CN116542170A - Drainage pipeline siltation disease dynamic diagnosis method based on SSAE and MLSTM - Google Patents
Drainage pipeline siltation disease dynamic diagnosis method based on SSAE and MLSTM Download PDFInfo
- Publication number
- CN116542170A CN116542170A CN202310375296.4A CN202310375296A CN116542170A CN 116542170 A CN116542170 A CN 116542170A CN 202310375296 A CN202310375296 A CN 202310375296A CN 116542170 A CN116542170 A CN 116542170A
- Authority
- CN
- China
- Prior art keywords
- siltation
- pipeline
- ssae
- mlstm
- disease
- 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
- 201000010099 disease Diseases 0.000 title claims abstract description 33
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 title claims abstract description 33
- 238000000034 method Methods 0.000 title claims abstract description 29
- 238000003745 diagnosis Methods 0.000 title claims abstract description 16
- 238000001514 detection method Methods 0.000 claims abstract description 16
- 230000008859 change Effects 0.000 claims abstract description 11
- 238000013528 artificial neural network Methods 0.000 claims abstract description 10
- 238000007781 pre-processing Methods 0.000 claims abstract description 7
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims abstract description 4
- 238000012549 training Methods 0.000 claims description 34
- 238000012360 testing method Methods 0.000 claims description 27
- 238000012545 processing Methods 0.000 claims description 11
- 238000004062 sedimentation Methods 0.000 claims description 9
- 230000004927 fusion Effects 0.000 claims description 6
- 238000007477 logistic regression Methods 0.000 claims description 5
- 230000009467 reduction Effects 0.000 claims description 4
- 230000000295 complement effect Effects 0.000 claims description 3
- 230000006835 compression Effects 0.000 claims description 3
- 238000007906 compression Methods 0.000 claims description 3
- 238000013135 deep learning Methods 0.000 abstract description 4
- 230000007547 defect Effects 0.000 abstract 1
- 230000006872 improvement Effects 0.000 description 5
- 230000008569 process Effects 0.000 description 4
- 238000012774 diagnostic algorithm Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000009826 distribution Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 238000010200 validation analysis Methods 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 238000012935 Averaging Methods 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 238000002790 cross-validation Methods 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 230000008021 deposition Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000004069 differentiation Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000009776 industrial production Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 239000010865 sewage Substances 0.000 description 1
- 239000010802 sludge Substances 0.000 description 1
- 238000013526 transfer learning Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/28—Design optimisation, verification or simulation using fluid dynamics, e.g. using Navier-Stokes equations or computational fluid dynamics [CFD]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/27—Regression, e.g. linear or logistic regression
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/10—Geometric CAD
- G06F30/18—Network design, e.g. design based on topological or interconnect aspects of utility systems, piping, heating ventilation air conditioning [HVAC] or cabling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
- G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
- G06N3/0455—Auto-encoder networks; Encoder-decoder networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/047—Probabilistic or stochastic networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/088—Non-supervised learning, e.g. competitive learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/09—Supervised learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2113/00—Details relating to the application field
- G06F2113/14—Pipes
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A10/00—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE at coastal zones; at river basins
- Y02A10/40—Controlling or monitoring, e.g. of flood or hurricane; Forecasting, e.g. risk assessment or mapping
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Geometry (AREA)
- Computer Hardware Design (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Probability & Statistics with Applications (AREA)
- Mathematical Analysis (AREA)
- Fluid Mechanics (AREA)
- Algebra (AREA)
- Computational Mathematics (AREA)
- Computer Networks & Wireless Communication (AREA)
- Medical Informatics (AREA)
- Activated Sludge Processes (AREA)
Abstract
The invention discloses a drainage pipeline siltation disease dynamic diagnosis method based on SSAE and MLSTM, which comprises the following steps: carrying out sample acquisition on the data of the fouling degree, the fouling length index and the inlet and outlet flow velocity and flow change index of the pipeline by utilizing fluent, and carrying out data preprocessing; constructing a pipeline siltation detection model based on a stacked sparse self-encoder SSAE and an SVM classifier, extracting characteristics and detecting siltation of relevant parameters of water flow of an underground drainage pipeline, and outputting siltation time; constructing a multi-element long and short memory neural network MLSTM, carrying out siltation disease condition identification on pipeline siltation time data, and outputting a siltation condition category; the invention uses the stack type sparse self-encoder and the deep learning method of the MLSTM model to rapidly and effectively diagnose the dynamic defect of the underground drainage pipeline.
Description
Technical Field
The invention relates to the technical field of artificial intelligence such as deep learning, transfer learning and the like, in particular to a drainage pipeline siltation disease dynamic diagnosis method based on SSAE and MLSTM.
Background
With the continuous expansion of the scale of the infrastructure of China, urban underground drainage pipelines exist in a large quantity, and are one of important urban infrastructure engineering facilities, and are responsible for collecting sewage such as urban life, industrial production and the like and timely draining rainwater falling in urban areas and flowing through urban areas. The quality and scientificity of the design and construction of the urban underground drainage pipe network directly determine the urban development level, influence urban landscapes and sanitation, and even relate to urban safety. Urban underground drainage pipelines are wide in distribution and various in types, and pipeline sludge is deposited after many years of operation, so that the normal operation of the drainage pipelines is seriously influenced.
In the prior art, the drainage pipeline sedimentation disease detection technology is mainly modeled by using a neural network, but the traditional neural network model is easy to sink into problems of local minimum values and the like, and the data such as the flow rate and the flow velocity of a pipeline can be changed to a certain extent in the drainage pipeline sedimentation process, but the change characteristics are not obvious, so that the problems of difficult dynamic diagnosis flow of the pipeline sedimentation disease and the like are easy to cause.
Disclosure of Invention
In order to solve the problems in the prior art, the invention aims to provide a drainage pipeline siltation disease dynamic diagnosis method based on SSAE and MLSTM.
In order to achieve the above purpose, the invention adopts the following technical scheme: a drainage pipeline fouling disease dynamic diagnosis method based on SSAE and MLSTM comprises the following steps:
s1, carrying out sample acquisition on the data of the sedimentation degree and the sedimentation length index of the pipeline, the flow velocity of an inlet and an outlet and the flow change index by utilizing fluent, and carrying out data preprocessing;
s2, constructing a pipeline siltation detection model based on a stacked sparse self-encoder SSAE and an SVM classifier, extracting characteristics and detecting siltation of relevant parameters of water flow of an underground drainage pipeline, and outputting siltation time;
s3, constructing a multi-element long and short memory neural network MLSTM, identifying the fouling disease condition of the pipeline fouling time data, and outputting the type of the fouling condition.
As a further improvement of the present invention, the step S1 specifically includes the steps of:
s11, simulating the flow and flow velocity change condition of a pipeline inlet and outlet under the pipeline siltation condition by utilizing fluent, and collecting a series of flow and flow velocity sample data with siltation degree result labels, wherein a normal sample, a siltation sample and a label-free sample are obtained through sliding window processing;
s12, preprocessing a sample by using a data complement method and 0-1 standardization, and randomly grouping to obtain a training set and a testing set according to K layers of cross tests; the treatment method is as follows:
wherein y is p Is the result of AND processing, y max And y min Is the maximum value and the minimum value of the set AND processing result, x max And x min Maximum and minimum values of raw data.
As a further improvement of the present invention, the step S2 specifically includes the steps of:
s21, performing unsupervised training on a stacked sparse self-encoder SSAE by using all training samples to obtain deep features;
s22, performing supervised training on a stacked sparse self-encoder SSAE and a logistic regression classifier by using a labeled sample to obtain supervised deep features;
s23, performing feature compression dimension reduction on the supervised deep features by using PCA to obtain compressed deep features;
and S24, sending the compressed deep features into an SVM classifier, training, and testing by using a testing set to obtain a final siltation detection network.
As a further improvement of the present invention, the step S3 specifically includes the steps of:
s31, dividing the training set and the testing set according to the LSTM input dimension to obtain sub-time sequences at a plurality of moments, wherein the sub-time sequences are used as inputs of an LSTM network;
s32, training the models of the LSTM and softmax classifiers by using a training set, extracting and fusing time characteristics of the two information samples, and testing by using a testing set;
s33, sending the extracted characteristic full-connection layer into an SVM classifier, classifying the final siltation disease condition, and outputting the siltation condition category.
As a further improvement of the present invention, in step S32, three LSTM subnetworks form a network body, and time features of flow velocity and flow data of the pipeline are extracted respectively; and the flow and flow velocity sequence of the pipeline is divided into subsequences according to a certain step length.
As a further improvement of the present invention, the step S33 is specifically as follows:
the information of the whole time sequence is acquired by taking the output of the last moment as a characteristic through an LSTM network, and is integrated;
before the fusion layer, adding a weight layer, wherein the number of the weight layer units is consistent with that of the previous layer, the units are connected in a one-to-one correspondence manner, and the sizes of the weight layer units are consistent;
and adding a BN layer and a Dropout layer after the fusion layer to solve the problem that a network is easy to be over-fitted due to insufficient data of the siltation disease record.
The principle of BN is to keep each batch of inputs of hidden layers in the neural network equally distributed and finally to perform scale and shift operations on the data. Assume that the samples of a batch are p= { x 1 ,x 2 ,...,x m The algorithm is as follows:
y i =γx i +β
wherein mu p Is the average of the inputs of one batch of hidden layer elements at the next,is the variance of the input of a sample lot at the next hidden layer unit,/for>Is the input normalized value, y i Is the value that is finally manipulated by the equal ratio change and offset.
The addition of BN allows all samples in each small sample set to be correlated together so that the network training results do not bias fit to some samples. When the same sample and different samples form a small sample set, the output of the small sample set is different, and the operation can be understood as data enhancement, so that the problem of over fitting is solved to a certain extent by adding BN.
The basic principle of Dropout is that during forward propagation, neurons operate normally with probability P, i.e. cease to operate with probability (1-P). Through such processing, the overall training process can be seen as training a number of different network structures, with the final classification result being the average of those network classification results. These networks may have over-fitting phenomena, respectively, but since the final results are averaged, the over-fitting phenomena may cancel each other out in the averaging process, so that the over-fitting problem of the overall network may be alleviated to some extent.
The invention combines a stack type sparse self-encoder and a long-short-term memory model, provides an MLSTM (Multi-LSTM) network model based on Multi-element information input, extracts and combines Multi-source information features, adds a weight layer for different correlation degrees of fouling conditions, flow speed and the like, and finally utilizes an SVM classifier to identify pipeline fouling disease conditions.
The beneficial effects of the invention are as follows:
according to the invention, sparse learning and a deep neural network are tightly combined, so that the detection and classification of the siltation disease can be accurately identified, and technical support is provided for the diagnosis of the siltation disease in the underground drainage pipeline scene.
Drawings
FIG. 1 is a flow chart of an embodiment of the present invention;
FIG. 2 is a flow chart of dynamic detection of pipeline fouling disease in an embodiment of the invention;
FIG. 3 is a flow chart of dynamic diagnosis of pipe fouling disease in an embodiment of the invention;
FIG. 4 is a schematic diagram of a basic fault detection network structure based on a stacked sparse self-encoder and a logistic regression classifier in an embodiment of the present invention;
FIG. 5 is a block diagram of a pipeline fouling disease diagnosis network based on an MLSTM in an embodiment of the invention.
Detailed Description
Embodiments of the present invention will be described in detail below with reference to the accompanying drawings.
Examples
As shown in fig. 1, a drainage pipeline flow disease dynamic diagnosis method based on SSAE and MLSTM includes: sample data are collected and preprocessed, a pipeline siltation monitoring model based on a stacked sparse self-encoder and SVM is built, and a data mining network model of a multi-element long and short memory neural network is built. The specific implementation method is as follows:
as shown in fig. 2 and 3, the method for dynamically diagnosing the sedimentation disease of the drainage pipeline based on the stack-type sparse self-encoder and the multi-element long and short memory neural network provided by the invention comprises the following steps:
s1: and (3) performing sample acquisition on indexes such as the fouling degree and the fouling length of the pipeline and the inlet and outlet flow velocity and flow change index data by utilizing fluent, and performing data preprocessing.
S11: the flow and flow velocity sample data with a fouling degree result label are acquired by using fluent to simulate the flow and flow velocity change condition of a pipeline inlet and outlet under the fouling condition of a pipeline, wherein a normal sample, a fouling sample and a label-free sample are acquired through sliding window processing;
s12, preprocessing samples by using a data complement method and 0-1 standardization, and obtaining a training set and a testing set by random grouping according to K-layer cross test (K-fold Cross Validation). The treatment method is as follows:
wherein y is p Is the result of AND processing, y max And y min Is the maximum value and the minimum value of the set AND processing result, x max And x min Maximum and minimum values of raw data. Here, y max And y min Set to 1 and 0. The preprocessed data has the same dimension, and the original internal change condition is not changed.
According to different stages of the deep learning algorithm, the pipeline inlet and outlet flow velocity and flow data sequence is divided into a training set, a verification set and a test set according to the proportion of approximately 3:1:1. The detailed numbers of training sets, validation sets and test sets are shown in table 1.
TABLE 1 training set, validation set and test set distribution
While the running platform of the pipeline fouling diagnostic algorithm is an intelligent flow rate and flow measurement device. The configuration of the device is shown in table 2, and compared with the conventional method, the Tensor RT 5.1.5 accelerator and the adaptive moment estimation (Adam) optimizer in the device improve the training convergence speed and accuracy by 16% and 3%, respectively.
Table 2 deep learning hardware platform configuration
This example designed 4 training set combining schemes as shown in table 3. After a part of sample sets are collected by the full-scale test system, the data set is expanded by utilizing the GAN technology. The combination scheme in table 3 has the best diagnostic effect on the test set in the experiment of the algorithm test set by the model obtained by the third training set. Therefore, the ratio of real data to extended data is 1:2, and the total number of samples is 6000.
Table 3 combined training sample set configuration scheme
S2, constructing a pipeline siltation detection model based on a stack type sparse self-encoder and an SVM, carrying out feature extraction and siltation detection on parameters related to water flow of an underground drainage pipeline, and outputting siltation time; basic fault detection network structures based on a stack-type sparse self-encoder and a logistic regression classifier are shown in fig. 4;
s21: performing unsupervised training on the SSAE by using all training samples (an unlabeled sample and a labeled sample) to obtain deep features;
s22: performing supervised training on the SSAE and the logistic regression classifier by using the labeled sample to obtain supervised deep features;
s23: performing feature compression and dimension reduction on the supervised deep features by using PCA to obtain compressed deep features;
s24: and sending the compressed deep features (with label samples) into an SVM classifier, training, and testing by using a testing set to obtain a final siltation detection network.
S3: an MLSTM (Multi-LSTM) network model is built, the fouling disease condition identification is carried out on the pipeline fouling time data, and the type of the fouling condition is output. The structural diagram of the pipeline fouling disease diagnosis network based on the MLSTM is shown in figure 5.
S31: dividing the training set and the testing set according to the LSTM input dimension to obtain sub-time sequences of a plurality of moments, wherein the sub-time sequences are used as the inputs of an LSTM network;
s32: training the models of a plurality of LSTM and softmax classifiers by using a training set, extracting and fusing time characteristics of two information samples, and testing by using a testing set; the method comprises the following specific substeps:
(1) and the three LSTM subnetworks form a network main body, and the time characteristics of the flow velocity and flow data of the pipeline are extracted respectively.
(2) Dividing the flow and flow velocity sequence of the pipeline into subsequences according to a certain step length.
S33: and (3) sending the extracted features (the full connection layer) into an SVM classifier, classifying the final siltation disease condition, and outputting the siltation condition category. The method comprises the following specific substeps:
(1) and acquiring the information of the whole time sequence by using the output of the last moment as a characteristic through the LSTM network, and integrating the information.
(2) Before the fusion layer, a weight layer is added, the number of the weight layer units is consistent with that of the previous layer, the units are connected in a one-to-one correspondence manner, and the sizes of the weight layer units are consistent. For the same LSTM subnetwork, the weight layer can be regarded as multiplying the characteristics by a weight coefficient, and the weight layer is used for reflecting the importance of the pipeline flow and flow velocity information on the influence of pipeline siltation, so that the differentiation degree of different information is improved, and the fused characteristics are reacted.
(3) And adding a BN layer and a Dropout layer after the fusion layer to solve the problem that a network is easy to be over-fitted due to insufficient data of the siltation disease record.
The principle of BN is to keep each batch of inputs of hidden layers in the neural network equally distributed and finally to perform scale and shift operations on the data. Assume that the samples of a batch are p= { x 1 ,x 2 ,...,x m The algorithm is as follows:
y i =γx i +β
wherein mu p Is the average of the inputs of one batch of hidden layer elements at the next,is the variance of the input of a sample lot at the next hidden layer unit,/for>Is the input normalized value, y i Is the value that is finally manipulated by the equal ratio change and offset.
To verify the effect of the above method, the deposition accuracy under different algorithm configuration parameters is optimized, and the optimizing process is shown in table 4. The present example investigated 6 combined parameter setting scenarios of a fouling diagnostic algorithm. The results showed that the optimal accuracy of the fouling diagnosis was 91.06%. The algorithm optimal momentum, batch size, learning rate, weight decay, and maximum number of iterations are 0.90, 32, 0.0005, 0.0002, and 35000, respectively.
Table 4 different hyper-parameter combination configuration schemes
In this embodiment, a pipeline fouling detection model based on a stacked sparse self-encoder+svm (gaussian kernel) and a multi-element long and short memory neural network+svm is experimentally compared with other self-encoders, and comparison results of different pipeline fouling detection and diagnosis algorithms are shown in table 5. SSAE introduced in this embodiment can compress and extract key features of drain line fouling. Therefore, the accuracy and F1 score of the SSAE-PCA-SVM were 97.18% and 0.935, respectively, superior to the DBN-SVM and the SSAE-SVM. In addition, the dimension reduction processing module of the SSAE-PCA-SVM makes the siltation decision time far shorter than that of the DBN-SVM and the SSAE-SVM. The evaluation index mAP, MAE, RMSE and the average decision time of the MLSTM-SVM are superior to those of the existing BPNN and RNN-SVM algorithms. From this, the effect of this model is superior to other models.
TABLE 5 comparison of Performance of different fouling detection and diagnostic algorithms
The foregoing examples merely illustrate specific embodiments of the invention, which are described in greater detail and are not to be construed as limiting the scope of the invention. It should be noted that it will be apparent to those skilled in the art that several variations and modifications can be made without departing from the spirit of the invention, which are all within the scope of the invention.
Claims (6)
1. A drainage pipeline fouling disease dynamic diagnosis method based on SSAE and MLSTM is characterized by comprising the following steps:
s1, carrying out sample acquisition on the data of the sedimentation degree and the sedimentation length index of the pipeline, the flow velocity of an inlet and an outlet and the flow change index by utilizing fluent, and carrying out data preprocessing;
s2, constructing a pipeline siltation detection model based on a stacked sparse self-encoder SSAE and an SVM classifier, extracting characteristics and detecting siltation of relevant parameters of water flow of an underground drainage pipeline, and outputting siltation time;
s3, constructing a multi-element long and short memory neural network MLSTM, identifying the fouling disease condition of the pipeline fouling time data, and outputting the type of the fouling condition.
2. The method for dynamically diagnosing a fouling disease of a drain pipe based on SSAE and MLSTM according to claim 1, wherein said step S1 specifically comprises the steps of:
s11, simulating the flow and flow velocity change condition of a pipeline inlet and outlet under the pipeline siltation condition by utilizing fluent, and collecting a series of flow and flow velocity sample data with siltation degree result labels, wherein a normal sample, a siltation sample and a label-free sample are obtained through sliding window processing;
s12, preprocessing a sample by using a data complement method and 0-1 standardization, and randomly grouping to obtain a training set and a testing set according to K layers of cross tests; the treatment method is as follows:
wherein y is p Is the result of AND processing, y max And y min Is the maximum value and the minimum value of the set AND processing result, x max And x min Maximum and minimum values of raw data.
3. The method for dynamically diagnosing a fouling disease of a drain pipe based on SSAE and MLSTM according to claim 2, wherein said step S2 specifically comprises the steps of:
s21, performing unsupervised training on a stacked sparse self-encoder SSAE by using all training samples to obtain deep features;
s22, performing supervised training on a stacked sparse self-encoder SSAE and a logistic regression classifier by using a labeled sample to obtain supervised deep features;
s23, performing feature compression dimension reduction on the supervised deep features by using PCA to obtain compressed deep features;
and S24, sending the compressed deep features into an SVM classifier, training, and testing by using a testing set to obtain a final siltation detection network.
4. The method for dynamically diagnosing a fouling disease of a drain pipe based on SSAE and MLSTM according to claim 3, wherein said step S3 specifically comprises the steps of:
s31, dividing the training set and the testing set according to the LSTM input dimension to obtain sub-time sequences at a plurality of moments, wherein the sub-time sequences are used as inputs of an LSTM network;
s32, training the models of the LSTM and softmax classifiers by using a training set, extracting and fusing time characteristics of the two information samples, and testing by using a testing set;
s33, sending the extracted characteristic full-connection layer into an SVM classifier, classifying the final siltation disease condition, and outputting the siltation condition category.
5. The dynamic diagnosis method for sedimentation disease of drainage pipeline based on SSAE and MLSTM according to claim 4, wherein in step S32, three LSTM sub-networks form a network main body, and the time characteristics of the flow velocity and flow data of the pipeline are extracted respectively; and the flow and flow velocity sequence of the pipeline is divided into subsequences according to a certain step length.
6. The method for dynamically diagnosing a fouling disease of a drain pipe based on SSAE and MLSTM as set forth in claim 5, wherein said step S33 is specifically as follows:
the information of the whole time sequence is acquired by taking the output of the last moment as a characteristic through an LSTM network, and is integrated;
before the fusion layer, adding a weight layer, wherein the number of the weight layer units is consistent with that of the previous layer, the units are connected in a one-to-one correspondence manner, and the sizes of the weight layer units are consistent;
and adding a BN layer and a Dropout layer after the fusion layer to solve the problem that a network is easy to be over-fitted due to insufficient data of the siltation disease record.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310375296.4A CN116542170A (en) | 2023-04-10 | 2023-04-10 | Drainage pipeline siltation disease dynamic diagnosis method based on SSAE and MLSTM |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310375296.4A CN116542170A (en) | 2023-04-10 | 2023-04-10 | Drainage pipeline siltation disease dynamic diagnosis method based on SSAE and MLSTM |
Publications (1)
Publication Number | Publication Date |
---|---|
CN116542170A true CN116542170A (en) | 2023-08-04 |
Family
ID=87455137
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202310375296.4A Pending CN116542170A (en) | 2023-04-10 | 2023-04-10 | Drainage pipeline siltation disease dynamic diagnosis method based on SSAE and MLSTM |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116542170A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117314900A (en) * | 2023-11-28 | 2023-12-29 | 诺比侃人工智能科技(成都)股份有限公司 | Semi-self-supervision feature matching defect detection method |
CN117332722A (en) * | 2023-11-23 | 2024-01-02 | 江西省水利科学院(江西省大坝安全管理中心、江西省水资源管理中心) | Intelligent drainage pipeline siltation diagnosis method based on hard constraint projection |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106124212A (en) * | 2016-06-16 | 2016-11-16 | 燕山大学 | Based on sparse coding device and the Fault Diagnosis of Roller Bearings of support vector machine |
CN113486291A (en) * | 2021-06-18 | 2021-10-08 | 电子科技大学 | Petroleum drilling machine micro-grid fault prediction method based on deep learning |
CN113743537A (en) * | 2021-09-26 | 2021-12-03 | 东南大学 | Deep sparse memory model-based highway electromechanical system fault classification method |
CN115712817A (en) * | 2022-10-27 | 2023-02-24 | 华苏数联科技有限公司 | Fault diagnosis method of industrial motor pump based on convolutional neural network |
-
2023
- 2023-04-10 CN CN202310375296.4A patent/CN116542170A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106124212A (en) * | 2016-06-16 | 2016-11-16 | 燕山大学 | Based on sparse coding device and the Fault Diagnosis of Roller Bearings of support vector machine |
CN113486291A (en) * | 2021-06-18 | 2021-10-08 | 电子科技大学 | Petroleum drilling machine micro-grid fault prediction method based on deep learning |
CN113743537A (en) * | 2021-09-26 | 2021-12-03 | 东南大学 | Deep sparse memory model-based highway electromechanical system fault classification method |
CN115712817A (en) * | 2022-10-27 | 2023-02-24 | 华苏数联科技有限公司 | Fault diagnosis method of industrial motor pump based on convolutional neural network |
Non-Patent Citations (2)
Title |
---|
DANYANG DI: "An automatic and integrated self-diagnosing system for the silting disease of drainage pipelines based on SSAE-TSNE and MS-LSTM", 《TUNNELLING AND UNDERGROUND SPACE TECHNOLOGY》, vol. 136, 9 March 2023 (2023-03-09), pages 1 - 16, XP087292392, DOI: 10.1016/j.tust.2023.105076 * |
王毅星: "基于深度学习和迁移学习的电力数据挖掘技术研究", 《中国博士学位论文全文数据库 工程科技Ⅱ辑》, no. 8, 15 August 2019 (2019-08-15), pages 042 - 1 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117332722A (en) * | 2023-11-23 | 2024-01-02 | 江西省水利科学院(江西省大坝安全管理中心、江西省水资源管理中心) | Intelligent drainage pipeline siltation diagnosis method based on hard constraint projection |
CN117332722B (en) * | 2023-11-23 | 2024-02-23 | 江西省水利科学院(江西省大坝安全管理中心、江西省水资源管理中心) | Intelligent drainage pipeline siltation diagnosis method based on hard constraint projection |
CN117314900A (en) * | 2023-11-28 | 2023-12-29 | 诺比侃人工智能科技(成都)股份有限公司 | Semi-self-supervision feature matching defect detection method |
CN117314900B (en) * | 2023-11-28 | 2024-03-01 | 诺比侃人工智能科技(成都)股份有限公司 | Semi-self-supervision feature matching defect detection method |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US11507049B2 (en) | Method for detecting abnormity in unsupervised industrial system based on deep transfer learning | |
CN111709448B (en) | Mechanical fault diagnosis method based on migration relation network | |
CN116542170A (en) | Drainage pipeline siltation disease dynamic diagnosis method based on SSAE and MLSTM | |
CN110728654B (en) | Automatic pipeline detection and classification method based on deep residual error neural network | |
WO2023109370A1 (en) | Fault detection method and equipment for drainage pipe network, server, and storage medium | |
CN111273623B (en) | Fault diagnosis method based on Stacked LSTM | |
CN114358123B (en) | Generalized open set fault diagnosis method based on deep countermeasure migration network | |
CN109086926B (en) | Short-time rail transit passenger flow prediction method based on combined neural network structure | |
CN109491338A (en) | A kind of relevant method for diagnosing faults of multimode procedure quality based on sparse GMM | |
CN113505655A (en) | Bearing fault intelligent diagnosis method for digital twin system | |
CN112147432A (en) | BiLSTM module based on attention mechanism, transformer state diagnosis method and system | |
CN112949189A (en) | Modeling method for multi-factor induced landslide prediction based on deep learning | |
CN115906949B (en) | Petroleum pipeline fault diagnosis method and system, storage medium and petroleum pipeline fault diagnosis equipment | |
CN113032917A (en) | Electromechanical bearing fault detection method based on generation countermeasure and convolution cyclic neural network and application system | |
CN105241665A (en) | Rolling bearing fault diagnosis method based on IRBFNN-AdaBoost classifier | |
CN113719764A (en) | Pipeline leakage detection method | |
Li et al. | Intelligent fault diagnosis of aeroengine sensors using improved pattern gradient spectrum entropy | |
CN113450562B (en) | Road network traffic state discrimination method based on clustering and graph convolution network | |
CN104537383A (en) | Massive organizational structure data classification method and system based on particle swarm | |
CN113343123A (en) | Training method and detection method for generating confrontation multiple relation graph network | |
CN117574262A (en) | Underwater sound signal classification method, system and medium for small sample problem | |
CN114817856B (en) | Beam-pumping unit fault diagnosis method based on structural information retention domain adaptation network | |
CN115659258B (en) | Power distribution network fault detection method based on multi-scale graph roll-up twin network | |
CN116484219A (en) | Water supply network water quality abnormal pollution source identification method based on gate control graph neural network | |
CN113970073B (en) | ResNet-based water supply network leakage accurate positioning method |
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 |