CN110887899B - Turbine blade water erosion defect monitoring and identifying method - Google Patents
Turbine blade water erosion defect monitoring and identifying method Download PDFInfo
- Publication number
- CN110887899B CN110887899B CN201911183911.1A CN201911183911A CN110887899B CN 110887899 B CN110887899 B CN 110887899B CN 201911183911 A CN201911183911 A CN 201911183911A CN 110887899 B CN110887899 B CN 110887899B
- Authority
- CN
- China
- Prior art keywords
- water erosion
- defect
- blade
- characteristic
- water
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N29/00—Investigating or analysing materials by the use of ultrasonic, sonic or infrasonic waves; Visualisation of the interior of objects by transmitting ultrasonic or sonic waves through the object
- G01N29/44—Processing the detected response signal, e.g. electronic circuits specially adapted therefor
- G01N29/4409—Processing the detected response signal, e.g. electronic circuits specially adapted therefor by comparison
- G01N29/4418—Processing the detected response signal, e.g. electronic circuits specially adapted therefor by comparison with a model, e.g. best-fit, regression analysis
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N29/00—Investigating or analysing materials by the use of ultrasonic, sonic or infrasonic waves; Visualisation of the interior of objects by transmitting ultrasonic or sonic waves through the object
- G01N29/04—Analysing solids
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N29/00—Investigating or analysing materials by the use of ultrasonic, sonic or infrasonic waves; Visualisation of the interior of objects by transmitting ultrasonic or sonic waves through the object
- G01N29/44—Processing the detected response signal, e.g. electronic circuits specially adapted therefor
- G01N29/4481—Neural networks
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N2291/00—Indexing codes associated with group G01N29/00
- G01N2291/02—Indexing codes associated with the analysed material
- G01N2291/028—Material parameters
- G01N2291/0289—Internal structure, e.g. defects, grain size, texture
Landscapes
- Physics & Mathematics (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- Pathology (AREA)
- Immunology (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Chemical & Material Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Signal Processing (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Acoustics & Sound (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention discloses a method for monitoring and identifying water erosion defects of a turbine blade. And the sampling modeling and warehousing comprises experimental measurement, numerical simulation and data training. Firstly, performing an accelerated water erosion experiment on a material consistent with an actual operation working condition to obtain a functional relation of impact angles, impact speeds and the like on the characteristic size of a water erosion defect; establishing a blade vibration model with water erosion defects and establishing a water erosion defect combination X-vibration signal oscillogram Y database based on the relational expression; integrating and dividing the data set, performing neural network training, and establishing an evaluation mechanism of the water erosion defect. And in actual operation analysis, a blade vibration signal oscillogram needs to be acquired on site, an analysis library is compared, the unit operation safety is evaluated, and a corresponding operation and maintenance scheme is formulated. The invention can monitor the abnormal vibration of the blade in real time during the actual operation, provides data support for the fault diagnosis of the unit, ensures the safe operation of the unit and reduces the huge economic loss caused by the blade accident.
Description
Technical Field
The invention belongs to the technical field of industrial equipment, and particularly relates to a method for monitoring and identifying water erosion defects of turbine blade materials under low-load operation.
Background
The steam turbine is used as the essential main power of power generation in a power plant and is the life line of the power generation industry in China. With the rapid increase of the single-machine power of the steam turbine, especially the development of ultra-supercritical and nuclear power high-power steam turbines, the adoption of longer last-stage blades is an inevitable means for improving the efficiency of the steam turbine, and meanwhile, the working conditions of the blades are more severe. In the early 70 to 80 s, Utility and West House Inc. had 50 units of final stage blades that experienced cracking and breakage. Meanwhile, according to incomplete statistics, at least 1061 leaf accidents occur in China. Today, the last stage blades (including the penultimate stage) of modern high power steam turbine sets typically operate in the wet steam region, and under normal operating conditions, the presence of wet steam can cause bucket erosion, resulting in reduced stage efficiency. Under the working conditions of deep peak regulation and zero-power heat supply, the humidity in the turbine stage is higher, the water erosion condition of the blades is more serious, and the blades are easy to break and lose efficacy, so that the safe operation of a unit is endangered. Meanwhile, simulation of a material water erosion process and detection of water erosion defects of an actual unit are very difficult, at present, the water erosion can only be observed by opening a cylinder during overhaul of the unit to determine whether the water erosion occurs, and whether blades need to be replaced is judged through human experience, so that huge economic loss is inevitably caused. Therefore, a quantitative real-time blade fault monitoring method must be established, and technical support is provided for the large repair cycle of the steam turbine unit and the formulation of the blade repair and replacement scheme.
