CN116663213A - Ocean platform flow pipeline sensor layout optimization method, ocean platform flow pipeline sensor layout optimization system and storage medium - Google Patents

Ocean platform flow pipeline sensor layout optimization method, ocean platform flow pipeline sensor layout optimization system and storage medium Download PDF

Info

Publication number
CN116663213A
CN116663213A CN202310631967.9A CN202310631967A CN116663213A CN 116663213 A CN116663213 A CN 116663213A CN 202310631967 A CN202310631967 A CN 202310631967A CN 116663213 A CN116663213 A CN 116663213A
Authority
CN
China
Prior art keywords
damage
pipeline
data
sensor layout
sensor
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202310631967.9A
Other languages
Chinese (zh)
Inventor
吴磊
陈迩齐
徐世霖
肖文生
刘超
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
China University of Petroleum East China
Original Assignee
China University of Petroleum East China
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by China University of Petroleum East China filed Critical China University of Petroleum East China
Priority to CN202310631967.9A priority Critical patent/CN116663213A/en
Publication of CN116663213A publication Critical patent/CN116663213A/en
Pending legal-status Critical Current

Links

Classifications

    • 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
    • G06F18/00—Pattern recognition
    • G06F18/10—Pre-processing; Data cleansing
    • G06F18/15—Statistical pre-processing, e.g. techniques for normalisation or restoring missing data
    • 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
    • G06F18/2131—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on a transform domain processing, e.g. wavelet transform
    • 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
    • G06F30/23—Design optimisation, verification or simulation using finite element methods [FEM] or finite difference methods [FDM]
    • 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
    • G06N3/004—Artificial life, i.e. computing arrangements simulating life
    • G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
    • 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
    • G06N3/0464—Convolutional networks [CNN, ConvNet]
    • 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
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2111/00—Details relating to CAD techniques
    • G06F2111/10—Numerical modelling
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2113/00—Details relating to the application field
    • G06F2113/14—Pipes
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2123/00—Data types
    • G06F2123/02—Data types in the time domain, e.g. time-series data
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/02—Preprocessing
    • G06F2218/04—Denoising
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/08—Feature extraction
    • 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
    • Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
    • Y02T90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Software Systems (AREA)
  • Biomedical Technology (AREA)
  • General Health & Medical Sciences (AREA)
  • Mathematical Physics (AREA)
  • Computing Systems (AREA)
  • Molecular Biology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Health & Medical Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Geometry (AREA)
  • Computer Hardware Design (AREA)
  • Probability & Statistics with Applications (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Mathematical Analysis (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Computational Mathematics (AREA)
  • Medical Informatics (AREA)
  • Mathematical Optimization (AREA)
  • Pure & Applied Mathematics (AREA)
  • Pipeline Systems (AREA)

Abstract

The application relates to a distribution optimizing method, a system and a storage medium of an ocean platform flow transmission pipeline sensor, wherein the optimizing method comprises the following steps: acquiring parameter information of a marine platform flow pipeline, constructing a finite element simulation model, acquiring strain mode data, acquiring the strain mode data, and constructing a training data set; performing noise simulation, data interpolation and linear normalization on the training data set; performing time-frequency analysis on the continuous sequence data set by utilizing wavelet packet analysis to obtain a damage characteristic information set of strain modal data of a sensor corresponding to the ocean platform flow transmission pipeline; establishing a damage characteristic-damage position mapping relation set through a convolutional neural network; and an improved whale optimization algorithm is adopted to optimally adjust the sensor layout positions. Compared with the prior art, the application considers the sensor layout scheme and the pipeline damage identification coupling, effectively improves the sensor layout efficiency and the damage identification accuracy and precision, and improves the unmanned and intelligent level of the health monitoring of the active ocean platform.

Description

Ocean platform flow pipeline sensor layout optimization method, ocean platform flow pipeline sensor layout optimization system and storage medium
Technical Field
The application relates to the field of machinery and ocean engineering, in particular to a method, a system and a storage medium for optimizing layout of a sensor of an ocean platform flow delivery pipeline.
Background
On ocean platforms, the flow pipeline plays an important role. They are responsible for transporting fluid media such as water, oil, gas, etc. Due to the particularity of the working environment of the ocean engineering, the high temperature, high humidity and corrosiveness of the ocean environment are faced, and the pipeline is extremely easy to damage, so that the leakage of the pipeline content is caused, and the normal service of the ocean engineering and the life safety of workers are threatened. At present, data such as pipeline acceleration, stress strain, vibration frequency and the like can be acquired by arranging sensors on the surface of a pipeline, and the data is compared with the data under the condition that the pipeline is not damaged through modal identification. The method has two defects, namely, the sensor layout scheme is mostly based on experience, and the data in an undamaged state cannot be collected on an active platform. This brings great difficulty to the damage recognition accuracy of ocean platform flow pipeline.
CN114857504a discloses a pipeline safety monitoring method based on a distributed optical fiber sensor and deep learning, which comprises the following steps: s1, collecting optical fiber vibration signals of all positions of a pipeline through a distributed optical fiber signal collecting device; s2, performing data cleaning, wavelet packet noise reduction and normalization processing on the collected optical fiber vibration signals, and dividing the processed data into a training set, a testing set and a prediction set; s3, constructing a convolutional neural network model, inputting training set and test set data for learning training, inputting prediction set data for evaluating the performance of the model, and then taking the convolutional neural network as a characteristic extractor and a support vector machine as a classification selector to establish a convolutional neural network and support vector machine joint model; s4, inputting the collected real-time data into a convolutional neural network and a support vector machine joint model for identification, and classifying types according to the output digital label to realize safety monitoring of the pipeline. Although the accuracy of pipeline safety detection is improved by deep learning, the current ocean platform flow pipeline sensor layout is difficult to optimize, technical guidance is still lacking, and research and development personnel are still required to further solve the problem.
Disclosure of Invention
The application aims to overcome the defects of the prior art and provide a method, a system and a storage medium for optimizing the layout of the ocean platform flow transmission pipeline sensor.
The aim of the application can be achieved by the following technical scheme:
the application provides a method for optimizing layout of sensors of a flow transmission pipeline of an ocean platform, which comprises the following steps:
s1: acquiring parameter information of a ocean platform flow transmission pipeline, constructing a finite element simulation model, constructing pipeline damage information in the finite element simulation model, setting a data acquisition starting point, acquiring strain mode data on an axial path of the pipeline, acquiring strain mode data on each sensor position based on an original layout scheme of the sensor, and constructing a training data set by taking the pipeline damage position as a label of the data;
s2: performing noise simulation, data interpolation and linear normalization on the training data set to obtain a preprocessed continuous sequence data set;
s3: performing time-frequency analysis on the continuous sequence data set by utilizing wavelet packet analysis to obtain a damage characteristic information set of sensor strain modal data corresponding to an ocean platform flow transmission pipeline;
s4: establishing a damage characteristic-damage position mapping relation set through a convolutional neural network based on the damage characteristic information set;
s5: and based on the damage characteristic-damage position mapping relation set, adopting an improved whale optimization algorithm to optimally adjust the sensor layout positions.
Further, in S1, the parameter information of the ocean platform flow pipeline includes length, outer diameter, inner diameter, fluid, material, and surrounding working environment parameters.
Further, in S1, the process of constructing the pipe damage information in the finite element simulation model includes:
and (3) constructing pipeline damage: obtaining pipeline damage with the length of L mm, the width of W mm and the depth of H mm at the inner wall of the pipeline simulation model through rotary cutting;
obtaining the position mark of the damage: and obtaining the distance X mm between one end point of the damage and the end part of the pipeline corresponding to the end point, and taking the X mm as the position mark of the damage.
Further, in S1, the process of constructing the training data set includes:
setting a Y mm position at one end of the ocean platform flow transmission pipeline simulation model as a data acquisition initial point, setting a (L-30) mm position at one end of the ocean platform flow transmission pipeline simulation model as a data acquisition end point, obtaining strain mode data on an axial path of the pipeline through finite element calculation, then taking out the strain mode data at each sensor layout position according to a sensor layout scheme, namely single training data, and taking a pipeline damage position Xmm as a label of the data.
Further, in the step S2, in the noise simulation process, white noise is doped into the simulated strain mode data to simulate the generation of noise in the actual engineering;
wherein, the generation of white noise is realized by adopting random numbers based on standard normal distribution:
x noise ＝x+normal(0，1) (1)
wherein normal (0, 1) is a standard normal distribution, x is input data, x noise Data after adding noise;
in the data interpolation process, three times of data interpolation are carried out on the basis of the original data characteristics;
in the linear normalization process, the value range of each order strain mode is mapped between [0,1], the dimension of the real sensor strain mode data of any order is unified, and the normalization formula is as follows:
wherein min (x) is a minimum function and max (x) is a maximum function.
Further, in S3, the process of wavelet packet analysis includes:
the iteration number of each data in the continuous sequence data set is num iter Is decomposed by the dichotomy of (5) when num is iter In the sub-decomposition, the frequency domain is divided equally intoEach node corresponds to a frequency band, each frequency band contains frequency domain characteristic information of an original signal in the range, and the wavelet packet analysis process is shown in the following formula:
in the formula ,p(i，j) The frequency domain coefficient of the jth node of the ith iteration is a group of wavelet basis functions;
the wavelet packet analysis results in coefficient vector p of all nodes of the last iteration (last，j) The definition is as follows:
p (last，j) ＝[fre 1 ，Fre 2 ，...，fre len ]
wherein last is the last iteration, j is the j node, fre is the frequency characteristic, and len is the number of the frequency characteristics;
coefficient vector p based on all nodes (last，j) And constructing and obtaining a damage characteristic information set of the sensor strain modal data corresponding to the ocean platform flow transmission pipeline.
Further, in S4, in the process of establishing the damage characteristic-damage location mapping relation set through the convolutional neural network:
stacking a two-dimensional matrix formed by a series of damage features into a three-dimensional space in a coefficient stacking mode, wherein the stacking height is the number of channels of the input convolutional neural network;
in the stacking process, the number of stacking layers is set as fold, and the specific process is as follows:
wherein ,representing the frequency characteristics of all nodes after the last iteration in wavelet packet analysis, +.>Represented as a length num res Two-dimensional matrix of width len, +.>Representing folded frequenciesThe rate characteristic matrix is used for the data processing,represents a high fold, long (num res Fold), a three-dimensional matrix of width len,
wherein → is expressed as a folding process, the implementation mode is that an open source library numpy.reshape function is called, and the specific process is that
The convolutional neural network is divided into an input layer, an implicit layer and an output layer;
the output layer is a full-connection layer and a LogSoftmax layer, the hidden layer structure is three layers, each layer structure is a convolution layer-BN layer-activation layer-pooling layer-regularization, the convolution kernel size is (5, 5), the pooling layer size is 2, the activation layer function is a ReLU () function, the regularization ratio is 0.1, the optimizer adopts a RMSprop function, the damage function adopts a cross entropy CrossEntrop function, the initial learning rate is set to 0.05, the learning rate is reduced by 75% every 5 iterations, the maximum iteration times is set to 200, and a damage characteristic-damage position mapping relation set is established through a convolution neural network.
Further, in S4, the process of optimally adjusting the sensor layout position includes:
the damage identification and positioning accuracy is taken as an objective function, and the sensor layout quantity and the sensor layout interval are taken as optimization variables, as follows:
var＝[num，inter]
wherein uar is an optimization variable, num is the number of sensors to be arranged, and inter is the distance between the sensors to be arranged;
then, using whale algorithm to find the optimized variable which makes the damage identification and positioning accuracy highest, comprising the following steps:
initializing global variables of a whale algorithm, determining a specific evolution strategy executed by a current whale individual according to a uniform random number r after iteration, executing an impact search strategy when r is less than or equal to 0.5, and executing a spiral search strategy when r is more than 0.5;
in the direct search strategy, determining whale individuals to be searched directly according to the following formula;
wherein ,a random number with a value range of 0 to 1;
when (when)When in use, a whale individual is randomly selected from the population for searching, and the specific formula is as follows:
wherein ,for randomly selected whale individuals;
when (when)When the current global optimal individual is selected as the object to execute the spiral search strategy, the specific formula is as follows：
wherein ,is the globally optimal individual;
the spiral search strategy is specifically as follows:
after the current whale individual executes the evolution strategy, checking whether the offspring crosses the boundary and repairing the variable crossing the boundary;
when the sensor layout quantity exceeds the variable interval, repairing the out-of-range sensor layout quantity by adopting the following process:
firstly, determining a specific repair strategy executed by a current whale individual according to a random decimal p, when p is less than or equal to 0.5, selecting the sensor layout number of the current optimal individual to replace the current out-of-range sensor layout number, and when p is more than 0.5, selecting the sensor layout number of the global optimal individual to replace the current out-of-range sensor layout number;
then, randomly selecting the sensor layout number of one whale individual in the population, and randomly perturbing the sensor layout number of the offspring of the current iteration individual when the sensor layout number of the random individual is equal to the sensor layout number of the global optimal individual, wherein the specific formula is as follows:
else：num＝num+round(normal(0，1))
wherein ,for globally optimal individuals, ++>For the currently best individual, round () is a rounding function, normal (0, 1) is a standard normal distribution;
finally, after each whale individual executes the evolution strategy, executing the global elite strategy, comparing the offspring of the current iteration individual with their parents, and comparing the offspring of the current iteration individual with the historical optimal individual and the current group optimal individual, so as to keep the optimal individual used by each whale iteration instant, and further accelerate the group convergence speed;
and after the iteration is finished, taking the optimal variable of the historical optimal individual as the optimal sensor layout parameter.
The application provides a distribution optimizing system for ocean platform flow transmission pipeline sensors, which comprises a simulation and data construction acquisition module, a data post-processing module, a characteristic analysis module, a convolution calculation module and an optimizing calculation module, wherein the simulation and data construction acquisition module comprises the following steps of:
the simulation and data construction acquisition module acquires parameter information of a ocean platform flow transmission pipeline, a finite element simulation model is constructed, pipeline damage information is constructed in the finite element simulation model, a data acquisition starting point is set, strain mode data on an axial path of the pipeline is acquired, strain mode data on each sensor position is acquired based on an original layout scheme of the sensor, and the pipeline damage position is used as a label of the data, so that a training data set is constructed;
the data post-processing module performs noise simulation, data interpolation and linear normalization on the training data set to obtain a preprocessed continuous sequence data set;
the characteristic analysis module performs time-frequency analysis on the continuous sequence data set by utilizing wavelet packet analysis to obtain a damage characteristic information set of sensor strain modal data corresponding to the ocean platform flow transmission pipeline;
the convolution calculation module establishes a damage characteristic-damage position mapping relation set through a convolution neural network based on the damage characteristic information set;
and the optimization calculation module is used for optimizing and adjusting the sensor layout positions based on the damage characteristic-damage position mapping relation set and by adopting an improved whale optimization algorithm.
A third aspect of the present application provides a storage medium containing computer executable instructions which when executed by a computer processor are used to perform the above-described ocean platform flow pipeline sensor layout optimization method.
Compared with the prior art, the application has the following technical advantages:
1) The application provides a new method for repairing out-of-range variables. The conventional variable patching method randomly selects or replaces a variable with a boundary of a variable interval. In the aspect of sensor layout optimization of ocean platform flow transmission pipelines, the traditional variable repair method is inapplicable to the sensor layout quantity and the sensor layout interval, and if the sensor layout quantity and the sensor layout interval are reset on a certain boundary, the actual engineering requirements are definitely not met; if the number of sensor arrangements and the sensor arrangement spacing are randomly reset, the continuous optimization of the existing optimization scheme can be influenced. Therefore, the application provides a new variable resetting method according to the characteristics of the distribution optimization problem of the ocean platform flow transmission pipeline sensor. Aiming at the characteristics that the sensor layout number and the layout spacing in the optimization problem show obvious local search characteristics, the method enables the layout number and the layout spacing to be searched randomly in a certain interval around a history optimal individual meeting the requirements according to a certain strategy, and therefore the sensor layout optimization effect is improved.
2) Aiming at the problem of too slow convergence of an original whale algorithm, the application provides a novel elite method. The method expands the comparison of the current individuals to the historical global optimal individuals, and ensures that each iterative individual can learn the optimization information carried by the historical global optimal individuals in real time by comparing the current whale individuals with the historical global optimal individuals in real time. Compared with the traditional elite method, the method can effectively improve the convergence speed and the optimization efficiency of the optimization algorithm, and is beneficial to saving the optimization time when the population quantity is set to be large.
3) Aiming at the problems that the interval between the sensor layout number and the sensor layout interval is too large, the numerical types are different, and the optimization effect is affected, the application provides a variable isolation strategy method, wherein the sensor layout number and the sensor layout interval are isolated in the iteration process, the two are not interfered with each other in the process of executing the evolution strategy, and the position information of the sensor layout number and the sensor layout interval is kept so as to be transmitted to next generation whale individuals. If the sensor layout number and the sensor layout interval are interacted through an evolutionary strategy, the layout number and the layout interval can quickly reach an optimization boundary, and optimization information of the previous generation cannot be well transmitted to subsequent whale individuals.
4) The wavelet packet analysis folding technology can effectively improve convolutional neural network. The traditional method is to input the wavelet packet analysis result into the convolutional neural network as a two-dimensional matrix, and the convolutional neural network can only sense the frequency characteristic information on adjacent frequencies and adjacent coordinates, so that the convolutional neural network has smaller receptive field. Therefore, the application provides a wavelet packet analysis folding technology, which folds the frequency characteristics of all nodes of wavelet packet analysis into a three-dimensional matrix in equal proportion, so that the convolutional neural network can feel the frequency characteristics of more nodes, the receptive field of the convolutional neural network is improved, and the damage function of the convolutional neural network is helped to be reduced more quickly.
Drawings
Fig. 1 is a schematic diagram of wavelet packet analysis principle in an embodiment of the present application.
Detailed Description
In specific implementation, the ocean platform flow delivery pipeline sensor layout optimization method adopts the following technical scheme that the method comprises the following steps:
step one, model simulation
For ocean platform flow pipelines needing to be provided with sensors, parameters such as length, outer diameter, inner diameter, fluid, materials, surrounding working environments and the like of the ocean platform flow pipelines are collected, and a simulation model is built in finite element software. And obtaining the pipeline damage with the length of Lmm, the width of Wmm and the depth of Hmm at the inner wall of the pipeline simulation model by rotary cutting, wherein the position of the damage is marked by the distance Xmm between the left end of the damage and the left end of the pipeline. Because the front end and the rear end of the pipeline are arranged as fixed supports in the finite element calculation model, the strain mode calculation result of the front end and the rear end can generate larger vibration which is inconsistent with the actual situation, the position of 30mm at the left end of the ocean platform flow pipeline simulation model is set as a data acquisition initial point, and the position of (L-30) mm at the left end of the ocean platform flow pipeline simulation model is set as a data acquisition end point. And calculating strain modal data on the axial path of the pipeline by finite element calculation software. And then according to the sensor layout scheme, strain mode data at each sensor layout position is taken out according to the position, namely single training data. The location of the pipe damage Xmm was used as a label for this data.
Step two, data preprocessing
(1) And (5) noise simulation. In consideration of the fact that data acquired by a sensor in actual engineering can be interfered by noise, white noise is required to be doped into sensor simulation strain mode data to simulate noise generation in the actual engineering. The application simulates the generation of white noise by adopting random numbers based on standard normal distribution. The formula is as follows:
x noise ＝x+normal(0，1) (1)
wherein normal (0, 1) is a standard normal distribution, x is input data, x noise After adding noiseData. Notably, the random noise varies from one discrete data point to another of the sensor simulated strain modal data herein.
(2) Cubic spline interpolation. Because the number of the sensors is usually small, the data length of the training set is short, and the too short data can cause difficulty in extracting damage characteristics, so that the damage identification accuracy of the neural network is affected. Therefore, here, cubic spline interpolation is employed to interpolate data on the basis of ensuring the original data characteristics as much as possible.
Since the cubic spline interpolation method is a mature interpolation mode with wide open source, and the method is not a main innovation point of the application, the specific mathematical process is not repeated here. The specific procedure uses the interpolate class in the python. Scipy library.
(3) And (5) linear normalization. Because the value ranges of the strain modes of all the orders are different, the value ranges of the strain modes of all the orders need to be mapped to between [0,1], and the dimension of the real sensor strain mode data of any order is unified. The normalization formula is as follows:
wherein min (x) is a minimum function and max (x) is a maximum function.
Step three, wavelet packet analysis
The data after the noise simulation-cubic spline interpolation-normalization process is regarded as a set of continuous sequence data, which is subjected to time-frequency analysis by wavelet packet analysis. The wavelet packet analysis is an iterative process, and each iteration wavelet packet analysis can convert input data into low-frequency coefficients and high-frequency coefficients of the input data, wherein the current low-frequency coefficients and high-frequency coefficients are the input data of the next wavelet packet analysis. Therefore, the number of iterations of the wavelet packet analysis needs to be determined according to the specific embodiment, and the number of results of the wavelet packet analysis is determined by the following formula:
wherein ,numres Num is the number of wavelet packet analysis results iter Is the number of iterations of the wavelet packet.
An iterative schematic of a wavelet packet is shown in fig. 1. Taking a 4-iteration wavelet packet analysis as an example in fig. 1, it is clear that the essence of the wavelet packet transformation is to iterate the original signal for a number of num iter Is decomposed by the dichotomy of (5) when num is iter In the sub-decomposition, the frequency domain is divided equally intoEach node corresponds to a frequency band, and each frequency band contains frequency domain characteristic information of the original signal in the range. The wavelet packet analysis process is shown in the following formula:
in the formula ,p(i，j) And the frequency domain coefficient of the jth node of the ith iteration is H and G which are a group of wavelet basis functions. The choice of wavelet basis functions will depend on the particular implementation.
The result of the wavelet packet analysis is the coefficient vector p for all nodes of the last iteration (last，j) The definition is as follows:
p (last，j) ＝[fre 1 ，fre 2 ，...，fre len ]
wherein last is the last iteration, j is the j-th node, fre is the frequency characteristic, and len is the frequency characteristic number. After the wavelet packet analysis result is obtained, stacking is needed, the stacking layer number is set as fold, and the specific process is as follows:
wherein ,representing the frequency characteristics of all nodes after the last iteration, < >>Indicating that this is a length num res A two-dimensional matrix of width len; />Representing the folded frequency characteristic matrix, +.>Represents a high fold, long (num res Fold), a three-dimensional matrix of width len. This is a folding procedure, which is implemented by calling the open source library numpy>
Step four, pipeline damage identification based on convolutional neural network
And identifying and positioning damage of the ocean platform flow delivery pipeline based on the convolutional neural network. After wavelet packet analysis extracts a series of damage characteristics of real sensor strain modal data of the ocean platform flow transmission pipeline, a mapping relation between the damage characteristics and damage positions is established through a convolutional neural network. The application provides a coefficient stacking technology, wherein a two-dimensional matrix formed by a series of damage features is stacked into a three-dimensional space, and the stacking height is the number of channels of an input convolutional neural network. The convolutional neural network is divided into an input layer, an implicit layer and an output layer. The output layer is a full connection layer and a LogSoftmax layer and is used for obtaining the probability of damage to each position of the pipeline. The hidden layer structure is three layers, each layer structure is a convolution layer-BN layer-activation layer-pooling layer-regularization, the convolution kernel size is (5, 5), the pooling layer size is 2, the activation layer function is a ReLU () function, the regularization proportion is 0.1, the optimizer adopts a RMSprop function, the damage function adopts a cross entropy Entrop function, the initial learning rate is set to 0.05, the learning rate is reduced by 75% every 5 iterations, and the maximum iteration number is set to 200.
Step five, optimizing sensor layout positions based on improved whale optimization algorithm
Based on the identification and positioning of the damage of the ocean platform flow delivery pipeline of a certain group of sensor layout schemes can be obtained through the previous steps, the optimization of the sensor layout positions is carried out by improving a whale optimization algorithm. The damage identification and positioning accuracy is taken as an objective function, and the sensor layout quantity and the sensor layout interval are taken as optimization variables, as follows:
var＝[num，inter]
wherein var is an optimization variable, num is the number of sensors to be laid, and inter is the distance between the sensors to be laid. Then, using whale algorithm to find the optimized variable which makes the damage identification and positioning accuracy rate highest. First, global variables of the whale algorithm, such as whale number, variable interval, maximum number of iterations, parameters b and l, are initialized. After iteration, determining a specific evolution strategy executed by the current whale individual according to a uniform random number r, executing an impact search strategy when r is less than or equal to 0.5, and executing a spiral search strategy when r is more than 0.5. In the direct search strategy, whale individuals for direct search were determined according to the following formula.
wherein ,is a random number with a value range of 0 to 1. Notably, here +.>And->Are vectors, meaning that different dimensions of the current whale individual may have different search objects. When->When in use, a whale individual is randomly selected from the population for searching, and the specific formula is as follows:
wherein ,is a randomly selected whale individual. When->When the method is used, the current global optimal individual is selected as an object to execute a direct search strategy, and the specific formula is as follows:
wherein ,is the globally optimal individual. The spiral search strategy is specifically as follows:
after the current whale individuals have performed the evolution strategy, they need to check if their offspring cross the boundary and repair the out-of-boundary variables. The application provides a novel out-of-range repair method to better accord with the optimization of the layout of the ocean platform flow transmission pipeline sensor. The number of the sensors is an integer, and the number of the sensors is a floating point number, so that the two sensors are required to be isolated, and a targeted out-of-range repair method is respectively used. When the number of the sensor layout exceeds the variable interval, repairing the out-of-range number of the sensor layout by using the following method: firstly, determining a specific repair strategy executed by a current whale individual according to a random decimal p, when p is less than or equal to 0.5, selecting the sensor layout number of the current optimal individual to replace the current out-of-range sensor layout number, and when p is more than 0.5, selecting the sensor layout number of the global optimal individual to replace the current out-of-range sensor layout number. And randomly selecting the sensor layout number of one whale individual in the population, and randomly perturbing the sensor layout number of the offspring of the current iteration individual when the sensor layout number of the random individual is equal to the sensor layout number of the global optimal individual. The specific formula is as follows:
else：num＝num+round(normal(0，1))
wherein ,for globally optimal individuals, ++>For the currently best individual, round () is a rounding function and normal (0, 1) is a standard normal distribution. When the sensor arrangement interval exceeds the variable interval, the sensor arrangement interval and the sensor arrangement quantity are related, so that the sensor arrangement interval cannot be simply replaced by the sensor arrangement interval of the global or local optimal individual or the sensor arrangement interval of the disturbance global or local optimal individual is considered. Therefore, the sensor arrangement interval for repairing the exceeding interval is divided into two parts, namely a sensor arrangement interval replacement object and a sensor arrangement interval replacement method. The method comprises the steps of determining replacement objects of sensor arrangement intervals, firstly, searching individuals with the same sensor arrangement quantity from a history global optimal individual according to a small-to-large fitness value, and if the same individuals exist, determining the history global optimal individual as the replacement objects of the sensor arrangement intervals; if the number of the individuals with the same sensor arrangement number is not found from small to large in the current population according to the fitness value, if the number of the individuals is found, the individuals are determined to be replacement objects of the sensor arrangement interval, and attention is paid to the fact that the individuals cannot be the same as the current iteration individuals; if not, there is no replacement object. The sensor arrangement interval replacement method is changed according to the presence or absence of a replacement object, and when the sensor arrangement interval replacement object is the optimal historical individual, the sensor arrangement interval to be repaired is replaced by the sensor arrangement interval of the current repair object; when the sensor layout interval replacement object is a current group of individuals, replacing the sensor layout interval to be repaired with the sensor layout interval of the replacement object under random disturbance, namely adding a normal distribution random number; when no sensor layout interval is used for replacing the object, the random number in one variable interval is used for repairing.
After each whale individual has performed the evolution strategy, a global elite strategy needs to be performed. Unlike previous elite strategies, in global elite strategies, offspring of current iterative individuals need to be compared with not only their parents but also historical optimal individuals and current population optimal individuals, so that optimal individuals used by each whale iteration are kept instant, and by the method, population convergence speed can be effectively accelerated when population number becomes large.
After the iteration is finished, the sensors are laid out using the optimal variables of the historic best individuals.
The application will now be described in detail with reference to the drawings and specific examples. In the technical scheme, the characteristics such as the names of structures/modules, control modes, algorithms, process procedures or composition ratios and the like which are not explicitly described are regarded as common technical characteristics disclosed in the prior art.
Example 1
The application provides a concrete implementation case of an ocean platform flow pipeline damage positioning and sensor layout optimizing method.
Step one: and (3) carrying out ocean platform flow transmission pipeline simulation according to an actual engineering environment, wherein the pipeline simulation model is a steel pipe with the length of 1000mm, the outer diameter of 60mm and the wall thickness of 3mm, and the steel pipe is made of structural steel. The initial position of the damage simulation is 30mm, and the final position is 970mm. The lesion was 3mm long, 1mm wide and 2mm deep. The damage simulation was set at intervals of 3.5mm starting from 30mm for the pipe. Together, valid data 278 sets are collected. The starting position of the damage is used as a label, and strain mode data is used as a training set.
Step two:
(1) And (3) carrying out noise simulation on the strain mode data obtained in the step (1) according to the formula to obtain noise strain mode data.
(2) And performing cubic spline interpolation on the noise strain modal data by adopting an spline type splrep function in an interpolation class in a Python. Scipy library to obtain interpolation strain modal data.
(3) And (3) carrying out normalization processing on the interpolation strain modal data according to the formula (2) to obtain the normalization strain modal data.
Step three: aiming at the data of the return-to-strain mode obtained in the second step, a Pywt.WaveletPacket function in Python is adopted to realize wavelet packet analysis, db1 series functions are selected as wavelet packet basis functions, the number of decomposition layers is 9, and 512 frequency coefficients are obtained. A specific command is pywt.wavelet packet (mode= 'symmetry', wavelet= 'db1', maxlevel=9). The result of the wavelet packet analysis is a two-dimensional matrix (512, 128), i.e., each node has a coefficient feature length of 128. The two-dimensional matrix is converted into a three-dimensional matrix of (4, 128, 128) using wavelet packet stacking techniques.
Step four: and processing 278 groups of data obtained in the first step through the second step and the third step. And inputting the processed data into a convolutional neural network. In the specific implementation case, the structural parameters of the convolutional neural network are as follows:
TABLE 1 convolutional neural network structural parameters
The training set accounts for 90% of the total sample number, the testing set accounts for 10% of the total sample number, the iteration times are 1000, the loss function is a cross entropy function, the convolutional neural network is built based on Pytorch, the operation environment is Python3.8, numpy1.23, scipy1.9.3, pytorch1.12 and cuda11.3, and the hardware environment is: intel (R) Xeon (R) Silver4210R [email protected], nvidia 3090, 128GB DDR4 2400,batch_size is 16, learning rate is 0.05. The recognition results are shown in the following table:
TABLE 2 lesion location identification results
It can be seen from the table that the convolutional neural network identification accuracy for the lesion locations is 85% for 20 test samples. For the sample working condition with the identification error, the identification error is 5.9% at the highest, which indicates that the identification result of a single damage position is more accurate.
Step five: setting global parameters of an improved whale optimization algorithm: the number of whales is set to be 50, the number interval of sensors is [50,200], the arrangement interval of the sensors is [5,20], b is set to be 1, l is set to be 1, the maximum iteration number is set to be 100, and the fitness is the damage position identification accuracy. After the improved whale optimization algorithm iteration is completed, the historical best individual is taken as the sensor layout setting.
After the improved whale optimization algorithm iteration is completed, the historical optimal individual is taken as the sensor layout parameter. The optimization results are shown in table 3:
TABLE 3 optimized identification results for lesion locations
Table 4 sensor layout
The accuracy of the convolutional neural network for identifying the damage position is 95% for 20 test samples according to the optimized sensor layout mode. For the sample working condition with the identification error, the identification error is 0.2% at most, which shows that the improved whale optimizing algorithm has obvious optimizing effect on the sensor layout mode.
The previous description of the embodiments is provided to facilitate a person of ordinary skill in the art in order to make and use the present application. It will be apparent to those skilled in the art that various modifications can be readily made to these embodiments and the generic principles described herein may be applied to other embodiments without the use of the inventive faculty. Therefore, the present application is not limited to the above-described embodiments, and those skilled in the art, based on the present disclosure, should make improvements and modifications without departing from the scope of the present application.

Claims (10)

1. The ocean platform flow delivery pipeline sensor layout optimization method is characterized by comprising the following steps of:
s1: acquiring parameter information of a ocean platform flow transmission pipeline, constructing a finite element simulation model, constructing pipeline damage information in the finite element simulation model, setting a data acquisition starting point, acquiring strain mode data on an axial path of the pipeline, acquiring strain mode data on each sensor position based on an original layout scheme of the sensor, and constructing a training data set by taking the pipeline damage position as a label of the data;
s2: performing noise simulation, data interpolation and linear normalization on the training data set to obtain a preprocessed continuous sequence data set;
s3: performing time-frequency analysis on the continuous sequence data set by utilizing wavelet packet analysis to obtain a damage characteristic information set of sensor strain modal data corresponding to an ocean platform flow transmission pipeline;
s4: establishing a damage characteristic-damage position mapping relation set through a convolutional neural network based on the damage characteristic information set;
s5: and based on the damage characteristic-damage position mapping relation set, adopting an improved whale optimization algorithm to optimally adjust the sensor layout positions.
2. The ocean platform flow pipeline sensor layout optimization method according to claim 1, wherein in S1, the parameter information of the ocean platform flow pipeline comprises length, outer diameter, inner diameter, fluid, material and surrounding working environment parameters.
3. The ocean platform flow pipeline sensor layout optimization method according to claim 2, wherein in S1, the process of constructing pipeline damage information in the finite element simulation model includes:
and (3) constructing pipeline damage: obtaining pipeline damage with the length of L mm, the width of W mm and the depth of H mm at the inner wall of the pipeline simulation model through rotary cutting;
obtaining the position mark of the damage: and obtaining the distance X mm between one end point of the damage and the end part of the pipeline corresponding to the end point, and taking the X mm as the position mark of the damage.
4. The ocean platform flow pipeline sensor layout optimization method according to claim 3, wherein in S1, the training data set constructing process includes:
setting a Y mm position at one end of the ocean platform flow transmission pipeline simulation model as a data acquisition initial point, setting a (L-30) mm position at one end of the ocean platform flow transmission pipeline simulation model as a data acquisition end point, obtaining strain mode data on an axial path of the pipeline through finite element calculation, then taking out the strain mode data at each sensor layout position according to a sensor layout scheme, namely single training data, and taking a pipeline damage position Xmm as a label of the data.
5. The ocean platform flow pipeline sensor layout optimization method according to claim 1, wherein in the step S2, white noise is doped into simulation strain mode data to simulate noise generation in actual engineering in the noise simulation process;
wherein, the generation of white noise is realized by adopting random numbers based on standard normal distribution:
x noise ＝x+normal(0,1) (1)
wherein normal (0, 1) is a standard normal distribution, x is input data, x noise Data after adding noise;
in the data interpolation process, three times of data interpolation are carried out on the basis of the original data characteristics;
in the linear normalization process, the value range of each order strain mode is mapped between [0,1], the dimension of the real sensor strain mode data of any order is unified, and the normalization formula is as follows:
wherein min (x) is a minimum function and max (x) is a maximum function.
6. The ocean platform flow pipeline sensor layout optimization method according to claim 1, wherein in S3, the wavelet packet analysis process comprises:
the iteration number of each data in the continuous sequence data set is num iter Is decomposed by the dichotomy of (5) when num is iter In the sub-decomposition, the frequency domain is divided equally intoEach node corresponds to a frequency band, each frequency band contains frequency domain characteristic information of an original signal in the range, and the wavelet packet analysis process is shown in the following formula:
in the formula ,p(i,j) The frequency domain coefficient of the jth node of the ith iteration is a group of wavelet basis functions;
the wavelet packet analysis results in coefficient vector p of all nodes of the last iteration (last,j) The definition is as follows:
p (last,j) ＝[fre 1 ,fre 2 ,…,fre len ]
wherein last is the last iteration, j is the j node, fre is the frequency characteristic, and len is the number of the frequency characteristics;
coefficient vector p based on all nodes (last,j) And constructing and obtaining a damage characteristic information set of the sensor strain modal data corresponding to the ocean platform flow transmission pipeline.
7. The ocean platform flow pipeline sensor layout optimization method according to claim 6, wherein in the step S4, in the process of establishing the damage characteristic-damage position mapping relation set through a convolutional neural network:
stacking a two-dimensional matrix formed by a series of damage features into a three-dimensional space in a coefficient stacking mode, wherein the stacking height is the number of channels of the input convolutional neural network;
in the stacking process, the number of stacking layers is set as fold, and the specific process is as follows:
wherein ,representing the frequency characteristics of all nodes after the last iteration in wavelet packet analysis, +.>Represented as a length num res Two-dimensional matrix of width len, +.>Representing the folded frequency characteristic matrix,represents a high fold, long (num res Fold), a three-dimensional matrix of width len,
wherein → is expressed as a folding process, the implementation mode is that an open source library numpy.reshape function is called, and the specific process is that
The convolutional neural network is divided into an input layer, an implicit layer and an output layer;
the output layer is a full-connection layer and a LogSoftmax layer, the hidden layer structure is three layers, each layer structure is a convolution layer-BN layer-activation layer-pooling layer-regularization, the convolution kernel size is (5, 5), the pooling layer size is 2, the activation layer function is a ReLU () function, the regularization ratio is 0.1, the optimizer adopts a RMSprop function, the damage function adopts a cross entropy CrossEntrop function, the initial learning rate is set to 0.05, the learning rate is reduced by 75% every 5 iterations, the maximum iteration times is set to 200, and a damage characteristic-damage position mapping relation set is established through a convolution neural network.
8. The optimization method for sensor layout of ocean platform flow pipeline according to claim 1, wherein in S4, the process of optimizing and adjusting the sensor layout position comprises:
the damage identification and positioning accuracy is taken as an objective function, and the sensor layout quantity and the sensor layout interval are taken as optimization variables, as follows:
var＝[num,inter]
wherein var is an optimization variable, num is the number of sensors to be distributed, and inter is the distance between the sensors to be distributed;
then, using whale algorithm to find the optimized variable which makes the damage identification and positioning accuracy highest, comprising the following steps:
initializing global variables of a whale algorithm, determining a specific evolution strategy executed by a current whale individual according to a uniform random number r after iteration, executing an impact search strategy when r is less than or equal to 0.5, and executing a spiral search strategy when r is more than 0.5;
in the direct search strategy, determining whale individuals to be searched directly according to the following formula;
wherein ,a random number with a value range of 0 to 1;
when (when)When in use, a whale individual is randomly selected from the population for searching, and the specific formula is as follows:
wherein ,for randomly selected whale individuals;
when (when)When the method is used, the current global optimal individual is selected as an object to execute a spiral search strategy, and the specific formula is as follows:
wherein ,is the globally optimal individual;
the spiral search strategy is specifically as follows:
after the current whale individual executes the evolution strategy, checking whether the offspring crosses the boundary and repairing the variable crossing the boundary;
when the sensor layout quantity exceeds the variable interval, repairing the out-of-range sensor layout quantity by adopting the following process:
firstly, determining a specific repair strategy executed by a current whale individual according to a random decimal p, when p is less than or equal to 0.5, selecting the sensor layout number of the current optimal individual to replace the current out-of-range sensor layout number, and when p is more than 0.5, selecting the sensor layout number of the global optimal individual to replace the current out-of-range sensor layout number;
then, randomly selecting the sensor layout number of one whale individual in the population, and randomly perturbing the sensor layout number of the offspring of the current iteration individual when the sensor layout number of the random individual is equal to the sensor layout number of the global optimal individual, wherein the specific formula is as follows:
else:num＝num+round(normal(0,1))
wherein ,for globally optimal individuals, ++>For the currently best individual, round () is a rounding function, normal (0, 1) is a standard normal distribution;
finally, after each whale individual executes the evolution strategy, executing the global elite strategy, comparing the offspring of the current iteration individual with their parents, and comparing the offspring of the current iteration individual with the historical optimal individual and the current group optimal individual, so as to keep the optimal individual used by each whale iteration instant, and further accelerate the group convergence speed;
and after the iteration is finished, taking the optimal variable of the historical optimal individual as the optimal sensor layout parameter.
9. An ocean platform flow pipeline sensor layout optimization system, comprising:
the simulation and data construction acquisition module acquires parameter information of a ocean platform flow transmission pipeline, a finite element simulation model is constructed, pipeline damage information is constructed in the finite element simulation model, a data acquisition starting point is set, strain mode data on an axial path of the pipeline is acquired, strain mode data on each sensor position is acquired based on an original layout scheme of the sensor, and the pipeline damage position is used as a label of the data, so that a training data set is constructed;
the data post-processing module is used for carrying out noise simulation, data interpolation and linear normalization on the training data set to obtain a preprocessed continuous sequence data set;
the characteristic analysis module is used for carrying out time-frequency analysis on the continuous sequence data set by utilizing wavelet packet analysis to obtain a damage characteristic information set of sensor strain modal data corresponding to the ocean platform flow transmission pipeline;
the convolution calculation module is used for establishing a damage characteristic-damage position mapping relation set through a convolution neural network based on the damage characteristic information set;
and the optimization calculation module is used for optimizing and adjusting the sensor layout positions based on the damage characteristic-damage position mapping relation set and by adopting an improved whale optimization algorithm.
10. A storage medium containing computer executable instructions which, when executed by a computer processor, are for performing the ocean platform flow pipeline sensor layout optimization method of any one of claims 1-8.
CN202310631967.9A 2023-05-30 2023-05-30 Ocean platform flow pipeline sensor layout optimization method, ocean platform flow pipeline sensor layout optimization system and storage medium Pending CN116663213A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310631967.9A CN116663213A (en) 2023-05-30 2023-05-30 Ocean platform flow pipeline sensor layout optimization method, ocean platform flow pipeline sensor layout optimization system and storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310631967.9A CN116663213A (en) 2023-05-30 2023-05-30 Ocean platform flow pipeline sensor layout optimization method, ocean platform flow pipeline sensor layout optimization system and storage medium

Publications (1)

Publication Number Publication Date
CN116663213A true CN116663213A (en) 2023-08-29

Family

ID=87713210

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310631967.9A Pending CN116663213A (en) 2023-05-30 2023-05-30 Ocean platform flow pipeline sensor layout optimization method, ocean platform flow pipeline sensor layout optimization system and storage medium

Country Status (1)

Country Link
CN (1) CN116663213A (en)

Similar Documents

Publication Publication Date Title
GB2547816B (en) Actually-measured marine environment data assimilation method based on sequence recursive filtering three-dimensional variation
CN115018021B (en) Machine room abnormity detection method and device based on graph structure and abnormity attention mechanism
CN108595916B (en) Gene expression full-spectrum inference method based on generation of confrontation network
CN110363337B (en) Oil measuring method and system of oil pumping unit based on data driving
CN116448419A (en) Zero sample bearing fault diagnosis method based on depth model high-dimensional parameter multi-target efficient optimization
CN114492675B (en) Intelligent fault cause diagnosis method for capacitor voltage transformer
CN112132102B (en) Intelligent fault diagnosis method combining deep neural network with artificial bee colony optimization
CN114254460B (en) Turbomachine pneumatic robustness optimization method based on graph convolution neural network
CN116451123A (en) Transformer early fault diagnosis method based on ISSA-MKELM
CN111625992B (en) Mechanical fault prediction method based on self-optimal deep learning
CN115983374A (en) Cable partial discharge database sample expansion method based on optimized SA-CACGAN
CN111597753B (en) Data depth change characteristic self-adaptive two-dimensional resistivity inversion method and system
CN112417764A (en) K nearest neighbor regression prediction method for boiler special equipment steam flow prediction
CN113868765B (en) Ship main scale parameter optimization method based on approximate model
CN116663213A (en) Ocean platform flow pipeline sensor layout optimization method, ocean platform flow pipeline sensor layout optimization system and storage medium
CN113111923A (en) Water supply network leakage detection and positioning method based on one-dimensional migration learning convolutional neural network integrated model
CN112862063A (en) Complex pipe network leakage positioning method based on deep belief network
CN116861256A (en) Furnace temperature prediction method, system, equipment and medium for solid waste incineration process
CN116451133A (en) Heavy-duty robot speed reducer fault diagnosis method based on graph structure information
CN114021422B (en) Underground structure internal defect identification method based on cross-hole radar and deep learning
CN111444614B (en) Flow field reconstruction method based on graph convolution
CN115564989A (en) Random forest algorithm for land use classification
CN115455853A (en) Three-order compact reconstruction method based on finite volume method on unstructured grid
EP2442279B1 (en) Design optimization of physical bodies
Bhaduri et al. Efficient surrogate modeling for turbine blade row cyclic symmetric mode shapes

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