CN114611564A - Steam pipeline drain valve fault detection method based on convolutional neural network and temperature - Google Patents

Steam pipeline drain valve fault detection method based on convolutional neural network and temperature Download PDF

Info

Publication number
CN114611564A
CN114611564A CN202011335224.XA CN202011335224A CN114611564A CN 114611564 A CN114611564 A CN 114611564A CN 202011335224 A CN202011335224 A CN 202011335224A CN 114611564 A CN114611564 A CN 114611564A
Authority
CN
China
Prior art keywords
neural network
data
drain valve
temperature data
temperature
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.)
Pending
Application number
CN202011335224.XA
Other languages
Chinese (zh)
Inventor
陈振兴
张新生
刘根
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Sinomach Internet Research Institute Henan Co ltd
Original Assignee
Sinomach Internet Research Institute Henan Co ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Sinomach Internet Research Institute Henan Co ltd filed Critical Sinomach Internet Research Institute Henan Co ltd
Priority to CN202011335224.XA priority Critical patent/CN114611564A/en
Publication of CN114611564A publication Critical patent/CN114611564A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00Pattern recognition
    • G06F18/20Analysing
    • G06F18/24Classification techniques
    • G06F18/241Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • FMECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F16ENGINEERING ELEMENTS AND UNITS; GENERAL MEASURES FOR PRODUCING AND MAINTAINING EFFECTIVE FUNCTIONING OF MACHINES OR INSTALLATIONS; THERMAL INSULATION IN GENERAL
    • F16KVALVES; TAPS; COCKS; ACTUATING-FLOATS; DEVICES FOR VENTING OR AERATING
    • F16K37/00Special means in or on valves or other cut-off apparatus for indicating or recording operation thereof, or for enabling an alarm to be given
    • F16K37/0075For recording or indicating the functioning of a valve in combination with test equipment
    • F16K37/0083For recording or indicating the functioning of a valve in combination with test equipment by measuring valve parameters
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00Pattern recognition
    • G06F18/20Analysing
    • G06F18/21Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/04Architecture, e.g. interconnection topology
    • G06N3/045Combinations of networks

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Computational Linguistics (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Mechanical Engineering (AREA)
  • Examining Or Testing Airtightness (AREA)

Abstract

The invention discloses a steam pipeline trap fault detection method based on a convolutional neural network and temperature data, which comprises the steps of firstly, respectively arranging four temperature sensor probes on pipelines at an inlet end and an outlet end of a trap, and acquiring temperature data of a trap in a normal working state, a blocking state and an air leakage state, wherein 10000 groups of data in each state are acquired; 80% of the data is used for training the neural network, and 20% of the data is used for evaluating the error degree of the neural network; and finally, inputting the temperature data of the pipeline into the trained neural network, and automatically mapping the temperature data into the running state of the drain valve by the neural network through internal processing. The invention uses the neural network algorithm, so that the detection system can judge the fault condition of the trap only according to the temperature. Compared with other methods for detecting the drain valve, the method firstly introduces the neural network algorithm to analyze the temperature data of the drain valve, and has the advantages of simple used sensor and detection device, no need of manual participation, high detection accuracy and the like.

Description

Steam pipeline drain valve fault detection method based on convolutional neural network and temperature
Technical Field
The invention relates to the field of pipeline valves, in particular to a fault detection method for a steam pipeline drain valve.
Background
High-temperature steam output by the steam boiler generates certain condensed water in a pipeline, the purity of the steam is reduced due to the existence of the condensed water, the steam containing water can affect the use, and the quality of products needing to be heated and baked by the steam is affected. In order to keep the dryness of the steam and reduce the moisture, a steam trap for blocking and draining the water is arranged on the steam pipeline at intervals to drain the condensed water in the pipeline. Because the aquatic contains impurity, through certain time, impurity is at pipeline wall and the inside deposit of trap to and because the corrosion of high temperature steam environment to pipeline wall and trap, the trap is malfunctioning very easily, makes moisture can not in time discharge and reduces the steam quality, or makes steam leak, causes the loss.
The fault detection method of the drain valve used in the overhaul comprises the following steps: 1. detecting the internal structure of the drain valve by using ultrasonic waves; 2. a small transparent window is arranged at the rear end of the drain valve, and the drainage condition of the drain valve can be seen through the small window; 3. the method comprises the steps of comprehensively acquiring and analyzing signals such as temperature, sound and the like to judge the working condition of the drain valve. The detection methods all need to manually carry detection equipment to check on site, the technology used by the detection equipment is complex, various sensor probes need to be installed on the drain valve, the fault detection accuracy rate needs to be judged according to the experience of people, and the accuracy rate is not high.
With the development and progress of machine learning algorithms such as neural networks, the analysis understanding and learning capability of the algorithms on data are enhanced. The advanced data analysis algorithm is introduced, decision judgment can be made instead of manual work, the data learning and analysis understanding capacity is enhanced, the requirement on data is lowered, and only a single type of signal data is needed to generate an accurate decision through machine learning. In the fault diagnosis of the drain valve, a neural network algorithm is introduced to replace the experience of people to judge the fault condition of the drain valve, the requirement on signal data is reduced, the accuracy of fault diagnosis is improved, and related research is lacked in the aspect.
Disclosure of Invention
The invention mainly aims to provide steam pipeline steam trap fault detection based on a convolutional neural network and temperature data. The invention mainly aims to provide a steam pipeline drain valve fault detection method based on a convolutional neural network and temperature data, which realizes fault diagnosis of a drain valve by analyzing the temperature data of front and rear pipelines of the drain valve, so as to overcome the problems that the existing drain valve detection method needs to manually carry detection equipment to check on site, the detection equipment has complex technology, various sensor probes need to be installed on the drain valve, the fault detection accuracy rate needs to be judged according to the experience of people, the accuracy rate is low, and the like, and the method is very practical.
The invention is realized by adopting the following technical scheme and technical measures.
The invention provides a steam pipeline drain valve fault detection method based on a convolutional neural network and temperature data,
firstly, a plurality of temperature sensors are respectively and uniformly distributed and installed at the front end and the rear end of a steam pipeline drain valve, and the temperature sensors are used for acquiring temperature data of the drain valve in different running states, including a blocking state of the drain valve, an air leakage state of the drain valve, a normal working state of the drain valve and the like. 80% of these data were used to train the neural network and 20% were used to test the error of the neural network.
And secondly, building a convolutional neural network model, using the temperature data of the training sample data as input data of the neural network, using the drain valve operation state code of the training sample data as output data of the neural network, and training the neural network. The test data is used to test the error level of the neural network.
And finally, taking the temperature data of the front end and the rear end of the steam pipeline steam trap as the input of a neural network, and mapping the temperature data into a working state code of the steam trap by the trained neural network.
The pipeline is a steam pipeline output by the steam boiler, the temperature data of the pipeline is the temperature data of a plurality of points at the front end and the rear end of the steam trap measured by temperature sensors, and the data are used as the input of a neural network. The normal and fault states of the trap are represented by working state codes, 1000, 0100, 0010, 0001 represent the normal working, blocking, leakage and semi-blocking states of the trap respectively, and the working state codes are used as the output of the neural network.
The convolutional neural network input is
Figure 100002_DEST_PATH_IMAGE001
The output of the neural network is Y =
Figure 100002_DEST_PATH_IMAGE002
The convolution layer is output as
Figure 100002_DEST_PATH_IMAGE003
Where T represents the transpose of the matrix,
Figure 100002_DEST_PATH_IMAGE004
to is that
Figure 100002_DEST_PATH_IMAGE005
Representing the temperature of 8 sites as inputs to the neural network,
Figure 100002_DEST_PATH_IMAGE006
to
Figure 100002_DEST_PATH_IMAGE007
Representing the output value of the neural network,
Figure 100002_DEST_PATH_IMAGE008
the value is 0 or 1, and the like,
Figure 100002_DEST_PATH_IMAGE009
representing a neural network
Figure 100002_DEST_PATH_IMAGE010
The output of the layer(s) is,
Figure 100002_DEST_PATH_IMAGE011
represents the Relu activation function, th
Figure 51357DEST_PATH_IMAGE010
The number of the layer convolution kernels is
Figure 100002_DEST_PATH_IMAGE012
Wherein
Figure 499656DEST_PATH_IMAGE010
Is as follows
Figure 279393DEST_PATH_IMAGE010
The number of the layers is one of the volume layers,
Figure 100002_DEST_PATH_IMAGE013
is a convolution kernel that is a function of the convolution kernel,
Figure 100002_DEST_PATH_IMAGE014
is the offset value.
The convolutional neural network pooling layer adopts a maximum pooling method.
The transfer function of the convolutional neural network full-connection layer is a sigmoid function.
And the convolutional neural network obtains a fault type classification result by adopting a softmax classifier.
Compared with the prior art, the invention has at least the following advantages and beneficial effects.
Aiming at steam pipeline drain valve equipment, the fault condition of the drain valve can be judged only through temperature data by the steam pipeline drain valve fault detection method based on the convolutional neural network and the temperature data, and the method has the advantages of replacing human experience to judge the fault condition of the drain valve, reducing the requirements on signal data, reducing the complexity of detection equipment, improving the accuracy of fault diagnosis and the like.
Drawings
FIG. 1 is a diagram of a steam pipeline trap fault detection model based on a convolutional neural network and temperature data.
Fig. 2 is a schematic diagram of a convolutional neural network structure.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be obtained by a person skilled in the art without inventive effort based on the embodiments of the present invention, are within the scope of the present invention.
The convolutional neural network can automatically adjust the weight and bias of the network through training, forward propagation of training data and backward propagation of errors, so that the error between input data and expected output is gradually reduced, and the mapping relation from the input to the output is established.
According to the method, a convolutional neural network algorithm is introduced into the steam pipeline drain valve fault diagnosis process, a plurality of temperature data collected at the front end and the rear end of the steam pipeline drain valve are used as the input of the neural network by utilizing the outstanding advantages of the convolutional neural network, the common fault type of the drain valve is used as the output, and the mapping relation between the input and the output is established. The steam trap fault diagnosis method based on the convolutional neural network and the temperature data, which is provided by the inventor, firstly trains the neural network by using the existing temperature data and the corresponding steam trap fault data, and then uses the trained neural network for analyzing the acquired temperature data in a use stage.
The convolutional neural network is mainly used for processing image data, and the invention uses temperature data of the front end and the rear end of the steam pipeline steam trap and corresponding fault codes. The convolutional neural network of the present invention is a convolutional neural network for processing one-dimensional sequence data, as shown in fig. 2, the number of input layers is 50, the number of output layers is 30, and the network structure sequentially includes 3 convolutional layers, 1 pooling layer, and 200 full-link layers.
In the embodiment of the invention, firstly, a plurality of temperature sensors are respectively and uniformly distributed and installed at the front end and the rear end of the steam pipeline drain valve, and the temperature sensors are used for acquiring temperature data of the drain valve in different running states, including a blocking state of the drain valve, an air leakage state of the drain valve, a normal working state of the drain valve and the like. 80% of these data were used to train the neural network and 20% were used to test the error of the neural network.
And secondly, building a convolutional neural network model, using the temperature data of the training sample data as input data of the neural network, using the drain valve operation state code of the training sample data as output data of the neural network, and training the neural network. The test data is used to test the error level of the neural network.
And finally, taking the temperature data of the front end and the rear end of the steam pipeline steam trap as the input of a neural network, and mapping the temperature data into a working state code of the steam trap by the trained neural network.
The temperature data of the pipeline is the temperature data of a plurality of point positions at the front end and the rear end of the trap measured by the temperature sensor, and the data are used as the input of the neural network. The normal state and the fault state of the drain valve are represented by working state codes, 1000, 0100, 0010 and 0001 respectively represent the normal working state, the blocking state, the gas leakage state and the semi-blocking state of the drain valve, and the working state codes are used as the output of a neural network.
The convolutional neural network input is
Figure DEST_PATH_IMAGE015
The output of the neural network is Y =
Figure 736920DEST_PATH_IMAGE002
The convolution layer is output as
Figure 82450DEST_PATH_IMAGE003
Where T represents the transpose of the matrix,
Figure 904913DEST_PATH_IMAGE004
to is that
Figure 703105DEST_PATH_IMAGE005
Representing the temperature of 8 sites as inputs to the neural network,
Figure 433163DEST_PATH_IMAGE006
to
Figure 102042DEST_PATH_IMAGE007
Representing the output value of the neural network,
Figure 128029DEST_PATH_IMAGE008
the value is 0 or 1, and the like,
Figure 882358DEST_PATH_IMAGE009
representing a neural network
Figure 150529DEST_PATH_IMAGE010
The output of the layer(s) is,
Figure 673914DEST_PATH_IMAGE011
represents the Relu activation function, th
Figure 634917DEST_PATH_IMAGE010
The number of the layer convolution kernels is
Figure 876542DEST_PATH_IMAGE012
Wherein
Figure 682824DEST_PATH_IMAGE010
Is as follows
Figure 326295DEST_PATH_IMAGE010
The number of the layers is one of the volume layers,
Figure 192620DEST_PATH_IMAGE013
is a convolution kernel that is a function of the convolution kernel,
Figure 452700DEST_PATH_IMAGE014
is the offset value.
The convolutional neural network pooling layer adopts a maximum pooling method.
The transfer function of the convolutional neural network fully-connected layer is a sigmoid function.
And the convolutional neural network obtains a fault type classification result by adopting a softmax classifier.

Claims (6)

1. The method for detecting the fault of the steam pipeline drain valve based on the convolutional neural network and the temperature data is characterized in that firstly, a plurality of temperature sensors are uniformly distributed and installed at the front end and the rear end of the steam pipeline drain valve respectively, and the temperature sensors are used for acquiring the temperature data of the drain valve in different running states, including the blocking state of the drain valve, the air leakage state of the drain valve, the normal working state of the drain valve and the like; 80% of these data were used to train the neural network and 20% were used to test the error of the neural network; secondly, building a convolution neural network model, using temperature data of training sample data as input data of the neural network, using drain valve operation state codes of the training sample data as output data of the neural network, training the neural network, and using test data to test the error degree of the neural network; and finally, taking the temperature data of the front end and the rear end of the steam pipeline steam trap as the input of a neural network, and mapping the temperature data into a working state code of the steam trap by the trained neural network.
2. The method of claim 1, wherein the steam pipe is a steam pipe of a steam boiler, the temperature data of the pipe is the temperature data of a plurality of points at the front end and the rear end of the steam trap measured by the temperature sensors, the data are used as the input of the neural network, the normal and fault states of the steam trap are represented by operation state codes, 1000, 0100, 0010, 0001 respectively represent the normal operation, blockage, air leakage, and semi-blockage states of the steam trap, and the operation state codes are used as the output of the neural network.
3. The method of claim 1, wherein the convolutional neural network is used as an input to the convolutional neural network for steam pipe trap fault detection based on convolutional neural network and temperature data
Figure DEST_PATH_IMAGE001
The output of the neural network is Y =
Figure DEST_PATH_IMAGE002
The convolution layer is output as
Figure DEST_PATH_IMAGE003
Where T represents the transpose of the matrix,
Figure DEST_PATH_IMAGE004
to
Figure DEST_PATH_IMAGE005
Representing the temperature of 8 sites as inputs to the neural network,
Figure DEST_PATH_IMAGE006
to
Figure DEST_PATH_IMAGE007
Represents the output value of the neural network and,
Figure DEST_PATH_IMAGE008
the value is 0 or 1, and the like,
Figure DEST_PATH_IMAGE009
representing a neural network
Figure DEST_PATH_IMAGE010
The output of the layer(s) is,
Figure DEST_PATH_IMAGE011
represents the Relu activation function, th
Figure 682210DEST_PATH_IMAGE010
The number of the layer convolution kernels is
Figure DEST_PATH_IMAGE012
Wherein
Figure 221600DEST_PATH_IMAGE010
Is as follows
Figure 993247DEST_PATH_IMAGE010
The number of the layers is one of the volume layers,
Figure DEST_PATH_IMAGE013
is a convolution kernel that is a function of the convolution kernel,
Figure DEST_PATH_IMAGE014
is the offset value.
4. The method of claim 1, wherein the convolutional neural network pooling layer employs a max pooling method.
5. The method of claim 1, wherein the transfer function of the convolutional neural network fully-connected layer is a sigmoid function.
6. The convolutional neural network and temperature data based steam pipe trap fault detection method of claim 1, wherein a softmax classifier is used to derive the fault type classification result.
CN202011335224.XA 2020-11-24 2020-11-24 Steam pipeline drain valve fault detection method based on convolutional neural network and temperature Pending CN114611564A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011335224.XA CN114611564A (en) 2020-11-24 2020-11-24 Steam pipeline drain valve fault detection method based on convolutional neural network and temperature

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011335224.XA CN114611564A (en) 2020-11-24 2020-11-24 Steam pipeline drain valve fault detection method based on convolutional neural network and temperature

Publications (1)

Publication Number Publication Date
CN114611564A true CN114611564A (en) 2022-06-10

Family

ID=81856915

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011335224.XA Pending CN114611564A (en) 2020-11-24 2020-11-24 Steam pipeline drain valve fault detection method based on convolutional neural network and temperature

Country Status (1)

Country Link
CN (1) CN114611564A (en)

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1487247A (en) * 2003-07-31 2004-04-07 上海交通大学 Cooling air conditioner unit fault simulating and diagnosing system
DE102008008096A1 (en) * 2008-02-08 2009-08-13 Siemens Aktiengesellschaft Distributed system for e.g. evaluation of destruction of infrastructure, in power stations, has recording and alarming unit announcing danger situations, where objects, danger situations and extent of announcing are determined
CN104712542A (en) * 2015-01-12 2015-06-17 北京博华信智科技股份有限公司 Reciprocating compressor sensitive characteristic extracting and fault diagnosis method based on internet of things
CN111027260A (en) * 2019-12-24 2020-04-17 大连圣力来监测技术有限公司 Reciprocating compressor fault diagnosis method based on neural network
CN111191693A (en) * 2019-12-18 2020-05-22 广西电网有限责任公司电力科学研究院 Method for identifying thermal fault state of high-voltage switch cabinet based on convolutional neural network
CN111562110A (en) * 2020-05-25 2020-08-21 南京理工大学 Fault diagnosis model based on convolutional neural network and cross-component fault diagnosis method

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1487247A (en) * 2003-07-31 2004-04-07 上海交通大学 Cooling air conditioner unit fault simulating and diagnosing system
DE102008008096A1 (en) * 2008-02-08 2009-08-13 Siemens Aktiengesellschaft Distributed system for e.g. evaluation of destruction of infrastructure, in power stations, has recording and alarming unit announcing danger situations, where objects, danger situations and extent of announcing are determined
CN104712542A (en) * 2015-01-12 2015-06-17 北京博华信智科技股份有限公司 Reciprocating compressor sensitive characteristic extracting and fault diagnosis method based on internet of things
CN111191693A (en) * 2019-12-18 2020-05-22 广西电网有限责任公司电力科学研究院 Method for identifying thermal fault state of high-voltage switch cabinet based on convolutional neural network
CN111027260A (en) * 2019-12-24 2020-04-17 大连圣力来监测技术有限公司 Reciprocating compressor fault diagnosis method based on neural network
CN111562110A (en) * 2020-05-25 2020-08-21 南京理工大学 Fault diagnosis model based on convolutional neural network and cross-component fault diagnosis method

Similar Documents

Publication Publication Date Title
CN111664365B (en) Oil and gas pipeline leakage detection method based on improved VMD and 1DCNN
Wang et al. Prediction model of natural gas pipeline crack evolution based on optimized DCNN-LSTM
CN109828089B (en) DBN-BP-based water quality parameter nitrous acid nitrogen online prediction method
CN109838696A (en) Pipeline fault diagnostic method based on convolutional neural networks
CN108596260A (en) Water pipe network leakage positioning method and device
CN112665801B (en) Gas pipeline valve internal leakage identification device and method based on convolutional neural network
CN105221933A (en) A kind of pipeline network leak detecting method in conjunction with resistance identification
CN101246467A (en) Leakage locating method combining self-adapting threshold value leak detection and multi-dimension fast delay time search
CN111853555A (en) Water supply pipe network blind leakage identification method based on dynamic process
CN111695465B (en) Pipe network fault diagnosis and positioning method and system based on pressure wave mode identification
CN106197886A (en) Hydraulic cylinder interior leakage leakage fault diagnosis appraisal procedure
CN112446091A (en) Artificial neural network-based pulsating pressure prediction method
CN111750283A (en) Deep learning-based gas pipeline leakage identification method in strong background noise environment
CN103742794A (en) Simulating device and simulating method for pipeline leakage acoustic emission signals
CN112377817A (en) Municipal pipe network pipe explosion monitoring system and method
CN115824519A (en) Valve leakage fault comprehensive diagnosis method based on multi-sensor information fusion
CN111402249B (en) Image evolution analysis method based on deep learning
CN114858207A (en) Soft measurement-based gridding source tracing investigation method for drain outlet of river channel
CN111365624A (en) Intelligent terminal and method for detecting leakage of brine transportation pipeline
CN117216703A (en) Water delivery pipe network operation data anomaly detection and correction method
CN117073933B (en) Pipeline pressure monitoring method and system for gas pipe network
CN114611564A (en) Steam pipeline drain valve fault detection method based on convolutional neural network and temperature
CN104730027B (en) The method that puccinia striiformis uredospores germination rate is determined using near-infrared spectrum technique
CN116975763A (en) Water supply network abnormality diagnosis method based on bispectrum and convolutional neural network
CN115980286B (en) Method for detecting wastewater at different stages of sewage treatment plant by utilizing electronic nose

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