The water erosion problem of the turbine blade is a complex problem comprising multiple action processes of shock wave expansion, fluid-solid coupling, material failure, medium phase change and the like, the blade water erosion problem in actual operation of the turbine blade is monitored in real time, and how to establish an effective blade water erosion defect evaluation mechanism is very critical. At present, two main means for researching the water erosion problem are available: 1. the actual operation condition of the steam turbine is simulated by adopting an experimental method, the sample material is eroded, and the water erosion process and the water erosion mechanism are further researched, because the experiment is closer to the real condition; 2. the key parameters of the turbine blade and the material are selected, the water erosion process of the blade material is simulated based on the current numerical simulation conditions, and the water erosion process and the water erosion mechanism are further researched, because the numerical values can more conveniently generate different conditions, and the simulation can save a large amount of research time and cost. However, due to the complexity of the water erosion problem, it is difficult to completely describe the water erosion problem with a single water erosion experiment or numerical simulation method. Therefore, combining the water erosion experiment with the numerical simulation is the most effective means for studying the water erosion problem, which is the fundamental basis of the water erosion evaluation mechanism. The real-time monitoring and identifying method for the water erosion defect of the turbine blade based on experimental measurement, numerical simulation and data training analysis is adopted, data support can be provided for fault prediction and diagnosis of a unit, and the method plays an important role in preventing the water erosion problem of the turbine blade.
Disclosure of Invention
The invention aims to ensure that the normal operation of a steam turbine is not interfered, and simultaneously, the water erosion problem of a steam turbine blade material can be monitored. Therefore, the invention provides a turbine blade material water erosion defect monitoring and identifying method applied to low-load operation, which is based on a water erosion experiment system, a water erosion numerical simulation method and a water erosion defect evaluation mechanism established through neural network training, can effectively monitor abnormal vibration in the actual operation process of a turbine blade, provides technical support for fault diagnosis of a unit, ensures safe operation of the unit and reduces huge economic loss caused by blade accidents.
In order to achieve the purpose, the invention adopts the technical scheme that:
a turbine blade water erosion defect monitoring and identifying method comprises the following steps:
step one, sampling, modeling and warehousing
The sampling modeling and warehousing comprises three stages of experimental measurement, numerical simulation and data training, and specifically comprises the following steps:
(1-1) firstly, determining a target material, then processing a sample meeting the requirement, carrying out a water erosion experiment on the sample, and obtaining weight loss data and a water erosion defect three-dimensional microscopic morphology map of the sample at each moment; obtaining the influence degree and the functional relation of an impact angle, an impact speed, a droplet size and sample roughness on the water erosion weight loss and the characteristic size of the water erosion defect through regression analysis of experimental data, and simultaneously comparing a three-dimensional microscopic topography map of the water erosion defect obtained through the experiment with the water erosion topography of an actual steam turbine and correcting the functional relation;
(1-2) based on a function relation obtained by experiments, establishing a blade vibration model with water erosion defects by using finite element analysis software, carrying out numerical simulation analysis to obtain blade vibration models with water erosion defects at different positions and different sizes, and establishing a water erosion defect combination X-vibration signal oscillogram Y database;
(1-3) integrating and dividing a data set based on a water erosion defect combination X-vibration signal oscillogram Y database, constructing a ResNet-GRU network for neural network training, dividing an evaluation mechanism of water erosion degree according to different water erosion defect combinations, and integrally evaluating the safety level of the operation of the steam turbine;
step two, actual operation analysis
The field device is used for collecting a blade vibration signal oscillogram, and the safety level of the unit operation is integrally evaluated by utilizing an evaluation mechanism in an analysis library, so that whether shutdown maintenance is carried out or not is determined, and a blade or a strengthening layer is replaced.
The invention has the further improvement that in the experimental measurement stage (1-1) of the step I, a sample subjected to an accelerated water erosion test is obtained through a high-speed water erosion rotation experiment; measuring weight loss data of the sample by a precision balance; shooting by a single lens reflex digital camera to obtain the macroscopic erosion morphology of the surface of the sample; shooting by a 3D super-depth-of-field microscope to obtain the three-dimensional microscopic morphology of the water erosion defect of the sample; and defining the characteristic width and the characteristic depth of the defect by intercepting the characteristic section of the three-dimensional microscopic topography map of the water erosion defect.
The invention has the further improvement that the weight loss data of the sample obtained by the experiment and the characteristic section size of the sample are subjected to data regression analysis to obtain the change rule of the characteristic size of the material water erosion defect appearance along with time and a function relation formula reflecting the influence rule of all factors:
water erosion area A ═ WS 2 (1)
Depth of feature
Characteristic width
Wherein A is the water erosion area of the material surface, WSIs the characteristic width of the water erosion defect; dSIs the characteristic depth of the water erosion defect, Emax is the maximum erosion rate of the material,
is the water erosion state coefficient, d is the weighted mean diameter of the water droplet, n is the droplet size index, V is the impact velocity, V is the water erosion state coefficient0Is the threshold velocity, m is the velocity index, θ is the impact angle, l angle index, Ra is the sample roughness, z is the roughness index, t is the water drop action time, ρ is the density of the material; k is the experimental coefficient, C is the sound velocity in water;
comparing actual turbines simultaneouslyThe water erosion shape of (1) is obtained by correcting the functional relation with a correction coefficient of mu, so that the characteristic width W is equal to mu WSThe characteristic depth D ═ μ DS。
The invention has the further improvement that in the numerical simulation stage (1-2) of the step one, a finite element analysis software is utilized to establish a dry friction damping blade vibration characteristic finite element model with different defect combinations, a three-dimensional finite element entity unit is adopted to carry out mesh division on the blade, and a spring damping unit is adopted to establish a local finite element mesh model between contact surfaces; a plurality of blades with different water erosion defect combinations are subjected to vibration response simulation to obtain a large number of blade vibration signal oscillograms in one-to-one correspondence, and therefore a water erosion defect combination X-vibration signal oscillogram Y database is established.
