CN117554000A - Intelligent detection method and system for leakage of tailing conveying pipeline - Google Patents
Intelligent detection method and system for leakage of tailing conveying pipeline Download PDFInfo
- Publication number
- CN117554000A CN117554000A CN202311413210.9A CN202311413210A CN117554000A CN 117554000 A CN117554000 A CN 117554000A CN 202311413210 A CN202311413210 A CN 202311413210A CN 117554000 A CN117554000 A CN 117554000A
- Authority
- CN
- China
- Prior art keywords
- pressure
- leakage
- data
- time sequence
- pipeline
- 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
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 106
- 238000012549 training Methods 0.000 claims abstract description 28
- 238000011156 evaluation Methods 0.000 claims abstract description 5
- 238000000034 method Methods 0.000 claims description 40
- 230000008569 process Effects 0.000 claims description 22
- 238000010276 construction Methods 0.000 claims description 15
- 238000012545 processing Methods 0.000 claims description 11
- 238000012544 monitoring process Methods 0.000 claims description 10
- 238000005070 sampling Methods 0.000 claims description 10
- 238000005259 measurement Methods 0.000 claims description 9
- 238000001914 filtration Methods 0.000 claims description 7
- 238000007781 pre-processing Methods 0.000 claims description 6
- 230000011218 segmentation Effects 0.000 claims description 6
- 230000005540 biological transmission Effects 0.000 claims description 4
- 238000012407 engineering method Methods 0.000 claims description 3
- 230000009467 reduction Effects 0.000 claims description 3
- 238000013210 evaluation model Methods 0.000 claims description 2
- 238000010586 diagram Methods 0.000 description 9
- 230000006870 function Effects 0.000 description 9
- 238000004590 computer program Methods 0.000 description 7
- 230000004913 activation Effects 0.000 description 6
- 238000011161 development Methods 0.000 description 5
- 230000018109 developmental process Effects 0.000 description 5
- 238000009826 distribution Methods 0.000 description 5
- 230000015654 memory Effects 0.000 description 5
- 238000005065 mining Methods 0.000 description 5
- 239000013598 vector Substances 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 4
- 238000010606 normalization Methods 0.000 description 4
- 230000002159 abnormal effect Effects 0.000 description 3
- 230000005856 abnormality Effects 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 3
- 238000000354 decomposition reaction Methods 0.000 description 3
- 229910052500 inorganic mineral Inorganic materials 0.000 description 3
- 239000011707 mineral Substances 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 238000003860 storage Methods 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 230000006978 adaptation Effects 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000005272 metallurgy Methods 0.000 description 2
- 238000013508 migration Methods 0.000 description 2
- 230000005012 migration Effects 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 238000002759 z-score normalization Methods 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 239000012530 fluid Substances 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 238000013526 transfer learning Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M3/00—Investigating fluid-tightness of structures
- G01M3/02—Investigating fluid-tightness of structures by using fluid or vacuum
- G01M3/26—Investigating fluid-tightness of structures by using fluid or vacuum by measuring rate of loss or gain of fluid, e.g. by pressure-responsive devices, by flow detectors
- G01M3/28—Investigating fluid-tightness of structures by using fluid or vacuum by measuring rate of loss or gain of fluid, e.g. by pressure-responsive devices, by flow detectors for pipes, cables or tubes; for pipe joints or seals; for valves ; for welds
- G01M3/2807—Investigating fluid-tightness of structures by using fluid or vacuum by measuring rate of loss or gain of fluid, e.g. by pressure-responsive devices, by flow detectors for pipes, cables or tubes; for pipe joints or seals; for valves ; for welds for pipes
- G01M3/2815—Investigating fluid-tightness of structures by using fluid or vacuum by measuring rate of loss or gain of fluid, e.g. by pressure-responsive devices, by flow detectors for pipes, cables or tubes; for pipe joints or seals; for valves ; for welds for pipes using pressure measurements
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Examining Or Testing Airtightness (AREA)
Abstract
The invention relates to an intelligent detection method and system for leakage of a tailing conveying pipeline, comprising the following steps: s1, collecting pressure data of nodes at two ends of pressure sensors at the head end and the tail end of a tailing conveying pipeline in real time, S2, constructing a leakage detection model based on an generated countermeasure network by taking pressure time sequence characteristic sequence sample data of the pressure time sequence characteristic sequence data in the normal state at the head end and the tail end of the pipeline as a training set, and establishing a pressure fluctuation mode of the normal state of the tailing conveying pipeline, so that a pseudo pressure trend generated based on the leakage detection model of the generated countermeasure network is consistent with a real pressure trend in the normal state, and accurate detection and identification of pipeline leakage are realized; s3, constructing a leakage detection model effectiveness evaluation method. By adopting the mode to detect the leakage of the pipeline, the leakage problem can be found in time, so that the leakage is prevented from causing larger loss.
Description
Technical Field
The application relates to the technical field of pipeline detection, and more particularly relates to an intelligent detection method and system for leakage of a tailing conveying pipeline.
Background
The green, deep and intelligent mining construction of mineral resources is a key for guaranteeing sustainable and efficient development of mineral resources, wherein tailings transportation is an important link of green and sustainable mining. In the tailings conveying process, the practical problems of long distance of a conveying pipeline, complex surrounding environment of a crossing area, easiness in being influenced by artificial and natural disasters and the like are generally faced, so that the leakage accidents of the tailings conveying pipeline frequently occur. The method not only causes huge economic loss for mine enterprises, but also brings huge potential safety hazards for surrounding environment and personnel.
In recent years, artificial intelligence technology is rapidly developed, and intelligent data processing, analysis and decision model represented by a deep learning algorithm are endless aiming at the problem of generalization and interpretability of a neural network which is the focus of heat in the field of machine learning. However, due to the limitation of the actual condition of the mining and metallurgy site, the acquired actual data quantity of the site is limited, compared with the normal running state, the leakage is less, the data quantity of the leakage sample is seriously insufficient, and the simulation data acquired in the ideal physical test environment is greatly different from the actual condition, so that the requirement of large-scale balanced training samples for machine learning cannot be met, and the real-time application of the artificial intelligence technology in the mining and metallurgy field is greatly limited. Therefore, how to realize intelligent detection of the leakage of the tailing conveying pipeline under the condition of a small amount of normal state samples and no leakage samples is more and more paid attention to expert students in relevant fields at home and abroad, and pipeline leakage detection research based on small samples and no negative samples is developed by gradually exploring and applying models such as threshold processing, a self-encoder, an countermeasure generation network, a convolutional neural network, a recurrent neural network, a transfer learning model, a Transformer model and the like.
Disclosure of Invention
The invention aims to provide an intelligent detection method and system for leakage of a tailing conveying pipeline, which can be used for migration and expansion application to the aspects of abnormal detection of large-scale industrial equipment and the like on the basis of the application of leakage detection of the tailing conveying pipeline, assist the intelligent development of the industry and have positive promotion effects on the development of a novel real-time, quantitative and accurate method, a novel theory and a novel technology for abnormal intelligent detection.
The object of the present invention is thus achieved.
The invention discloses an intelligent detection method for leakage of a tailing conveying pipeline, which is characterized by comprising the following steps of:
s1, collecting pressure data of nodes at the two ends of a tailing conveying pipeline in real time, carrying out data filtering and data standardization on the collected pressure data of the nodes at the two ends, and constructing normal state pressure time sequence characteristic sequence sample data of the conveying pipeline at the two ends;
s2, taking pressure time sequence characteristic sequence sample data in the normal state of the head end and the tail end of the pipeline as a training set, constructing a leakage detection model based on the generation of an countermeasure network, and establishing a pressure fluctuation mode of the normal state of the tailing conveying pipeline, so that a pseudo pressure trend generated based on the leakage detection model of the generation of the countermeasure network is consistent with a real pressure trend in the normal state, and accurate detection and identification of pipeline leakage are realized;
s3, constructing a leakage detection model effectiveness evaluation method, evaluating model identification accuracy by using an area index and an F1-score index under a working characteristic curve of a subject, and evaluating reconstruction quality of a pseudo pressure time sequence characteristic sequence under a normal state by using a root mean square error index.
Preferably, the pressure sensors at the head and tail ends of the tailing conveying pipeline in S1 collect pressure data of nodes at the two ends in real time, and construct sample data of pressure time sequence characteristic sequences at the head and tail ends by combining characteristic engineering methods such as data filtering and data standardization, and the method comprises the following steps:
installing pressure sensors and GPS timing modules at the head end and the tail end of a conveying pipeline, and remotely setting a leakage detection and central monitoring machine to store and analyze data so as to acquire pipeline information of detection point numbers, time codes and pressure;
preprocessing the collected actual measurement pressure data of the pipeline; wherein the preprocessing comprises data noise reduction processing and data standardization processing;
and (3) establishing a time sequence segmentation model based on sampling, so as to ensure that the original pressure trend is unchanged, and respectively constructing pressure time sequence characteristic sequence sample data of the head end and the tail end of the pipeline in the tailing transmission process based on the preprocessed pipeline actual measurement pressure data through the time sequence segmentation model.
Preferably, the data denoising process adopts a wavelet threshold denoising algorithm, and the data normalization process adopts a Z-score normalization process method.
Preferably, the step of respectively constructing pressure time sequence characteristic sequence sample data of the head end and the tail end of the pipeline in the tailing transmission process based on the preprocessed pipeline actual measurement pressure comprises the following steps:
according to the size of the sampling interval, the original pressure time sequence characteristic sequence is divided into a plurality of blocks, a point is sequentially selected from each block, and finally, the blocks are spliced into a sampling sample with a specific length, so that a plurality of sample data can be obtained from one actually measured sample, and the construction of a pressure time sequence characteristic sequence sample data set for model training is completed.
