CN106021880B - Jacket platform structural response calculation method based on BP neural network - Google Patents
Jacket platform structural response calculation method based on BP neural network Download PDFInfo
- Publication number
- CN106021880B CN106021880B CN201610312508.4A CN201610312508A CN106021880B CN 106021880 B CN106021880 B CN 106021880B CN 201610312508 A CN201610312508 A CN 201610312508A CN 106021880 B CN106021880 B CN 106021880B
- Authority
- CN
- China
- Prior art keywords
- neural network
- layer
- network system
- platform
- input
- 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.)
- Active
Links
Classifications
-
- 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
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16Z—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS, NOT OTHERWISE PROVIDED FOR
- G16Z99/00—Subject matter not provided for in other main groups of this subclass
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2219/00—Indexing scheme relating to application aspects of data processing equipment or methods
- G06F2219/10—Environmental application, e.g. waste reduction, pollution control, compliance with environmental legislation
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention relates to the structure-design technique fields of ocean platform, specifically disclose a kind of jacket platform structural response calculation method based on BP neural network, it includes the following steps: step 1, establish BP neural network system, the BP neural network system includes: input layer, hidden layer and output layer, it is mutually linked as the first stage of jacket platform stress between input layer and hidden layer, the second stage of jacket platform stress is mutually linked as between hidden layer and output layer;Step 2, training BP neural network system;Step 3 inputs BP neural network system using arbitrary wind, wave and stream environmental load combined value as input vector, then obtains corresponding platform structure response using BP neural network system.It is applied to calculate the structural response of jacket platform using the artificial neural network system that the present invention is built, any environmental load is combined, the structural response very close with numerical simulation can be then instantly available by being inputted nerve network system, accurately, quickly.
Description
Technical field
The present invention relates to the structure-design technique field of ocean platform more particularly to this kind of ocean platforms of jacket platform
Structural response calculation method.
Background technique
Ocean platform is the marine heavy construction structure object for offshore oil and gas production and operation.Offshore platform structure is multiple
It is miscellaneous, bulky, involve great expense, Service Environment condition is sufficiently complex and severe, not only wind-engaging, wave, stream and ice joint make
With, while also by the threat of the extreme loads such as earthquake, once there is accident, generated economy, environmental hazard are very huge.
In the structure design of ocean platform and relevant fail-safe analysis, need to carry out the environmental loads such as a large amount of stormy waves stream
Simulation calculates, to obtain the structural response of platform.Computer and finite element software in spite of modernization carry out calculating analysis, but
Software analysis mode process is still very cumbersome, and calculating process is also long.
Specifically, for the structural response analysis of ocean platform, most common technological means be using ANSYS,
The Large-scale professionals finite element software such as SEASAM carries out numerical simulation, and analytic process needs first to use ocean platform
The 3D such as AutoCAD graphics software carries out simplifying modeling, then imports finite element software and is calculated, and the main of the technological means lacks
Point has:
(1) to carry out 3D modeling sufficiently complex as this kind of large complicated marine structure of ocean platform, have modeling tired
Difficult disadvantage;
(2) FEM software operation requires computer hardware high, while buying professional finite element software and being cost
With high and need to grasp by system training and how to operate.For same structure, using different cell types, grid
Division methods, there may be a great difference, inappropriate operation likely results in calculating not receiving obtained calculated result to occur
Holding back causes operation to fail, and debugging early period of program is very complicated, the comprehensive finite element of the suitable system of related personnel's needs and software
Operative knowledge could complete relevant calculation, have the shortcomings that technical threshold is high;
(3) finite element algorithm belongs to iterative algorithm, for large scale structure, even if using advanced finite elements division side
Method, during single iteration needed for calculation amount it is also very huge, using the technological means although available accurate structure
Response, but have that calculate the speed of service slow, the shortcomings that inefficiency;
(4) ocean platform this kind of for jacket, the environmental variances such as numerical simulation calculation such as wave act on works
On, theoretically simplified calculation method is to calculate the size of wave force first, then, in works, is analyzed by action of wave force
The load effect that a certain key position is generated.The related software co-ordination for needing to analyze fluid and structure could complete numerical value
Analysis, further increases difficulty in computation, it is often necessary to while calling more than two finite element programs, structural response calculate to
The present is also a relatively difficult problem.
In short, needing to consider a large amount of different load cases combinations, operating condition since ocean platform environmental load combined situation is complicated
Combination is up to ten thousand easily, is such as applied to calculate offshore platform structure response analysis, relevant design portion using numerical simulation technology means
Door generally requires one or two months or even could complete to calculate for more time.