The invention is further improved in that the numerical training stage (1-3) of the first step comprises four sub-steps:
firstly integrating data, packaging a water erosion defect combination X and a corresponding blade vibration signal oscillogram Y, and then dividing a picture set into training set data { X } according to the proportion of a training set/a verification set which is 3.0t}train、{Yt}trainAnd verification set data { Xt}validation、{Yt}validation；
Then establishing a ResNet-GRU neural network structure, after a characteristic blade vibration signal oscillogram obtained by numerical simulation is transmitted into the ResNet neural network, forming data containing vibration signal oscillogram characteristics through the processing of a characteristic extractor, and transmitting the processed input information into the GRU neural network for iterative learning to obtain effective characteristics;
the network is next trained: setting a variable learning rate optimization cross-loss function (MSE) to train the network through an Adam gradient descent algorithm;
and finally, establishing a water erosion defect evaluation mechanism based on a water erosion characteristic depth distribution result obtained by neural network prediction, and further classifying the vibration signal oscillogram of the blade with different water erosion defects under the safety level of the operation of the steam turbine, wherein the vibration signal oscillogram is divided into three levels: abnormal, inefficient, and hazardous operation.
The invention is further improved in that in the water erosion defect evaluation mechanism in the sub-step four of the numerical training stage (1-3), the size of the defect in each water erosion defect combination is characterized by the characteristic dimension thereof, wherein the characteristic depth of the water erosion defect is used as an evaluation reference value of the water erosion degree, and whether the characteristic depth D of the water erosion defect is smaller than a threshold D or not is specifically judgedr1If so, the unit is considered to be safe to operate; if not, judging whether the characteristic dimension D of the water erosion defect is smaller than the threshold D or notr2If so, determining that the unit is abnormal in operation; if not, judging whether the characteristic dimension D of the water erosion defect is smaller than the threshold D or notr3If so, the unit is considered to be low in operation efficiency; if not, the unit is considered to be in dangerous operation; wherein, 0<Dr1<Dr2<Dr3And D isr1＝0.5mm，Dr2＝1mm，Dr3＝2mm。
The invention has the further improvement that in the step two, corresponding operation and maintenance schemes are made for abnormal, inefficient or dangerous operation according to the feedback information; if monitoring of other turbine blades is to be realized, the blade water erosion defect monitoring, identifying and analyzing library suitable for other turbines can be customized in a targeted manner only by repeating the step one.
Compared with the prior art for preventing water erosion, the invention has the advantages that:
firstly, the number of secondary water drops is reduced by a dehumidification device and the like in the traditional active water erosion prevention measure, so that the liquid-solid impact phenomenon is reduced, and the service life of the blade is prolonged; the traditional passive water erosion prevention measures are that the water erosion resistance of the blade material is enhanced by strengthening the surface of the blade material, so that the water erosion problem is delayed, and the service life of the blade is prolonged. However, even so still can not avoid the emergence of blade water erosion, and in the unit operation process, the staff does not know whether the blade takes place water erosion, does not know the impaired position and the impaired degree of present blade yet, when the blade seriously water erodees, with crisis unit safety operation, causes huge economic loss. At present, the steam turbine is usually overhauled and the blades are replaced regularly, and the shutdown overhaul also means economic loss. However, if the technology of the invention is adopted, the vibration signal waveform diagram of the blade of the steam turbine is monitored in real time during normal operation, and when the surface of the blade has water erosion defects, the vibration signal waveform diagram changes along with the water erosion defects. Based on the blade water erosion defect monitoring, identifying and analyzing library, a worker can judge the damaged position and the damaged degree of the current blade through a vibration signal and decide whether to replace the related blade or the surface strengthening layer, so that the position of the blade fault can be efficiently diagnosed, the serious concurrent accident of the steam turbine is avoided, the economic loss is reduced to the minimum, and the safety and the economy of the unit operation are greatly improved.
In summary, the present invention has the following advantages:
1. on the basis of prolonging the service life of the blade by the existing water erosion preventing process, the vibration signal oscillogram of the blade can be monitored in real time, and the occurrence of serious water erosion is effectively avoided;
2. by monitoring the oscillogram of the vibration signal of the blade in real time, the damaged position and the damaged degree of the blade with water erosion can be effectively judged, and the blade material or the blade strengthening layer can be replaced more purposefully with higher replacement efficiency;
3. the current safety of the blade can be comprehensively evaluated through the water erosion defect monitoring, identifying and analyzing library, if the blade reaches the dangerous service life, the damaged blade needs to be replaced in time, and therefore economic loss caused by regular shutdown maintenance is saved;
4. the water erosion defect monitoring, identifying and analyzing library can customize corresponding vibration models according to different blade materials, different impact speeds and angles and different water drop sizes, and can be generally applied to water erosion monitoring and identification of the blades of the steam turbine in actual operation of various thermal power plants and nuclear power plants;
5. the method is simple to use, safe and stable, and can achieve the purpose of monitoring the water erosion of the blade without influencing the normal operation of the steam turbine.
Drawings
FIG. 1 is a schematic flow chart of the method for monitoring and identifying the water erosion defect of the blade according to the present invention;
FIG. 2 is a schematic diagram of the process of impacting a sample by a jet in a water erosion test in an embodiment of the invention;
FIG. 3 is a water erosion defect macro-topography of a sample in an example of the invention;
FIG. 4 is a three-dimensional micrograph of a water erosion defect of a test specimen in an example of the present invention;
FIG. 5 is a schematic representation of the characteristic dimension of a sample water erosion defect as a function of time in an embodiment of the present invention;
FIG. 6 is a macroscopic topographical view of a water erosion defect of an actual turbine blade according to an embodiment of the present invention;
FIG. 7 is a graph comparing vibration signals of simulated blades with different defects in an embodiment of the invention;
FIG. 8 is a diagram of a ResNet-GRU neural network architecture constructed in an embodiment of the present invention.
Detailed Description
In order to make the purpose, technical effect and technical solution of the embodiments of the present invention clearer, the following clearly and completely describes the technical solution of the embodiments of the present invention with reference to the drawings in the embodiments of the present invention; it is to be understood that the described embodiments are only some of the embodiments of the present invention.
Referring to fig. 1 to 8, a method for monitoring and identifying a water erosion defect of a turbine blade according to an embodiment of the present invention is mainly based on experimental measurement, numerical simulation, and data training analysis, and a basic implementation flow includes two steps of sampling, modeling, warehousing, and actual operation analysis (see fig. 1).
Step one is sampling modeling warehousing, which can be divided into three stages of experimental measurement, numerical simulation and data training, and the specific implementation mode is as follows:
1. in the stage of water erosion experiment measurement, the measurement is based on the built water erosion rotary experiment system, the blade material can be subjected to accelerated water erosion test (see fig. 2), the original water erosion process for ten thousand hours is shortened to dozens of minutes, and the water erosion appearance of the blade surface in different stages is determined through experiment means. The method can be divided into the following substeps:
1-1) collecting key parameters of operation of a certain steam turbine, specifically comprising material parameters of blades used by a unit and range working condition parameters of the unit operation, wherein the range working condition parameters mainly comprise droplet size, impact angle and impact speed, so that a target material and a target range working condition of an experimental test can be determined. Because the actual operating condition of the steam turbine changes, the invention can select a plurality of operating conditions within the range of the actual operating condition of the steam turbine to carry out the water erosion experiment, and thus, the influence of each factor on the water erosion process of the material is researched.
1-2) processing a target material blank to make the target material blank accord with the size, roughness and impact angle set by an experimental scheme, cleaning the target material blank by using an ultrasonic cleaner, weighing the target material blank by using a precision balance, and recording the initial weight.
1-3) installing a sample, adjusting the working condition of an experiment table to be consistent with the working condition of the actual blade operation range, carrying out a water erosion experiment in a time-sharing manner, recording the weight loss data of the sample, respectively acquiring the macro morphology of the water erosion defect and the corresponding three-dimensional microscopic morphology (see figures 3 and 4) of the sample at each moment by using a single-lens reflex digital camera and a 3D super-depth-of-field microscope, averaging and defining the characteristic width and the characteristic depth of the defect by intercepting the characteristic section of the three-dimensional microscopic morphology of the water erosion defect for multiple times, and observing and recording. After a group of experiments are completed, the working conditions of the experiment table are changed, and the experiment operation is carried out one by one according to a plurality of working conditions formulated by the experiment scheme.
1-4) carrying out data regression analysis on the sample weight loss data and the sample characteristic section size obtained by the experiment to obtain the change rule of the characteristic size (representing the size of the defect and reflecting the water erosion degree) of the material water erosion defect appearance along with time (see figure 5) and a function relation reflecting the influence rule of each factor:
water erosion area A ═ WS 2 (1)
Depth of feature
Characteristic width
Wherein A is the water erosion area of the material surface, WSIs the characteristic width of the water erosion defect; dSIs the characteristic depth of the water erosion defect, Emax is the maximum erosion rate of the material,
is the water erosion state coefficient, d is the weighted mean diameter of the water droplet, n is the droplet size index, V is the impact velocity, V is the water erosion state coefficient0Is the threshold velocity, m is the velocity index, θ is the impact angle, l angle index, Ra is the sample roughness, z is the roughness index, t is the water drop action time, ρ is the density of the material; k is the experimental coefficient and C is the sound velocity in water.
Meanwhile, comparing the water erosion shapes of the actual blades in the characteristic periods (latency period, acceleration period and stabilization period) (see fig. 6), correcting the functional relation with a correction coefficient of mu, wherein mu is f { impact working condition experiment parameter, unit operation working condition parameter, test time and accumulated operation time }, so that the characteristic width W is mu WSThe characteristic depth D ═ μ DS。
2. In the stage of simulating the water erosion numerical value, the numerical value can better generate defect combinations (such as a single large-size water erosion defect at the blade top, two small-size water erosion defects at the blade root and the like) at different positions and different sizes, and the simulation can save a large amount of research time and cost. The method can be divided into the following substeps:
2-1) based on a functional relation of influence of various factors obtained by water erosion experiment measurement on the appearance of a water erosion defect of the blade, establishing a dry friction damping blade vibration characteristic finite element model with different defect combinations (different positions and sizes) by using finite element analysis software, carrying out meshing on the blade by using a three-dimensional finite element entity unit, and establishing a local finite element mesh model between contact surfaces by using a spring damping unit. And adjusting parameters to meet the requirement of collecting actual turbine blade material parameters and operation range working condition parameters in an experimental measurement stage, and then carrying out vibration numerical simulation analysis on the defective blades.
2-2) changing the water erosion defect combination, and obtaining a specific vibration signal waveform diagram of the blade with water erosion defects of different positions and sizes through numerical simulation (see fig. 7). In short, blades with different water erosion defects have different vibration characteristics. For example, in fig. 7, in embodiment 1, the waveform of the vibration signal at the position of the defect near the top of the blade is different from the waveform of the vibration signal at the position near the middle of the blade, while the waveform of the vibration signal at a shallower defect level (smaller size) is different from the waveform of the vibration signal at a deeper defect level (larger size), and the amplitude and the frequency of the vibration signal are different. In order to compare the vibration signal wave diagrams of the defects at different positions, the phase of the normal vibration signal wave diagram is adjusted, although the vibration signal wave diagrams of the defects have differences, the overall rule is still unchanged, the change value is small, particularly the frequency, and the difference is difficult to distinguish by naked eyes.
And 2-3) establishing a water erosion defect combination X-vibration signal oscillogram Y database based on a large number of defect combinations obtained by numerical simulation and corresponding blade vibration signal oscillograms thereof.
3. In the data training and analyzing stage, neural network training is to establish a set of end-to-end (blade vibration signal waveform diagram-water erosion defect feature depth distribution) prediction method, and because the vibration signal waveform diagrams with different defective blades are difficult to artificially define feature differences, machine learning and image recognition need to be introduced, which is a classification problem essentially. And the complex theoretical calculation of material failure is skipped, so that a large amount of analysis time can be saved, the traditional experience judgment result is replaced by the quantitative water erosion degree analysis result, and the technical support can be provided for the actual operation safety of the steam turbine. The method can be divided into the following substeps:
3-1) integrating data, packaging the water erosion defect combination X and the corresponding blade vibration signal oscillogram Y, and dividing a picture set into training set data { X { according to the proportion of a training set/verification set being 3.0t}train、{Yt}trainAnd verification set data { Xt}validation、{Yt}validationAnd randomly shuffles the training set data.
3-2) building a GRU-based neural network structure (see fig. 8): the whole neural network consists of two parts, namely for the original pictureThe ResNet neural network is used for carrying out feature extraction and dimension reduction, and the GRU neural network is used for analyzing and predicting data to obtain the position and the size of the water erosion defect. In the ResNet network part, the step size is reduced in the partial network layer based on the traditional ResNet-34 network architecture. In addition, a jump layer connection is introduced, and a 112 × 112 × 256 tensor is formed to capture low-layer details such as corners and high-layer information at the same time. Operations within a particular network layer also include 3x3 convolution kernels, group normalization, ReLU, max pooling, and the like. After a characteristic blade vibration signal oscillogram obtained by numerical simulation is transmitted into a ResNet neural network, data containing vibration signal oscillogram characteristics are formed through the processing of a characteristic extractor; the processed input information is transmitted into a GRU neural network, and iterative learning is carried out to obtain effective characteristics; adding an activation function as a single neuron L of a linear activation function at the tail end of the model to calculate the predicted value of the GRU neural network, and finally outputting the output of the full-connection layer as different water erosion defect position and degree combinations Xt. Each training data signal Yt}trainObtaining the water erosion defect combination X displayed by the picture through the main structure output of the neural networkt。
3-3) training the network: and (3) setting a variable learning rate optimization cross loss function (MSE) to train the network through an Adam self-adaptive gradient descent algorithm, stopping learning when the loss function is not updated within 100 steps or the number of iteration steps exceeds 1000, and storing the network model. And a dropout method is added in the training process, so that the overfitting phenomenon which possibly occurs is reduced, and the network prediction precision reaches more than 0.85, so that the training requirement is considered to be met, and the method can be put into practical operation process.
