CN111914490A - Pump station unit state evaluation method based on deep convolution random forest self-coding - Google Patents
Pump station unit state evaluation method based on deep convolution random forest self-coding Download PDFInfo
- Publication number
- CN111914490A CN111914490A CN202010895695.XA CN202010895695A CN111914490A CN 111914490 A CN111914490 A CN 111914490A CN 202010895695 A CN202010895695 A CN 202010895695A CN 111914490 A CN111914490 A CN 111914490A
- Authority
- CN
- China
- Prior art keywords
- waveform
- vibration
- coding
- pump station
- station unit
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/24323—Tree-organised classifiers
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/02—Reliability analysis or reliability optimisation; Failure analysis, e.g. worst case scenario performance, failure mode and effects analysis [FMEA]
Abstract
The invention discloses a pump station unit state evaluation method based on deep convolution random forest self-coding, which comprises the steps of collecting an upper guide vibration X-direction waveform, an upper guide vibration Y-direction vibration waveform, a water guide vibration X-direction vibration waveform, a water guide vibration Y-direction vibration waveform, an upper frame X-direction vibration waveform, an upper frame Y-direction vibration waveform, an upper frame vertical vibration waveform, impeller shell X-direction vibration, impeller shell Y-direction vibration, a large-axis swing waveform and key phase signals of all waveform data of a pump station unit in the whole life cycle to obtain a historical waveform health sample library; selecting information of each measuring point of a pump station unit in a health state period from a historical waveform health sample library as a historical normal data sample; and the like. The advantages are that: based on the on-line monitoring data, the shutdown and manual arrangement of measuring points are not needed, the economic benefit of the power plant is increased, the modeling is performed by utilizing the multi-source data of the units, the hidden characteristics are automatically mined, and the method is suitable for the units with different forms and operation time.
Description
Technical Field
The invention relates to the technical field of pump station state evaluation, in particular to a pump station unit state evaluation method based on deep convolutional random forest self-coding.
Background
With the continuous construction of the south-to-north water transfer project, the pump station unit is an important power device of the water transfer project, and the reliability of the pump station unit directly influences the safe and economic operation of the water transfer project. Meanwhile, as the water demand of the user side is continuously variable, the adjusting frequency and the adjusting depth are greatly improved compared with the prior art, and the pump station unit works under severe conditions of high temperature, high speed and the like and under the influence of various random factors, so that various faults of the unit are easy to occur. The performance of each component is gradually degraded with the increase of the service time, thereby causing a vicious accident. Therefore, the running state of the unit needs to be accurately mastered in real time, so that an accurate and scientific decision basis is made for the running of the unit.
Aiming at the evaluation of the mechanical performance state of a pump station unit, China mostly adopts the method of evaluating according to vibration measurement results of key points of unit components and analyzing the health condition of the unit through a performance test. Because the pump station unit is influenced by various factors such as water, electricity and the like, various performance tests are adopted to evaluate the state, and the running stability state of the pump station unit is represented by the abnormity or standard exceeding of vibration and water pressure pulsation. However, traditional health assessment requires manual sensor placement, data collection, and manual data processing and analysis, and manual testing usually requires downtime or after being scheduled and overhauled, is limited by construction period and causes some economic loss. Meanwhile, corresponding judgment standards for the operation health of the pump station unit are lacked at home and abroad at present, and the evaluation is generally carried out according to related regulations. Because the standard or the regulation only defines a certain limit value, the unit types can be various in the actual operation process of the power station, and meanwhile, because the operation working condition of the unit changes frequently, the evaluation standard under the single limit value cannot accurately and reasonably measure the health state of the unit. Meanwhile, the performance degradation of the unit may only be reflected in the trend change of the performance index along with the time, and the condition that the monitoring value exceeds the limit value does not occur yet, but the unit operation is obviously abnormal at the moment. Therefore, the health state of the unit running state cannot be effectively evaluated by simply adopting a limit evaluation method under the current single working condition point.
With the continuous development of the industrial internet technology, the real-time online evaluation and the residual life estimation of the unit are possible by combining multi-source online information with a data driving model. However, conventional life estimation algorithms often require that a failure curve, such as exponential, linear, be assumed and the health indicators be extrapolated with the help of the assumed model to predict the residual life of the unit. Because the pump station unit is adjusted frequently, the assumption is far away from the actual situation, on one hand, the failure curve cannot follow a fixed shape alone, and is difficult to express by a simple trend model; on the other hand, the time to failure is often not the same. Meanwhile, the method also has the defects of sensor noise, small sample number and the like, and is difficult to use in an actual system.