Preferably, S2, the training process of the constructed leak detection model (GANomaly) based on the generation of the countermeasure network is only trained by using the normal state pressure time sequence characteristic sequence sample data, and no leak sample data is required, and the leak detection model includes:
a generator network that generates a pseudo timing sequence by encoding-decoding input timing sequence data using an Auto Encoder (AE);
the encoder network is used for obtaining the code representation of the generated pressure time sequence characteristic sequence and is also used as a classifier for leakage detection, and judging whether the current input is leaked or not according to the error between the code representation of the generated sequence and the code representation of the input sequence;
and the discriminator network discriminates the true and false between the input pressure time sequence characteristic sequence and the generated pseudo pressure time sequence characteristic sequence, so as to guide the training of the generator network.
Preferably, the training construction is based on generating a leak detection model of the countermeasure network by using the obtained pressure time sequence characteristic sequence sample data in the normal state, and a pressure fluctuation mode of the normal state of the tailing conveying pipeline is established, wherein the pressure time sequence characteristic sequence sample data at the head end and the tail end are respectively used as input to perform the training of generating the countermeasure network.
Preferably, the model parameters of the generator network and the encoder network are updated and optimized according to the discrimination loss obtained by the discriminator network, the coding loss between the input pressure time sequence characteristic sequence sample data and the coding representation of the pseudo pressure time sequence characteristic sequence, the Manhattan distance measurement loss between the input pressure time sequence characteristic sequence sample data and the pseudo pressure time sequence characteristic sequence, the three losses are given different weight proportions, and the self-updating of the parameters is carried out by the discriminator network according to the pseudo pressure time sequence characteristic sequence newly generated by the generator network.
Preferably, in the step S4, the accuracy is identified by using an area index under a curve of a subject working characteristic curve (Receiver Operating Characteristic curve, ROC) and an F1-score index evaluation model, and the difference between the input pressure time sequence characteristic sequence and the generated pseudo pressure time sequence characteristic sequence is evaluated by using a root mean square error index, so as to effectively evaluate the reconstruction quality of the pseudo pressure time sequence characteristic sequence in a normal state.
The invention discloses an intelligent detection system for leakage of a tailing conveying pipeline, which is characterized by comprising pressure sensors, GPS timing modules and a leakage detection and central monitoring machine, wherein the pressure sensors are arranged at the head end and the tail end;
the pressure sensor is used for collecting measured pressure data of the conveying pipeline in the tailing conveying process;
the GPS timing module is used for acquiring pipeline information comprising a detection point number, a time code number and pressure, and providing necessary data base for realizing construction of pressure time sequence characteristic sequence sample data;
the leakage detection and central monitoring machine is used for receiving the measured pressure data and the pipeline information obtained by the GPS timing module, and integrating the intelligent detection method for the leakage of the tailing conveying pipeline to realize intelligent and accurate detection and identification of the leakage of the tailing conveying pipeline.
The invention has the advantages that:
compared with the prior art, the intelligent detection method for leakage of the tailing conveying pipeline has the advantages that the related measured data are derived from pressure sensors which are installed at monitoring nodes at the head end and the tail end of the pipeline on site, and pressure time sequence characteristic sequence sample data at the head end and the tail end are respectively constructed according to pressure information after data preprocessing operations such as wavelet threshold denoising and standardization; then, constructing a leakage detection model based on a generated countermeasure network, and respectively taking pressure time sequence characteristic sequence sample data of the former pipeline in a normal state at the head end and the tail end as a training set to train the network model; finally, the constructed leak detection model can compare the difference between the input normal state pressure time sequence characteristic sequence sample data and the pseudo pressure time sequence characteristic sequence of the generating module, calculate the mean square error of the two on the hidden layer coding representation and take the mean square error as the score of the leak state. In the practical leakage detection application process, given pressure input, the model gives out the score of the input state in real time, namely the confidence coefficient of the leakage state of the tailing conveying pipeline is used, and intelligent detection of leakage of the tailing filling pipeline is realized. Thus, the leakage problem can be found in time by detecting the leakage of the pipeline in the mode, so that the leakage is prevented from causing larger loss.
Drawings
Fig. 1 is a schematic flow chart of an intelligent detection method for leakage of a tailing conveying pipeline, which is provided by an embodiment of the invention;
FIG. 2 is a block diagram of an intelligent detection system for leakage of a tailing conveying pipeline, which is provided by the embodiment of the invention;
FIG. 3 is a schematic diagram of a process for constructing a sequence of pressure timing features according to an embodiment of the present invention;
FIG. 4 is a schematic diagram of a leak detection model based on a generation countermeasure network provided by an embodiment of the invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the present invention more apparent, the embodiments of the present invention will be described in further detail with reference to the accompanying drawings.
First embodiment
As shown in fig. 1 to fig. 4, the embodiment provides an intelligent detection method for leakage of a tailing conveying pipeline, which can be implemented by electronic equipment, and the electronic equipment can be a terminal or a server, aiming at the problems of insufficient data volume of abnormal samples and the like widely existing in the state abnormality detection of the current industrial equipment.
The invention discloses an intelligent detection method for leakage of a tailing conveying pipeline, which is characterized by comprising the following steps of:
s1, collecting pressure data of nodes at the two ends of a tailing conveying pipeline in real time, carrying out data filtering and data standardization on the collected pressure data of the nodes at the two ends, and constructing normal state pressure time sequence characteristic sequence sample data of the conveying pipeline at the two ends;
specifically, the implementation procedure of S1 in this embodiment is as follows:
s1.1, the intelligent detection system for the leakage of the tailing conveying pipeline in the embodiment consists of a pressure sensor at the head end and the tail end, a GPS timing module and a leakage detection and center monitoring machine, as shown in FIG. 2; the central monitoring machine receives data through a network according to a self-defined communication protocol, and acquires actual measurement pressure data of the pipeline in the tailing conveying process through pressure sensors at the head end and the tail end of the tailing conveying pipeline, so that pipeline information comprising a detection point number, a time code number and pressure is acquired, and a necessary data basis is provided for realizing construction of a pressure time sequence characteristic sequence.
When a pipeline leaks on a certain position of a pipeline to be detected, the pressure signal of the fluid also changes correspondingly, the collected negative pressure wave signal is displayed in a curve form in real time by detection software of a central control station, and after the detection method in the embodiment judges that the pipeline leaks, the leakage in the pipeline is timely alarmed according to the signals collected at two ends of the pipeline on the basis of the theory of a negative pressure wave method.
S1.2, for the collected pressure data of the two end nodes
Firstly extracting leakage sample data from the acquired pipeline actual measurement pressure data, ensuring that the leakage sample only appears in a model test set, and carrying out noise reduction treatment by utilizing a data filtering algorithm; the data Z-score is standardized, the dimension influence of measured pressure data is eliminated, the data can be better characterized, and meanwhile, the data distribution trend is reserved;
specifically, in this embodiment, the step S1.2 includes the following steps:
s1.2.1, carrying out wavelet decomposition on pressure signals acquired by an information pressure sensor;
s1.2.2, selecting a threshold value for each layer of signals after wavelet decomposition, and performing soft thresholding on the high-frequency coefficient of the layer;
s1.2.3, performing wavelet reconstruction of the pressure signal according to the low-frequency coefficient of wavelet decomposition and the high-frequency coefficient after soft thresholding;
s1.2.4, performing a Z-score normalization process on the pressure data after the wavelet threshold denoising process according to the mean μ and the variance σ, as shown in the following formula:
s1.3, a time sequence characteristic sequence segmentation model based on sampling is established, so that the original pressure trend is kept unchanged, and pressure time sequence characteristic sequence sample data at the head end and the tail end of the pipeline in the tailing transmission process are respectively constructed based on the preprocessed pipeline actual measurement pressure data through the time sequence characteristic sequence segmentation model.
Specifically, as shown in fig. 4, the construction process of the pressure time sequence characteristic sequence sample is that the sample sequence length l is set according to the sample amount required by model training, and the sampling interval size is given that the original sequence length D of the pipeline measured pressure after pretreatment is NWherein->Representing rounding down, dividing the original sequence into l blocks according to the sampling interval, sequentially selecting a point in each block, and finally splicing into a sampling sample with the length of l, thereby obtaining i sample data, and completingThe construction of the pressure time sequence characteristic sequence data set for model training, the expression of the new sample set S is shown in the following formula,
S=D[[0,1,…,l-1]*i+j],j=[0,1,…,i],
it should be noted that, for convenience, the sample sequence length l set in this embodiment is a factor of the original sequence length, and in addition, the setting of the sample sequence length l affects the size of the data volume and the data distribution trend. When the l setting is too large, the data amount is too small, and the model may have difficulty in learning the pressure normal state fluctuation mode; when l is set too small, it may be difficult to preserve the raw data distribution trend, and there may be multiple repeated samples, affecting model training. Thus, during subsequent training, the sample sequence length l is tested as an important super-parameter.
S2, taking pressure time sequence characteristic sequence sample data in the normal state of the head end and the tail end of the pipeline as a training set, constructing a leakage detection model based on the generation of an countermeasure network, and establishing a pressure fluctuation mode of the normal state of the tailing conveying pipeline, so that a pseudo pressure trend generated based on the leakage detection model of the generation of the countermeasure network is consistent with a real pressure trend in the normal state, and accurate detection and identification of pipeline leakage are realized;
specifically, in the present embodiment, a leak detection model (GANomaly) based on generation of an countermeasure network is shown in fig. 4. The basic structure of each module based on the leak detection model training process for generating the countermeasure network is as follows:
s2.1, only taking the constructed pressure time sequence characteristic sequence sample data in the normal state as training input, so that the model can only capture a pressure fluctuation mode in the normal state;
s2.2, constructing a leak detection model based on a generation countermeasure network, wherein the model framework consists of three different sub-networks, namely:
the generator network G generates a pseudo timing sequence by encoding-decoding input timing sequence data using an Auto Encoder (AE). Firstly, the generator G reads the input sequence x and passes it to the encoder network Ge of the generator G; after downsampling one-dimensional convolutional layers, batch normalization and weak RELU activation functions activation, ge compresses x ε R (wXh×c) into a vector z ε R (d), d representing the hidden layer vector dimension, which in this embodiment is used as a super parameter; the decoder Gd then references the structure of a generator in the Deep Convolutional GAN (DCGAN) model, changes the two-dimensional convolutional layer to a one-dimensional convolutional layer, and uses the convolutional transpose layer, RELU activation, batch normalization, and tanh activation functions to obtain x '∈r (w×h×c), i.e., z=ge (x), x' =gd (z).
The encoder network obtains the encoded representation of the generated time series sequence, which is also used as a classifier for leakage detection, and judges whether the current input is leaked or not according to the error between the encoded representation of the generated sequence and the encoded representation of the input sequence. The encoder network is identical to the architecture details of Ge, and the functions implemented by the two are to compress x 'and x to obtain vectors z' and z, respectively. But the parameterization method of the two is different, the former minimizes the potential vector through the bottleneck feature, and the latter explicitly learns the minimum distance through parameterization, and the dimensions of z and z' generated by the two are the same.
The discriminator network D discriminates the true and false between the input pressure timing sequence and the generated pseudo timing sequence, thereby guiding the training of the generator network, referring to the DCGAN-introduced discriminator (5 layers of neural networks, each layer using convolution, batch normalization and LeakyReLU activation, wherein the activation function of the last layer is sigmoid) for discriminating the true and false of the input x and the generated x'.
S2.3, respectively taking pressure time sequence characteristic sequence sample data at the head end and the tail end as input, and training a network model: first, a discrimination loss L is obtained from a discriminator network adv Coding loss L between coded representations of input sequence and generated sequence enc Manhattan distance metric loss L between input sequence and generated sequence con The model parameters of the optimization generator and encoder subnetwork are updated giving three different weight scales (1:1:10) to the losses. The discriminator network then calculates the cross entropy loss to perform self-update of the parameters based on the data newly generated by the generator network.
In the present embodiment, the model meshWith reference to the DCGAN network, an Adam optimization method is adopted, and the self-adaptive learning rate is lr=2e -3 Momentum parameter beta 1 =0.5,β 2 =0.999. When the cross entropy loss of the discriminator network is below 1e -6 At this point, the discriminator network parameters are reinitialized. In addition, the weight ratio of the three losses is set so that the model is more concerned about Manhattan losses, and the model can better learn the context information of a given input sequence.
Specifically, the leak detection model training process based on generating the challenge network is as follows:
s3, constructing a leakage detection model effectiveness evaluation method.
In this embodiment, in order to evaluate the leak detection recognition accuracy of the above model, the model adopts the area under the curve AUC index of the subject work characteristic curve (Receiver Operating Characteristic curve, ROC) and the F1-score (threshold of 0.2) index to evaluate the model recognition accuracy. Further, in order to effectively evaluate the reconstruction quality of the pressure time series characteristic sequence sample data in the normal state, the embodiment adopts a root mean square error (Root Mean Square Error, RMSE) index to evaluate the difference between the input pressure time series characteristic sequence sample data and the pseudo pressure time series characteristic sequence.
Under different hyper-parameter settings, the comparison of the results of the leak detection models of the pressures at the head end and the tail end is shown in tables 1 and 2 respectively.
TABLE 1 leak detection model Performance for head end pressure with different parameter selections
TABLE 2 leak detection model Performance of tail pressure under different parameter selections
In tables 1 and 2, l represents the sequence length size corresponding to the interval of different samples, d represents the hidden layer vector coding dimension size, and normal_rmse and abnormal_rmse represent the reconstruction performance of the model on the pressure input time sequence under normal/leakage states, respectively. As can be seen from the RMSE index in the table, the model has the capability of simulating the pressure fluctuation modes at the head end and the tail end under the normal state, and can well distinguish the normal and leakage states of the pressure fluctuation, and the accuracy of AUC and F1-score in the table is 1, so that the practicability and effectiveness of the intelligent detection method for the leakage of the tailing conveying pipeline are fully illustrated.
In this embodiment, the threshold value is set to 0.2 at the time of calculating the F1-score index. This is because the pressure input sequence in the normal state is consistent with the distribution trend between the pseudo pressure sequences obtained by the reconstruction of the leak detection model, and the mean square error between the two is small. For the pressure input sequence of the leakage state, the model is difficult to reconstruct the distribution trend, so that the model output score is larger. Therefore, the threshold size should be set low.
In summary, the present embodiment provides an intelligent detection method capable of performing real-time pipeline leak detection under the condition that the data volume of the past leak sample is small. The related measured data are derived from pressure sensors which are arranged on the monitoring nodes at the head end and the tail end of the pipeline on site, and pressure time sequence characteristic sequences at the head end and the tail end are respectively constructed aiming at pressure information after data preprocessing operations such as wavelet threshold denoising and standardization; then, constructing a leakage detection model based on a generated countermeasure network, and respectively taking pressure time sequence sequences of the former pipeline in a normal state at the head end and the tail end as training sets to train the network model; finally, the constructed leakage detection model can compare the difference between the input real sequence characteristics and the generation characteristics of the generation module, calculate the mean square error of the two on the hidden layer coding representation and take the mean square error as the score of the leakage state. In the practical leakage detection application process, a given pressure is input into a real sequence, and a model gives a score of the state of the input sequence in real time, namely, the confidence coefficient of the tailing conveying pipeline belonging to the leakage state is used for realizing intelligent detection of the leakage of the tailing filling pipeline. The intelligent detection method for the leakage of the tailing conveying pipeline provides necessary technical support for quantitative, accurate and efficient leakage detection under complex working conditions, and can be used for well serving green, deep and intelligent mining construction of mineral resources.
The detection method of the embodiment can be used for migration and expansion application to the aspects of abnormality detection of large-scale industrial equipment and the like on the basis of the leakage detection application of the tailing conveying pipeline, and helps the intelligent development of the power industry, and has positive promotion effects on the development of a novel real-time, quantitative and accurate abnormality intelligent detection method, a novel theory and a novel technology.
Second embodiment
The embodiment provides an intelligent detection system for leakage of a tailing conveying pipeline, which comprises the following modules:
the system comprises a pipeline head-tail two-end time sequence characteristic sequence sample construction module, a tail-tail two-end pressure sensor, a data filtering and data standardization characteristic engineering method and the like, wherein the pressure sensors at the head end and the tail end of a tailing conveying pipeline collect pressure data of two end nodes in real time, and the head-tail two-end time sequence characteristic sequence sample data are constructed;
the leakage detection model construction module based on the generated countermeasure network is used for training and constructing a generated countermeasure network model by utilizing the obtained pressure time sequence characteristic sample data in the normal state, and a pressure fluctuation mode of the normal state of the tailing conveying pipeline is established, so that a pseudo pressure trend generated by the model is consistent with a real pressure trend in the normal state, and accurate detection and identification of a leakage sample are further realized based on the model characteristics;
the leakage detection model effectiveness evaluation method construction module evaluates model identification precision by using an area index and an F1-score index under a test subject working characteristic curve, and evaluates pseudo pressure sequence characteristic data reconstruction quality under a normal state by using a root mean square error index.
The intelligent detection system for the leakage of the tailing conveying pipeline of the embodiment corresponds to the intelligent detection method for the leakage of the tailing conveying pipeline of the first embodiment; the functions realized by the functional modules in the intelligent detection system for the leakage of the tailing conveying pipeline correspond to the flow steps in the intelligent detection method for the leakage of the tailing conveying pipeline in the first embodiment one by one; therefore, the description is omitted here.
Third embodiment
The embodiment provides an electronic device, which comprises a processor and a memory; wherein the memory stores at least one instruction that is loaded and executed by the processor to implement the method of the first embodiment.
The electronic device may vary considerably in configuration or performance and may include one or more processors (Central Processing Units, CPU) and one or more memories having at least one instruction stored therein that is loaded by the processors and performs the methods described above.
Fourth embodiment
The present embodiment provides a computer-readable storage medium having stored therein at least one instruction that is loaded and executed by a processor to implement the method of the first embodiment. The computer readable storage medium may be, among other things, ROM, random access memory, CD-ROM, magnetic tape, floppy disk, optical data storage device, etc. The instructions stored therein may be loaded by a processor in the terminal and perform the methods described above.
Furthermore, it should be noted that the present invention can be provided as a method, an apparatus, or a computer program product. Accordingly, embodiments of the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, embodiments of the invention may take the form of a computer program product on one or more computer-usable storage media having computer-usable program code embodied therein.
Embodiments of the present invention are described with reference to flowchart illustrations and/or block diagrams of methods, terminal devices (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks. These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
It should also be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or terminal that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or terminal. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article or terminal device comprising the element.
It is finally pointed out that the above description of the preferred embodiments of the invention, it being understood that although preferred embodiments of the invention have been described, it will be obvious to those skilled in the art that, once the basic inventive concepts of the invention are known, several modifications and adaptations can be made without departing from the principles of the invention, and these modifications and adaptations are intended to be within the scope of the invention. It is therefore intended that the following claims be interpreted as including the preferred embodiment and all such alterations and modifications as fall within the scope of the embodiments of the invention.
Claims (9)
1. The intelligent detection method for the leakage of the tailing conveying pipeline is characterized by comprising the following steps of:
s1, collecting pressure data of nodes at the two ends of a tailing conveying pipeline in real time, carrying out data filtering and data standardization on the collected pressure data of the nodes at the two ends, and constructing normal state pressure time sequence characteristic sequence sample data of the conveying pipeline at the two ends;
s2, taking pressure time sequence characteristic sequence sample data in the normal state of the head end and the tail end of the pipeline as a training set, constructing a leakage detection model based on the generation of an countermeasure network, and establishing a pressure fluctuation mode of the normal state of the tailing conveying pipeline, so that a pseudo pressure trend generated based on the leakage detection model of the generation of the countermeasure network is consistent with a real pressure trend in the normal state, and accurate detection and identification of pipeline leakage are realized;
s3, constructing a leakage detection model effectiveness evaluation method, evaluating model identification accuracy by using an area index and an F1-score index under a working characteristic curve of a subject, and evaluating reconstruction quality of a pseudo pressure time sequence characteristic sequence under a normal state by using a root mean square error index.
2. The intelligent detection method for leakage of the tailing conveying pipeline according to claim 1, wherein the step S1 of collecting pressure data of the nodes at the front and the tail ends of the tailing conveying pipeline in real time by using pressure sensors at the front and the tail ends, and constructing pressure time sequence characteristic sequence sample data at the front and the tail ends by combining characteristic engineering methods such as data filtering and data standardization, comprises the following steps:
installing pressure sensors and GPS timing modules at the head end and the tail end of a conveying pipeline, and remotely setting a leakage detection and central monitoring machine to store and analyze data so as to acquire pipeline information of detection point numbers, time codes and pressure;
preprocessing the collected actual measurement pressure data of the pipeline; wherein the preprocessing comprises data noise reduction processing and data standardization processing;
and (3) establishing a time sequence segmentation model based on sampling, so as to ensure that the original pressure trend is unchanged, and respectively constructing pressure time sequence characteristic sequence sample data of the head end and the tail end of the pipeline in the tailing transmission process based on the preprocessed pipeline actual measurement pressure data through the time sequence segmentation model.
3. The intelligent detection method for leakage of the tailing conveying pipeline according to claim 2, wherein the data denoising treatment adopts a wavelet threshold denoising algorithm, and the data standardization treatment adopts a Z-score standardization treatment method.
4. The intelligent detection method for leakage of the tailing conveying pipeline according to claim 2, wherein the construction of pressure time sequence characteristic sequence sample data of the head end and the tail end of the pipeline in the tailing conveying process based on the pretreated pipeline actual measured pressure comprises the following steps:
according to the size of the sampling interval, the original pressure time sequence characteristic sequence is divided into a plurality of blocks, a point is sequentially selected from each block, and finally, the blocks are spliced into a sampling sample with a specific length, so that a plurality of sample data can be obtained from one actually measured sample, and the construction of a pressure time sequence characteristic sequence sample data set for model training is completed.
5. The intelligent detection method for leakage of a tailings conveying pipeline according to claim 1, wherein in the training process of the constructed leakage detection model (GANomaly) based on the generation of an countermeasure network, only the sample data of a normal state pressure time sequence characteristic sequence is used for training, and the leakage sample data is not used, and the leakage detection model comprises the following steps:
a generator network that generates a pseudo timing sequence by encoding-decoding input timing sequence data using an Auto Encoder (AE);
the encoder network is used for obtaining the code representation of the generated pressure time sequence characteristic sequence and is also used as a classifier for leakage detection, and judging whether the current input is leaked or not according to the error between the code representation of the generated sequence and the code representation of the input sequence;
and the discriminator network discriminates the true and false between the input pressure time sequence characteristic sequence and the generated pseudo pressure time sequence characteristic sequence, so as to guide the training of the generator network.
6. The intelligent detection method for leakage of a tailings conveying pipeline according to claim 5, wherein the training construction is based on generating a leakage detection model of an countermeasure network by using the obtained pressure time sequence characteristic sequence sample data in a normal state, and a pressure fluctuation mode of the tailings conveying pipeline in the normal state is established, wherein the pressure time sequence characteristic sequence sample data at the head end and the tail end are respectively used as input for training of generating the countermeasure network.
7. The intelligent detection method for leakage of tailings conveying pipelines according to claim 5, wherein model parameters of a generator network and an encoder network are updated and optimized according to the distinguishing loss obtained by a discriminator network, the coding loss between the input pressure time sequence characteristic sequence sample data and the coding representation of the pseudo pressure time sequence characteristic sequence, the Manhattan distance metric loss between the input pressure time sequence characteristic sequence sample data and the pseudo pressure time sequence characteristic sequence, the three losses are given different weight proportions, and the discriminator network calculates the cross entropy loss to perform self-updating of the parameters according to the pseudo pressure time sequence characteristic sequence newly generated by the generator network.
8. The intelligent detection method for leakage of a tailings conveying pipeline according to claim 1, wherein in the step S4, the accuracy is identified by using an Area index (AUC) Under a subject work characteristic Curve (Receiver Operating Characteristic Curve, ROC) Curve and an F1 fraction (F1-score) index evaluation model, and the difference between an input pressure time sequence characteristic sequence and a generated pseudo pressure time sequence characteristic sequence is evaluated by using a root mean square error index, so that the reconstruction quality of the pseudo pressure time sequence characteristic sequence in a normal state is effectively evaluated.
9. The intelligent detection system for the leakage of the tailing conveying pipeline is characterized by comprising pressure sensors, GPS timing modules and a leakage detection and central monitoring machine, wherein the pressure sensors are arranged at the head end and the tail end;
the pressure sensor is used for collecting measured pressure data of the conveying pipeline in the tailing conveying process;
the GPS timing module is used for acquiring pipeline information comprising a detection point number, a time code number and pressure, and providing necessary data base for realizing construction of pressure time sequence characteristic sequence sample data;
the leakage detection and central monitoring machine is used for receiving the measured pressure data and the pipeline information obtained by the GPS timing module, and integrating the intelligent detection method for the leakage of the tailing conveying pipeline to realize intelligent and accurate detection and identification of the leakage of the tailing conveying pipeline.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311413210.9A CN117554000A (en) | 2023-10-27 | 2023-10-27 | Intelligent detection method and system for leakage of tailing conveying pipeline |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311413210.9A CN117554000A (en) | 2023-10-27 | 2023-10-27 | Intelligent detection method and system for leakage of tailing conveying pipeline |
Publications (1)
Publication Number | Publication Date |
---|---|
CN117554000A true CN117554000A (en) | 2024-02-13 |
Family
ID=89819457
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202311413210.9A Pending CN117554000A (en) | 2023-10-27 | 2023-10-27 | Intelligent detection method and system for leakage of tailing conveying pipeline |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN117554000A (en) |
-
2023
- 2023-10-27 CN CN202311413210.9A patent/CN117554000A/en active Pending
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111914883B (en) | Spindle bearing state evaluation method and device based on deep fusion network | |
CN115293280A (en) | Power equipment system anomaly detection method based on space-time feature segmentation reconstruction | |
CN113822201B (en) | Deep learning method for underwater object shape recognition based on flow field velocity component time course | |
CN111695465A (en) | Pipe network fault diagnosis and positioning method and system based on pressure wave mode identification | |
CN115035913B (en) | Sound abnormity detection method | |
Rosso et al. | Review on deep learning in structural health monitoring | |
CN115184054B (en) | Mechanical equipment semi-supervised fault detection and analysis method, device, terminal and medium | |
CN113987908A (en) | Natural gas pipe network leakage early warning method based on machine learning method | |
CN116805061A (en) | Leakage event judging method based on optical fiber sensing | |
CN113719764A (en) | Pipeline leakage detection method | |
CN117521512A (en) | Bearing residual service life prediction method based on multi-scale Bayesian convolution transducer model | |
CN113901927B (en) | Underwater object shape recognition method based on flow field pressure time course | |
CN117056865B (en) | Method and device for diagnosing operation faults of machine pump equipment based on feature fusion | |
CN114298144A (en) | Hydropower station equipment oil leakage identification method based on deep convolutional neural network | |
CN116541771A (en) | Unbalanced sample bearing fault diagnosis method based on multi-scale feature fusion | |
CN117671480A (en) | Landslide automatic identification method, system and computer equipment based on visual large model | |
CN117454100A (en) | Missing data reconstruction method for engineering structure | |
CN116383723A (en) | Debris flow data anomaly identification method, computer equipment and medium | |
CN115356599B (en) | Multi-mode urban power grid fault diagnosis method and system | |
CN117554000A (en) | Intelligent detection method and system for leakage of tailing conveying pipeline | |
CN110348094A (en) | Petroleum pipeline leakage detection method and system based on influence network | |
CN112836591B (en) | Method for extracting optical fiber early warning signal characteristics of oil gas long-distance pipeline | |
CN109945075A (en) | A kind of water supply line leakiness detection method and device | |
CN114139648B (en) | Intelligent detection method and system for abnormity of tailing filling pipeline | |
CN114199992B (en) | Method and system for detecting corrosion of tank wall of oil storage tank |
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 |