And 3-4) establishing a water erosion defect evaluation mechanism based on a water erosion characteristic depth distribution result obtained by neural network prediction, and integrally evaluating the operation safety of the steam turbine. The water erosion defect evaluation mechanism is further explained: the size of the defect in each water erosion defect combination is characterized by the characteristic dimension, wherein the characteristic depth of the water erosion defect is used as an evaluation reference value of the water erosion degree, and whether the characteristic depth D of the water erosion defect is smaller than a threshold value D or not is specifically judgedr1If so, the unit is considered to be safe to operate; if not, then a determination is madeWhether the characteristic dimension D of the water-break corrosion defect is less than the threshold value Dr2If so, determining that the unit is abnormal in operation; if not, judging whether the characteristic dimension D of the water erosion defect is smaller than the threshold D or notr3If so, the unit is considered to be low in operation efficiency; if not, the unit is considered to be in dangerous operation. Briefly, the characteristic water erosion defects corresponding to different vibration signal oscillograms are combined and further classified under the safety level of the operation of the steam turbine, and the method is mainly divided into three levels: abnormal, inefficient, and hazardous operation. Wherein, 0<Dr1<Dr2<Dr3And D isr1＝0.5mm，Dr2＝1mm，Dr3The threshold value is a universality limit value provided on the basis of summarizing a large number of blade water erosion accidents, can effectively distinguish the water erosion state of the last stage blade of the steam turbine, realizes accurate prediction and early warning of the water erosion state of the blade, and is an industry-recognized evaluation index.
And finally, basically completing sampling modeling of the blade water erosion defect monitoring and identifying method corresponding to the actual operating condition of a certain steam turbine and inputting the sampling modeling into an analysis library.
Step two is actual operation analysis, and the specific implementation mode is as follows:
before the method for monitoring and identifying the water erosion defects of the blades is applied to actual operation analysis, firstly measuring points are arranged on an actual unit, a sensor and field signal acquisition equipment are used for acquiring a vibration signal oscillogram of the blades, the oscillogram is input into a model trained by a neural network and output to obtain a corresponding water erosion defect combination, and then the safety level of the unit operation is evaluated according to an evaluation mechanism in a defect feature depth comparison analysis library. If the feedback grade is abnormal or low-efficiency, a worker needs to improve the alertness, increase the monitoring force, make a corresponding operation scheme and a corresponding maintenance scheme, and carry out replacement preparation aiming at the model of the damaged blade, so that the subsequent replacement preparation time is saved, the replacement efficiency is improved, and the economic loss is reduced. And if the feedback grade is dangerous operation, the staff need to immediately prepare for shutdown maintenance and replace a specific blade or a strengthening layer so as to avoid major accidents and eliminate huge economic loss hidden dangers.
And if monitoring of other turbine blades is to be realized, the blade water erosion defect monitoring, identifying and analyzing library suitable for other turbines can be customized in a targeted manner only by repeating the steps.
Claims (6)
1. A turbine blade water erosion defect monitoring and identification method is characterized by comprising the following steps:
step one, sampling, modeling and warehousing
The sampling modeling and warehousing comprises three stages of experimental measurement, numerical simulation and data training, and specifically comprises the following steps:
(1-1) firstly, determining a target material, then processing a sample meeting the requirement, carrying out a water erosion experiment on the sample, and obtaining weight loss data and a water erosion defect three-dimensional microscopic morphology map of the sample at each moment; obtaining the influence degree and the functional relation of an impact angle, an impact speed, a droplet size and sample roughness on the water erosion weight loss and the characteristic size of the water erosion defect through regression analysis of experimental data, and simultaneously comparing a three-dimensional microscopic topography map of the water erosion defect obtained through the experiment with the water erosion topography of an actual steam turbine and correcting the functional relation;
(1-2) based on a function relation obtained by experiments, establishing a blade vibration model with water erosion defects by using finite element analysis software, carrying out numerical simulation analysis to obtain blade vibration models with water erosion defects at different positions and different sizes, and establishing a water erosion defect combination X-vibration signal oscillogram Y database;
(1-3) integrating and dividing a data set based on a water erosion defect combination X-vibration signal oscillogram Y database, constructing a ResNet-GRU network for neural network training, dividing an evaluation mechanism of water erosion degree according to different water erosion defect combinations, and integrally evaluating the safety level of the operation of the steam turbine; the method specifically comprises the following four substeps:
firstly integrating data, packaging a water erosion defect combination X and a corresponding blade vibration signal oscillogram Y, and then dividing a picture set into training set data { X } according to the proportion of a training set/a verification set which is 3.0t}train、{Yt}trainAnd verification set data { Xt}validation、{Yt}validation；
Then establishing a ResNet-GRU neural network structure, after a characteristic blade vibration signal oscillogram obtained by numerical simulation is transmitted into the ResNet neural network, forming data containing vibration signal oscillogram characteristics through the processing of a characteristic extractor, and transmitting the processed input information into the GRU neural network for iterative learning to obtain effective characteristics;
the network is next trained: setting a variable learning rate optimization cross loss function to train the network through an Adam gradient descent algorithm;
and finally, establishing a water erosion defect evaluation mechanism based on a water erosion characteristic depth distribution result obtained by neural network prediction, and further classifying the vibration signal oscillogram of the blade with different water erosion defects under the safety level of the operation of the steam turbine, wherein the vibration signal oscillogram is divided into three levels: abnormal, inefficient, hazardous operation;
step two, actual operation analysis
The field device is used for collecting a blade vibration signal oscillogram, and the safety level of the unit operation is integrally evaluated by utilizing an evaluation mechanism in an analysis library, so that whether shutdown maintenance is carried out or not is determined, and a blade or a strengthening layer is replaced.