Disclosure of Invention
The invention aims to provide a pump station unit state evaluation method based on deep convolutional random forest self-coding, so that the problems in the prior art are solved.
In order to achieve the purpose, the technical scheme adopted by the invention is as follows:
a pump station unit state evaluation method based on deep convolutional random forest self-coding comprises the following steps,
s1, collecting the upward vibration X-direction waveform, upward vibration Y-direction waveform, water-guide vibration X-direction waveform, water-guide vibration Y-direction waveform, upper frame X-direction vibration waveform, upper frame Y-direction vibration waveform, upper frame vertical vibration waveform, impeller shell X-direction vibration, impeller shell Y-direction vibration, large-axis swing waveform and key phase signals of the waveform data of the pump station unit in the whole life cycle to obtain a historical waveform health sample library;
s2, selecting information of each measuring point in the health state period of the pump station unit from the historical waveform health sample library as a historical normal data sample;
s3, selecting the historical normal data samples according to the key phase signals, selecting the whole-period data of each waveform, and forming a two-dimensional matrix with a horizontal axis as a time point and a vertical axis as a channel number according to corresponding time scales;
s4, extracting the features of the two-dimensional matrix through a deep convolution neural network to obtain sample features;
s5, inputting the sample characteristics into a random forest self-coding model for further compression to obtain a health characteristic sample set;
s6, monitoring the pump station unit on line, and acquiring an upward vibration X-direction waveform, an upward vibration Y-direction vibration waveform, a water-guide vibration X-direction vibration waveform, a water-guide vibration Y-direction vibration waveform, an upper frame X-direction vibration waveform, an upper frame Y-direction vibration waveform, an upper frame vertical vibration waveform, an impeller shell X-direction vibration, an impeller shell Y-direction vibration, a large-axis swing waveform and key phase signals of all waveform data of the pump station unit in real time to obtain a waveform sample to be tested;
s7, executing the steps S2 to S5 on the waveform sample to be tested, and acquiring a feature sample set to be tested;
s8, performing Euclidean distance calculation on the to-be-tested feature sample set and the health feature sample set to obtain a distance deviation degree;
and S9, evaluating the health state of the pump station unit by using the distance deviation degree.
Preferably, the deep convolutional neural network comprises a positive convolutional model and a reverse convolutional model, wherein the positive convolutional model adopts 5 layers of convolution-pooling layers, and convolution kernels are 128 × 3, 64 × 3, 32 × 3, 16 × 3 and 8 × 3 in sequence; the deconvolution model adopts 5 convolution-pooling layers, and convolution kernels are 8 × 3, 16 × 3, 32 × 3, 64 × 3 and 128 × 3 in sequence; and the forward convolution model and the deconvolution model perform feature extraction on the two-dimensional matrix to obtain sample features.
Preferably, the random forest self-coding model is a tree integration model of a random forest to construct a self-coding process; the random forest self-coding model comprises forward coding and reverse coding; after receiving the sample features, the forward coding sends the sample features to each root node of the tree in the integration, and after traversing all the root nodes of the tree, the sample features return a first result vector; the reverse coding utilizes the characteristic that each leaf node in the tree corresponds to a unique path from the root to the leaf, a rule base is established to obtain each path from the leaf node to the root node of the tree, a plurality of rules in the rule base are simplified according to the logic that the range of each constraint in the rules can not be expanded, finally, the mapping between each leaf node to the root node of all the trees is obtained, namely, a second result vector mapped with the first result vector is obtained, and the second result vector is a healthy feature sample set.
The invention has the beneficial effects that: 1. the failure trend of the unit is obtained by analyzing a plurality of full life cycle examples from beginning to failure of the unit, and when a new failure trend of the unit occurs, the most possible trend change can be calculated through a model and the residual life can be estimated. 2. According to the invention, the deep learning can automatically represent the health state of the unit by automatically analyzing the inherent fault characteristics of the excavator group through mass data, modeling the data of the normal behavior of the unit and abstracting the implicit characteristics of the data step by step, after the model is established, the health degree of the unit can be measured by calculating the similarity between the current sensor data and the normal behavior state, and the more accurate residual life can be obtained through an extrapolation algorithm. 3. The state evaluation method is based on the on-line monitoring data, does not need to shut down and manually arrange measuring points, thereby greatly increasing the economic benefit of the power plant, can comprehensively utilize the multi-source data of the units to model, automatically excavate the hidden characteristics of the units, does not need experts to manually judge, is suitable for the units with different forms and different running times, and has self-adaptability and objectivity.
Drawings
FIG. 1 is a schematic flow chart of an evaluation method in an embodiment of the invention;
FIG. 2 is a backward decoding map of tree nodes in an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail below with reference to the accompanying drawings. It should be understood that the detailed description and specific examples, while indicating the invention, are intended for purposes of illustration only and are not intended to limit the scope of the invention.
Example one
As shown in fig. 1, in this embodiment, a method for evaluating a pump station unit state based on deep convolutional random forest self-coding is provided, which includes the following steps,
s1, collecting the upward vibration X-direction waveform, upward vibration Y-direction waveform, water-guide vibration X-direction waveform, water-guide vibration Y-direction waveform, upper frame X-direction vibration waveform, upper frame Y-direction vibration waveform, upper frame vertical vibration waveform, impeller shell X-direction vibration, impeller shell Y-direction vibration, large-axis swing waveform and key phase signals of the waveform data of the pump station unit in the whole life cycle to obtain a historical waveform health sample library;
s2, selecting information of each measuring point in the health state period of the pump station unit from the historical waveform health sample library as a historical normal data sample;
s3, selecting the historical normal data samples according to the key phase signals, selecting the whole-period data of each waveform, and forming a two-dimensional matrix with a horizontal axis as a time point and a vertical axis as a channel number according to corresponding time scales;
s4, extracting the features of the two-dimensional matrix through a deep convolution neural network to obtain sample features;
s5, inputting the sample characteristics into a random forest self-coding model for further compression to obtain a health characteristic sample set;
s6, monitoring the pump station unit on line, and acquiring an upward vibration X-direction waveform, an upward vibration Y-direction vibration waveform, a water-guide vibration X-direction vibration waveform, a water-guide vibration Y-direction vibration waveform, an upper frame X-direction vibration waveform, an upper frame Y-direction vibration waveform, an upper frame vertical vibration waveform, an impeller shell X-direction vibration, an impeller shell Y-direction vibration, a large-axis swing waveform and key phase signals of all waveform data of the pump station unit in real time to obtain a waveform sample to be tested;
s7, executing the steps S2 to S5 on the waveform sample to be tested, and acquiring a feature sample set to be tested;
s8, performing Euclidean distance calculation on the to-be-tested feature sample set and the health feature sample set to obtain a distance deviation degree;
and S9, evaluating the health state of the pump station unit by using the distance deviation degree.
In this embodiment, in the two-dimensional matrix, the horizontal axis includes 4096 time points; the vertical axis includes 10 channel numbers.
In this embodiment, the deep convolutional neural network includes a positive convolutional model and a reverse convolutional model, the positive convolutional model employs 5 layers of convolutional-pooling layers, and the convolutional kernels are 128 × 3, 64 × 3, 32 × 3, 16 × 3 and 8 × 3 in sequence; the deconvolution model adopts 5 convolution-pooling layers, and convolution kernels are 8 × 3, 16 × 3, 32 × 3, 64 × 3 and 128 × 3 in sequence; and the forward convolution model and the deconvolution model perform feature extraction on the two-dimensional matrix to obtain sample features.
In this example, the positive convolution model was constructed using 5 convolution-pooling layers, and the convolution kernels of the positive convolution model were (128 × 3), (64 × 3), (32 × 3), (16 × 3), and (8 × 3), respectively. The deconvolution model also needs to be deconvoluted by deconvolution kernels of (8 × 3), (16 × 3), (32 × 3), (64 × 3), and (128 × 3). And modeling and training by adopting a random forest self-coding model between the positive convolution and the deconvolution.
And (3) carrying out a reverse coding process of a random forest on the extracted feature sample to obtain a root node sequence of the feature sample, using the root node sequence as an input of the deconvolution, and carrying out the deconvolution according to deconvolution kernels of (8 × 3), (16 × 3), (32 × 3), (64 × 3) and (128 × 3) to obtain a deconvolution output.