And the structural response of physical experiments Measuring Oceanic platform is used, practical operation is time-consuming and laborious, only applies at present
It is compared with numerical simulation result.
The shortcomings that for numerical simulation calculation low efficiency, many experts and scholars both at home and abroad propose that many empirical equations are used for
The structural response of computing platform.Its technological means is for specific ocean platform, and by its structural response, (such as base shear inclines
Cover torque, deck displacement etc.) it is considered as the function of the environmental elements such as wave height, wind speed and flow velocity.But due to ocean platform itself
The complex nonlinear of structure, the structural response and truth error obtained using the technological means is larger, general relative error
It is more satisfactory as a result, can not be widely applied in practice for can control 30%.
Summary of the invention
The technical problem to be solved in the present invention is to find a kind of computational accuracy close to numerical simulation, but simultaneously convenient for behaviour
Make, low technical threshold, calculates rapid, efficient technological means to realize that the structural response of jacket platform calculates.
In order to solve the above-mentioned technical problem, the technical scheme adopted by the invention is that: a kind of leading based on BP neural network
Pipe support platform structure method of response calculation comprising following steps:
Step 1 establishes BP neural network system, which includes: input layer, hidden layer and output layer, defeated
The first stage for entering mutually to be linked as jacket platform stress between layer and hidden layer, jacket platform is mutually linked as between hidden layer and output layer
The second stage of stress, every layer includes several nodes, is not attached between every node layer, if the input vector of input layer is wind, wave
With the combination of stream, the output vector of output layer is that platform structure responds FX, FY, FZ and MX, MY and MZ, of input layer
Number is the dimension of input vector, and the number for exporting node layer is the dimension of output vector;
Step 2, training BP neural network system, using collecting obtained long-term actual measurement wind, wave and the flow data in sea area,
According to the distribution situation of wind, wave and flow data, several characteristic values is respectively taken to be combined, obtains feature load cases combination X as defeated
Incoming vector obtains the expectation platform in the case of each feature load cases combination X using physical experiments or using method for numerical simulation
X and corresponding W are inputted BP neural network system to be trained by structural response data W, with BP neural network system output vector
The root-mean-square error of actual platform structural response data Y and W are as network performance function, BP neural network system in training process
In weight and deviation be adjusted according to the error performance of network, correct the obtained actual platform of BP neural network system repeatedly
Structural response data Y makes it finally reach minimum with the root-mean-square error of W;
Step 3, using BP neural network system, using arbitrary wind, wave and stream environmental load combined value as input to
Amount input BP neural network system then obtains the corresponding response of output vector platform structure FX, FY, FZ and MX, MY and MZ.
Preferably, which only has 1 hidden layer, which includes 6 nodes.
Preferably, input layer excitation function is tansig,
Hidden layer excitation function is logsig,
Output layer excitation function is purelin, purelin (x)=x.
The technical solution of the present invention brings about beneficial effects:
(1) BP neural network topological structure applied by the present invention is decomposed according to two stages stress and is divided, by structural mechanics
Basic principle inspires, it is contemplated that various environmental load collective effects are when jacket platform, between combined load and structural response
Complicated non-linear relation;Lengthy and jumbled structural mechanics theory is avoided, the structural response calculating of jacket platform is concise, greatly
The professional threshold of relevant design personnel is reduced greatly.The determination of the topological structure simultaneously, it is excessively complicated both to have avoided neural network
Caused training is slow, and the accuracy for structural response simulation has also been effectively ensured;
(2) due to the related basic program of artificial neural network be built in the form of program bag extensively at present as
In the common numerical value software for calculation such as Matlab, relevant design personnel, can be directly sharp after specifying neural network topology structure
BP neural network system is built with it, is provided convenience condition for popularization of the invention;
(3) it is applied to calculate the structural response of jacket platform using the artificial neural network system that the present invention is built,
Any environmental load is combined, the structure very close with numerical simulation can be then instantly available by being inputted nerve network system
Response, it is accurate, quick.
Detailed description of the invention
The BP neural network topology diagram of the mono- hidden layer of Fig. 1.
Fig. 2 one embodiment of the invention BP neural network topology diagram.
The orientation Fig. 3-1 one embodiment of the invention N FX--BP neural network fitting result schematic diagram.
The orientation Fig. 3-2 one embodiment of the invention N FY--BP neural network fitting result schematic diagram.
The orientation Fig. 3-3 one embodiment of the invention N FZ--BP neural network fitting result schematic diagram.
The orientation Fig. 3-4 one embodiment of the invention N MX--BP neural network fitting result schematic diagram.
The orientation Fig. 3-5 one embodiment of the invention N MY--BP neural network fitting result schematic diagram.
The orientation Fig. 3-6 one embodiment of the invention N MZ--BP neural network fitting result schematic diagram.
Specific embodiment
The present invention will be further explained below with reference to the attached drawings and specific examples, so that those skilled in the art can be with
It better understands the present invention and can be practiced, but illustrated embodiment is not as a limitation of the invention.
The present invention is based on artificial neural network technologies.Artificial neural network is a kind of machine learning hand more mature at present
Section obtains good effect in the application in other many fields, therefore the present invention has solid reason in terms of model calculating
By for support.
Artificial neural network is on the basis of the modern biotechnology successes achieved in research, by nervous physiology science, Information Center
What the research achievement of the related sciences such as, mathematical and physical science and computational science was established, artificial neural network system is by largely handling list
First (i.e. neuron) interconnection composition, can imitate human brain information processing mechanism.Its main Types includes: BP, radial base, self-organizing
And feedback neural network, it is most widely used at present for using the multilayer feedforward artificial neural network of error backpropagation algorithm
Network, i.e. BP neural network.
There is BP neural network good None-linear approximation ability, generalization ability and easy adaptive, distinguishing feature to include:
(1) distributed information storage means
BP neural network is that information is stored with the state of each processor itself and the type of attachment between them, one
Information is not stored in a place, but over the entire network by content distribution.Certain is not only storage one on network
External information, but store the partial content of multiple information.Whole network is to just storage is each to network after multiple Information processions
Place, therefore, it is a kind of distributed storage mode.
(2) MPP
Since BP neural network is made of in a unique way a large amount of artificial neurons, can receive simultaneously multiple defeated
Enter information, and can also simultaneous transmission, artificial neuron can respond in the form of voting, therefore artificial neural network
Output can automatically achieve the effect of " the minority is subordinate to the majority " the result is that by most artificial neurons voting simultaneously.This function
Substantially maximally utilise space complexity, and effectively reduces time complexity.
(3) self study and adaptivity
Artificial neural network can develop various functions by posteriori study and training, be similar to biology
Neural network.The direct connection weight of each layer of BP neural network has certain adjustability, and network can pass through training and study
It determines the weight of network, shows very strong to the adaptive of environment and to the self-learning capability of extraneous things.
(4) stronger robustness and fault-tolerance
The distributed information storage means of BP neural network, makes it have stronger fault-tolerance and function of associate memory, this
If the information of certain a part of sample is lost or damage, network remains to recover original complete information, and system remains to run.
BP neural network principle
BP neural network is usually made of input layer, hidden layer and output layer, totally interconnected between layers, between every node layer
It is not attached to.The number of its input layer usually takes the dimension of input vector, export node layer number usually take output to
The dimension of amount, hidden node number there is no determining standard at present, then need to most be terminated by trying the method gathered repeatedly
Fruit.According to Kolmogor theorem, three layers of BP neural network with a hidden layer (hidden node is enough) can in closed set with
Arbitrary accuracy approaches any non-linear continuous function.By taking the BP neural network of single hidden layer as an example, topological structure is as shown in Figure 1.
BP neural network can regard one as from the mapping for being input to output, i.e. F:Rn→Rm, f (X)=Y.For sample
Set: input value xi(∈Rn) and output valve be yi(∈Rm), it is believed that there are a certain mapping g to make g (xi)=yi(i=
1,2,…,n).Neural network show that approximate function f is most preferably approaching for g by being fitted for several times to simple function.
The learning process of BP algorithm is made of forward-propagating and backpropagation.During forward-propagating, input pattern is from defeated
Enter layer and successively handled through hidden layer, and be transmitted to output layer, the state of each layer of neuron only under the influence of one layer of neuron state.Such as
Fruit output layer cannot obtain desired output, then be transferred to backpropagation, and error signal is returned along original connecting path, is passed through
The connection weight of each neuron is modified, so that error signal is minimum.Iterate the desired value Y for obtaining neural computingL
It is minimum with real response value T root-mean-square error.
General L layers of BP neural network, note input layer are the 0th layer, and output layer is L layers, and middle layer (i.e. hidden layer) is successively
It is the 1st layer to mono- 1 layers of L.The neuron number of kth layer is nk, one 1 layers of the kth weight matrix to kth layer be
WhereinIndicate the connection weight of kth -1 layer of i-th of neuron and j-th of neuron of kth layer.
Assuming that the input vector of artificial neural network is X=(X1,X2,…,Xn0)T, then its 1st layer received vector is Z1
=W1 TX, output vector Y1=(Y1 1,Y2 1,…,Yn1 1)T;The received vector that its kth layer (k >=2) can equally be obtained isOutput vector isWherein:
In formula, fi k() is the excitation function of i-th of neuron of kth layer.There are many forms, such as common S for excitation function
Type function or sigmoid function are as follows:
The study of network seeks to determine weight matrix Wk(k=1,2 ..., L), so that network ideal exports YL=(Y1 L,
Y2 L,…,YnL L)T, reach and the smallest effect of the error of reality output T.
The present invention is to calculate BP artificial neural network applied to jacket platform structural response.
The present invention is based on an embodiment of the jacket platform structural response calculation method of BP neural network is as follows.
Consider that the load effect of jacket platform is mainly generated by wind, wave, stream effect, needs when actual design to calculate given
One group of wind, wave, stream compound action are when jacket platform, if the response at platform mud face is FX, FY, FZ and MX, MY and MZ.
Step1: BP neural network system is established
The BP neural network system of the jacket platform has 3 inputs, 6 outputs, in conjunction with classical structural capacity Xue Zhi
Know, the true stress physical process of jacket platform be decomposed into two stages:
(1) when the environmental loads such as wind, wave, stream act solely on works, due to its active position, the size of active force is not
Together, different platform response FX, FY, FZ and MX, MY and MZ can be generated, and when environmental load acts on jacket platform simultaneously
When, the same type structural response that each environmental load generates can influence each other, and finally be changed into the actual structure of jacket platform
Response;
(2) between each response of jacket platform there is also the relationship that one kind mutually restricts, i.e. FX can to FY, FZ and
MX, MY and MZ have an impact, similarly with other kinds of structural response.
Decomposed based on the above two stages stress, BP neural network system be designed to the topological structure such as Fig. 2, input layer and
It is the first stage of jacket platform stress between hidden layer, hidden layer to the stress that second stage is considered between output layer is closed
System.
For excitation function, tansig and logsig function are S type function, when differentiating to it, can use itself
Certain form indicate.This point is critically important when doing numerical experimentation, and in training neural network, the backpropagation of weight is needed
The derivative of excitation function is used, multilayer then be needed using multiple derivatives, use S type function then can be with as excitation function at this time
It reduces the storage space of computer, improve operation and convergence rate.In addition, no matter excitation function is for discrimination or convergence rate
There is significant impact.When approaching high order curve, sigmoid function ratio of precision linear function wants much higher, but calculation amount is also big
Much.In conclusion input layer excitation function is set to tansig by the present invention, play the role of expanding codomain, make each in hidden layer
Non-linear joint between branch is more extensive, and hidden layer uses logsig, and output layer uses purelin.
Net=newff (in, ou, [6 6], ' tansig ' ' logsig ' ' purelin ', ' trainlm ');
The expression formula of each excitation function is as follows;
Purelin (x)=x (5)
Step2: training BP neural network system
BP neural network belongs to the learning process for having supervision, needs to utilize long-term actual measurement wind, the wave for collecting obtained sea area
And flow data respectively takes several characteristic values to be combined according to the distribution situation of wind, wave and flow data, obtains feature operating condition group
X is closed, physical experiments can be used or measures the platform structure response in the case of each load cases combination using method for numerical simulation
Data W, training process are that input vector X and corresponding desired output response W are inputted BP neural network system to be trained, with
Power of the root-mean-square error of BP neural network output response Y and W as network performance function, in training process in BP neural network
Value and deviation are adjusted according to the error performance of network, correct the platform structure response Y that BP neural network system obtains repeatedly,
Making it, most the root-mean-square error of expected response W reaches minimum finally;
Net=train (net, input, output2);%input is input data X, and output2 is that target exports W
In the present embodiment, long-term actual measurement 10min mean wind speed, the significant wave height, surface layer for collecting certain obtained sea area are utilized
Flow speed data respectively takes 8,6,8 characteristic values according to wind speed, wave height, the distribution situation of flow velocity, see the table below 1, considers that stormy waves stream is equal
From same orientation, there are 384 kinds of composite conditions, physical experiments can be used and measure corresponding 6 structural responses of every group of operating condition,
The technological means that numerical simulation can also be used when conditions permit simulates its result.
1 stormy waves stream characteristic value of table
1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | |
Wind speed | 5 | 10 | 20 | 30 | 40 | 50 | 60 | 65 |
Significant wave height | 2 | 5 | 8 | 11 | 14 | 17 | ||
Surface current speed | 0.6 | 0.9 | 1.2 | 1.5 | 1.8 | 2.1 | 2.4 | 2.7 |
Note: wind speed, flow rate take m/s, and significant wave height unit is m.
Step3: BP neural network is applied
Using trained BP neural network, arbitrary environmental load is combined, wind, wave, fluxion value need to only be inputted
BP neural network system can then be instantly available corresponding structural response.
Simout=sim (net, input);%input is input data X, and simout is neural computing response
In order to verify accuracy of the invention, Fig. 3-1 to 3-6 is N to structural response BP neural network fitted figure.It provides and works as
The jacket platform structural response being fitted when stormy waves stream is all from the orientation N using BP neural network.It uses as the result is shown
BP neural network fitting platform response effect is very ideal, and network output valve is almost identical with true value.
2 N of table is to Fitness Test table
RMSE | MaxRSE | MinRSE | MRSE | DRSE | PSE (%) | |
FX | 0.0115 | 0.0528 | 1.8897e-5 | 0.0071 | 0.0034 | 0 |
FY | 0.0027 | 0.0300 | 1.8728e-6 | 0.0014 | 8.1096e-4 | 0 |
FZ | 5.8882e-5 | 9.8062e-5 | 8.9855e-7 | 5.1852e-5 | 5.9299e-5 | 0 |
MX | 0.0071 | 0.1276 | 1.1146e-5 | 0.0015 | 5.9918e-4 | 0.2604 |
MY | 6.7219e-4 | 0.0014 | 1.3212e-6 | 5.7532e-4 | 5.7648e-4 | 0 |
MZ | 0.0686 | 1.2822 | 8.2276e-7 | 0.0087 | 0.0015 | 1.0417 |
Note: RMSE is that average deviation quadratic sum opens radical sign, and MaxRSE is relative error maximum, and MinRSE is relative error
Minimum, MRSE are relative error mean value, and DRSE is relative error median, and PSE is the percentage that relative error is greater than 10%.
Embodiment described above is only to absolutely prove preferred embodiment that is of the invention and being lifted, protection model of the invention
It encloses without being limited thereto.Those skilled in the art's made equivalent substitute or transformation on the basis of the present invention, in the present invention
Protection scope within.Protection scope of the present invention is subject to claims.
Claims (1)
1. a kind of jacket platform structural response calculation method based on BP neural network, which is characterized in that it includes following step
It is rapid:
Step 1 establishes BP neural network system, which includes: input layer, hidden layer and output layer, input layer
It is mutually linked as the first stage of jacket platform stress between hidden layer, jacket platform stress is mutually linked as between hidden layer and output layer
Second stage, every layer includes several nodes, is not attached between every node layer, if the input vector of input layer is wind, wave and stream
Combination, the output vector of output layer is that platform structure responds FX, FY, FZ and MX, MY and MZ, the number of input layer and is
The dimension of input vector, the number for exporting node layer is the dimension of output vector;
Step 2, training BP neural network system, using collecting obtained long-term actual measurement wind, wave and the flow data in sea area, according to
The distribution situation of wind, wave and flow data respectively takes several characteristic values to be combined, obtain feature load cases combination X as input to
Amount obtains the expectation platform structure in the case of each feature load cases combination X using physical experiments or using method for numerical simulation
X and corresponding W are inputted BP neural network system to be trained by response data W, practical with BP neural network system output vector
The root-mean-square error of platform structure response data Y and W are as network performance function, in training process in BP neural network system
Weight and deviation are adjusted according to the error performance of network, correct the actual platform structure that BP neural network system obtains repeatedly
Response data Y makes it finally reach minimum with the root-mean-square error of W;
Step 3, it is using BP neural network system, arbitrary wind, wave and stream environmental load combined value is defeated as input vector
Enter BP neural network system, then obtains the corresponding response of output vector platform structure FX, FY, FZ and MX, MY and MZ;
The BP neural network system only has 1 hidden layer, which includes 6 nodes;
Input layer excitation function is tansig,
Hidden layer excitation function is logsig,
Output layer excitation function is purelin, purelin (x)=x.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610312508.4A CN106021880B (en) | 2016-05-12 | 2016-05-12 | Jacket platform structural response calculation method based on BP neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610312508.4A CN106021880B (en) | 2016-05-12 | 2016-05-12 | Jacket platform structural response calculation method based on BP neural network |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106021880A CN106021880A (en) | 2016-10-12 |
CN106021880B true CN106021880B (en) | 2019-03-26 |
Family
ID=57100066
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610312508.4A Active CN106021880B (en) | 2016-05-12 | 2016-05-12 | Jacket platform structural response calculation method based on BP neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106021880B (en) |
Families Citing this family (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107908817A (en) * | 2017-10-17 | 2018-04-13 | 江苏大学 | A kind of sea transfer platform device structure stress monitoring and diagnostic system |
CN109480872B (en) * | 2018-11-08 | 2021-09-17 | 哈尔滨工业大学 | Driving fatigue detection method based on electroencephalogram signal frequency band energy ratio characteristics |
CN109871609B (en) * | 2019-02-18 | 2020-10-27 | 中国海洋大学 | Method for predicting response of marine floating platform mooring system based on BP-FEM |
CN113283138B (en) * | 2021-05-25 | 2024-03-26 | 大连理工大学 | Deep-learning-based dynamic response analysis method for deep-sea culture platform |
-
2016
- 2016-05-12 CN CN201610312508.4A patent/CN106021880B/en active Active
Non-Patent Citations (4)
Title |
---|
BP网络的结构动力响应预测;郭剑虹;《泰州职业技术学院学报》;20070630;第7卷(第3期);第15-18页 |
基于SESAM的导管架平台结构动力响应分析;渠基顺 等;《江苏船舶》;20130228;第30卷(第1期);第1-5页 |
基于神经网络算法的海上风机结构状态监测研究;王春 等;《中国水利水电科学研究院学报》;20151231;第13卷(第5期);第344-351页 |
神经网络在结构响应预测中的应用研究;韩卫华 等;《信息技术》;20040531;第28卷(第5期);第51-52,55页 |
Also Published As
Publication number | Publication date |
---|---|
CN106021880A (en) | 2016-10-12 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106021880B (en) | Jacket platform structural response calculation method based on BP neural network | |
CN111272334A (en) | Particle swarm optimization BP neural network-based multidimensional force sensor calibration decoupling method | |
CN104077438B (en) | Power network massive topologies structure construction method and system | |
CN112446091A (en) | Artificial neural network-based pulsating pressure prediction method | |
CN104049536B (en) | The building method of chassis nonlinear system fuzzy-neural network generalized inverse controller | |
CN108090629A (en) | Load forecasting method and system based on nonlinear auto-companding neutral net | |
CN104408773B (en) | Method for interpolating structured grid non-matching interface | |
Djavareshkian et al. | Neuro-fuzzy based approach for estimation of Hydrofoil performance | |
CN109753690A (en) | Nonlinear unsteady aerodynamics order reducing method based on Fluid Mechanics Computation | |
CN111709350B (en) | Low-frequency oscillation modal parameter identification method and system based on FCM clustering | |
CN112883522A (en) | Micro-grid dynamic equivalent modeling method based on GRU (generalized regression Unit) recurrent neural network | |
CN103559340A (en) | Irrelevant multisource frequency domain load identification method based on COMI-PSO (comprehensive improved particle swarm optimization) algorithm | |
Cui et al. | Fuel savings model after aero-engine washing based on convolutional neural network prediction | |
CN115146538A (en) | Power system state estimation method based on message passing graph neural network | |
CN111914488B (en) | Data area hydrologic parameter calibration method based on antagonistic neural network | |
CN101540504B (en) | Current analytical device and method on basis of step-length variable neural network | |
Zhou et al. | Stability prediction of tailing dam slope based on neural network pattern recognition | |
CN116910513A (en) | Centrifugal pump rotor fault diagnosis method based on CWGAN-GP and double-flow CNN model | |
CN113283138A (en) | Deep sea culture platform dynamic response analysis method based on deep learning | |
CN111177975A (en) | Aviation equipment availability prediction method based on artificial intelligence | |
Zhu et al. | Application of artificial neural network in the prediction of output in oilfield | |
CN117436317B (en) | Wave current load simulation calculation method, system and equipment based on offshore wind power pile foundation | |
Li et al. | Parameter Estimation of Particle Flow Model for Soils Using Neural Networks. | |
Jiang et al. | A multi-fidelity prediction model for vertical bending moment and total longitudinal stress of a ship based on composite neural network | |
Shao et al. | Positive kinematics solution of a six-degree-of-freedom parallel platform based on neural network algorithm |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |