identify impact characteristics caused by eccentric wear, an SVR machine learning method is adopted to identify abnormal bearing bush temperature, and the identification, early warning and detection of the eccentric wear of the guide bearing bush of the set are realized by comparing impact azimuth and corresponding bearing bush temperature change.

Drawings

FIG. 1 is a waveform of a wobble signal after a partial weak friction occurs on a hydraulic guide bearing of a hydraulic turbine of a power station;

FIG. 2 is a schematic diagram of a SVR model;

FIG. 3 is a detailed flowchart of step S1;

FIG. 4 is a graph comparing predicted temperature data of a prediction model obtained after training by the method with actual measured tile temperature;

FIG. 5 is an error distribution diagram of predicted and true temperatures;

FIG. 6 is a detailed flowchart of step S2;

FIG. 7 is an original top cover vibration signal for a unit;

FIG. 8 is a wavelet packet filtered signal;

FIG. 9 is an envelope signal of an impingement signal;

FIG. 10 is a one cycle envelope signal;

FIG. 11 is an envelope spectrum;

Fig. 12 is a detailed flowchart of step S4.

Detailed Description

It is noted that relational terms such as "first" and "second", and the like, are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.

It should be noted that, the method of identifying rub-impact fault mainly searches for the original vibration signal to meet the above characteristics. Based on the characteristics, the method provides and adopts the impact vibration detection method based on wavelet packet filtering and combines the tile temperature anomaly identification based on SVR machine learning method to perform early warning identification on the bearing bush eccentric wear fault.

1. Basic SVR method and wavelet packet filtering method

1.1 Support Vector Regression (SVR) machine learning algorithm

First, a Support Vector Regression (SVR) machine learning algorithm is adopted, and the Support Vector Regression (SVR) is a machine learning method for regression based on a support vector machine (Support Vector Machine). Compared with a common support vector classifier (SVR), the SVR predicts the output value corresponding to the newly observed input variable based on the relation between the input variable and the output variable.

As shown in fig. 2, a schematic diagram of the SVR model is shown. Unlike traditional regression methods, SVR allows a certain amount of deviation between model predictions and true references during fitting, noted as. Fitting-basedToFor the center, construct a width ofSample data falling within this interval band is not required to calculate a loss, and prediction is considered to be correct, whereas data outside the interval band is considered to have a false calculation loss.

When presentFor each input variable, a nonlinear estimation model is assumed:

Wherein: represents the deviation and the deviation is calculated, As a weight vector of the weight vector,For the purpose of its transposition,Nonlinear functions are high-dimensional feature space; Is the input vector.

SVR is realized by maximizing the interval band width and minimizing the total loss to optimize the model, i.e. solving to obtainIn (a) and (b)AndCertain constraint conditions need to be met, namely:

To output the vector, in practice, Setting too small to ensure that all sample points are in the spacer, setting too large to ensure that all sample points are in the spacer, the regression hyperplane will be biased by some outlier bands, for which SVR allows each sampleAdding relaxation variablesTo avoid that the optimization problem is not feasible to solve, the SVR objective function is expressed as:

in the formula, The upper and lower bound relaxation variables of the interval are respectively represented; To balance model complexity and loss, a too large value will result in too narrow a spacing band, a smaller or approaching 0 will result in too wide a spacing band, all of which will affect the fit, and can be taken generally 。

When the solution in the sample space is impossible, a kernel function is introducedThe form, now solved, is as follows:

In the middle of AndThe lagrangian method is used to solve for the coefficient to be determined,Is a symmetric positive real function. Commonly available kernel functions include linear kernel functions, polynomial kernel functions, radial basis function kernels, gaussian kernel functions, and the like.

Second, wavelet packet filtering is adopted

1.2 Wavelet packet filtering

Conventional vibration signal analysis and processing methods typically employ fourier transforms (Fourier transform). The basic idea of fourier transform is to decompose the signal into a series of cosine functions, while wavelet transform (Wavelet transform) decomposes the signal into a multi-scale wavelet function (basic wavelet). The Fourier analysis is an analysis method with a fixed window function, and can not reflect the characteristics of non-stationary signal, short duration, time domain and frequency domain localization and the like. Whereas wavelet analysis (WAVELET ANALYSIS) is a time-frequency localized analysis method where the window area is fixed but its shape can be changed, i.e. both time and frequency windows can be changed.

Wavelet Packet Decomposition (WPD) was further developed on the basis of discrete wavelet transforms. In wavelet transformation, a signal is decomposed into a series of high frequency and low frequency components, but only the low frequency components are iteratively decomposed. In contrast, WPD performs iterative decomposition of high-frequency and low-frequency components simultaneously in each stage of decomposition. Meaning that it is able to analyze the frequency content of the signal in more detail.

A key advantage of wavelet packet decomposition over traditional wavelet decomposition is its full band decomposition capability on signals. Conventional DWT ignores decomposition in the high frequency band, which may lose important signal details in some applications. This full band analysis capability of WPD is particularly useful for applications where the signal characteristics are not limited to the low frequency range alone, and wavelet packet decomposition shows its unique advantages in a variety of applications due to its fine-grained analysis characteristics.

Wavelet packet filtering is a denoising filtering method realized by utilizing wavelet packet transformation, and is generally completed by three steps: wavelet packet transform decomposition and nonlinear processing of wavelet coefficients to filter out noise and wavelet packet inverse transform (signal reconstruction).

(1) Wavelet packet transform (Wavelet transform): wavelet coefficient of decomposing signal into different scale and frequency band). The wavelet packet transform provides local information of the signal in the time and frequency domains and thus has better time-frequency resolution.

(2) And (3) threshold processing: for wavelet coefficientsThresholding is performed to suppress noise by zeroing out smaller coefficients. The threshold is typically calculated based on the energy or statistical properties of the coefficients.

(3) Wavelet packet inverse transformation: the processed wavelet coefficients are inversely converted back to the original signal. The inverse transform produces a denoised signal.

Advantages of wavelet/wavelet packet filtering include strong adaptability to non-stationary signals, and better processing of transient signals and local features. Whereas fourier filtering focuses more on global frequency domain characteristics, wavelet/wavelet packet filtering focuses more on local time-frequency characteristics, especially for nonlinear signals with widely varying transient characteristics, whereas impulse signals are typically time-varying nonlinear signals. In addition, wavelet packet analysis has full band decomposition capability of the signal, and more high frequency signal details can be preserved compared to wavelet analysis. Therefore, the method adopts wavelet packet filtering to realize band-pass filtering and noise filtering.

The features and capabilities of the present invention are described in further detail below in connection with examples.

Example 1

According to the method, the bearing bush eccentric wear detection is realized through impact signal characteristic extraction and temperature anomaly detection based on vibration, swing degree, temperature and other signals acquired by an on-line monitoring system or an off-line measuring system arranged on the hydroelectric generating set. These signals include in particular:

table 1 guide bearing eccentric wear warning signal meter

In addition to the above signals, in order to identify the temperature change in the thermal stabilization process under cold start-up conditions, it is also necessary to introduce "post grid-tie time" as an auxiliary characteristic signal.

In this embodiment, specifically, a method for pre-warning eccentric wear of a guide bearing bush of a hydroelectric generating set specifically includes the following steps:

step S1: training according to the historical health data sample to obtain a prediction model of the temperature change of each guide bearing bush and the maximum allowable bush temperature deviation; the actual flow chart is shown in fig. 3:

the specific substeps are as follows:

Further, the step S1 includes:

step S11: extracting feature index data which are in a steady state after grid connection from a stored historical data record; these feature index data include:

table 2 guide bearing bushing temperature prediction characteristic index

That is, the time series data of the acquired original input characteristic index data is set as(，) WhereinRepresents the first of Table 2The characteristic index is input in a number of ways,The total number of feature indexes for all participating predictions, i.e., the total number of input variables. By way of example, the above pilot bearing is predicted by a bearing shell temperature, such asRepresenting the time series of upward-directed X-ray wobble peak-to-peak values,Representing the time series of upward-directed Y-direction waviness peaks,Representing the time series of the peak-to-peak value of the X-direction vibration of the upper frame; Time series representing a certain characteristic parameter Data corresponding to the true side value of the characteristic index at a certain moment,Representing the length of the time series of the characteristic parameter acquired, i.e. the total number of real measured data. The corresponding predicted characteristic index time sequence isI.e. a certain watt-temperature time sequence of a certain guide bearing.

Step S12: preprocessing the feature index data extracted in the step S11, removing abnormal points, and obtaining a new time sequence of the feature index dataAnd a time sequence of predicted features; Wherein, ，，For the total number of feature indicators that all participate in the prediction,The length of the time sequence of each characteristic index; wherein, the pretreatment comprises:

the time series of each feature indicator was pre-processed with 99.7% confidence intervals using gaussian filtering.

Step S13: normalizing; mapping all time sequences of all characteristic indexes to [ -1,1] range to form new time sequences; I.e. for each input characteristic indexFrom the slaveFrom sequences of individual characteristic valuesMaximum value of (2)And minimum valueThen define:

Wherein:

Step S14: for new time seriesTime series of predicted featuresRandomly scrambling the sequence, and then extracting the first 70% of data as training set time sequenceAndThe remaining 30% of the data was used as test set time seriesAnd。

Step S15: invoking training set time seriesAndHandleAs an independent variable,As dependent variable input SVR predictive model for training, obtaining predictive model vector data after training; Wherein, the parameters of the SVR prediction model are set as follows: the penalty coefficient c=1 and the kernel function is chosen to be gaussian.

Step S16: predictive model vector data obtained by training in step S15Based on the time series of test setsAndCarrying out SVR prediction model, calculating prediction error and penalty coefficientAlso known as the decision coefficient:

In the above-mentioned method, the step of, For the actual measurements in the training set,As an average of the actual measured values in the training set,The predicted value is represented by a value of the prediction,To test the length in the time series of the set,A value closer to 1 indicates a better fitting effect with less model error, e.gIf the training set is smaller than 0.8, the training set is required to be readjusted, SVR training parameters (including penalty coefficient, kernel function selection and the like) are required to be adjusted, the step S11 is re-executed, and otherwise, the next step is executed;

the comparison graph of the temperature data predicted by the prediction model obtained after training by the method and the actual measured tile temperature is shown in fig. 4, the predicted temperature well tracks the actual measured tile temperature, and the error distribution graph of the predicted temperature and the actual temperature is shown in fig. 5, and the maximum is about 0.2%.

Step S17: predictive model vector data obtained by training in step S15Based on, willAndThe SVR prediction model is carried in, calculating an error point by point:

Step S18: repeating the steps S11-S17, traversing all the upper guide bearing bush temperature, the lower guide bearing bush temperature and the water guide bearing bush temperature to obtain SVR prediction model vector data and maximum allowable deviation of each bush temperature, and recordingVector data for a certain watt-temperature SVR predictive model,For the maximum allowable deviation value of the watt temperature,，The total temperature of the upper guide, the lower guide and the water guide shoe.

Step S2: based on wavelet belting pass filtering, extracting impact characteristic indexes of guide bearing swing degree signals and vibration signals acquired in real time through an impact envelope demodulation algorithm; the actual flow is shown in fig. 6, and the specific sub-steps are as follows:

step S21: acquiring the acquired real-time swing and vibration original time domain waveform signals ，A key phase mark is recorded in the memory; i.e. settingThe obtained unit guide bearing swing degree and vibration time domain continuous waveform signal is recorded with key phase mark,. In the above-mentioned example of the guidance,For the up-directed X-direction wobble time domain signal,For the upstroke Y-direction wobble time domain signal,For the upper gantry X-direction vibratory time domain signal,The lower guide and the water guide are similar for the X-direction vibration time domain signal of the upper frame. Setting upThe sampling frequency of the time domain waveform isIn the usual vibration, swing degree signal acquisition,Selected to be 1024Hz, setIs thatSampling point number not lower than 4096 points, andThe sampling waveform contains signals with no less than 8 complete rotation periods, and the period number is recorded as。

Step S22: morlet is chosen as the wavelet basis function, setting the decomposition layer number to 4, forPerforming wavelet packet transformation to extract wavelet packet coefficients of each scale，；

According to the principle of wavelet packet decomposition, after 4 layers of decomposition,Is decomposed into signals consisting of 16 different wavelet packets, each wavelet packet having a wavelet coefficient strengthAnd each wavelet packet corresponds to a bandpass filter. Selection of=1024 Hz, then the frequency ranges to which the 16 wavelet packets correspond respectively are:

Table 3 corresponding bands of each wavelet packet after 4-layer wavelet packet decomposition

In order to remove the influence of the frequency conversion and the main multiple frequency thereof on the identification of the impact signals, signals below 20X (20 times of rotating speed frequency) are filtered, and meanwhile, the measurement noise is considered to be concentrated in a high frequency band, so that signals above 300Hz are needed to be filtered, only signals within the range of 20X-300 Hz are reserved, namely, the impact signals caused by the eccentric wear of the bearing are detected and identified mainly through the signals with vibration and swing within the frequency range;

the method is based on the wavelet packet decomposition result of step S22, and based on the design of the band-pass filter, the method comprises the steps of Different filter coefficients are set to realize band-pass filtering, and the specific steps are as follows:

In the above-mentioned method, the step of, Is thatA center frequency of the corresponding band pass frequency; assuming that the rated rotation speed of a certain unit is 75r/min, the rated rotation frequency of the unit is 1.25Hz, the corresponding frequency of 20X is 25Hz, and the unit is calculated according to the above formula、、、、、、、All take 0 s, others；

Representing the maximum value.

Step S24: according to wavelet packet coefficients after filteringPerforming wavelet packet inverse transformation ‌ to obtain reconstruction signal to form new vibration and swing degree signal; The signal is filtered to remove low frequency signals and high frequency interference signals, and noise filtering is performed, so that instantaneous impact signals caused by local weak collision and friction and the like can be highlighted. FIG. 7 shows the original top cover vibration signal of a certain unit; as shown in fig. 8, the vibration signal is subjected to wavelet packet filtering treatment, and a clear impact signal can be observed.

The step S25 includes:

Setting up For the analysis signal of the signal envelope signal, there are:

fig. 9 shows the envelope signal of the impulse signal.

Step S26: extracting impact characteristics;

Specifically, the step S26 includes:

as in the envelope signal of fig. 10 for one cycle, from The waveform signal between two adjacent key phases (such as key phase signal 1 and key phase signal 2 in fig. 10) is intercepted to generate the firstOf a cycle ofThe signal is transmitted to the host computer via the communication network,；

For a pair ofSorting to remove the largest 50% data, and obtaining the maximum value from the rest 50% data as the noise baseline value in the signal；

For envelope waveform signals Performing full-period FFT to form an envelope spectrum (see FIG. 11 below), extracting the dominant frequency, and recording the maximum amplitude dominant frequency as。

Step S3: based on wavelet band pass filtering, through a shock envelope demodulation algorithm, selecting a plurality of historical health data to calculate an envelope average peak value basic value in the historical sample data in an accumulated modeThe upper guide X-direction swing degree signal, the upper guide Y-direction swing degree signal, the upper frame X-direction vibration signal and the upper frame X-direction vibration signal are respectively corresponding; the method is the same as step S2, and will not be described in detail here.

As shown in fig. 12, the specific steps are as follows, and the step S4 includes:

condition II: judging vibration impact indexes:

step S43: detecting abnormal temperature of a bearing bush:

Bearing bush capable of being collected in real timeTemperature, settingIs a real-time bearing bushA measured value of temperature; meanwhile, other input parameters (such as X/Y-direction swing degree peak value, upper frame X/Y-direction vibration peak value, two adjacent tile temperatures, opposite tile temperatures, active power, reactive power, environment temperature, time after grid connection and the like) used by the bearing bush temperature prediction model are collected; predictive model vector data trained in step S1Based on the input parameters, the SVR prediction model is carried, the predicted temperature is calculated and recorded asRecordingBearing bush obtained for training according to step S1Allowable deviation of watt-temperature, then:

The above examples merely illustrate specific embodiments of the application, which are described in more detail and are not to be construed as limiting the scope of the application. It should be noted that it is possible for a person skilled in the art to make several variants and modifications without departing from the technical idea of the application, which fall within the scope of protection of the application.

This background section is provided to generally present the context of the present invention and the work of the presently named inventors, to the extent it is described in this background section, as well as the description of the present section as not otherwise qualify as prior art at the time of filing, are neither expressly nor impliedly admitted as prior art against the present invention.

Claims

1. A method for early warning of eccentric wear of guide bearing bushes of a hydropower unit, characterized by comprising:

Step S1: training based on historical health data samples to obtain a prediction model for the temperature change of each guide bearing bush and a maximum allowable bush temperature deviation;

Step S2: Based on wavelet packet bandpass filtering, the impact characteristic index of the guide bearing swing signal and vibration signal collected in real time is extracted through the impact envelope demodulation algorithm;

Step S3: Based on wavelet packet bandpass filtering, through the impact envelope demodulation algorithm, select multiple historical health data to accumulate the envelope average peak value in the historical sample data , which correspond to the upper guide X-direction swing signal, the upper guide Y-direction swing signal, the upper frame X-direction vibration signal, and the upper frame X-direction vibration signal respectively;

Step S4: Perform early warning detection of eccentric wear of the guide bearing bushing of the unit.

2. A method for early warning of eccentric wear of guide bearing bushes of a hydropower unit according to claim 1, characterized in that the guide bearing shafts of each part include: an upper guide bearing, a lower guide bearing, and a water guide bearing.

3. A method for early warning of eccentric wear of guide bearings of a hydropower unit according to claim 2, characterized in that said step S1 comprises:

Step S11: extracting characteristic index data in a steady state after grid connection from the stored historical data records;

Step S12: Preprocess the characteristic index data extracted in step S11, remove abnormal points, and obtain the time series of new characteristic index data and the time series of the predicted features ;in, , , is the total number of feature indicators involved in the prediction, is the length of the time series of each characteristic indicator;

Step S13: Normalization processing: map all time series of feature indicators to the range of [-1, 1] to form a new time series ;

Step S14: For the new time series and the time series of the predicted features Randomly shuffle the order, and then extract the first 70% of the data as the training set time series and , the remaining 30% of the data is used as the test set time series and ;

Step S15: Call the training set time series and ,Bundle As independent variable, As the dependent variable, the SVR prediction model is input for training, and the prediction model vector data is obtained after training. ; The parameters of the SVR prediction model are set as follows: the penalty coefficient C=1, and the kernel function is selected as the Gaussian kernel function;

Step S16: The prediction model vector data obtained by training in step S15 Based on the test set time series and , bring it into the SVR prediction model, calculate the prediction error and the penalty coefficient;

Step S17: The prediction model vector data obtained by training in step S15 Based on and Bring in the SVR prediction model and calculate the error point by point;

Step S18: Repeat steps S11 to S17, traverse all upper guide bearing pad temperatures, lower guide bearing pad temperatures, and water guide bearing pad temperatures, obtain the SVR prediction model vector data and maximum allowable deviation of each pad temperature, and record them. is the vector data of a certain watt temperature SVR prediction model, is the maximum allowable deviation of the watt temperature, , It is the sum of the upper, lower and water-conducting tile temperatures.

4. A method for early warning of eccentric wear of guide bearings of a hydropower unit according to claim 3, characterized in that the pre-processing in step S12 comprises:

The time series of each characteristic indicator was preprocessed using Gaussian filtering with a 99.7% confidence interval;

The step S13 comprises:

in:

For the The maximum value of the characteristic index time series, For the The minimum value of the time series of the characteristic indicators.

5. The method for early warning of eccentric wear of guide bearings of a hydroelectric generator set according to claim 4, characterized in that said step S16 comprises:

In the above formula, is the actual measured value in the training set, is the average value of the actual measured values in the training set, represents the predicted value, is the length of the test set time series, The closer the value is to 1, the better the fitting effect is and the smaller the model error is. If it is less than 0.8, it is necessary to readjust the training set, adjust the SVR training parameters, and re-execute step S11, otherwise, execute the next step.

6. A method for early warning of eccentric wear of guide bearings of a hydroelectric generator set according to claim 5, characterized in that said step S17 comprises:

In the above formula, is the actual measured value in the time series of the entire sample set, represents the predicted value, is the error value;

Then according to 1.25 times of the distribution is taken as the maximum allowable temperature deviation, that is:

In the above formula, Indicates that from m Find it in the data Width range, record This is the allowable deviation of the bearing temperature.

7. A method for early warning of eccentric wear of guide bearings of a hydroelectric generator set according to claim 6, characterized in that said step S2 comprises:

Step S21: Obtain the collected real-time swing and vibration original time domain waveform signals , Key phase marks are recorded in the

Step S22: Select Morlet as the wavelet basis function, set the decomposition level to 4, Perform wavelet packet transform to extract wavelet packet coefficients of each scale , ;

Step S23: bandpass filtering based on wavelet packets to filter out low-frequency and high-frequency noise;

Step S24: Based on the wavelet packet coefficients after filtering Perform inverse wavelet packet transform to obtain the reconstructed signal and form new vibration and swing signals ;

Step S25: , the envelope waveform signal of the impulse signal is obtained by using the Hilbert transform digital envelope demodulation technology ;

Step S26: Extracting impact features.

8. The method for early warning of eccentric wear of guide bearings of a hydroelectric generator set according to claim 7, characterized in that said step S23 comprises:

set up is the wavelet packet coefficient after filtering, then Calculated by the following formula:

In the above formula, for The corresponding center frequency of the passband frequency;

Then, in order to reduce the noise impact of signals in the range of 20X to 300Hz, the following method is used to perform noise filtering:

is the adaptive threshold after wavelet transformation, is an estimate of the noise level, calculated as:

in, is the number of data points in the data window used in wavelet transform, The function represents the median value. is the soft threshold function:

in, is the threshold parameter, is the coefficient after transformation, The input value is , The input value is ; The function is a symbolic function and its values are as follows:

Indicates finding the maximum value.

9. The method for early warning of eccentric wear of guide bearings of a hydroelectric generator set according to claim 8, characterized in that said step S25 comprises:

set up is the analytical signal of the impact signal envelope signal, then:

in, is an imaginary unit, Represents the Hilbert transform:

According to the impulse signal envelope waveform, use fast Fourier transform to perform envelope spectrum recognition and extract envelope signal ,in, is the complex conjugate of the analytic signal:

The step S26 comprises:

In obtaining Afterwards, according to the key phase signal in the waveform, the data of each cycle is intercepted, and the envelope peak value and azimuth of each cycle are calculated according to the data of each cycle;

from The waveform signal between two adjacent key phases is intercepted to generate the first Cycle Signal;

right Sort the data, remove the largest 50% of the data, and find the maximum value from the remaining 50% of the data as the noise baseline value in the signal ;

Constructing new time domain waveform ,right The signal is pulse peak detected to calculate the pulse amplitude caused by the impact, and the phase corresponding to the pulse peak is synchronously calculated according to the key phase signal;

Repeated calculation Record the number of shock pulses, pulse amplitude and pulse orientation in all cycles; is the total number of pulses, For the amplitude of each pulse, is the pulse direction, find:

For envelope waveform signal Perform full-cycle FFT transformation to form an envelope spectrum, extract the main frequency, and record the maximum amplitude main frequency as .

10. A method for early warning of eccentric wear of guide bearings of a hydroelectric generator set according to claim 9, characterized in that said step S4 comprises:

Step S41: Based on the extracted envelope peak and orientation of each cycle, calculate the average pulse amplitude and basic pulse amplitude growth coefficient of each guide bearing swing and vibration ;

Step S42: All the impact signals of the guide bearing swing and vibration are judged as follows:

Condition 1: Judgment of swing impact index: , and the main frequency for The frequency multiplication, where is an integer, and , is the number of pads of the corresponding guide bearing;

Condition 2: Vibration and shock index judgment:

For the vibration and shock characteristic indicators of displacement output: , and the main frequency for Frequency doubling;

Vibration shock characteristic indicators for speed output: , and the main frequency for Frequency doubling;

Vibration and shock characteristic indicators for acceleration output: , and the main frequency for Frequency doubling;

If all the above conditions are not met, there is no eccentric wear of the guide bearing bush, and the process returns. Otherwise, proceed to the next step.

Step S43: Detection of abnormal bearing bush temperature:

Read the phase of the corresponding impact pulse from the vibration or swing signal that meets the above conditions ,according to Calculate the bearing number of the corresponding position and set the bearing number to ;

Real-time collection of bearings Temperature, setting For real-time bearing The measured value of temperature; at the same time, other input parameters used in the bearing temperature prediction model are collected; the prediction model vector data obtained by training in step S1 Based on the above input parameters, the SVR prediction model is used to calculate the predicted temperature and record it as ,remember The bearing obtained by training in step S1 The allowable deviation of the watt temperature is:

if , then the bearing Eccentric wear warning, bearing number is ;

Otherwise the bearing If eccentric wear is not established, step S43 is repeated until the impact pulse features in all directions are detected.