And calculating errors between original data (positive convolution input) and deconvolution output, and when the obtained errors are minimum after multiple times of cyclic iterative training, indicating that the neural network model and the random forest self-coding model are successfully trained, and testing the sample to be tested by using the neural network model and the random forest self-coding model.
In this embodiment, 1024-dimensional sample features are generated through convolution-pooling operations.
In this embodiment, the sample characteristics are input into a random forest self-coding model for further compression, and 100-dimensional characteristics are obtained, that is, a health characteristic sample set.
In this embodiment, the random forest self-coding model is a tree integration model of a random forest to construct a self-coding process; the random forest self-coding model comprises forward coding and reverse coding; after receiving the sample features, the forward coding sends the sample features to each root node of the tree in the integration, and after traversing all the root nodes of the tree, the sample features return a first result vector; the reverse coding utilizes the characteristic that each leaf node in the tree corresponds to a unique path from the root to the leaf, a rule base is established to obtain each path from the leaf node to the root node of the tree, a plurality of rules in the rule base are simplified according to the logic that the range of each constraint in the rules can not be expanded, finally, the mapping between each leaf node to the root node of all the trees is obtained, namely, a second result vector mapped with the first result vector is obtained, and the second result vector is a healthy feature sample set.
In this embodiment, the Euclidean distance calculation is performed between the feature sample set to be tested and the healthy feature sample set to obtain a distance deviation degree, which is specifically calculated as follows,
wherein d is12The distance deviation degree between the characteristic sample set to be tested and the health characteristic sample set is obtained; x is the number of1And y1Characteristic of a healthy sample, x2And y2Is the sample feature to be tested.
In this embodiment, the distance deviation degree is used as a state evaluation index of the pump station unit, and the specific evaluation mode is that the built-in characteristic dimension is set to be 50 dimensions, and a 50-dimensional characteristic vector, namely a health characteristic sample set, is obtained by bringing the health sample of the historical pump station unit into the model; when the on-line data of the pump station unit is acquired, the calculation is carried out in the same way to acquire a characteristic sample set to be tested, after the deviation degree calculation and normalization are carried out on the characteristic sample set to be tested and the health characteristic sample set, the result value is 1 to represent complete failure, the result value is 0 to represent health, the comprehensive health degree of the pump station unit can be estimated through the change of the result value from 0 to 1, and particularly when the health degree shows trend rising, a certain fault is indicated.
Example two
In this embodiment, an axial flow pump station unit of a river-Jihuai project is taken as an example to specifically describe the implementation process of the evaluation method of the present invention.
(1) The rated rotating speed and the structural form of the unit are investigated firstly, the implementation object adopts two guide type arrangement (an upper guide bearing and a water guide bearing), only an upper frame is adopted for supporting, meanwhile, the rotating speed of the unit is 1200r/min, the calculated main frequency is about 20Hz, and therefore, a speed type sensor is adopted for collection. The pump station unit was arranged with sensors as shown in the following table:
(2) and (3) performing equal-period sampling on the waveforms, and acquiring and storing the waveforms of all parts by adopting a sampling rate of more than 3000Hz according to the fragrance concentration law.
After storing the online waveform, combining with expert experience, carrying out expert marking on the historical waveform, collecting normal samples to form a historical waveform health sample library, capturing waveform data of a positive period through the 0-1 state of a key phase signal, and aligning the waveforms.
(3) And selecting the data according to the key phase signals, selecting the whole-period data of respective waveforms, and forming a two-dimensional matrix with a horizontal axis as 4096 time points (based on the actual sampling rate) and a vertical axis as channel numbers (10) according to corresponding time scales.
(4) The construction was performed using 5 convolution-pooling layers with convolution kernels of (128 × 3), (64 × 3), (32 × 3), (16 × 3) (8 × 3), respectively, and the resulting output was expanded to generate features. The deconvolution model also needs to be deconvoluted by deconvolution kernels of (8 × 3), (16 × 3), (32 × 3), (64 × 3), and (128 × 3). And performing modeling training by adopting random forest self-coding between convolution and deconvolution.
(5) And extracting features through a deep convolutional neural network, and generating 1024-dimensional sample features through a convolution-pooling operation. And (4) bringing the sample characteristics into a random forest self-coding model, and further compressing to obtain 100-dimensional characteristics as a health characteristic sample set.
The random forest self-coding adopts a tree integration model of random forests to construct a self-coding process which comprises a forward coding part and a backward coding part, the forward coding process receives input data and then sends the data to each root node of the trees in the integration, and when the data traverses leaf nodes of all the trees, the process returns a result vector. The process from root to leaf node is forward coding. And (3) backward decoding, namely establishing a rule base by utilizing the characteristic that each leaf node in the tree corresponds to a unique path from the root to the leaf, finding each path from the leaf node to the root node of the tree, simplifying a plurality of rules according to a maximum complete rule, namely logic that the range of each constraint in the rule can not be expanded any more, and finally obtaining the mapping from each leaf node to the root node, thereby realizing the backward decoding.
The symbol rule corresponding to backward decoding is:
RULE1:(x1≥0)∧(x2≥1.5)∧(x3==RED)
according to the rule being the most complete rule, i.e. the rule cannot be expanded any more within a constraint range, if expanded, a conflict arises.
(6) And (3) after the samples are obtained through online monitoring, obtaining a characteristic sample set to be tested according to the steps (2) to (3), performing Euclidean distance calculation with the health characteristic sample set, and taking the obtained distance deviation degree as a health index.
The degree of distance deviation is:
by adopting the technical scheme disclosed by the invention, the following beneficial effects are obtained:
the invention provides a pump station unit state evaluation method based on deep convolution random forest self-coding, which is characterized in that a unit failure trend is obtained by analyzing a plurality of full life cycle examples from the beginning to the failure of a unit, and when a new failure trend appears in the unit, the most possible trend change can be calculated through a model and the residual life can be estimated. The method comprises the steps of deeply learning the inherent fault characteristics of the excavator group which can be automatically realized through mass data, representing the health state of the excavator group by modeling the data of the normal behavior of the excavator group and abstracting the implicit characteristics of the data step by step, measuring the health degree of the excavator group by calculating the similarity between the current sensor data and the normal behavior state after the model is built, and obtaining the more accurate residual life through an extrapolation algorithm. The state evaluation method is based on online monitoring data, does not need to shut down and manually arrange measuring points, greatly increases the economic benefit of the power plant, can comprehensively utilize multi-source data of the units to model, automatically excavates the hidden characteristics of the units, does not need experts to manually judge, is suitable for the units with different forms and different running times, and has self-adaptability and objectivity.
The foregoing is only a preferred embodiment of the present invention, and it should be noted that, for those skilled in the art, various modifications and improvements can be made without departing from the principle of the present invention, and such modifications and improvements should also be considered within the scope of the present invention.
Claims (3)
1. A pump station unit state evaluation method based on deep convolutional random forest self-coding is characterized by comprising the following steps: comprises the following steps of (a) carrying out,
s1, collecting the upward vibration X-direction waveform, upward vibration Y-direction waveform, water-guide vibration X-direction waveform, water-guide vibration Y-direction waveform, upper frame X-direction vibration waveform, upper frame Y-direction vibration waveform, upper frame vertical vibration waveform, impeller shell X-direction vibration, impeller shell Y-direction vibration, large-axis swing waveform and key phase signals of the waveform data of the pump station unit in the whole life cycle to obtain a historical waveform health sample library;
s2, selecting information of each measuring point in the health state period of the pump station unit from the historical waveform health sample library as a historical normal data sample;
s3, selecting the historical normal data samples according to the key phase signals, selecting the whole-period data of each waveform, and forming a two-dimensional matrix with a horizontal axis as a time point and a vertical axis as a channel number according to corresponding time scales;
s4, extracting the features of the two-dimensional matrix through a deep convolution neural network to obtain sample features;
s5, inputting the sample characteristics into a random forest self-coding model for further compression to obtain a health characteristic sample set;
s6, monitoring the pump station unit on line, and acquiring an upward vibration X-direction waveform, an upward vibration Y-direction vibration waveform, a water-guide vibration X-direction vibration waveform, a water-guide vibration Y-direction vibration waveform, an upper frame X-direction vibration waveform, an upper frame Y-direction vibration waveform, an upper frame vertical vibration waveform, an impeller shell X-direction vibration, an impeller shell Y-direction vibration, a large-axis swing waveform and key phase signals of all waveform data of the pump station unit in real time to obtain a waveform sample to be tested;
s7, executing the steps S2 to S5 on the waveform sample to be tested, and acquiring a feature sample set to be tested;
s8, performing Euclidean distance calculation on the to-be-tested feature sample set and the health feature sample set to obtain a distance deviation degree;
and S9, evaluating the health state of the pump station unit by using the distance deviation degree.
2. The pump station unit state evaluation method based on the deep convolutional random forest self-coding as claimed in claim 1, wherein: the deep convolution neural network comprises a positive convolution model and a reverse convolution model, wherein the positive convolution model adopts 5 layers of convolution-pooling layers, and convolution kernels are 128 × 3, 64 × 3, 32 × 3, 16 × 3 and 8 × 3 in sequence; the deconvolution model adopts 5 convolution-pooling layers, and convolution kernels are 8 × 3, 16 × 3, 32 × 3, 64 × 3 and 128 × 3 in sequence; and the forward convolution model and the deconvolution model perform feature extraction on the two-dimensional matrix to obtain sample features.
3. The pump station unit state evaluation method based on the deep convolutional random forest self-coding as claimed in claim 1, wherein: the random forest self-coding model is a tree integration model of a random forest to construct a self-coding process; the random forest self-coding model comprises forward coding and reverse coding; after receiving the sample features, the forward coding sends the sample features to each root node of the tree in the integration, and after traversing all the root nodes of the tree, the sample features return a first result vector; the reverse coding utilizes the characteristic that each leaf node in the tree corresponds to a unique path from the root to the leaf, a rule base is established to obtain each path from the leaf node to the root node of the tree, a plurality of rules in the rule base are simplified according to the logic that the range of each constraint in the rules can not be expanded, finally, the mapping between each leaf node to the root node of all the trees is obtained, namely, a second result vector mapped with the first result vector is obtained, and the second result vector is a healthy feature sample set.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010895695.XA CN111914490B (en) | 2020-08-31 | 2020-08-31 | Pump station unit state evaluation method based on depth convolution random forest self-coding |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010895695.XA CN111914490B (en) | 2020-08-31 | 2020-08-31 | Pump station unit state evaluation method based on depth convolution random forest self-coding |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111914490A true CN111914490A (en) | 2020-11-10 |
CN111914490B CN111914490B (en) | 2023-10-13 |
Family
ID=73267657
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010895695.XA Active CN111914490B (en) | 2020-08-31 | 2020-08-31 | Pump station unit state evaluation method based on depth convolution random forest self-coding |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111914490B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112539183A (en) * | 2020-11-25 | 2021-03-23 | 南京河海南自水电自动化有限公司 | Fault diagnosis method and system for water pump shut-off device failure |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7013223B1 (en) * | 2002-09-25 | 2006-03-14 | The Board Of Trustees Of The University Of Illinois | Method and apparatus for analyzing performance of a hydraulic pump |
CN108241873A (en) * | 2018-01-10 | 2018-07-03 | 中国水利水电科学研究院 | A kind of intelligent failure diagnosis method towards pumping plant main equipment |
CN108869145A (en) * | 2018-04-26 | 2018-11-23 | 中国水利水电科学研究院 | Pumping plant unit diagnostic method based on compound characteristics index and depth limit learning machine |
CN109063782A (en) * | 2018-08-16 | 2018-12-21 | 中国水利水电科学研究院 | A kind of adaptive pumping plant intelligent fault diagnosis method |
CN109102032A (en) * | 2018-09-03 | 2018-12-28 | 中国水利水电科学研究院 | A kind of pumping plant unit diagnostic method based on depth forest and oneself coding |
WO2019153388A1 (en) * | 2018-02-12 | 2019-08-15 | 大连理工大学 | Power spectral entropy random forest-based aeroengine rolling bearing fault diagnosis method |
-
2020
- 2020-08-31 CN CN202010895695.XA patent/CN111914490B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7013223B1 (en) * | 2002-09-25 | 2006-03-14 | The Board Of Trustees Of The University Of Illinois | Method and apparatus for analyzing performance of a hydraulic pump |
CN108241873A (en) * | 2018-01-10 | 2018-07-03 | 中国水利水电科学研究院 | A kind of intelligent failure diagnosis method towards pumping plant main equipment |
WO2019153388A1 (en) * | 2018-02-12 | 2019-08-15 | 大连理工大学 | Power spectral entropy random forest-based aeroengine rolling bearing fault diagnosis method |
CN108869145A (en) * | 2018-04-26 | 2018-11-23 | 中国水利水电科学研究院 | Pumping plant unit diagnostic method based on compound characteristics index and depth limit learning machine |
CN109063782A (en) * | 2018-08-16 | 2018-12-21 | 中国水利水电科学研究院 | A kind of adaptive pumping plant intelligent fault diagnosis method |
CN109102032A (en) * | 2018-09-03 | 2018-12-28 | 中国水利水电科学研究院 | A kind of pumping plant unit diagnostic method based on depth forest and oneself coding |
Non-Patent Citations (6)
Title |
---|
TAOTAO ZHOU等: "Degradation Assessment and Remaining Useful Life Prognostic of Centrifugal Pump using Multi-Sensor Process Monitoring", 《2020 ASIA-PACIFIC INTERNATIONAL SYMPOSIUM ON ADVANCED RELIABILITY AND MAINTENANCE MODELING (APARM)》, pages 1 - 4 * |
侯锋等: "基于AdaBoost算法的水质监测数据清洗方", 《水电站机电技术》, vol. 46, no. 5, pages 109 - 111 * |
殷振兴: "基于稀疏自编码支持向量机泵站故障检测", 《水科学与工程技术》, no. 06, pages 52 - 55 * |
洪健: "基于自编码支持向量机泵站故障识别", 《中国水利学会2018学术年会论文集第五分册》, no. 6 * |
薛萍: "调水工程水位数据清洗及预测模型研究", 《中国优秀硕士学位论文全文数据库》, pages 037 - 99 * |
郑和震等: "梯级泵站输水系统预测控制方法研究", 《人民黄河》, vol. 40, no. 2, pages 140 - 143 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112539183A (en) * | 2020-11-25 | 2021-03-23 | 南京河海南自水电自动化有限公司 | Fault diagnosis method and system for water pump shut-off device failure |
CN112539183B (en) * | 2020-11-25 | 2022-06-17 | 南京河海南自水电自动化有限公司 | Fault diagnosis method and system for water pump shut-off device failure |
Also Published As
Publication number | Publication date |
---|---|
CN111914490B (en) | 2023-10-13 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107480705B (en) | Tap water pipe water leakage detection method | |
CN111401584B (en) | Automatic equipment fault diagnosis method and system | |
CN111337237B (en) | Equipment fault diagnosis method and system | |
CN113570138B (en) | Method and device for predicting residual service life of equipment of time convolution network | |
CN109670714B (en) | Ship gas turbine comprehensive state evaluation method based on membership degree analysis | |
CN112465124A (en) | Twin depth space-time neural network model acquisition/fault diagnosis method and device | |
CN113757093B (en) | Flash steam compressor unit fault diagnosis method | |
CN116450399A (en) | Fault diagnosis and root cause positioning method for micro service system | |
CN116720112A (en) | Machine pump fault diagnosis method and system based on machine learning | |
CN111612149A (en) | Main network line state detection method, system and medium based on decision tree | |
CN116306806A (en) | Fault diagnosis model determining method and device and nonvolatile storage medium | |
CN111914490A (en) | Pump station unit state evaluation method based on deep convolution random forest self-coding | |
CN103852525A (en) | Acoustic emission signal identification method based on AR-HMM | |
CN112926686B (en) | BRB and LSTM model-based power consumption anomaly detection method and device for big power data | |
CN117314020A (en) | Wetland carbon sink data monitoring system of plankton | |
CN112202630A (en) | Network quality abnormity detection method and device based on unsupervised model | |
CN111880957A (en) | Program error positioning method based on random forest model | |
CN116739829A (en) | Big data-based power data analysis method, system and medium | |
CN114354666B (en) | Soil heavy metal spectral feature extraction and optimization method based on wavelength frequency selection | |
CN114783538A (en) | Coal ash content prediction method and device | |
CN111124816B (en) | HDP algorithm-based server log analysis method and system | |
CN111695634A (en) | Data abnormal mutation point detection algorithm based on limited accompanying censoring mechanism | |
CN112990524A (en) | Residual error correction-based residual life prediction method for rolling bearing | |
CN111044176A (en) | Method for monitoring temperature abnormity of generator | |
CN116738759B (en) | Method and device for designing and operating equipment, computer equipment and readable storage medium |
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 |