2. The turbine blade water erosion defect monitoring and identifying method according to claim 1, characterized in that in the experimental measurement stage (1-1) of the first step, a sample subjected to an accelerated water erosion test is obtained through a high-speed water erosion rotation experiment; measuring weight loss data of the sample by a precision balance; shooting by a single lens reflex digital camera to obtain the macroscopic erosion morphology of the surface of the sample; shooting by a 3D super-depth-of-field microscope to obtain the three-dimensional microscopic morphology of the water erosion defect of the sample; and defining the characteristic width and the characteristic depth of the defect by intercepting the characteristic section of the three-dimensional microscopic topography map of the water erosion defect.
3. The method for monitoring and identifying the water erosion defect of the steam turbine blade according to claim 2, wherein data regression analysis is performed on the sample weight loss data and the sample characteristic section size obtained by the experiment to obtain the change rule of the characteristic size of the material water erosion defect morphology along with time and a functional relation reflecting the influence rule of each factor:
water erosion area A ═ WS 2 (1)
Depth of feature
Characteristic width
Wherein A is the water erosion area of the material surface, WSIs the characteristic width of the water erosion defect; dSIs the characteristic depth of the water erosion defect, Emax is the maximum erosion rate of the material,
is the water erosion state coefficient, d is the weighted mean diameter of the water droplet, n is the droplet size index, V is the impact velocity, V is the water erosion state coefficient0Is the threshold velocity, m is the velocity index, θ is the impact angle, l angle index, Ra is the sample roughness, z is the roughness index, t is the water drop action time, ρ is the density of the material; k is the experimental coefficient, C is the sound velocity in water;
meanwhile, the water erosion appearance of the actual steam turbine is compared, the functional relation is corrected, the correction coefficient is mu, and therefore the characteristic width W is equal to mu WSThe characteristic depth D ═ μ DS。
4. The turbine blade water erosion defect monitoring and identification method according to claim 3, characterized in that in the numerical simulation stage (1-2) of the first step, a finite element analysis software is used for establishing a dry friction damping blade vibration characteristic finite element model with different defect combinations, a three-dimensional finite element entity unit is used for carrying out mesh division on the blade, and a spring damping unit is used for establishing a local finite element mesh model between contact surfaces; a plurality of blades with different water erosion defect combinations are subjected to vibration response simulation to obtain a large number of blade vibration signal oscillograms in one-to-one correspondence, and therefore a water erosion defect combination X-vibration signal oscillogram Y database is established.
5. The method for monitoring and identifying the water erosion defect of the steam turbine blade according to claim 1, wherein in the water erosion defect evaluation mechanism in the sub-step four of the numerical training stage (1-3), the size of the defect in each water erosion defect combination is represented by the characteristic dimension thereof, wherein the characteristic depth of the water erosion defect is used as an evaluation reference value of the water erosion degree, and whether the characteristic depth D of the water erosion defect is smaller than a threshold D or not is specifically judgedr1If so, the unit is considered to be safe to operate; if not, judging whether the characteristic dimension D of the water erosion defect is smaller than the threshold D or notr2If so, determining that the unit is abnormal in operation; if not, judging whether the characteristic dimension D of the water erosion defect is smaller than the threshold D or notr3If so, the unit is considered to be low in operation efficiency; if not, the unit is considered to be in dangerous operation; wherein, 0<Dr1<Dr2<Dr3And D isr1＝0.5mm，Dr2＝1mm，Dr3＝2mm。
6. The turbine blade water erosion defect monitoring and identification method according to claim 5, wherein in the second step, corresponding operation and maintenance schemes are made for abnormal, inefficient or dangerous operation according to the feedback information; if monitoring of other turbine blades is to be realized, the blade water erosion defect monitoring, identifying and analyzing library suitable for other turbines can be customized in a targeted manner only by repeating the step one.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911183911.1A CN110887899B (en) | 2019-11-27 | 2019-11-27 | Turbine blade water erosion defect monitoring and identifying method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911183911.1A CN110887899B (en) | 2019-11-27 | 2019-11-27 | Turbine blade water erosion defect monitoring and identifying method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110887899A CN110887899A (en) | 2020-03-17 |
| CN110887899B true CN110887899B (en) | 2020-11-10 |
Family
ID=69749100
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911183911.1A Active CN110887899B (en) | 2019-11-27 | 2019-11-27 | Turbine blade water erosion defect monitoring and identifying method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110887899B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111474300B (en) * | 2020-04-15 | 2021-04-30 | 同济大学 | Structure local defect detection method based on space-time regression model |
| CN113204841B (en) * | 2021-04-28 | 2023-09-05 | 西安热工研究院有限公司 | Turbine moving blade numerical value checking and analyzing method |
| CN114893258B (en) * | 2022-05-05 | 2023-05-12 | 华北电力大学（保定） | Steam turbine last stage stationary blade surface vapor film measurement and dehumidification integrated system |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6448668B1 (en) * | 1999-06-30 | 2002-09-10 | Armand Robitaille | Vertical-axis wind mill supported by a fluid |
| JP4630745B2 (en) * | 2005-07-05 | 2011-02-09 | バブコック日立株式会社 | Calculation method of flow accelerated corrosion thinning rate and remaining life diagnosis method |
| CN201653838U (en) * | 2010-03-05 | 2010-11-24 | 湖南省湘电锅炉压力容器检验中心有限公司 | Water erosion test apparatus |
| KR20120077899A (en) * | 2010-12-31 | 2012-07-10 | 중앙산업 (주) | Scraper device with circular blade |
| CN103233222A (en) * | 2013-04-17 | 2013-08-07 | 武汉点金激光科技有限公司 | Laser cladding method of steam turbine last-stage blade inlet edge surface |
| CN104502214B (en) * | 2014-12-23 | 2017-08-15 | 西安交通大学 | Turbine blade material and strengthened coat Anti-erosion and corrosive nature pilot system |
-
2019
- 2019-11-27 CN CN201911183911.1A patent/CN110887899B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN110887899A (en) | 2020-03-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110887899B (en) | Turbine blade water erosion defect monitoring and identifying method | |
| US8116990B2 (en) | Method and system for real-time prognosis analysis and usage based residual life assessment of turbine engine components and display | |
| CA2604118C (en) | A system and method for real-time prognostics analysis and residual life assessment of machine components | |
| JP5595000B2 (en) | Method and system for monitoring the condition of a wind turbine | |
| CN106870298A (en) | Blade root bolt fracture fault detection method based on machine learning | |
| CN114619292B (en) | Milling cutter wear monitoring method based on fusion of wavelet denoising and attention mechanism with GRU network | |
| EP2300887A2 (en) | Methods, apparatus and computer readable storage mediums for model-based diagnosis of gearboxes | |
| CN110006552B (en) | Method for detecting abnormal temperature of equipment | |
| CN111881574B (en) | Wind turbine generator set key component reliability modeling method based on distribution function optimization | |
| CN102588210A (en) | Filtering method for preprocessing fitting data of power curve | |
| Vatani et al. | Health monitoring and degradation prognostics in gas turbine engines using dynamic neural networks | |
| CN103925155A (en) | Self-adaptive detection method for abnormal wind turbine output power | |
| CN112067701A (en) | Fan blade remote auscultation method based on acoustic diagnosis | |
| CN112711850A (en) | Unit online monitoring method based on big data | |
| CN110513336B (en) | Method for determining offline water washing time of gas turbine of power station | |
| CN114004059B (en) | Health portrait method for hydroelectric generating set | |
| CN115163424A (en) | Wind turbine generator gearbox oil temperature fault detection method and system based on neural network | |
| CN112926698A (en) | Vibration prediction and assembly evaluation method for large-scale rotating equipment | |
| CN109325310B (en) | High-speed train intermittent fault detection method based on multiple T-square control diagram | |
| CN116628976A (en) | Comprehensive evaluation method for state change of hydraulic turbine unit equipment | |
| CN113987871B (en) | Online identification method for damage of aero-engine blade | |
| CN104252573A (en) | Stepwise mode-based mechanical multi-fault diagnosis method | |
| Velasco et al. | Wind turbine blade damage detection using data-driven techniques | |
| CN116460653A (en) | Numerical control machine tool data processing method based on sensor | |
| CN117605660A (en) | Dynamic energy-saving control method and system for air compressor |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |