CN109800627A - The method for detecting abnormality and device of petroleum pipeline signal, equipment and readable medium - Google Patents
The method for detecting abnormality and device of petroleum pipeline signal, equipment and readable medium Download PDFInfo
- Publication number
- CN109800627A CN109800627A CN201811466072.XA CN201811466072A CN109800627A CN 109800627 A CN109800627 A CN 109800627A CN 201811466072 A CN201811466072 A CN 201811466072A CN 109800627 A CN109800627 A CN 109800627A
- Authority
- CN
- China
- Prior art keywords
- petroleum pipeline
- feature
- signal
- sample data
- forecast sample
- 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
Landscapes
- Testing And Monitoring For Control Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention provides the method for detecting abnormality and device, equipment and readable medium of a kind of petroleum pipeline signal.Its method includes: the petroleum pipeline signal acquired in real time according to sensor, generates forecast sample data；Feature extraction processing is carried out to the forecast sample data, obtains forecast sample feature；The forecast sample feature is input to the petroleum pipeline abnormal signal detection model trained and obtained, obtains the output result of the petroleum pipeline abnormal signal detection model；According to the output of the petroleum pipeline abnormal signal detection model as a result, determining whether petroleum pipeline signal is abnormal.Technical solution of the present invention can effectively improve the accuracy of abnormality detection, and not influenced by the complicated shape of petroleum pipeline, can effectively guarantee detection efficiency.
Description
Technical field
The present invention relates to the method for detecting abnormality of computer application technology more particularly to a kind of petroleum pipeline signal and
Device, equipment and readable medium.
Background technique
The abnormality detection of petroleum pipeline signal directly affects the safety of petroleum transportation, is the signal detection of oil field
In a very important job.
Whether the abnormal signal detection technique of existing oil field, usually sentence petroleum pipeline signal manually to lay down a regulation
It is abnormal.Specifically, signal data is detected from petroleum pipeline by sensor, by the rule artificially formulated come to these signals
Data are classified, and judge whether to belong to abnormal conditions.Such as common artificial rule can be the statistics side of some classics
Method, such as the methods of the standard deviation judgement of maximum value, the judgement of minimum value or normal distribution.
But the statistical method of the prior art, due to the property of itself have the defects that it is natural, can only be from data point
Do abnormal judgement on a priori assumption of cloth, be unable to adaptation to local conditions, for example, petroleum pipeline field it is many abnormal signal and value it is big
It is small unrelated, the abnormal signal of the complicated shape of some specific areas can not be coped with, so that existing petroleum pipeline signal
The accuracy of abnormality detection is poor.
Summary of the invention
The present invention provides the method for detecting abnormality and device, equipment and readable medium of a kind of petroleum pipeline signal, for mentioning
The accuracy of the abnormality detection of high petroleum pipeline signal.
The present invention provides a kind of method for detecting abnormality of petroleum pipeline signal, comprising:
The petroleum pipeline signal acquired in real time according to sensor generates forecast sample data；
Feature extraction processing is carried out to the forecast sample data, obtains forecast sample feature；
The forecast sample feature is input to the petroleum pipeline abnormal signal detection model trained and obtained, described in acquisition
The output result of petroleum pipeline abnormal signal detection model；
According to the output of the petroleum pipeline abnormal signal detection model as a result, determining whether petroleum pipeline signal is abnormal.
The present invention also provides a kind of training methods of the abnormality detection model of petroleum pipeline signal, comprising:
According to the historic state information of the petroleum pipeline signal of sensor historic acquisition and corresponding petroleum pipeline, instruction is generated
Practice sample data set；
Feature extraction processing is carried out to the training sample data collection, obtains training sample feature set；
Based on the training sample feature set, the historic state information of acquisition and machine learning algorithm, training petroleum
Pipeline abnormal signal detection model.
The present invention also provides a kind of abnormal detectors of petroleum pipeline signal, comprising:
Generation module, the petroleum pipeline signal for being acquired in real time according to sensor generate forecast sample data；
Abstraction module obtains forecast sample feature for carrying out feature extraction processing to the forecast sample data；
Detection module has trained obtained petroleum pipeline abnormal signal to detect for the forecast sample feature to be input to
Model obtains the output result of the petroleum pipeline abnormal signal detection model；
Determining module, for the output according to the petroleum pipeline abnormal signal detection model as a result, determining petroleum pipeline
Whether signal is abnormal.
The present invention also provides a kind of training devices of the abnormality detection model of petroleum pipeline signal, comprising:
Generation module, the history shape of petroleum pipeline signal and corresponding petroleum pipeline for being acquired according to sensor historic
State information generates training sample data collection；
Abstraction module obtains training sample feature set for carrying out feature extraction processing to the training sample data collection；
Training module, for the historic state information and machine learning based on the training sample feature set, acquisition
Algorithm, training petroleum pipeline abnormal signal detection model.
The present invention also provides a kind of calculating equipment, comprising:
Processor；And
Memory is stored thereon with executable code, when the executable code is executed by the processor, makes described
Processor executes method described in any one as above.
The present invention also provides a kind of non-transitory machinable mediums, are stored thereon with executable code, when described
When executable code is executed by the processor of electronic equipment, the processor is made to execute as above described in any item methods.
The method for detecting abnormality and device, equipment and readable medium of petroleum pipeline signal of the invention, through the above scheme,
Petroleum pipeline abnormal signal detection model can be trained, and is realized using trained petroleum pipeline abnormal signal detection model
The petroleum pipeline signal acquired in real time is carried out abnormality detection, compared with prior art, since petroleum pipeline abnormal signal detects
Model is realized using neural network model, can effectively improve the accuracy of abnormality detection, and do not answered by petroleum pipeline
Miscellaneous shape influences, and can effectively guarantee detection efficiency.
Detailed description of the invention
Disclosure illustrative embodiments are described in more detail in conjunction with the accompanying drawings, the disclosure above-mentioned and its
Its purpose, feature and advantage will be apparent, wherein in disclosure illustrative embodiments, identical reference label
Typically represent same parts.
Fig. 1 is the flow chart of the method for detecting abnormality embodiment one of petroleum pipeline signal of the invention.
Fig. 2 is the structural schematic diagram of LSTM provided by the invention.
Fig. 3 is the exemplary diagram of the abnormality detection of ECG data of the invention.
Fig. 4 is the flow chart of the method for detecting abnormality embodiment two of petroleum pipeline signal of the invention.
Fig. 5 is the flow chart of the training method embodiment of the abnormality detection model of petroleum pipeline signal of the invention.
Fig. 6 is the structure chart of the abnormal detector embodiment one of petroleum pipeline signal of the invention.
Fig. 7 is the structure chart of the abnormal detector embodiment two of petroleum pipeline signal of the invention.
Fig. 8 is the structure chart of the training device embodiment of the abnormality detection model of petroleum pipeline signal of the invention.
Fig. 9 shows the structural schematic diagram that an embodiment according to the present invention can be used for realizing the calculating equipment of the above method.
Specific embodiment
The preferred embodiment of the disclosure is more fully described below with reference to accompanying drawings.Although showing the disclosure in attached drawing
Preferred embodiment, however, it is to be appreciated that may be realized in various forms the disclosure without the embodiment party that should be illustrated here
Formula is limited.On the contrary, these embodiments are provided so that this disclosure will be more thorough and complete, and can be by the disclosure
Range is completely communicated to those skilled in the art.
Fig. 1 is the flow chart of the method for detecting abnormality embodiment one of petroleum pipeline signal of the invention.As shown in Figure 1, this
The method for detecting abnormality of the petroleum pipeline signal of embodiment, can specifically include following steps:
S100, the petroleum pipeline signal acquired in real time according to sensor generate forecast sample data；
The executing subject of the method for detecting abnormality of the petroleum pipeline signal of the present embodiment is the abnormal inspection of petroleum pipeline signal
Device is surveyed, the abnormal detector of the petroleum pipeline signal, which uses, has trained obtained petroleum pipeline abnormal signal to detect mould under line
Type, to be carried out abnormality detection to the collected petroleum pipeline signal of sensor.
Specifically, in practical applications, it in order to be carried out abnormality detection to petroleum pipeline signal, can be set in petroleum pipeline
It sets sensor and acquires petroleum pipeline signal in real time, for example, the sensor of 500 sample rates can be set, per second available 500
A numerical value, as the petroleum pipeline signal of this second.During actual samples, it can select according to the preset sampling period
The petroleum pipeline signal in the sampling period is taken, as the petroleum pipeline signal acquired in real time.The sampling period of the present embodiment can be with
5 seconds, 3 seconds or other predetermined time periods are taken according to actual needs, and no longer citing repeats one by one herein.Sensor acquisition is worked as
Signal in preceding predetermined time period is the petroleum pipeline signal acquired in real time.For the letter in every predetermined time period
Number, all it is used as a real-time forecast sample data.For example, 500 sample rate is corresponding pre- when predetermined time period is 5 seconds
It include 2500 sampled datas in test sample notebook data.
S101, feature extraction processing is carried out to forecast sample data, obtains forecast sample feature；
It include all sampled datas in predetermined time period in the forecast sample data of the present embodiment, data are more rich
Richness can directly extract some features, as forecast sample feature from all sampled datas.It can also be from all sampled datas
It is middle to extract a part of feature, secondary operation processing then is carried out to this Partial Feature of extraction, the feature that obtains that treated, as
Forecast sample feature.Or the feature of above two type can also be obtained simultaneously, it is integrated into forecast sample feature together.
For example, step S101 " carries out feature extraction processing to forecast sample data, obtains forecast sample in the present embodiment
Feature " can specifically include following steps:
(1) forecast sample data are carried out with the processing of following at least one mode:
(a1) numerical value category feature is extracted from forecast sample data；
For example, numerical value category feature is extracted in the present embodiment from forecast sample data, it specifically can be from forecast sample data
At least one of middle acquisition maximum value, minimum value, mean variance and quantile, as numerical value category feature.In particular it is required that
It selects which feature as numerical value category feature, can be selected according to concrete application scene.
(b1) frequency domain category feature is extracted from forecast sample data；
For example, in specific implementation, first Fast Fourier Transform (FFT) processing can be carried out to forecast sample data；Then it takes fast
Multiple numerical value of fast Fourier transformation treated real part；In the quantity and forecast sample data of the numerical value of the real part
Including petroleum pipeline signal strength indication quantity it is equal；Finally extract maximum value, the minimum in multiple numerical value of real part
At least one of value, mean variance and quantile, as frequency domain category feature.Which similarly, need to select feature as frequency
Domain category feature can be selected according to concrete application scene.
(c1) itself variation tendency feature is extracted from forecast sample data；
For example, the extracting mode of Self-variation trend feature may include the following two kinds mode in the present embodiment, actually answer
In, one of which can choose according to specific requirements to realize.First way are as follows: divide forecast sample data from centre
It opens, obtains anter segmentation and the segmentation of rear piece；The numerical value category feature of anter segmentation and the segmentation of rear piece is obtained respectively；The numerical value category feature
Acquisition modes can refer to aforesaid way (a1), details are not described herein.Then, the numerical value category feature of anter segmentation is obtained with after
The difference of the numerical value category feature of piece segmentation, as Self-variation trend feature.The second way: forecast sample data are fitted to
Binary Gaussian Profile；The fit procedure of existing binary Gaussian Profile can be referred in detail, and details are not described herein.Finally, obtaining
The difference of the mean value of binary Gaussian Profile and the difference of variance, as Self-variation trend feature.
(d1) forecast sample data are compared with the phase of history petroleum pipeline signal of arest neighbors, obtain historical data
Comparative feature；And
For example, the numerical value category feature of forecast sample data and a Duan Li of arest neighbors can be obtained respectively when specific implementation
The numerical value category feature of history petroleum pipeline signal；Then by comparing the difference for the numerical value category feature for obtaining the two, as history number
According to comparative feature.
Specifically, forecast sample data can take the sample of signal A of current sample period, if the sampling period can be 5 seconds.
The phase of history petroleum pipeline signal of arest neighbors can take the signal in the phase of history time cycle before current sample period
Data B in practical application, can come from for example, the historical time period of the present embodiment can be 1 minute according to specific business
The length in historical time period is defined, different time length can be set in different business.For example, can be set to 2 minutes, 3
Minute, 5 minutes or other times length.The numerical value category feature of forecast sample data and the phase of history of arest neighbors are obtained respectively
The purpose of the numerical value category feature of petroleum pipeline signal is: when extracting the last period of numerical value category feature to indicate forecast sample data sum
Between history petroleum pipeline signal difference.
Therefore, A and B as the sample of signal of two different lengths, corresponding numerical value category feature and frequency domain are extracted respectively
Category feature, extraction process can refer to aforesaid way (a1) and mode (b1).Then the numerical value category feature of the two is done into difference,
Obtain the variation of current sample and historical sample.Difference is bigger to indicate that the sample changed of current slot is more violent.
(e1) petroleum pipeline signal corresponding with the period locating for forecast sample data is predicted using machine learning model,
The petroleum pipeline signal predicted is compared with forecast sample data, obtains prediction error character；
It, can be first with unsupervised machine learning method based on multiple petroleum pipelines before forecast sample in this kind of mode
Signal segment training machine learning model.For example, when training it is N+1 sections continuous that countless parts can be obtained first from historical data
History petroleum pipeline signal, as training data.When specific training, by preceding N sections of history petroleum pipelines in a training data
Signal is input in the machine learning model, N+1 sections of petroleum pipeline signal of machine learning model output prediction, so
Whether N+1 sections of petroleum pipeline signal of comparison prediction and the difference of true N+1 sections of petroleum pipeline signal are less than afterwards
Preset threshold, if it is not, the parameter of adjustment machine learning model, so that difference is less than preset threshold.Using countless parts of acquisition
Continuous N+1 sections of history petroleum pipeline signals, the machine learning model is trained according to above-mentioned training method, can be made
Obtaining machine learning model can be based on multiple petroleum pipeline signal segments before forecast sample, the Accurate Prediction forecast sample pair
The petroleum pipeline signal segment answered.The preset threshold of the present embodiment can be the numerical value of a very little.Finally based on predicting
Petroleum pipeline signal is compared with forecast sample data, obtains prediction error character.Such as it can be first based on the stone predicted
Oil-piping signal and forecast sample data obtain corresponding feature, such as this feature can be the numerical value in aforesaid way (a1)
Category feature, or the frequency domain category feature in aforesaid way (b1).Finally the feature of the two is compared, prediction is obtained and misses
Poor feature.
The machine learning model of the present embodiment can remember (Long Short-Term Memory for shot and long term；LSTM) net
Network model.When training, the signal data of history is inputed into LSTM model, defines a time step N (this i.e. current letter
Number value is judged by how many a signal values before), as N takes 300.After excessive ratation school habit, LSTM model can be according to given
300 signal values come the signal that occurs after successively predicting.
Why LSTM model can accomplish that the prediction of signal is because its structure summarizes historical information to working as well
The influence of preceding value.For example, Fig. 2 is the structural schematic diagram of LSTM provided by the invention: the input at each moment includes history packaging
Information and current information, and historical information can be adjusted by parameter to current influence.
For example, Fig. 3 is the exemplary diagram of the abnormality detection of ECG data of the invention.As shown in figure 3, being done with electrocardiogram
A example, to explain the abnormality detection ability of LSTM network model.The first behavior ECG signal of Fig. 3, the second behavior LSTM
According to the entire signal graph of 300 signal value predictions before each moment, the third line is then predicted value and true signal value
Mean square error (MSE).Although from figure 3, it can be seen that do not tell model which kind of be abnormal signal, it still finds well
The anomalies of signal.
Therefore, in the present embodiment, machine learning model can be based on more before forecast sample using LSTM network model
A petroleum pipeline signal segment accurately predicts the corresponding signal segment of the forecast sample, improves the accurate of prediction error character
Property.
It can be seen from the above, the step of the present embodiment (1), any one in (a1)-(e1) can specifically come through the above way
It realizes, and obtains corresponding feature.In practical application, according to the demand of different business scene, obtained feature can for one,
Two or more.
(2) feature obtained based on step (1) obtains forecast sample feature corresponding with forecast sample data；
For example, being obtained and pre- test sample if integrating obtained feature when the quantity for the feature that step (1) obtains is more than one
The corresponding forecast sample feature of notebook data；And if the feature quantity that step (1) obtains be one when, the feature that will be obtained, as
Forecast sample feature corresponding with forecast sample data.
S102, forecast sample feature is input to the petroleum pipeline abnormal signal detection model trained and obtained, obtains stone
The output result of oil-piping abnormal signal detection model；
S103, according to the output of petroleum pipeline abnormal signal detection model as a result, determining whether petroleum pipeline signal abnormal.
In the present embodiment, which is trained in advance.In use, by upper
The mode of step S101 is stated, it is available to arrive forecast sample feature.Then forecast sample feature can be input to the petroleum pipeline
In road abnormal signal detection model, forecast sample feature of the petroleum pipeline abnormal signal detection model based on input, output knot
Fruit.According to the output as a result, can determine whether petroleum pipeline signal is abnormal.
Specifically, training when, just known output result why when, indicate petroleum pipeline signal it is normal, output
As a result why when, indicate petroleum pipeline abnormal signal.For example, training when, can be set output result for number 0 when, table
Show that petroleum pipeline signal is normal, when exporting result 1 or other numerical value, indicates petroleum pipeline abnormal signal.These are abnormal and just
Normal mark can be arranged in training according to the use habit in business scenario.
The petroleum pipeline abnormal signal detection model of the present embodiment can use gradient boosted tree (Gradient
Boosting Decision Tree；GBDT network model) is realized, in practical application, can also use other nerve nets
Network model realizes that no longer citing repeats one by one herein.
The method for detecting abnormality of the petroleum pipeline signal of the present embodiment can be used and be trained in advance through the above scheme
Petroleum pipeline abnormal signal detection model, the petroleum pipeline signal acquired in real time is carried out abnormality detection to realize, and it is existing
Technology is compared, and since petroleum pipeline abnormal signal detection model is realized using neural network model, can be effectively improved different
The accuracy often detected, and do not influenced by the complicated shape of petroleum pipeline, it can effectively guarantee detection efficiency.
Fig. 4 is the flow chart of the method for detecting abnormality embodiment two of petroleum pipeline signal of the invention.As shown in figure 4, this
The method for detecting abnormality of the petroleum pipeline signal of embodiment, on the basis of the technical solution of above-mentioned embodiment illustrated in fig. 1, description
Another application scenarios of the invention.As shown in figure 4, the method for detecting abnormality of the petroleum pipeline signal of the present embodiment, specifically may be used
To include the following steps:
S200, the petroleum pipeline signal that sensor acquires in real time is obtained；
Whether the intensity value for the petroleum pipeline signal that S201, judgement acquire in real time is located at preset normal intensity interval range
It is interior；If so, executing step S202；It is no to then follow the steps 203；
S202, determine that the petroleum pipeline signal acquired in real time is located in normal intensity interval range；Execute step S204；
S203, determine that the corresponding petroleum pipeline signal of forecast sample data is abnormal, end.
In the present embodiment, the petroleum pipeline signal acquired in real time can be judged according to preset normal intensity interval range
It is normal or abnormal, if abnormal, can not be detected using subsequent step.And if in the normal range,
It needs further to carry out abnormality detection using subsequent step.Because numerical value is in the normal range, do not represent implicit in the presence of abnormal
Situation.
For example, big according to the normal petroleum pipeline vibration signal strength values that the technical specialist of related service experience provides
About between 24000 to 34000, therefore, in the present embodiment, a preset normal intensity interval range can be designed accordingly,
It such as can be set to 20000 to 40000.If the signal strength of certain section of sample is more than 40000 or less than 20000, it is believed that
The segment signal has exception.
S204, the petroleum pipeline signal acquired in real time according to sensor generate forecast sample data；Execute step S205；
S205, feature extraction processing is carried out to forecast sample data, obtains forecast sample feature；Execute step S206；
S206, forecast sample feature is input to the petroleum pipeline abnormal signal detection model trained and obtained, obtains stone
The output result of oil-piping abnormal signal detection model；Execute step S207；
In practical applications, trained petroleum pipeline abnormal signal detection model, can be deployed to own service
On device, or it is packaged into a network interface and responds network request.Whenever signal data of having newly arrived, it is returned to petroleum pipeline exception
The result of signal detection model judgement.
For example, when step S206 is implemented, it can be to the service for being deployed with petroleum pipeline abnormal signal detection model
Device sends forecast sample feature, and receives the output result that server returns；
Alternatively, can send to the corresponding network interface of petroleum pipeline abnormal signal detection model includes forecast sample feature
Network request, and receive comprising output result network response message.
S207, in a manner of clock signal, in webpage show petroleum pipeline abnormal signal detection model output result；
In practical application, the data area shown can be chosen by configuration file.Believe extremely getting petroleum pipeline
After the output result of number detection model, display module is run, waits the output result meter of petroleum pipeline abnormal signal detection model
It calculates and completes, generate the network address for showing output result, access the network address of generation, that is, may be viewed by output result.The webpage of the present embodiment
It shows using using flask frame, front end to use HighStock to realize the web displaying for carrying out clock signal.
S208, according to the output of the petroleum pipeline abnormal signal detection model of web page display as a result, determining petroleum pipeline letter
It is number whether abnormal.
The realization and the realization of the step S100-S102 of above-mentioned embodiment illustrated in fig. 1 of the step S204-S206 of the present embodiment
It is identical, it can be recorded in detail with reference to the related of above-mentioned embodiment illustrated in fig. 1, details are not described herein.
The method for detecting abnormality of the petroleum pipeline signal of the present embodiment can be used and be trained in advance through the above scheme
Petroleum pipeline abnormal signal detection model, the petroleum pipeline signal acquired in real time is carried out abnormality detection to realize, and it is existing
Technology is compared, and since petroleum pipeline abnormal signal detection model is realized using neural network model, can be effectively improved different
The accuracy often detected, and do not influenced by the complicated shape of petroleum pipeline, it can effectively guarantee detection efficiency.
Fig. 5 is the flow chart of the training method embodiment of the abnormality detection model of petroleum pipeline signal of the invention.Such as Fig. 5
Shown, the training method of the abnormality detection model of the petroleum pipeline signal of the present embodiment can specifically include following steps:
The historic state information of S300, the petroleum pipeline signal and corresponding petroleum pipeline that are acquired according to sensor historic,
Generate training sample data collection；
The executing subject of the training method of the abnormality detection model of the petroleum pipeline signal of the present embodiment is petroleum pipeline letter
Number abnormality detection model training device, the training device of the abnormality detection model of the petroleum pipeline signal is above-mentioned for training
The abnormality detection model of the petroleum pipeline signal of embodiment illustrated in fig. 1.
Firstly, the generating process of the training sample data collection in step S300, can specifically include following steps: will pass
Length is cut into multiple signal segments to the petroleum pipeline signal of sensor history acquisition at preset timed intervals；For each signal segment,
Using signal segment as sample, and using the historic state of the petroleum pipeline in the correspondence time of signal segment as label, obtain
One training sample data is put into training sample data concentration.
The predetermined time period of the present embodiment is the sampling period, and history is acquired continuous petroleum pipeline according to the sampling period
Road signal cutting, just obtains multiple signal segments.
For each signal segment, the historic state of the signal segment can be obtained from historical data, normal shape in this way
State or abnormality.Then using each signal segment historic state corresponding with the signal segment as a number of training
According to.For multiple signal segments, corresponding available a plurality of training sample data.And each training sample data that will be obtained
It is put into training sample data concentration.
S301, feature extraction processing is carried out to training sample data collection, obtains training sample feature set；
For example, step S301 can specifically include following steps when realizing:
(I) carries out following at least one processing to every training sample data that training sample data are concentrated:
(a2) numerical value category feature is extracted from every training sample data；
For example, can specifically extract maximum value, the minimum of petroleum pipeline signal strength indication from every training sample data
At least one of value, mean variance and quantile, as numerical value category feature.
(b2) frequency domain category feature is extracted from every training sample data；
For example, when specific implementation Fast Fourier Transform (FFT) processing can be carried out first to every training sample data；It takes pair
Multiple numerical value of the Fast Fourier Transform (FFT) answered treated real part；The quantity and this training of the numerical value of the real part
The quantity for the petroleum pipeline signal strength indication for including in sample data is equal；Finally extract multiple numerical value of corresponding real part
In at least one of maximum value, minimum value, mean variance and quantile, as frequency domain category feature.
(c2) itself variation tendency feature is extracted from every training sample data；
For example, the extraction of Self-variation trend feature can use any one following mode: first way is will be every
Training sample data are separated from the middle, and obtain anter segmentation and the segmentation of rear piece；Anter segmentation and the segmentation of rear piece are obtained respectively
Numerical value category feature；The difference for obtaining the numerical value category feature of anter segmentation and the numerical value category feature of rear piece segmentation, as Self-variation
Trend feature；The second way is that every training sample data are fitted to binary Gaussian Profile；Obtain binary Gaussian Profile
The difference of mean value and the difference of variance, as Self-variation trend feature.
(d2) every training sample data are compared with the phase of history petroleum pipeline signal of arest neighbors, obtain history
Data comparative feature；And
For example, the numerical value category feature of every training sample data and the phase of history petroleum pipeline of arest neighbors can be obtained respectively
The numerical value category feature of road signal；Then the value category feature for comparing the two, obtains historical data comparative feature.
(e2) petroleum pipeline corresponding with the period locating for every training sample data is predicted using machine learning model
The petroleum pipeline signal predicted is compared by signal with corresponding training sample data, obtains prediction error character；
Specifically, the program needs to be based on every number of training first with unsupervised machine learning method when implementing
According to multiple petroleum pipeline signal segment training machine learning models before.The machine learning model can be LSTM model.So
Petroleum pipeline signal corresponding with the period locating for every training sample data is predicted using the machine learning model afterwards, and will
The petroleum pipeline signal predicted is compared with corresponding training sample data, obtains prediction error character.
Specifically, the step of the present embodiment (a2)-(e2) realization principle, the step of with above-mentioned embodiment illustrated in fig. 1
(a1)-(e1) realization principle is identical, can record in detail with reference to the related of above-mentioned embodiment illustrated in fig. 1, details are not described herein.
The feature of (II) based on every obtained training sample data obtains the corresponding training of corresponding training sample data
Sample characteristics, and corresponding training sample feature is put into training sample feature set；
For example, if integrating obtained feature when the quantity of the feature of every obtained training sample data is more than one,
As the corresponding training sample feature of corresponding training sample data；
If the quantity of the feature of every obtained training sample data is one, by the spy of obtained training sample data
Sign, as the corresponding training sample feature of corresponding training sample data.
Processing based on the step, each training sample data memory feature that can be concentrated to training sample data mention
It takes, obtains the corresponding training sample feature of each training sample data.Each training sample that training sample data are concentrated
The corresponding training sample feature of notebook data is stored in training sample feature set.
S302, based on training sample feature set, the historic state information of acquisition and machine learning algorithm, training petroleum
Pipeline abnormal signal detection model.
By the processing of above-mentioned steps, available training sample feature set may include in the training sample feature set
A plurality of training sample feature.In addition, due to the corresponding one section of petroleum pipeline signal of each training sample feature, and each section of petroleum
Pipe signal corresponds to a historic state information again, therefore each training sample feature can correspond to a historic state letter
Breath.Before training, using each training sample feature and its one historic state information of correspondence as a training data.Training
Before, the abnormality detection model of a unbred petroleum pipeline signal can be first obtained, such as can be a GBDT model, so
Initial value is assigned afterwards for the parameter of the model.When training, a training data is taken, training sample feature therein is input to this
In the abnormality detection model of petroleum pipeline signal, the abnormality detection model of the petroleum pipeline signal exports the state of a prediction.
Then judge whether the state of prediction and the historic state acquired are consistent, if inconsistent, adjust the exception of petroleum pipeline signal
The parameter of detection model, so that the state of prediction is consistent with the historic state acquired.Using the training sample feature set of acquisition
In all training sample features and corresponding historic state information, according to above-mentioned training method, to the different of petroleum pipeline signal
Normal detection model is trained, until the state of prediction is consistent with the historic state of acquisition, training terminates, and determines that petroleum pipeline is believed
Number abnormality detection model parameter, so that it is determined that the abnormality detection model of petroleum pipeline signal.
More, the trained petroleum pipeline of the quantity for the training sample data that the training sample data acquired in the present embodiment are concentrated
The abnormality detection model of road signal is more accurate.Such as the number of training that the training sample data acquired in practical application are concentrated
According to quantity can achieve hundreds of thousands item or more.
The training process of the abnormality detection model of the petroleum pipeline signal of the present embodiment, in above-mentioned embodiment illustrated in fig. 1
Abnormality detecting process principle it is similar, in detail can also with reference to above-mentioned embodiment illustrated in fig. 1 correlation step record,
This is repeated no more.
The training method of the abnormality detection model of the petroleum pipeline signal of the present embodiment can train through the above scheme
One accurately petroleum pipeline abnormal signal detection model, and then it can be based on the petroleum pipeline abnormal signal detection model, it is right
Petroleum pipeline signal carries out abnormality detection, and compared with prior art, can effectively improve the accuracy of abnormality detection, and not by
The complicated shape of petroleum pipeline influences, and can effectively guarantee detection efficiency.
Fig. 6 is the structure chart of the abnormal detector embodiment one of petroleum pipeline signal of the invention.As shown in fig. 6, this
The abnormal detector of the petroleum pipeline signal of embodiment, can specifically include:
Generation module 10 is used for the petroleum pipeline signal acquired in real time according to sensor, generates forecast sample data；
The forecast sample data that abstraction module 11 is used to generate generation module 10 carry out feature extraction processing, are predicted
Sample characteristics；
Detection module 12 is used to the forecast sample feature that abstraction module 11 extracts being input to the petroleum pipeline trained and obtained
Road abnormal signal detection model obtains the output result of petroleum pipeline abnormal signal detection model；
The output of the petroleum pipeline abnormal signal detection model that determining module 13 is used to be obtained according to detection module 12 as a result,
Determine whether petroleum pipeline signal is abnormal.
The abnormal detector of the petroleum pipeline signal of the present embodiment realizes petroleum pipeline letter by using above-mentioned module
Number abnormality detection processing, it is identical as the realization principle of above-mentioned related method embodiment and technical effect, can refer in detail
The related of above method embodiment is recorded, and details are not described herein.
Fig. 7 is the structure chart of the abnormal detector embodiment two of petroleum pipeline signal of the invention.As shown in fig. 7,
On the basis of the technical solution of above-mentioned embodiment illustrated in fig. 6, technical solution of the present invention is further introduced in further detail.
In the abnormal detector of the petroleum pipeline signal of the present embodiment, generation module 10 for acquiring sensor in real time
Every predetermined time period in signal as a forecast sample data.
Still optionally further, in the abnormal detector of the petroleum pipeline signal of the present embodiment, abstraction module 11 is used for:
Following at least one processing is carried out to the forecast sample data that generation module 10 generates: being mentioned from forecast sample data
Access value category feature；Frequency domain category feature is extracted from forecast sample data；Itself variation tendency is extracted from forecast sample data
Feature；Forecast sample data are compared with the phase of history petroleum pipeline signal of arest neighbors, it is more special to obtain historical data
Sign；And petroleum pipeline signal corresponding with the period locating for forecast sample data is predicted using machine learning model, it will be pre-
The petroleum pipeline signal measured is compared with forecast sample data, obtains prediction error character；
Based on obtained feature, forecast sample feature corresponding with forecast sample data is obtained；
Further, based on obtained feature, forecast sample feature corresponding with forecast sample data is obtained, comprising:
If the quantity of obtained feature is more than one, obtained feature is integrated, is obtained corresponding with forecast sample data
Forecast sample feature；
If obtained feature quantity is one, the feature that will be obtained, as pre- test sample corresponding with forecast sample data
Eigen.
Still optionally further, in the abnormal detector of the petroleum pipeline signal of the present embodiment, abstraction module 11 is used for:
Extracted from forecast sample data the maximum value of petroleum pipeline signal strength indication, minimum value, mean variance and point
At least one of digit, as numerical value category feature.
Still optionally further, in the abnormal detector of the petroleum pipeline signal of the present embodiment, abstraction module 11 is used for:
Fast Fourier Transform (FFT) processing is carried out to forecast sample data；
Take Fast Fourier Transform (FFT) treated multiple numerical value of real part；The quantity and prediction of the numerical value of real part
The quantity for the petroleum pipeline signal strength indication for including in sample data is equal；
Extract at least one in maximum value, minimum value, mean variance and the quantile in multiple numerical value of real part
It is a, as frequency domain category feature.
Still optionally further, in the abnormal detector of the petroleum pipeline signal of the present embodiment, abstraction module 11 is used for:
Forecast sample data are separated from the middle, anter segmentation and the segmentation of rear piece are obtained；Anter segmentation is obtained respectively with after
The numerical value category feature of piece segmentation；The difference for obtaining the numerical value category feature of anter segmentation and the numerical value category feature of rear piece segmentation, as
Self-variation trend feature；
Alternatively,
Forecast sample data are fitted to binary Gaussian Profile；Obtain the difference and variance of the mean value of binary Gaussian Profile
Difference, as Self-variation trend feature.
Still optionally further, in the abnormal detector of the petroleum pipeline signal of the present embodiment, abstraction module 11 is used for:
The numerical value of the numerical value category feature of forecast sample data and the phase of history petroleum pipeline signal of arest neighbors is obtained respectively
Category feature；
The difference for obtaining the numerical value category feature of the two, as historical data comparative feature.
Still optionally further, as shown in fig. 7, in the abnormal detector of the petroleum pipeline signal of the present embodiment, further includes:
Training module 14 is used to believe using unsupervised machine learning method based on multiple petroleum pipelines before forecast sample
Number segment training machine learning model.
Accordingly, the machine learning model that abstraction module 11 can use that the training of training module 14 obtains is predicted and is predicted
The corresponding petroleum pipeline signal of period locating for sample data carries out the petroleum pipeline signal predicted and forecast sample data
Compare, obtains prediction error character.
Still optionally further, as shown in fig. 7, further including in the abnormal detector of the petroleum pipeline signal of the present embodiment
Judgment module 15.
Whether the intensity value for the petroleum pipeline signal that judgment module 15 is used to judge to acquire in real time is located at preset normal strong
It spends in interval range；
Determining module 13 is used for the judging result according to judgment module 15, determines that the petroleum pipeline signal acquired in real time is located at
In normal intensity interval range；
If further, it is determined that the intensity value for the petroleum pipeline signal that module 13 is also used to acquire in real time not it is preset just
Within the scope of normal intensity interval, determine that the corresponding petroleum pipeline signal of forecast sample data is abnormal.
Still optionally further, in the abnormal detector of the petroleum pipeline signal of the present embodiment, detection module 12 is used for:
Forecast sample feature is sent to the server for being deployed with petroleum pipeline abnormal signal detection model, and receives service
The output result that device returns；
Alternatively,
The network comprising forecast sample feature is sent to the corresponding network interface of petroleum pipeline abnormal signal detection model to ask
It asks, and receives the network response message comprising output result.
Still optionally further, in the abnormal detector of the petroleum pipeline signal of the present embodiment, further includes:
Display module 16 is used in a manner of clock signal, and the petroleum pipeline for showing that detection module 12 obtains in webpage is different
The output of regular signal detection model as a result, web page display using using flask frame, front end to use HighStock with reality
It is existing.
Accordingly, the output that determining module 13 can be shown according to display module 16 is as a result, determine that petroleum pipeline signal is
No exception.
The abnormal detector of the petroleum pipeline signal of the present embodiment realizes petroleum pipeline letter by using above-mentioned module
Number abnormality detection processing, it is identical as the realization principle of above-mentioned related method embodiment and technical effect, can refer in detail
The related of above method embodiment is recorded, and details are not described herein.
Fig. 8 is the structure chart of the training device embodiment of the abnormality detection model of petroleum pipeline signal of the invention.Such as Fig. 8
Shown, the training device of the abnormality detection model of the petroleum pipeline signal of the present embodiment can specifically include:
The history of petroleum pipeline signal and corresponding petroleum pipeline that generation module 20 is used to be acquired according to sensor historic
Status information generates training sample data collection；
The training sample data collection that abstraction module 21 is used to generate generation module 20 carries out feature extraction processing, is instructed
Practice sample characteristics collection；
The instruction that training module 22 is used to obtain training sample feature set based on abstraction module 21 and generation module 20 generates
Practice the historic state information and machine learning algorithm that sample data is concentrated, training petroleum pipeline abnormal signal detection model.
The training device of the abnormality detection model of the petroleum pipeline signal of the present embodiment is realized by using above-mentioned module
The training of the abnormality detection model of petroleum pipeline signal, realization principle and technical effect phase with above-mentioned related method embodiment
Together, the related record that reference can be made to the above method embodiment in detail, details are not described herein.
Still optionally further, in the training device of the abnormality detection model of the petroleum pipeline signal of the present embodiment, mould is generated
Block 20 is used for:
By the petroleum pipeline signal of sensor historic acquisition, length is cut into multiple signal segments at preset timed intervals；
For each signal segment, using signal segment as sample, and by the petroleum pipeline in the correspondence time of signal segment
The historic state in road obtains a training sample data and is put into training sample data concentration as label.
Still optionally further, in the training device of the abnormality detection model of the petroleum pipeline signal of the present embodiment, mould is extracted
Block 21 is used for:
Following at least one processing is carried out to every training sample data that training sample data are concentrated: from every trained sample
Numerical value category feature is extracted in notebook data；Frequency domain category feature is extracted from every training sample data；From every training sample data
Middle itself variation tendency feature of extraction；The phase of history petroleum pipeline signal of every training sample data and arest neighbors is compared
Compared with obtaining historical data comparative feature；And it is predicted and the time locating for every training sample data using machine learning model
The corresponding petroleum pipeline signal of section, the petroleum pipeline signal predicted is compared with corresponding training sample data, is obtained
Predict error character；
Based on the feature of every obtained training sample data, the corresponding training sample of corresponding training sample data is obtained
Feature, and corresponding training sample feature is put into training sample feature set；
Further, it is corresponding to obtain corresponding training sample data for the feature based on every obtained training sample data
Training sample feature, comprising:
If the quantity of the feature of every obtained training sample data is more than one, obtained feature is integrated, as
The corresponding training sample feature of corresponding training sample data；
If the quantity of the feature of every obtained training sample data is one, by the spy of obtained training sample data
Sign, as the corresponding training sample feature of corresponding training sample data.
Still optionally further, in the training device of the abnormality detection model of the petroleum pipeline signal of the present embodiment, mould is extracted
Block 21 is used for:
Extracted from every training sample data the maximum value of petroleum pipeline signal strength indication, minimum value, mean variance with
And at least one of quantile, as numerical value category feature.
Still optionally further, in the training device of the abnormality detection model of the petroleum pipeline signal of the present embodiment, mould is extracted
Block 21 is used for:
To every training sample data, Fast Fourier Transform (FFT) processing is carried out；
Take corresponding Fast Fourier Transform (FFT) treated multiple numerical value of real part；The quantity of the numerical value of real part
It is equal with the quantity of petroleum pipeline signal strength indication for including in this training sample data；
It extracts in maximum value, minimum value, mean variance and the quantile in multiple numerical value of corresponding real part
At least one, as frequency domain category feature.
Still optionally further, in the training device of the abnormality detection model of the petroleum pipeline signal of the present embodiment, mould is extracted
Block 21 is used for:
Every training sample data are separated from the middle, anter segmentation and the segmentation of rear piece are obtained；Anter segmentation is obtained respectively
The numerical value category feature being segmented with rear；The difference of the numerical value category feature of anter segmentation and the numerical value category feature of rear piece segmentation is obtained,
As Self-variation trend feature；
Alternatively,
Every training sample data are fitted to binary Gaussian Profile；Obtain difference and the side of the mean value of binary Gaussian Profile
The difference of difference, as Self-variation trend feature.
Still optionally further, in the training device of the abnormality detection model of the petroleum pipeline signal of the present embodiment, mould is extracted
Block 21 is used for:
The numerical value category feature of every training sample data and the phase of history petroleum pipeline signal of arest neighbors are obtained respectively
Numerical value category feature；
The difference for obtaining the numerical value category feature of the two, as historical data comparative feature.
Still optionally further, in the training device of the abnormality detection model of the petroleum pipeline signal of the present embodiment, training mould
Block 22 is also used to using unsupervised machine learning method based on multiple petroleum pipeline signal patch before every training sample data
Section training machine learning model.
Accordingly, abstraction module 21 is used to predict and every using the machine learning model that the training of training module 22 obtains
The corresponding petroleum pipeline signal of period locating for training sample data, by the petroleum pipeline signal predicted and corresponding trained sample
Notebook data is compared, and obtains prediction error character；
The training device of the abnormality detection model of the petroleum pipeline signal of above-described embodiment, it is real by using above-mentioned module
The training of the abnormality detection model of existing petroleum pipeline signal, realization principle and technical effect with above-mentioned related method embodiment
Related record identical, that reference can be made to the above method embodiment in detail, details are not described herein.
Fig. 9 shows the structural schematic diagram that an embodiment according to the present invention can be used for realizing the calculating equipment of the above method.
For example, the calculating equipment can be used to implement the method for detecting abnormality of above-mentioned petroleum pipeline signal or the exception of petroleum pipeline signal
The training method of detection model.
Referring to Fig. 9, calculating equipment 1000 includes memory 1010 and processor 1020.
Processor 1020 can be the processor of a multicore, also may include multiple processors.In some embodiments,
Processor 1020 may include a general primary processor and one or more special coprocessors, such as graphics process
Device (GPU), digital signal processor (DSP) etc..In some embodiments, the circuit reality of customization can be used in processor 1020
It is existing, such as application-specific IC (ASIC, Application Specific Integrated Circuit) or scene
Programmable gate array (FPGA, Field Programmable Gate Arrays).
Memory 1010 may include various types of storage units, such as Installed System Memory, read-only memory (ROM), and
Permanent storage.Wherein, static data that other modules that ROM can store processor 1020 or computer need or
Instruction.Permanent storage can be read-write storage device.Permanent storage can be after computer circuit breaking
The non-volatile memory device of the instruction and data of storage will not be lost.In some embodiments, permanent storage device is adopted
Use mass storage device (such as magnetically or optically disk, flash memory) as permanent storage.In other embodiment, permanently
Storage device can be removable storage equipment (such as floppy disk, CD-ROM drive).Installed System Memory can be read-write storage equipment or
The read-write storage equipment of person's volatibility, such as dynamic random access memory.Installed System Memory can store some or all processing
The instruction and data that device needs at runtime.In addition, memory 1010 may include the group of any computer readable storage medium
It closes, including various types of semiconductor memory chips (DRAM, SRAM, SDRAM, flash memory, programmable read only memory), disk
And/or CD can also use.In some embodiments, memory 1010 may include readable and/or write removable
Store equipment, such as laser disc (CD), read-only digital versatile disc (such as DVD-ROM, DVD-dual layer-ROM), read-only indigo plant
Light CD, super disc density, flash card (such as SD card, min SD card, Micro-SD card etc.), magnetic floppy disc etc..It calculates
Machine readable storage medium does not include carrier wave and the momentary electron signal by wirelessly or non-wirelessly transmitting.
It is stored with executable code on memory 1010, when executable code is handled by processor 1020, can make to locate
Reason device 1020 executes the method for detecting abnormality for the petroleum pipeline signal addressed above or the abnormality detection model of petroleum pipeline signal
Training method.
Above by reference to attached drawing be described in detail petroleum pipeline signal according to the present invention method for detecting abnormality or
The training of the abnormality detection model of petroleum pipeline signal.
In addition, being also implemented as a kind of computer program or computer program product, the meter according to the method for the present invention
Calculation machine program or computer program product include the calculating for executing the above steps limited in the above method of the invention
Machine program code instruction.
Alternatively, the present invention can also be embodied as a kind of (or the computer-readable storage of non-transitory machinable medium
Medium or machine readable storage medium), it is stored thereon with executable code (or computer program or computer instruction code),
When the executable code (or computer program or computer instruction code) by electronic equipment (or calculate equipment, server
Deng) processor execute when, so that the processor is executed each step according to the above method of the present invention.
Those skilled in the art will also understand is that, various illustrative logical blocks, mould in conjunction with described in disclosure herein
Block, circuit and algorithm steps may be implemented as the combination of electronic hardware, computer software or both.
The flow chart and block diagram in the drawings show the possibility of the system and method for multiple embodiments according to the present invention realities
Existing architecture, function and operation.In this regard, each box in flowchart or block diagram can represent module, a journey
A part of sequence section or code, a part of the module, section or code include one or more for realizing defined
The executable instruction of logic function.It should also be noted that in some implementations as replacements, the function of being marked in box can also
To be occurred with being different from the sequence marked in attached drawing.For example, two continuous boxes can actually be basically executed in parallel,
They can also be executed in the opposite order sometimes, and this depends on the function involved.It is also noted that block diagram and/or stream
The combination of each box in journey figure and the box in block diagram and or flow chart, can the functions or operations as defined in executing
Dedicated hardware based system realize, or can realize using a combination of dedicated hardware and computer instructions.
Various embodiments of the present invention are described above, above description is exemplary, and non-exclusive, and
It is not limited to disclosed each embodiment.Without departing from the scope and spirit of illustrated each embodiment, for this skill
Many modifications and changes are obvious for the those of ordinary skill in art field.The selection of term used herein, purport
In the principle, practical application or improvement to the technology in market for best explaining each embodiment, or make the art
Other those of ordinary skill can understand each embodiment disclosed herein.
Claims (10)
1. a kind of method for detecting abnormality of petroleum pipeline signal, comprising:
The petroleum pipeline signal acquired in real time according to sensor generates forecast sample data；
Feature extraction processing is carried out to the forecast sample data, obtains forecast sample feature；
The forecast sample feature is input to the petroleum pipeline abnormal signal detection model trained and obtained, obtains the petroleum
The output result of pipeline abnormal signal detection model；
According to the output of the petroleum pipeline abnormal signal detection model as a result, determining whether petroleum pipeline signal is abnormal.
2. according to the method described in claim 1, wherein, the petroleum pipeline signal acquired in real time according to sensor generates
Forecast sample data include:
Using the signal in every predetermined time period that the sensor acquires in real time as a forecast sample data.
3. according to the method described in claim 2, wherein, carrying out feature extraction processing to the forecast sample data, obtaining pre-
Test sample eigen includes:
Following at least one processing is carried out to forecast sample data: extracting numerical value category feature from forecast sample data；From prediction
Frequency domain category feature is extracted in sample data；Itself variation tendency feature is extracted from forecast sample data；By forecast sample data
It is compared with the phase of history petroleum pipeline signal of arest neighbors, obtains historical data comparative feature；And utilize machine learning
Model prediction goes out petroleum pipeline signal corresponding with the period locating for the forecast sample data, the petroleum pipeline that will be predicted
Road signal is compared with the forecast sample data, obtains prediction error character；
Based on obtained feature, the forecast sample feature corresponding with the forecast sample data is obtained；
Further, based on obtained feature, the forecast sample feature corresponding with the forecast sample data, packet are obtained
It includes:
If the quantity of obtained feature is more than one, obtained feature is integrated, is obtained corresponding with the forecast sample data
The forecast sample feature；
If obtained feature quantity is one, the feature that will be obtained, as institute corresponding with the forecast sample data
State forecast sample feature.
4. according to the method described in claim 3, wherein, numerical value category feature is extracted from forecast sample data, comprising:
Maximum value, minimum value, mean variance and the quantile of petroleum pipeline signal strength indication are extracted from forecast sample data
At least one of, as the numerical value category feature.
5. according to the method described in claim 3, wherein, frequency domain category feature is extracted from forecast sample data, comprising:
Fast Fourier Transform (FFT) processing is carried out to forecast sample data；
Take the Fast Fourier Transform (FFT) treated multiple numerical value of real part；The quantity of the numerical value of the real part with
The quantity for the petroleum pipeline signal strength indication for including in forecast sample data is equal；
Extract at least one in maximum value, minimum value, mean variance and the quantile in multiple numerical value of the real part
It is a, as the frequency domain category feature.
6. a kind of training method of the abnormality detection model of petroleum pipeline signal, comprising:
According to the historic state information of the petroleum pipeline signal of sensor historic acquisition and corresponding petroleum pipeline, training sample is generated
Notebook data collection；
Feature extraction processing is carried out to the training sample data collection, obtains training sample feature set；
Based on the training sample feature set, the historic state information of acquisition and machine learning algorithm, training petroleum pipeline
Abnormal signal detection model.
7. a kind of abnormal detector of petroleum pipeline signal, comprising:
Generation module, the petroleum pipeline signal for being acquired in real time according to sensor generate forecast sample data；
Abstraction module obtains forecast sample feature for carrying out feature extraction processing to the forecast sample data；
Detection module has trained obtained petroleum pipeline abnormal signal to detect mould for the forecast sample feature to be input to
Type obtains the output result of the petroleum pipeline abnormal signal detection model；
Determining module, for the output according to the petroleum pipeline abnormal signal detection model as a result, determining petroleum pipeline signal
It is whether abnormal.
8. a kind of training device of the abnormality detection model of petroleum pipeline signal, comprising:
Generation module, the historic state letter of petroleum pipeline signal and corresponding petroleum pipeline for being acquired according to sensor historic
Breath generates training sample data collection；
Abstraction module obtains training sample feature set for carrying out feature extraction processing to the training sample data collection；
Training module, for the historic state information and machine learning algorithm based on the training sample feature set, acquisition,
Training petroleum pipeline abnormal signal detection model.
9. a kind of calculating equipment, comprising:
Processor；And
Memory is stored thereon with executable code, when the executable code is executed by the processor, makes the processing
Device executes the method as described in any one of claim 1-6.
10. a kind of non-transitory machinable medium, is stored thereon with executable code, when the executable code is electric
When the processor of sub- equipment executes, the processor is made to execute the method as described in any one of claim 1-6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811466072.XA CN109800627B (en) | 2018-12-03 | 2018-12-03 | Petroleum pipeline signal abnormity detection method and device, equipment and readable medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811466072.XA CN109800627B (en) | 2018-12-03 | 2018-12-03 | Petroleum pipeline signal abnormity detection method and device, equipment and readable medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109800627A true CN109800627A (en) | 2019-05-24 |
| CN109800627B CN109800627B (en) | 2020-06-09 |
Family
ID=66556450
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811466072.XA Active CN109800627B (en) | 2018-12-03 | 2018-12-03 | Petroleum pipeline signal abnormity detection method and device, equipment and readable medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109800627B (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110378386A (en) * | 2019-06-20 | 2019-10-25 | 平安科技（深圳）有限公司 | Based on unmarked abnormality recognition method, device and the storage medium for having supervision |
| CN110555413A (en) * | 2019-09-05 | 2019-12-10 | 第四范式（北京）技术有限公司 | method and device for processing time sequence signal, equipment and readable medium |
| CN111027607A (en) * | 2019-11-29 | 2020-04-17 | 泰康保险集团股份有限公司 | Unsupervised high-dimensional data feature importance evaluation and selection method and unsupervised high-dimensional data feature importance evaluation and selection device |
| CN111598039A (en) * | 2020-05-24 | 2020-08-28 | 绍兴声科科技有限公司 | Abnormal sound monitoring method and monitoring device for petroleum pipeline monitoring |
| CN111639711A (en) * | 2020-05-29 | 2020-09-08 | 中国地质大学（武汉） | Oil pipeline leakage monitoring method based on pressure monitoring time sequence data |
| CN111832617A (en) * | 2020-06-05 | 2020-10-27 | 上海交通大学 | Engine cold state test fault diagnosis method |
| CN112016600A (en) * | 2020-08-14 | 2020-12-01 | 中国石油大学(北京) | Pipeline abnormity identification method, device and system |
| CN112163528A (en) * | 2020-09-30 | 2021-01-01 | 四川弘和通讯有限公司 | Oil pipe mopping identification method based on convolutional neural network |
| CN113362850A (en) * | 2020-03-03 | 2021-09-07 | 杭州海康威视数字技术股份有限公司 | Detection method and device of audio signal acquisition device and storage medium |
| CN116011183A (en) * | 2022-12-08 | 2023-04-25 | 中国石油大学(北京) | In-service oil and gas pipeline detection method, device, equipment and storage medium |
| CN117147807A (en) * | 2023-11-01 | 2023-12-01 | 中海(天津)能源科技有限公司 | Oil quality monitoring system and method for petroleum exploration |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102122351A (en) * | 2011-03-01 | 2011-07-13 | 哈尔滨工程大学 | Intelligent identification method for pipeline defect on basis of RBF (Radical Basis Function) neural network |
| CN102269972A (en) * | 2011-03-29 | 2011-12-07 | 东北大学 | Method and device for compensating pipeline pressure missing data based on genetic neural network |
| JP2013113366A (en) * | 2011-11-28 | 2013-06-10 | Honda Motor Co Ltd | Gas supply system |
| WO2014031749A1 (en) * | 2012-08-22 | 2014-02-27 | General Electric Company | Wireless system and method for measuring an operative condition of a machine |
| CN103939749A (en) * | 2014-04-28 | 2014-07-23 | 东北大学 | Oil pipeline network leakage intelligent self-adaptation monitoring system and method based on big data |
| CN108730776A (en) * | 2018-06-22 | 2018-11-02 | 大连理工大学 | A kind of Oil Pipeline Leakage Detection based on the fusion of extreme learning machine information |
-
2018
- 2018-12-03 CN CN201811466072.XA patent/CN109800627B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102122351A (en) * | 2011-03-01 | 2011-07-13 | 哈尔滨工程大学 | Intelligent identification method for pipeline defect on basis of RBF (Radical Basis Function) neural network |
| CN102269972A (en) * | 2011-03-29 | 2011-12-07 | 东北大学 | Method and device for compensating pipeline pressure missing data based on genetic neural network |
| JP2013113366A (en) * | 2011-11-28 | 2013-06-10 | Honda Motor Co Ltd | Gas supply system |
| WO2014031749A1 (en) * | 2012-08-22 | 2014-02-27 | General Electric Company | Wireless system and method for measuring an operative condition of a machine |
| CN103939749A (en) * | 2014-04-28 | 2014-07-23 | 东北大学 | Oil pipeline network leakage intelligent self-adaptation monitoring system and method based on big data |
| CN108730776A (en) * | 2018-06-22 | 2018-11-02 | 大连理工大学 | A kind of Oil Pipeline Leakage Detection based on the fusion of extreme learning machine information |
Cited By (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110378386A (en) * | 2019-06-20 | 2019-10-25 | 平安科技（深圳）有限公司 | Based on unmarked abnormality recognition method, device and the storage medium for having supervision |
| CN110555413A (en) * | 2019-09-05 | 2019-12-10 | 第四范式（北京）技术有限公司 | method and device for processing time sequence signal, equipment and readable medium |
| CN111027607A (en) * | 2019-11-29 | 2020-04-17 | 泰康保险集团股份有限公司 | Unsupervised high-dimensional data feature importance evaluation and selection method and unsupervised high-dimensional data feature importance evaluation and selection device |
| CN111027607B (en) * | 2019-11-29 | 2023-10-17 | 泰康保险集团股份有限公司 | Unsupervised high-dimensional data feature importance assessment and selection method and device |
| CN113362850A (en) * | 2020-03-03 | 2021-09-07 | 杭州海康威视数字技术股份有限公司 | Detection method and device of audio signal acquisition device and storage medium |
| CN113362850B (en) * | 2020-03-03 | 2024-05-28 | 杭州海康威视数字技术股份有限公司 | Method and device for detecting audio signal acquisition device and storage medium |
| CN111598039A (en) * | 2020-05-24 | 2020-08-28 | 绍兴声科科技有限公司 | Abnormal sound monitoring method and monitoring device for petroleum pipeline monitoring |
| CN111639711A (en) * | 2020-05-29 | 2020-09-08 | 中国地质大学（武汉） | Oil pipeline leakage monitoring method based on pressure monitoring time sequence data |
| CN111639711B (en) * | 2020-05-29 | 2023-04-18 | 中国地质大学（武汉） | Oil pipeline leakage monitoring method based on pressure monitoring time sequence data |
| CN111832617B (en) * | 2020-06-05 | 2022-11-08 | 上海交通大学 | Engine cold state test fault diagnosis method |
| CN111832617A (en) * | 2020-06-05 | 2020-10-27 | 上海交通大学 | Engine cold state test fault diagnosis method |
| CN112016600A (en) * | 2020-08-14 | 2020-12-01 | 中国石油大学(北京) | Pipeline abnormity identification method, device and system |
| CN112163528A (en) * | 2020-09-30 | 2021-01-01 | 四川弘和通讯有限公司 | Oil pipe mopping identification method based on convolutional neural network |
| CN116011183A (en) * | 2022-12-08 | 2023-04-25 | 中国石油大学(北京) | In-service oil and gas pipeline detection method, device, equipment and storage medium |
| CN116011183B (en) * | 2022-12-08 | 2023-09-15 | 中国石油大学(北京) | In-service oil and gas pipeline detection method, device, equipment and storage medium |
| CN117147807A (en) * | 2023-11-01 | 2023-12-01 | 中海(天津)能源科技有限公司 | Oil quality monitoring system and method for petroleum exploration |
| CN117147807B (en) * | 2023-11-01 | 2024-01-26 | 中海(天津)能源科技有限公司 | Oil quality monitoring system and method for petroleum exploration |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109800627B (en) | 2020-06-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109800627A (en) | The method for detecting abnormality and device of petroleum pipeline signal, equipment and readable medium | |
| CN109948669A (en) | A kind of abnormal deviation data examination method and device | |
| CN110348580B (en) | Method and device for constructing GBDT model, and prediction method and device | |
| CN109614997A (en) | A kind of stealing Risk Forecast Method and device based on deep learning | |
| CN107949812A (en) | For detecting the abnormal combined method in water distribution system | |
| CN109657697A (en) | Classified optimization method based on semi-supervised learning and fine granularity feature learning | |
| CN104112284B (en) | The similarity detection method and equipment of a kind of picture | |
| TW202014940A (en) | Training sample obtaining method, account prediction method, and corresponding devices | |
| RU2014153646A (en) | METHODS AND RELATED SYSTEMS FOR BUILDING MODELS AND FORECASTING OPERATIONAL RESULTS OF DRILLING OPERATION | |
| CN105095756A (en) | Method and device for detecting portable document format document | |
| CN109858683B (en) | Method, device, electronic equipment and storage medium for determining business state of store | |
| KR20190065015A (en) | Support method for responding to stream disaster, and support system for responding to stream disaster | |
| CN109918687B (en) | Train dynamics simulation method and simulation platform based on machine learning | |
| CN110659825A (en) | Cash demand prediction method and device for multiple learners of bank outlets | |
| CN109002810A (en) | Model evaluation method, Radar Signal Recognition method and corresponding intrument | |
| CN109190907A (en) | The small micro- power honesty risk index construction method of power supply station based on big data | |
| CN111191696A (en) | Deep learning algorithm-based steel bar layering method and system | |
| Ke et al. | New framework for automatic identification and quantification of freeway bottlenecks based on wavelet analysis | |
| CN109242165A (en) | A kind of model training and prediction technique and device based on model training | |
| CN113297044A (en) | Operation and maintenance risk early warning method and device | |
| KR20140137210A (en) | computing system for well placement optimization developed by SA/ANN and well placement optimization method using Thereof | |
| CN109508738A (en) | A kind of information processing method and relevant device | |
| CN104964736A (en) | Optical fiber invasion vibration source identification method based on time-frequency characteristic EM classification | |
| Eichinski et al. | Clustering acoustic events in environmental recordings for species richness surveys | |
| CN115712834A (en) | Alarm false alarm detection method, device, equipment and 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 |