CN117826658A - Control system of gas turbine impeller washing module - Google Patents
Control system of gas turbine impeller washing module Download PDFInfo
- Publication number
- CN117826658A CN117826658A CN202410072851.0A CN202410072851A CN117826658A CN 117826658 A CN117826658 A CN 117826658A CN 202410072851 A CN202410072851 A CN 202410072851A CN 117826658 A CN117826658 A CN 117826658A
- Authority
- CN
- China
- Prior art keywords
- blade
- cleaning
- surface state
- training
- image
- 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.)
- Granted
Links
- 238000005406 washing Methods 0.000 title abstract description 39
- 238000004140 cleaning Methods 0.000 claims abstract description 273
- 230000000694 effects Effects 0.000 claims abstract description 72
- 238000012545 processing Methods 0.000 claims abstract description 9
- 238000012549 training Methods 0.000 claims description 85
- 238000010586 diagram Methods 0.000 claims description 78
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 52
- 238000005259 measurement Methods 0.000 claims description 30
- 238000005728 strengthening Methods 0.000 claims description 30
- 239000013598 vector Substances 0.000 claims description 23
- 238000000605 extraction Methods 0.000 claims description 14
- 239000011159 matrix material Substances 0.000 claims description 10
- 238000001514 detection method Methods 0.000 claims description 9
- 238000004364 calculation method Methods 0.000 claims description 5
- 230000004913 activation Effects 0.000 claims description 2
- 230000003416 augmentation Effects 0.000 claims description 2
- 238000011156 evaluation Methods 0.000 abstract description 11
- 239000000428 dust Substances 0.000 abstract description 8
- 238000012423 maintenance Methods 0.000 abstract description 6
- 238000004458 analytical method Methods 0.000 abstract description 5
- 239000007789 gas Substances 0.000 description 51
- 238000000034 method Methods 0.000 description 15
- 230000000875 corresponding effect Effects 0.000 description 13
- 230000006870 function Effects 0.000 description 12
- 239000007788 liquid Substances 0.000 description 12
- 238000009826 distribution Methods 0.000 description 8
- 230000009471 action Effects 0.000 description 4
- 230000008901 benefit Effects 0.000 description 4
- 230000009286 beneficial effect Effects 0.000 description 3
- 238000005457 optimization Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 239000000356 contaminant Substances 0.000 description 2
- 239000003599 detergent Substances 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000001105 regulatory effect Effects 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000007792 addition Methods 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- QVGXLLKOCUKJST-UHFFFAOYSA-N atomic oxygen Chemical compound [O] QVGXLLKOCUKJST-UHFFFAOYSA-N 0.000 description 1
- 239000012459 cleaning agent Substances 0.000 description 1
- 238000002485 combustion reaction Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 239000000306 component Substances 0.000 description 1
- 230000001276 controlling effect Effects 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 239000008358 core component Substances 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 238000002790 cross-validation Methods 0.000 description 1
- -1 dirt Substances 0.000 description 1
- 238000005265 energy consumption Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000012530 fluid Substances 0.000 description 1
- 239000004519 grease Substances 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 239000001301 oxygen Substances 0.000 description 1
- 229910052760 oxygen Inorganic materials 0.000 description 1
- 238000010248 power generation Methods 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 230000003252 repetitive effect Effects 0.000 description 1
- 239000002689 soil Substances 0.000 description 1
- 238000005507 spraying Methods 0.000 description 1
Landscapes
- Structures Of Non-Positive Displacement Pumps (AREA)
- Cleaning In General (AREA)
Abstract
A control system of a gas turbine impeller washing module is characterized in that a camera is used for collecting images before blade washing and images after blade washing, an image processing and analyzing algorithm is introduced into the rear end to analyze and compare the images before blade washing and the images after blade washing, so that whether dirt and dust on blades are effectively removed or not is judged, and the control of the washing module is carried out based on the analysis result so as to judge when to stop the washing module. Therefore, the automatic evaluation and control of the blade cleaning effect of the gas compressor can be realized, so that the blade cleaning effect is improved, the operation and maintenance cost is reduced, the normal and efficient operation of the gas turbine is ensured, and the performance and the service life of the gas turbine are improved.
Description
Technical Field
The present application relates to the field of intelligent control technology, and more particularly, to a control system for a gas turbine impeller water wash module.
Background
The gas turbine is a thermodynamic machine using gas as working fluid, and is widely applied to the fields of power generation, aviation, ships and the like. One of the core components of a gas turbine is a compressor, which is responsible for compressing and delivering air to a combustor for providing oxygen for combustion. During operation of a gas turbine, contaminants such as dirt, dust, grease and the like are easily accumulated on the surfaces of blades of the gas turbine, and the contaminants can reduce the surface quality of the blades, affect the performance and service life of the gas turbine, increase oil consumption and emission, and even cause damage or falling of the blades. Therefore, to ensure proper and efficient operation of the gas turbine, the compressor blades need to be cleaned periodically.
A water wash module is a common compressor blade cleaning device that washes dirt and dust from a surface by spraying a water stream onto the blades. The cleaning effect of the water washing module directly influences the performance recovery degree of the air compressor, so that the cleaning effect of the water washing module needs to be effectively evaluated. At present, the cleaning effect evaluation of the water washing module mainly depends on manual observation and experience judgment, and the method is time-consuming and labor-consuming, and has subjectivity and inaccuracy.
Accordingly, an optimized control system for a gas turbine impeller water wash module is desired.
Disclosure of Invention
The present application has been made in order to solve the above technical problems. The embodiment of the application provides a control system of a gas turbine impeller washing module, which collects images before blade washing and images after blade washing through a camera, and introduces image processing and analysis algorithms at the rear end to analyze and compare the images before blade washing and the images after blade washing so as to judge whether dirt and dust on the blades are effectively removed or not, and controls the washing module based on analysis results so as to judge when to stop the washing module. Therefore, the automatic evaluation and control of the blade cleaning effect of the gas compressor can be realized, so that the blade cleaning effect is improved, the operation and maintenance cost is reduced, the normal and efficient operation of the gas turbine is ensured, and the performance and the service life of the gas turbine are improved.
In a first aspect, a control system for a gas turbine impeller water wash module is provided, comprising:
the blade cleaning image acquisition module is used for acquiring a blade cleaning front image and a blade cleaning rear image acquired by the camera;
the blade surface state feature extraction module is used for enabling the image before blade cleaning and the image after blade cleaning to pass through a blade surface state twin detector so as to obtain a blade surface state feature map before blade cleaning and a blade surface state feature map after blade cleaning;
the surface state characteristic strengthening expression module is used for enabling the surface state characteristic diagram before blade cleaning and the surface state characteristic diagram after blade cleaning to obtain a surface state characteristic diagram after blade cleaning through a surface state strengthening expression device based on a space mutual attention layer;
the cleaning effect semantic expression module is used for calculating cleaning front and back state semantic measurement coefficients between the front surface state characteristic diagram of the blade and the characteristic matrix of each group of corresponding channel dimensions of the surface state characteristic diagram of the reinforced blade so as to obtain cleaning effect semantic expression characteristics consisting of a plurality of cleaning front and back state semantic measurement coefficients;
and the dirt cleaning detection module is used for determining whether dirt is effectively cleaned based on the cleaning effect semantic expression characteristics.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are needed in the embodiments or the description of the prior art will be briefly introduced below, it being obvious that the drawings in the following description are only some embodiments of the present application, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 is a schematic flow diagram of a gas turbine according to an embodiment of the present application.
FIG. 2 is a block diagram of a control system for a gas turbine impeller water wash module according to an embodiment of the present application.
Fig. 3 is a flowchart of a control method of a gas turbine impeller water wash module according to an embodiment of the present application.
Fig. 4 is a schematic diagram of a control method architecture of a gas turbine impeller water wash module according to an embodiment of the present application.
Fig. 5 is an application scenario diagram of a control system of a gas turbine impeller water wash module according to an embodiment of the present application.
Detailed Description
The following description of the technical solutions in the embodiments of the present application will be made with reference to the drawings in the embodiments of the present application, and it is apparent that the described embodiments are some embodiments of the present application, but not all embodiments. All other embodiments, which can be made by one of ordinary skill in the art without undue burden from the present disclosure, are within the scope of the present disclosure.
Unless defined otherwise, all technical and scientific terms used in the examples of this application have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in the present application is for the purpose of describing particular embodiments only and is not intended to limit the scope of the present application.
In the description of the embodiments of the present application, unless otherwise indicated and defined, the term "connected" should be construed broadly, and for example, may be an electrical connection, may be a communication between two elements, may be a direct connection, or may be an indirect connection via an intermediary, and it will be understood by those skilled in the art that the specific meaning of the term may be understood according to the specific circumstances.
It should be noted that, the term "first\second\third" in the embodiments of the present application is merely to distinguish similar objects, and does not represent a specific order for the objects, it is to be understood that "first\second\third" may interchange a specific order or sequence where allowed. It is to be understood that the "first\second\third" distinguishing objects may be interchanged where appropriate such that the embodiments of the present application described herein may be implemented in sequences other than those illustrated or described herein.
Having described the basic principles of the present application, various non-limiting embodiments of the present application will now be described in detail with reference to the accompanying drawings.
During operation of the gas turbine, the blade surfaces of the compressor may be subject to dirt, dust, and the like, which may affect the performance and life of the compressor, so that a washing module is required to wash the blades.
The main components of the water washing module are as follows: the device comprises a water tank, a cleaning liquid tank, a filter, a vane pump, a ball valve, an ejector, a one-way valve and a float valve. The control system mainly comprises a detection instrument, a central control unit (including a PLC touch screen and the like), an electric valve and the like. The main instrument of the water washing module is as follows: level gauge, flowmeter, thermometer, pressure switch, flow switch. The whole system is arranged in a special container, and an air circulation system is formed by a space heater, a circulation ventilator and the like.
As shown in fig. 1, the operation principle flow of the gas turbine includes: the water tank is internally provided with an invasive heater, water is required to be heated to 85 ℃ before cleaning, the water in the water tank enters the vane pump through the filter, meanwhile, the detergent in the cleaning liquid tank also enters the high-pressure water pump group (vane pump) under the action of the ejector, and the detergent and the water enter the client through the water outlet pipeline after being mixed so as to clean the compressor vane of the gas turbine, and the outlet pressure is as follows: 87Psi, temperature: 85 ℃, outlet flow: 22.5 GPM, flow of cleaning liquid pipeline: 4.5GPM.
And (3) a control system: the water level of the water tank is controlled by the ball float valve, if the water level is too high, the ball float valve is closed, and the water supply of the water tank is terminated; if the water level is low, the float valve is opened and water is supplied into the tank. The liquid level of the cleaning liquid tank is controlled by a float valve, if the liquid level of the cleaning liquid is too high, the float valve is closed, and the cleaning liquid is stopped being provided; if the liquid level is low, the float valve is opened and cleaning liquid enters the cleaning liquid tank. The pressure switch is arranged at the front and the back of the pump, if the pressure is abnormal, a pressure signal can be transmitted into the control unit in the electrical cabinet, and the pump can stop working immediately.
In one embodiment of the present application, FIG. 2 is a block diagram of a control system for a gas turbine impeller water wash module according to an embodiment of the present application. As shown in fig. 2, a control system 100 of a gas turbine impeller washing module according to an embodiment of the present application includes: a blade cleaning image acquisition module 110 for acquiring a blade cleaning pre-image and a blade cleaning post-image acquired by the camera; the blade surface state feature extraction module 120 is configured to pass the image before blade cleaning and the image after blade cleaning through a blade surface state twin detector to obtain a blade surface state feature map before blade cleaning and a blade surface state feature map after blade cleaning; a surface state feature enhancement expression module 130, configured to obtain a surface state feature map after blade cleaning by using a surface state enhancement expression device based on a spatial mutual attention layer for the blade cleaning front surface state feature map and the blade cleaning rear surface state feature map; the cleaning effect semantic expression module 140 is configured to calculate a cleaning effect semantic expression feature composed of a plurality of cleaning effect semantic measurement coefficients, where the cleaning effect semantic measurement coefficients are between the feature matrices of each group of corresponding channel dimensions of the front surface state feature map and the back surface state feature map of the enhanced blade; the dirt cleaning detection module 150 is configured to determine whether dirt is effectively cleaned based on the cleaning effect semantic expression feature.
In the blade cleaning image acquisition module 110, it is ensured that the camera can accurately capture images before and after blade cleaning, and the definition and accuracy of the images are ensured. In this way, image data before and after cleaning can be provided, which provides a basis for subsequent surface state feature extraction and cleaning effect evaluation. In the blade surface state feature extraction module 120, the blade surface state twin detector is used to extract the surface state features of the blade before and after cleaning, so as to ensure that the algorithm can accurately extract dirt on the surface of the blade and the state features after cleaning. And converting images before and after blade cleaning into a surface state characteristic diagram, and providing input for subsequent characteristic strengthening and cleaning effect evaluation. In the surface state characteristic enhancement expression module 130, a surface state enhancement expression device based on a spatial mutual attention layer is used, so that the surface state characteristic of the blade after cleaning can be effectively enhanced, and the accuracy of the cleaning effect is improved. The method provides a surface state characteristic diagram after generating the enhanced blade cleaning, and provides more accurate input for cleaning effect semantic expression. In the cleaning effect semantic expression module 140, the cleaning front and back state semantic measurement coefficients between the feature matrices of each group of corresponding channel dimensions of the surface state feature graphs before and after cleaning are calculated, so that the semantic difference of the cleaning effect can be accurately measured. And generating semantic expression characteristics of the cleaning effect consisting of a plurality of semantic measurement coefficients of states before and after cleaning, and evaluating the cleaning effect. In the dirt cleaning detection module 150, it is determined whether the dirt is effectively cleaned based on the cleaning effect semantic expression feature, and an appropriate threshold or model judgment standard needs to be set. The dirt cleaning detection result is provided, the effectiveness of the cleaning effect is judged, and further cleaning work is guided.
Aiming at the technical problems, the technical concept of the application is that a camera is used for collecting a blade pre-cleaning image and a blade post-cleaning image, an image processing and analyzing algorithm is introduced at the rear end to analyze and compare the blade pre-cleaning image and the blade post-cleaning image, so as to judge whether dirt and dust on a blade are effectively removed, and the control of a water washing module is carried out based on an analysis result to judge when to stop the water washing module. Therefore, the automatic evaluation and control of the blade cleaning effect of the gas compressor can be realized, so that the blade cleaning effect is improved, the operation and maintenance cost is reduced, the normal and efficient operation of the gas turbine is ensured, and the performance and the service life of the gas turbine are improved.
Specifically, in the technical scheme of the application, first, a blade pre-cleaning image and a blade post-cleaning image acquired by a camera are acquired. Next, feature extraction of the pre-blade-cleaning image and the post-blade-cleaning image is performed separately using a convolutional neural network model having excellent performance in terms of implicit feature extraction of images. In particular, considering that the characteristic information of the two images needs to be analyzed and compared after the characteristic extraction, whether the dirt on the blade is cleaned is judged. Therefore, in the technical scheme of the application, the image before blade cleaning and the image after blade cleaning are passed through a blade surface state twin detector comprising a first image encoder and a second image encoder to obtain a blade surface state characteristic diagram before blade cleaning and a blade surface state characteristic diagram after blade cleaning. The blade surface state twin detector is used for processing, and the image before blade cleaning and the image after blade cleaning can be subjected to feature extraction and analysis so as to capture the surface state feature distribution information about the blade in the two images respectively. In particular, the first image encoder and the second image encoder have the same network result, so that feature distribution information which is not obvious at the source domain end of the two images is focused during image feature extraction, and therefore the change of the surface state of the blade, including the distribution condition of dirt and dust and the difference of cleaning effects, is more beneficial to capture.
In a specific embodiment of the present application, the blade surface state twinning detector comprises a first image encoder and a second image encoder.
Specifically, the blade surface state feature extraction module includes: the blade pre-cleaning image passes through a first image encoder of the blade surface state twin detector to obtain a blade pre-cleaning surface state characteristic diagram; and passing the blade cleaned image through a second image encoder of the blade surface state twin detector to obtain the blade cleaned surface state characteristic diagram.
During the cleaning of gas turbine blades, some fine dirt or non-uniformity may still exist in the surface condition of the blade after cleaning. Therefore, in order to describe the surface state of the blade more accurately, in the technical scheme of the application, the surface state characteristic diagram before cleaning the blade and the surface state characteristic diagram after cleaning the blade are further processed through a surface state strengthening expression device based on a spatial mutual attention layer to obtain a surface state characteristic diagram after cleaning the blade. It should be appreciated that the spatial mutual attention mechanism may focus on domain invariant features of the blade post-cleaning surface state feature map relative to the blade pre-cleaning surface state feature map, thereby reducing cross-validation of domain invariant features with domain variant features in the feature space itself, and thereby achieving the attention effect of the output feature map on domain invariant features. Thus, by processing the spatial mutual attention layer, the blade pre-cleaning surface state feature map and the post-cleaning surface state feature map can be spatially interacted and correlated. Therefore, the surface state characteristic diagram obtained by the surface state strengthening expression device after the blade is cleaned can better capture the details and state changes of the surface of the blade, thereby reflecting the cleaning effect of the blade more accurately.
In a specific embodiment of the present application, the surface state feature enhancement expression module is configured to: processing the surface state characteristic diagram before blade cleaning and the surface state characteristic diagram after blade cleaning through a surface state strengthening expression device based on a space mutual attention layer according to the following strengthening formula to obtain the surface state characteristic diagram after blade cleaning strengthening; wherein, the strengthening formula is:
wherein (1)>For the surface state profile after cleaning of the blade, < > for>Striving for spatial mutual awareness->Representing an activation function->Represents a convolution layer, and->Indicating an inexpensive augmentation of the surface state signature after cleaning of the blade when the size of the convolution kernel is greater than one,is a characteristic diagram of the state of the front surface of the blade before cleaning, < >>A surface state characteristic diagram after cleaning the reinforced blade,representing multiplication by location.
Further, since the surface state feature diagrams before and after cleaning all contain rich semantic information of the blade state, each feature matrix along the channel dimension in the feature diagrams represents each type of semantic feature information of the blade state respectively. Therefore, in order to quantify the state difference before and after the cleaning of the blade so as to evaluate the cleaning effect better, in the technical scheme of the application, the cleaning effect semantic expression feature vector composed of a plurality of cleaning effect semantic measurement coefficients is further calculated by the state semantic measurement coefficients before and after the cleaning between the feature matrixes of each group of corresponding channel dimensions of the surface state feature map before the cleaning of the blade and the surface state feature map after the cleaning of the reinforced blade. It should be understood that by calculating the semantic measurement coefficients between the feature matrices of each set of corresponding channel dimensions in the surface state feature diagrams before and after cleaning, a set of measurement values describing the cleaning effect can be obtained, and these measurement values can reflect the semantic feature difference degree of each type of the blade surface state before and after cleaning, and are used for reflecting the dirt removal degree of the blade surface, the improvement of the surface uniformity, and the like. And then integrating the semantic measurement coefficients of the states before and after cleaning into the semantic expression feature vector of the cleaning effect, which is used for describing the overall change and difference of blade cleaning, thereby being beneficial to more accurately judging the cleaning effect of the blade.
In a specific embodiment of the present application, the cleaning effect semantic expression module includes: calculating the cleaning front and back state semantic measurement coefficients between the characteristic matrixes of each group of corresponding channel dimensions of the blade cleaning front surface state characteristic diagram and the reinforced blade cleaning back surface state characteristic diagram by using the following semantic measurement formulas to obtain a cleaning effect semantic expression characteristic vector consisting of a plurality of cleaning front and back state semantic measurement coefficients as the cleaning effect semantic expression characteristic; wherein, the semantic measurement formula is:
wherein (1)>And->The characteristic diagram and the characteristic diagram of the state of the surface of the blade before cleaning are respectivelyThe characteristic values of the positions of the characteristic matrix of each corresponding channel dimension of the surface state characteristic diagram after the cleaning of the reinforced blade are +.>Is the width of the feature matrix of each group of corresponding channel dimensions of the surface state feature diagram before blade cleaning and the surface state feature diagram after blade cleaning strengthening, +.>Is the height of the feature matrix of each group of corresponding channel dimensions of the surface state feature diagram before blade cleaning and the surface state feature diagram after blade cleaning strengthening, +.>Is the feature value of each position in the cleaning effect semantic expression feature vector,/for each position in the cleaning effect semantic expression feature vector>A logarithmic function with a base of 2 is shown.
And then, the cleaning effect semantic expression feature vector is passed through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the dirt is effectively cleaned. That is, classification processing is performed using the cleaning effect semantic expression characteristic information of the blade surface of the compressor to determine whether dirt and dust on the blade has been effectively removed, and control of the washing module is performed based on the classification result to determine when to stop the washing module.
In a specific embodiment of the present application, the soil cleaning detection module is configured to: and the cleaning effect semantic expression feature vector is passed through a classifier to obtain a classification result, wherein the classification result is used for indicating whether dirt is effectively cleaned.
The classifier is used for classifying the semantic expression feature vectors of the cleaning effect, so that automatic dirt cleaning detection can be realized, and subjectivity and uncertainty of manual judgment are reduced. Through the classification result of the classifier, whether the dirt is effectively cleaned can be rapidly judged, the evaluation efficiency is improved, and images are not required to be compared one by one manually or complex calculation is not required, so that time and energy are saved. Through judging by using the classifier, the objectivity and consistency of the evaluation can be improved, the classifier judges based on training data and a model, the influence of subjective factors is avoided, and a consistent evaluation result can be obtained. The classifier can be trained and adjusted as required to adapt to different cleaning tasks and dirt types, so that the method has certain flexibility and expandability and can be applied to different scenes and application requirements.
In one embodiment of the present application, the control system of the gas turbine impeller washing module further comprises a training module for training the blade surface state twinning detector comprising the first image encoder and the second image encoder, the surface state intensification expression based on the spatial mutual attention layer, and the classifier. The training module comprises: the training blade cleaning image acquisition unit is used for acquiring a training blade cleaning image and a training blade cleaning image acquired by the camera; the training blade surface state feature extraction unit is used for enabling the image before cleaning the training blade and the image after cleaning the training blade to pass through the blade surface state twin detector so as to obtain a training blade surface state feature map before cleaning and a training blade surface state feature map after cleaning; the training surface state characteristic strengthening expression unit is used for enabling the surface state characteristic diagram before cleaning the training blade and the surface state characteristic diagram after cleaning the training blade to pass through the surface state strengthening expression device based on the spatial mutual attention layer so as to obtain a surface state characteristic diagram after cleaning the training strengthening blade; the training cleaning effect semantic expression unit is used for calculating the cleaning front and back state semantic measurement coefficients between the training blade cleaning front surface state feature map and the feature matrix of each group of corresponding channel dimensions of the training reinforced blade cleaning front surface state feature map so as to obtain a training cleaning effect semantic expression feature vector consisting of a plurality of cleaning front and back state semantic measurement coefficients; the training classification unit is used for enabling the training cleaning effect semantic expression feature vector to pass through the classifier to obtain a classification loss function value; a training calculation unit for calculating a specific loss function value between the surface state feature map before the training blade is cleaned and the surface state feature map after the training reinforced blade is cleaned; a training unit for training the blade surface state twin detector including the first image encoder and the second image encoder, the spatial mutual attention layer based surface state enhancement expressive and the classifier with a weighted sum of the classification loss function value and the specific loss function value as a loss function value.
In particular, in the above technical solution, the training vane pre-cleaning surface state feature map and the training vane post-cleaning surface state feature map represent image semantic features based on convolution encoding of the training vane pre-cleaning image and the training vane post-cleaning image, respectively, and respective local features of the training vane pre-cleaning surface state feature map and the training vane post-cleaning surface state feature map are distributed along the channel direction of the vane surface state twin detector including the first image encoder and the second image encoder. However, in consideration of the difference in source image between the pre-cleaning image of the training blade and the post-cleaning image of the training blade and the possible image noise, the obtained feature group density representations of the pre-cleaning surface state feature map of the training blade and the post-cleaning surface state feature map of the training blade in the distribution dimension of the overall feature map may also have a difference, and such feature group density differences may be increased due to the spatial dimension saliency expression of the spatial-mutual-attention-layer-based surface state enhancement expressive device, thereby affecting the calculation accuracy of the pre-cleaning and post-cleaning state semantic metric coefficients between the feature matrices of each set of corresponding channel dimensions of the pre-cleaning surface state feature map of the training blade and the post-cleaning surface state feature map of the training blade.
Thus, to promote consistency in the feature group density representation of the training blade pre-cleaning surface state feature map and the training enhanced blade post-cleaning surface state feature map, the applicant of the present application further introduced specific loss functions for the training blade pre-cleaning surface state feature map and the training enhanced blade post-cleaning surface state feature map expressed as: calculating a specific loss function value between the surface state characteristic diagram before cleaning the training blade and the surface state characteristic diagram after cleaning the training reinforced blade according to the following optimization formula; wherein, the optimization formula is:
wherein,and->The first feature vector and the second feature vector are respectively obtained by expanding the surface state feature map before cleaning the training blade and the surface state feature map after cleaning the training reinforced blade, and the first feature vector and the second feature vector are respectively obtained by expanding the surface state feature map before cleaning the training blade and the surface state feature map after cleaning the training reinforced blade>Is the length of the feature vector, andrepresenting the square of the two norms of the vector, +.>Representing a specific loss function value, ">Representing per-position subtraction.
Here, the loss function performs group count attention based on feature group density by performing adaptive attention of different density representation modes between the training blade pre-cleaning surface state feature map and the training enhanced blade post-cleaning surface state feature map by recursively mapping the group count as an output feature group density. By taking the model as a loss function to train the model, the model can avoid overestimation and underestimation aiming at different density modes under the characteristic distribution of the surface state characteristic diagram before the training blade is cleaned and the surface state characteristic diagram after the training strengthening blade is cleaned, and the corresponding relation between the characteristic value distribution and the group density distribution is learned, so that the semantic distribution consistency optimization between the surface state characteristic diagram before the training blade is cleaned and the surface state characteristic diagram after the training strengthening blade with different characteristic densities is realized. Therefore, the calculation accuracy of the state semantic measurement coefficients before and after training and cleaning is improved, the expression effect of the cleaning effect semantic expression feature vector formed by a plurality of state semantic measurement coefficients before and after training and cleaning is improved, and the accuracy of the classification result obtained by the classifier is improved. Therefore, the automatic evaluation and control of the blade cleaning effect of the gas compressor can be realized, so that the blade cleaning effect is improved, the operation and maintenance cost is reduced, the normal and efficient operation of the gas turbine is ensured, and the performance and the service life of the gas turbine are improved.
The control system of the gas turbine impeller washing module has the beneficial effects that:
1. the whole system is assembled in a container, the container is waterproof, dampproof and sound-proof, and an air circulation system with a temperature regulating function is arranged, so that the whole system is not influenced by external climate and environment factors to ensure the optimal working state.
2. Unique jet application. Compared with systems of other factories, the system uses the ejector to replace a low-pressure water pump, reduces energy consumption and enables the input amount of the cleaning agent to be accurately regulated. And the ejector is almost maintenance-free relative to the water pump set, so that the online running time of the unit is ensured, and the shutdown maintenance time is reduced and avoided.
3. The central control system detects and controls the self-checking, liquid level, temperature and flow of the system, so that the automation degree of the system is improved, and the unattended operation is realized.
In summary, the control system 100 of the gas turbine impeller washing module according to the embodiment of the present application is illustrated, which can realize automatic evaluation and control of the blade cleaning effect of the gas turbine, so as to improve the blade cleaning effect, reduce the operation and maintenance costs, and ensure the normal and efficient operation of the gas turbine, thereby improving the performance and the service life of the gas turbine.
As described above, the control system 100 of the gas turbine impeller water washing module according to the embodiment of the present application may be implemented in various terminal devices, such as a server or the like for control of the gas turbine impeller water washing module. In one example, the control system 100 of the gas turbine impeller water wash module according to embodiments of the present application may be integrated into the terminal device as a software module and/or hardware module. For example, the control system 100 of the gas turbine impeller water wash module may be a software module in the operating system of the terminal device, or may be an application developed for the terminal device; of course, the control system 100 of the turbine wheel wash module may also be one of the hardware modules of the terminal device.
Alternatively, in another example, the control system 100 of the gas turbine impeller water washing module and the terminal device may be separate devices, and the control system 100 of the gas turbine impeller water washing module may be connected to the terminal device through a wired and/or wireless network and transmit interactive information in a agreed data format.
In one embodiment of the present application, FIG. 3 is a flow chart of a method of controlling a gas turbine impeller water wash module according to an embodiment of the present application. Fig. 4 is a schematic diagram of a control method architecture of a gas turbine impeller water wash module according to an embodiment of the present application. As shown in fig. 3 and 4, the control method of the turbine impeller washing module comprises the following steps: 210, acquiring a blade pre-cleaning image and a blade post-cleaning image acquired by a camera; 220, passing the image before blade cleaning and the image after blade cleaning through a blade surface state twin detector to obtain a blade surface state characteristic diagram before blade cleaning and a blade surface state characteristic diagram after blade cleaning; 230, passing the blade cleaning front surface state characteristic diagram and the blade cleaning rear surface state characteristic diagram through a surface state strengthening expression device based on a space mutual attention layer to obtain a strengthening blade cleaning rear surface state characteristic diagram; 240, calculating a cleaning front and rear state semantic measurement coefficient between the characteristic matrixes of each group of corresponding channel dimensions of the blade cleaning front surface state characteristic diagram and the reinforced blade cleaning rear surface state characteristic diagram so as to obtain a cleaning effect semantic expression characteristic consisting of a plurality of cleaning front and rear state semantic measurement coefficients; 250, determining whether the dirt is effectively cleaned based on the cleaning effect semantic expression features.
It will be appreciated by those skilled in the art that the specific operation of the respective steps in the above-described control method of the gas turbine impeller water washing module has been described in detail in the above description of the control system of the gas turbine impeller water washing module with reference to fig. 2, and thus, repetitive description thereof will be omitted.
Fig. 5 is an application scenario diagram of a control system of a gas turbine impeller water wash module according to an embodiment of the present application. As shown in fig. 5, in this application scenario, first, a pre-blade-cleaning image (e.g., C1 as illustrated in fig. 5) and a post-blade-cleaning image (e.g., C2 as illustrated in fig. 5) acquired by a camera are acquired; the acquired pre-blade-cleaning image and post-blade-cleaning image are then input to a server (e.g., S as illustrated in fig. 5) that deploys a control algorithm of the gas turbine wheel washing module, wherein the server is capable of processing the pre-blade-cleaning image and the post-blade-cleaning image based on the control algorithm of the gas turbine wheel washing module to determine whether the dirt is effectively cleaned.
The basic principles of the present application have been described above in connection with specific embodiments, however, it should be noted that the advantages, benefits, effects, etc. mentioned in the present application are merely examples and not limiting, and these advantages, benefits, effects, etc. are not to be considered as necessarily possessed by the various embodiments of the present application. Furthermore, the specific details disclosed herein are for purposes of illustration and understanding only, and are not intended to be limiting, as the application is not intended to be limited to the details disclosed herein as such.
The previous description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other aspects without departing from the scope of the application. Thus, the present application is not intended to be limited to the aspects shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Finally, it is further 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 terminal 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 terminal. 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 terminal device comprising the element.
The foregoing description has been presented for purposes of illustration and description. Furthermore, this description is not intended to limit the embodiments of the application to the form disclosed herein. Although a number of example aspects and embodiments have been discussed above, a person of ordinary skill in the art will recognize certain variations, modifications, alterations, additions, and subcombinations thereof.
Claims (7)
1. A control system for a gas turbine impeller water wash module, comprising:
the blade cleaning image acquisition module is used for acquiring a blade cleaning front image and a blade cleaning rear image acquired by the camera;
the blade surface state feature extraction module is used for enabling the image before blade cleaning and the image after blade cleaning to pass through a blade surface state twin detector so as to obtain a blade surface state feature map before blade cleaning and a blade surface state feature map after blade cleaning;
the surface state characteristic strengthening expression module is used for enabling the surface state characteristic diagram before blade cleaning and the surface state characteristic diagram after blade cleaning to obtain a surface state characteristic diagram after blade cleaning through a surface state strengthening expression device based on a space mutual attention layer;
the cleaning effect semantic expression module is used for calculating cleaning front and back state semantic measurement coefficients between the front surface state characteristic diagram of the blade and the characteristic matrix of each group of corresponding channel dimensions of the surface state characteristic diagram of the reinforced blade so as to obtain cleaning effect semantic expression characteristics consisting of a plurality of cleaning front and back state semantic measurement coefficients;
the dirt cleaning detection module is used for determining whether dirt is effectively cleaned based on the cleaning effect semantic expression characteristics;
wherein, the surface state characteristic strengthening expression module is used for: processing the surface state characteristic diagram before blade cleaning and the surface state characteristic diagram after blade cleaning through a surface state strengthening expression device based on a space mutual attention layer according to the following strengthening formula to obtain the surface state characteristic diagram after blade cleaning strengthening;
wherein, the strengthening formula is:
wherein (1)>For the surface state profile after cleaning of the blade, < > for>Striving for spatial mutual awareness->Representing an activation function->Represents a convolution layer, and->Indicating an inexpensive augmentation of the surface state profile after cleaning of the blade when the size of the convolution kernel is greater than one, +.>Is a characteristic diagram of the state of the front surface of the blade before cleaning, < >>For the surface state characteristic map after cleaning of the reinforced blade, < > for the reinforced blade>Representing multiplication by location.
2. The control system of a gas turbine impeller water wash module of claim 1, wherein the blade surface state twinning detector comprises a first image encoder and a second image encoder.
3. The control system of a gas turbine impeller water wash module of claim 2, wherein the blade surface state feature extraction module comprises:
passing the pre-cleaning blade image through a first image encoder of the blade surface state twin detector to obtain the pre-cleaning blade surface state feature map;
and passing the blade cleaned image through a second image encoder of the blade surface state twin detector to obtain the blade cleaned surface state characteristic diagram.
4. A control system for a gas turbine impeller water wash module according to claim 3, wherein said cleaning effect semantic expression module comprises: calculating the cleaning front and back state semantic measurement coefficients between the characteristic matrixes of each group of corresponding channel dimensions of the blade cleaning front surface state characteristic diagram and the reinforced blade cleaning back surface state characteristic diagram by using the following semantic measurement formulas to obtain a cleaning effect semantic expression characteristic vector consisting of a plurality of cleaning front and back state semantic measurement coefficients as the cleaning effect semantic expression characteristic;
wherein, the semantic measurement formula is:
wherein (1)>And->The characteristic values of the respective positions of the characteristic matrix of each set of corresponding channel dimensions of the blade-cleaning-front surface state characteristic diagram and the enhanced blade-cleaning-rear surface state characteristic diagram are respectively->Is the width of the feature matrix of each group of corresponding channel dimensions of the surface state feature diagram before blade cleaning and the surface state feature diagram after blade cleaning strengthening, +.>Is the height of the feature matrix of each group of corresponding channel dimensions of the surface state feature diagram before blade cleaning and the surface state feature diagram after blade cleaning strengthening, +.>Is the feature value of each position in the cleaning effect semantic expression feature vector,/for each position in the cleaning effect semantic expression feature vector>A logarithmic function with a base of 2 is shown.
5. The control system of a gas turbine impeller water wash module of claim 4, wherein the dirty wash detection module is configured to: and the cleaning effect semantic expression feature vector is passed through a classifier to obtain a classification result, wherein the classification result is used for indicating whether dirt is effectively cleaned.
6. The control system of a gas turbine impeller water wash module of claim 5, further comprising a training module for training the blade surface state twinning detector comprising a first image encoder and a second image encoder, the spatial mutual attention layer based surface state intensification expression, and the classifier.
7. The control system of a gas turbine impeller water wash module of claim 6, wherein the training module comprises:
the training blade cleaning image acquisition unit is used for acquiring a training blade cleaning image and a training blade cleaning image acquired by the camera;
the training blade surface state feature extraction unit is used for enabling the image before cleaning the training blade and the image after cleaning the training blade to pass through the blade surface state twin detector so as to obtain a training blade surface state feature map before cleaning and a training blade surface state feature map after cleaning;
the training surface state characteristic strengthening expression unit is used for enabling the surface state characteristic diagram before cleaning the training blade and the surface state characteristic diagram after cleaning the training blade to pass through the surface state strengthening expression device based on the spatial mutual attention layer so as to obtain a surface state characteristic diagram after cleaning the training strengthening blade;
the training cleaning effect semantic expression unit is used for calculating training cleaning effect semantic expression feature vectors composed of a plurality of training cleaning effect semantic measurement coefficients, wherein the training cleaning effect semantic measurement coefficients are between the training cleaning front surface state feature diagram and the training cleaning back surface state feature diagram of each group of feature matrixes corresponding to channel dimensions;
the training classification unit is used for enabling the training cleaning effect semantic expression feature vector to pass through the classifier to obtain a classification loss function value;
a training calculation unit for calculating a specific loss function value between the surface state feature map before the training blade is cleaned and the surface state feature map after the training reinforced blade is cleaned;
a training unit for training the blade surface state twin detector including the first image encoder and the second image encoder, the spatial mutual attention layer based surface state enhancement expressive and the classifier with a weighted sum of the classification loss function value and the specific loss function value as a loss function value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410072851.0A CN117826658B (en) | 2024-01-18 | 2024-01-18 | Control system of gas turbine impeller washing module |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410072851.0A CN117826658B (en) | 2024-01-18 | 2024-01-18 | Control system of gas turbine impeller washing module |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117826658A true CN117826658A (en) | 2024-04-05 |
| CN117826658B CN117826658B (en) | 2024-07-05 |
Family
ID=90524118
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410072851.0A Active CN117826658B (en) | 2024-01-18 | 2024-01-18 | Control system of gas turbine impeller washing module |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117826658B (en) |
Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPS60166702A (en) * | 1984-02-10 | 1985-08-30 | Hitachi Ltd | Contamination detecting device for gas turbine compressor blades |
| JP2000274206A (en) * | 1999-03-24 | 2000-10-03 | Hitachi Ltd | Gas turbine |
| CN109115798A (en) * | 2018-08-14 | 2019-01-01 | 珠海格力电器股份有限公司 | Control device and method for detecting cleanliness of fan based on image recognition and fan |
| CN110378870A (en) * | 2019-06-06 | 2019-10-25 | 西安交通大学 | A kind of turbine blade erosion degree method of discrimination based on ResNet-GRU network |
| CN113847107A (en) * | 2021-09-26 | 2021-12-28 | 中国联合重型燃气轮机技术有限公司 | Method and device for monitoring accumulated dirt state of gas turbine compressor blade |
| CN218493902U (en) * | 2022-09-02 | 2023-02-17 | 重庆通用工业(集团)有限责任公司 | Cleaning structure of MVR large-tonnage steam centrifugal compressor |
| CN116482119A (en) * | 2023-03-17 | 2023-07-25 | 北京航力安太科技有限责任公司 | One-time imaging inspection system and method for aero-engine Kong Tan |
| CN116834244A (en) * | 2023-07-10 | 2023-10-03 | 滁州精镁装备模具制造有限公司 | Image monitoring alarm system and method for injection mold |
| CN117237367A (en) * | 2023-11-16 | 2023-12-15 | 江苏星火汽车部件制造有限公司 | Spiral blade thickness abrasion detection method and system based on machine vision |
| CN117233164A (en) * | 2023-09-13 | 2023-12-15 | 大连中远海运海事工程技术有限公司 | Machine vision-based waste gas turbine cleaning sewage water quality judging method and system |
| CN117372881A (en) * | 2023-12-08 | 2024-01-09 | 中国农业科学院烟草研究所(中国烟草总公司青州烟草研究所) | Intelligent identification method, medium and system for tobacco plant diseases and insect pests |
| WO2024042508A1 (en) * | 2022-08-24 | 2024-02-29 | Edgy Bees Ltd. | Geosynchronization of an aerial image using localizing multiple features |
-
2024
- 2024-01-18 CN CN202410072851.0A patent/CN117826658B/en active Active
Patent Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPS60166702A (en) * | 1984-02-10 | 1985-08-30 | Hitachi Ltd | Contamination detecting device for gas turbine compressor blades |
| JP2000274206A (en) * | 1999-03-24 | 2000-10-03 | Hitachi Ltd | Gas turbine |
| CN109115798A (en) * | 2018-08-14 | 2019-01-01 | 珠海格力电器股份有限公司 | Control device and method for detecting cleanliness of fan based on image recognition and fan |
| CN110378870A (en) * | 2019-06-06 | 2019-10-25 | 西安交通大学 | A kind of turbine blade erosion degree method of discrimination based on ResNet-GRU network |
| CN113847107A (en) * | 2021-09-26 | 2021-12-28 | 中国联合重型燃气轮机技术有限公司 | Method and device for monitoring accumulated dirt state of gas turbine compressor blade |
| WO2024042508A1 (en) * | 2022-08-24 | 2024-02-29 | Edgy Bees Ltd. | Geosynchronization of an aerial image using localizing multiple features |
| CN218493902U (en) * | 2022-09-02 | 2023-02-17 | 重庆通用工业(集团)有限责任公司 | Cleaning structure of MVR large-tonnage steam centrifugal compressor |
| CN116482119A (en) * | 2023-03-17 | 2023-07-25 | 北京航力安太科技有限责任公司 | One-time imaging inspection system and method for aero-engine Kong Tan |
| CN116834244A (en) * | 2023-07-10 | 2023-10-03 | 滁州精镁装备模具制造有限公司 | Image monitoring alarm system and method for injection mold |
| CN117233164A (en) * | 2023-09-13 | 2023-12-15 | 大连中远海运海事工程技术有限公司 | Machine vision-based waste gas turbine cleaning sewage water quality judging method and system |
| CN117237367A (en) * | 2023-11-16 | 2023-12-15 | 江苏星火汽车部件制造有限公司 | Spiral blade thickness abrasion detection method and system based on machine vision |
| CN117372881A (en) * | 2023-12-08 | 2024-01-09 | 中国农业科学院烟草研究所(中国烟草总公司青州烟草研究所) | Intelligent identification method, medium and system for tobacco plant diseases and insect pests |
Non-Patent Citations (2)
| Title |
|---|
| XUELI PENG, ET.AL.: "Optical Remote Sensing Image Change Detection Based on Attention Mechanism and Image Difference", IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, vol. 59, no. 9, 10 November 2020 (2020-11-10), pages 7296 - 7307, XP011874094, DOI: 10.1109/TGRS.2020.3033009 * |
| 杜俊翰等: "基于多尺度注意力特征与孪生判别的遥感影像变化检测及其抗噪性研究", 数据采集与处理, vol. 37, no. 1, 31 January 2022 (2022-01-31), pages 35 - 48 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117826658B (en) | 2024-07-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108959778B (en) | Method for predicting residual life of aircraft engine based on consistency of degradation modes | |
| Li et al. | An enhanced PCA-based chiller sensor fault detection method using ensemble empirical mode decomposition based denoising | |
| EP2317461A1 (en) | Turbine operation degradation determination system and method | |
| CN101726567B (en) | Method and system for detecting a corrosive deposit in a compressor | |
| CN110751385B (en) | Non-invasive load identification method, terminal device and storage medium | |
| CN112669305B (en) | Metal surface rust resistance test bench and rust resistance evaluation method | |
| CN110056544B (en) | Method for obtaining washing period of compressor | |
| CN110513336B (en) | Method for determining offline water washing time of gas turbine of power station | |
| CN111178602A (en) | Circulating water loss prediction method based on support vector machine and neural network | |
| CN110991701A (en) | Wind power plant fan wind speed prediction method and system based on data fusion | |
| CN110428100A (en) | A kind of blower short-term power generation power prediction technique | |
| CN117826658B (en) | Control system of gas turbine impeller washing module | |
| US11143056B2 (en) | System and method for gas turbine compressor cleaning | |
| CN110033181B (en) | Power generation equipment state evaluation method based on self-encoder | |
| CN113468732B (en) | System and method for determining production cost for heat supply of steam extraction heat supply unit | |
| EP3373233A1 (en) | Scheduling maintenance to reduce degradation of a power generation system | |
| CN116778301A (en) | Quantitative detection method and system for combustion state of hearth flame | |
| CN116384843A (en) | Energy efficiency evaluation model training method and monitoring method for digital energy nitrogen station | |
| CN109272154B (en) | Blower fault prediction method based on typical variable analysis and hidden Markov | |
| Song et al. | Improved CEEMDAN-based aero-engine gas-path parameter forecasting using SCINet | |
| KR100752765B1 (en) | A real-time calculation system and method on performance impact of controllable operation parameters for combined-cycle power plant | |
| CN111859773A (en) | Electric gate valve fault determination method and system based on regularization particle filtering | |
| CN116629029B (en) | Data-driven-based flow industry user flexibility assessment method and related equipment | |
| CN117784864B (en) | Self-adaptive state control photovoltaic panel cleaning robot environment adjusting method and system | |
| CN117968431B (en) | Method and device for controlling flue gas waste heat recovery of coal-fired power plant |
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 |