CN107292015B - Neural network algorithm-based underwater vehicle equilibrium submerged model evaluation method - Google Patents

Neural network algorithm-based underwater vehicle equilibrium submerged model evaluation method Download PDF

Info

Publication number
CN107292015B
CN107292015B CN201710452977.0A CN201710452977A CN107292015B CN 107292015 B CN107292015 B CN 107292015B CN 201710452977 A CN201710452977 A CN 201710452977A CN 107292015 B CN107292015 B CN 107292015B
Authority
CN
China
Prior art keywords
model
underwater vehicle
emp
neural network
error coefficient
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
Application number
CN201710452977.0A
Other languages
Chinese (zh)
Other versions
CN107292015A (en
Inventor
赵东明
柳欣
杨田田
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.)
Wuhan University of Technology WUT
Original Assignee
Wuhan University of Technology WUT
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 Wuhan University of Technology WUT filed Critical Wuhan University of Technology WUT
Priority to CN201710452977.0A priority Critical patent/CN107292015B/en
Publication of CN107292015A publication Critical patent/CN107292015A/en
Application granted granted Critical
Publication of CN107292015B publication Critical patent/CN107292015B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/10—Geometric CAD
    • G06F30/15—Vehicle, aircraft or watercraft design

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Geometry (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Computer Hardware Design (AREA)
  • Evolutionary Computation (AREA)
  • Computational Mathematics (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Optimization (AREA)
  • Pure & Applied Mathematics (AREA)
  • Aviation & Aerospace Engineering (AREA)
  • Automation & Control Theory (AREA)
  • Feedback Control In General (AREA)
  • Traffic Control Systems (AREA)

Abstract

The invention belongs to the field of computer simulation evaluation, and particularly relates to a neural network algorithm-based simulation evaluation method for an underwater vehicle equilibrium submerged model, which comprises the following steps of: setting input parameters under various working conditions, and testing a mathematical model of an underwater vehicle balanced submergence-floatation model; when the submerged model reaches a set state, recording various input and output parameters as a sample set; carrying out numerical analysis on the sample set by utilizing a neural network algorithm to obtain an error coefficient between the mathematical model and an ideal value; and if the error coefficient is within a specified range, evaluating the submerged model by using the error coefficient. The method can evaluate the accuracy of the underwater vehicle balanced submerging and surfacing model.

Description

Neural network algorithm-based underwater vehicle equilibrium submerged model evaluation method
Technical Field
The invention belongs to the field of computer simulation evaluation, and particularly relates to a neural network algorithm-based simulation evaluation method for an underwater vehicle equilibrium submerged model.
Background
In recent years, underwater vehicles have played an important role, both in civilian and military applications. As an important component in the defense industry, the control technology of the national defense industry develops rapidly. The control mode of the underwater vehicle is developed from the control mode of cooperation of four systems which are arranged dispersedly originally into the control mode of a 'control system' which integrates all control devices into a whole. Balanced submerging and surfacing is an important component of the underwater vehicle steering control technology, and can be subdivided into buoyancy balancing and trim balancing. With the rapid development of the underwater vehicle control technology, the control precision of the pitch balancing subsystem is higher and higher.
However, the structure of the underwater vehicle is complex, the number of devices on the underwater vehicle is large, the offshore environment is variable, a large amount of manpower and material resources are consumed in the test, and the test verification of the control technology of the underwater vehicle is difficult. The three-dimensional simulation verification method can effectively perform model simulation, and can more visually present test effects for testers through a visualization method. However, the built equilibrium latency model of the underwater vehicle has no related evaluation method, and the accuracy of the built equilibrium latency model of the underwater vehicle is unknown, so that no related effective method exists for experimental verification of the control technology of the underwater vehicle.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provide a simulation evaluation method of an underwater vehicle balanced submerging and surfacing model based on a neural network algorithm, which can evaluate the accuracy of the underwater vehicle balanced submerging and surfacing model.
The invention relates to a simulation evaluation method of an underwater vehicle equilibrium submergence-floatation model based on a neural network algorithm, which comprises the following steps:
setting input parameters under various working conditions, and testing a mathematical model of an underwater vehicle balanced submergence-floatation model;
when the submerged model reaches a set state, recording various input and output parameters as a sample set;
carrying out numerical analysis on the sample set by utilizing a neural network algorithm to obtain an error coefficient between the mathematical model and an ideal value;
and if the error coefficient is within a specified range, evaluating the submerged model by using the error coefficient.
Further, if the error coefficient exceeds a specified range, changing input parameters, substituting the changed input parameters into the mathematical model for re-testing:
when the submerged model reaches a set state, recording the changed input and output parameters as a new sample set;
carrying out numerical analysis on the new sample set by utilizing a neural network algorithm to obtain a new error coefficient between the mathematical model and an ideal value;
correcting the new error coefficient by using a weighting algorithm to obtain a corrected error coefficient;
evaluating the model using the corrected error coefficients.
Further, the numerical analysis specifically includes:
a sample set is arranged: (X)m，ym)，m＝1～M；
Obtaining the best parameteraSo that the risk function Re m p() Reaching a minimum value;
will be provided withaAnd a minimum value Re m pAs weight vector and threshold parameterAnd calculating to obtain the error coefficient.
Still further, said Re m p() The expression is as follows:
Figure BDA0001322976740000022
still further, in the above-described manner,
the optimal parameter is obtainedaSo that the risk function Re m p() Reaching a minimum value, including:
adding a steepest descent algorithm and an iterative algorithm;
according to the steepest descent algorithm and the iterative algorithmaAnd a minimum value Re m p。
Still further, the calculation formula of the steepest descent algorithm and the iterative algorithm is as follows:
(k+1)＝(k)+Δ(k) (2)
Figure BDA0001322976740000023
preferably, the first and second electrodes are formed of a metal,
obtaining the optimal parameter according to the steepest descent algorithm and the iterative algorithmaAnd a minimum value Re m pThe method specifically comprises the following steps:
let a take a smaller value, so that | | | Δ (k) | < 1, then:
Re m p((k+1))＝Re m p((k))+ΔRe m p((k)) (4)
as can be seen from the formula (4), Re m p((k)) decreases with increasing k, which tends towards Re m p() A local minimum of (a);
taking R corresponding to the local minimum pointe m p() Is an optimum parametera。
Preferably, the first and second electrodes are formed of a metal,
will be provided withaAnd a minimum value Re m pAs a weight vector and a threshold parameter, the error coefficient is obtained by calculation, which specifically includes:
will be provided withaAnd a minimum value Re m pAs weight vector and thresholdParameters, substituting the following equation:
Figure BDA0001322976740000031
Figure BDA0001322976740000032
in the formula (5) and the formula (6),
Figure BDA0001322976740000033
the ith output value of the 1 st layer is represented as an error coefficient; xl -1Represents the output value of layer 1-1;
Figure BDA0001322976740000034
representing the ith input value of layer 1.
Preferably, the first and second electrodes are formed of a metal,
Figure BDA0001322976740000035
Figure BDA0001322976740000036
in the invention, the neural network algorithm is utilized to carry out numerical analysis on the input parameters under different working conditions, so that the error coefficient between the mathematical model and the ideal value of the underwater vehicle balanced submergence and floatation model is obtained, and the underwater vehicle balanced submergence and floatation model is evaluated through the error coefficient. Thus, a channel for experimental verification of underwater vehicle control technology is provided.
Drawings
The invention will be further described with reference to the accompanying drawings and examples, in which:
FIG. 1 is a block diagram of a method of an embodiment of the invention;
fig. 2 is a flow chart of a method of an embodiment of the present invention.
Detailed Description
For a more clear understanding of the technical features, objects and effects of the present invention, embodiments of the present invention will now be described in detail with reference to the accompanying drawings.
As shown in fig. 1 and fig. 2, before the method for evaluating the equilibrium submerged-floating model of the underwater vehicle based on the neural network algorithm is invented, a mathematical model of the equilibrium submerged-floating model of the underwater vehicle is firstly established;
the balanced submerging and surfacing model mainly refers to the change rule of water transfer between the fore and the aft, and is interfered by various factors such as pressure and the like, and the mathematical model is complex. Here, a mathematical model in which the pressure after the pressure reducer is 0.579Mpa and the pitch angle is in the range of 0 ° to 35 ° is used, specifically as follows:
the mathematical model for balancing the submerged and floated forward and stern water transfer is as follows:
Figure BDA0001322976740000041
WS(t)＝WS(t-Δt)+WS(Δt)
the mathematical model for balancing the submerged and floated stern forward water transfer:
Figure BDA0001322976740000042
t represents the time to migrate, unit: s;
Δ t represents the water transfer time step, unit: s;
WS(Δ t) represents the amount of water removed over time Δ t in units of: m is3；
WS(t) represents the amount of water removed at time t, in units: m is3；
α1Represents the initial pitch angle in Δ t time, in units: (iv) DEG;
α2represents the end pitch angle in Δ t time, in units: degree.
Building a three-dimensional visual scene for an underwater vehicle; and establishing a three-dimensional model of each main part of the underwater vehicle by using a three-dimensional modeling tool. And (4) importing the model into a scene editing engine, adjusting parameters of each component according to a model algorithm, and establishing a complete virtual scene of the underwater vehicle. And converting the mathematical model into a scene script, binding the scene script to the model in the scene, and carrying out a simulation test in the scene.
The invention discloses a simulation evaluation method of an underwater vehicle equilibrium submergence-floatation model based on a neural network algorithm, which comprises the following steps:
101. setting input parameters under various working conditions, and testing a mathematical model of an underwater vehicle balanced submergence-floatation model;
102. when the submerged model reaches a set state, recording various input and output parameters as a sample set;
103. carrying out numerical analysis on the sample set by utilizing a neural network algorithm to obtain an error coefficient between the mathematical model and an ideal value;
104. and if the error coefficient is within a specified range, evaluating the submerged model by using the error coefficient.
201. If the error coefficient exceeds the specified range, changing the input parameters, and substituting the changed input parameters into the mathematical model for testing again;
202. when the submerged model reaches a set state, recording the changed input and output parameters as a new sample set;
203. carrying out numerical analysis on the new sample set by utilizing a neural network algorithm to obtain a new error coefficient between the mathematical model and an ideal value;
204. correcting the new error coefficient by using a weighting algorithm to obtain a corrected error coefficient;
205. evaluating the model using the corrected error coefficients.
The numerical analysis specifically includes:
a sample set is arranged: (X)m，ym)，m＝1～M；
Obtaining the best parameteraSo that the risk function Re m p() Reaching a minimum value;
will be provided withaAnd a minimum value Re m pAnd calculating to obtain the error coefficient as a weight vector and a threshold parameter.
The R ise m p() The expression is as follows:
Figure BDA0001322976740000061
the optimal parameter is obtainedaSo that the risk function Re m p() Reaching a minimum value, including:
adding a steepest descent algorithm and an iterative algorithm;
according to the steepest descent algorithm and the iterative algorithmaAnd a minimum value Re m p。
The calculation formula of the steepest descent algorithm and the iterative algorithm is as follows:
(k+1)＝(k)+Δ(k) (2)
Figure BDA0001322976740000062
obtaining the optimal parameter according to the steepest descent algorithm and the iterative algorithmaAnd a minimum value Re m pThe method specifically comprises the following steps:
let a take a smaller value, so that | | | Δ (k) | < 1, then:
Re m p((k+1))＝Re m p((k))+ΔRe m p((k)) (4)
as can be seen from the formula (4), Re m p((k)) decreases with increasing k, which tends towards Re m p() A local minimum of (a);
taking R corresponding to the local minimum pointe m p() Is an optimum parametera。
Will be provided withaAnd a minimum value Re m pAs a weight vector and a threshold parameter, the error coefficient is obtained by calculation, which specifically includes:
will be provided withaAnd a minimum value Re m pAs weight vector and threshold parameter, the following formula is substituted:
Figure BDA0001322976740000063
Figure BDA0001322976740000064
in the formula (5) and the formula (6),
Figure BDA0001322976740000065
the ith output value of the 1 st layer is represented as an error coefficient; xl -1Represents the output value of the l-1 layer;
Figure BDA0001322976740000066
representing the ith input value of layer 1.
Wherein:
Figure BDA0001322976740000071
Figure BDA0001322976740000072
the method for performing numerical analysis on the new sample set by using the neural network algorithm is completely the same as the method for performing numerical analysis on the sample set by using the neural network, and only the related data is different, so that the detailed description is omitted here.
In the embodiment, functions of control simulation and three-dimensional visualization of cabin equipment of each system in the underwater vehicle are realized in a three-dimensional model space by using a virtual reality technology; and a basic simulation platform is provided for services such as future scheme verification, equipment demonstration, simulation test, technical research and the like. The modeling method avoids the construction of an entity test boat, greatly reduces the cost of equipment development and research, and accords with the development trend of underwater vehicle research.
The evaluation method in the embodiment adds a neural network algorithm, can process information from various complex parameters, and can quickly calculate to obtain a required result. And an iterative algorithm is added, so that the influence of individual larger error points on the calculation of the overall error value is reduced, and the accuracy of the evaluation method is greatly improved.
While the present invention has been described with reference to the embodiments shown in the drawings, the present invention is not limited to the embodiments, which are illustrative and not restrictive, and it will be apparent to those skilled in the art that various changes and modifications can be made therein without departing from the spirit and scope of the invention as defined in the appended claims.

Claims (7)

1. An underwater vehicle equilibrium submergence and floatation model simulation evaluation method based on a neural network algorithm is characterized by comprising the following steps:
setting input parameters under various working conditions, and testing a mathematical model of an underwater vehicle balanced submergence-floatation model;
when the submerged model reaches a set state, recording various input and output parameters as a sample set;
carrying out numerical analysis on the sample set by utilizing a neural network algorithm to obtain an error coefficient between the mathematical model and an ideal value;
if the error coefficient is within a specified range, evaluating the submerged model by using the error coefficient;
the numerical analysis specifically includes:
a sample set is arranged: (X)m，ym)，m＝1～M；
Obtaining the best parameteraSo that the risk function Remp() Reaching a minimum value;
will be provided withaAnd a minimum value RempCalculating to obtain the error coefficient as a weight vector and a threshold parameter;
the R isemp() The expression is as follows:
Figure FDA0002571268020000011
2. the neural network algorithm-based underwater vehicle equilibrium submergence model simulation evaluation method according to claim 1, characterized in that,
if the error coefficient exceeds the specified range, changing the input parameters, and substituting the changed input parameters into the mathematical model for testing again;
when the submerged model reaches a set state, recording the changed input and output parameters as a new sample set;
carrying out numerical analysis on the new sample set by utilizing a neural network algorithm to obtain a new error coefficient between the mathematical model and an ideal value;
correcting the new error coefficient by using a weighting algorithm to obtain a corrected error coefficient;
evaluating the model using the corrected error coefficients.
3. The neural network algorithm-based underwater vehicle equilibrium submergence model simulation evaluation method according to claim 2, characterized in that,
the optimal parameter is obtainedaSo that the risk function Remp() Reaching a minimum value, including:
adding a steepest descent algorithm and an iterative algorithm;
according to the steepest descent algorithm and the iterative algorithmaAnd a minimum value Remp。
4. The simulation evaluation method for the underwater vehicle balanced submergence and floatation model based on the neural network algorithm is characterized in that the calculation formula of adding the steepest descent algorithm and the iterative algorithm is as follows:
(k+1)＝(k)+Δ(k) (2)
Figure FDA0002571268020000021
5. the neural network algorithm-based underwater vehicle equilibrium submergence model simulation evaluation method according to claim 4, characterized in that,
obtaining the optimal parameter according to the steepest descent algorithm and the iterative algorithmaAnd a minimum value RempThe method specifically comprises the following steps:
making a take the value, making | | | delta (k) | < 1, then:
Remp((k+1))＝Remp((k))+ΔRemp((k)) (4)
as can be seen from the formula (4), Remp((k)) decreases with increasing k, which tends towards Remp() A local minimum of (a);
taking the local minima points foremp() Is an optimum parametera。
6. The neural network algorithm-based underwater vehicle equilibrium submergence model simulation evaluation method according to claim 5, characterized in that,
will be provided withaAnd a minimum value RempAs a weight vector and a threshold parameter, the error coefficient is obtained by calculation, which specifically includes:
will be provided withaAnd a minimum value RempAs weight vector and threshold parameter, the following formula is substituted:
Figure FDA0002571268020000022
Figure FDA0002571268020000023
in the formula (5) and the formula (6),
Figure FDA0002571268020000031
the ith output value of the ith layer is represented, namely the error coefficient; xl-1Represents the output value of the l-1 layer;
Figure FDA0002571268020000032
represents the ith input value of the ith layer.
7. The neural network algorithm-based underwater vehicle equilibrium submergence model simulation evaluation method according to claim 6,
Figure FDA0002571268020000033
Figure FDA0002571268020000034
CN201710452977.0A 2017-06-15 2017-06-15 Neural network algorithm-based underwater vehicle equilibrium submerged model evaluation method Active CN107292015B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710452977.0A CN107292015B (en) 2017-06-15 2017-06-15 Neural network algorithm-based underwater vehicle equilibrium submerged model evaluation method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710452977.0A CN107292015B (en) 2017-06-15 2017-06-15 Neural network algorithm-based underwater vehicle equilibrium submerged model evaluation method

Publications (2)

Publication Number Publication Date
CN107292015A CN107292015A (en) 2017-10-24
CN107292015B true CN107292015B (en) 2020-09-01

Family

ID=60097059

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710452977.0A Active CN107292015B (en) 2017-06-15 2017-06-15 Neural network algorithm-based underwater vehicle equilibrium submerged model evaluation method

Country Status (1)

Country Link
CN (1) CN107292015B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109625333B (en) * 2019-01-03 2021-08-03 西安微电子技术研究所 Spatial non-cooperative target capturing method based on deep reinforcement learning

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102679982A (en) * 2012-04-06 2012-09-19 西北工业大学 Route planning method for autonomous underwater vehicle aiming at undetermined mission time
CN106780434A (en) * 2016-11-15 2017-05-31 天津大学 Underwater picture visual quality evaluation method

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140107839A1 (en) * 2012-10-16 2014-04-17 Massachusetts Institute Of Technology High efficiency, smooth robot design

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102679982A (en) * 2012-04-06 2012-09-19 西北工业大学 Route planning method for autonomous underwater vehicle aiming at undetermined mission time
CN106780434A (en) * 2016-11-15 2017-05-31 天津大学 Underwater picture visual quality evaluation method

Also Published As

Publication number Publication date
CN107292015A (en) 2017-10-24

Similar Documents

Publication Publication Date Title
CN109345875B (en) Estimation method for improving measurement accuracy of automatic ship identification system
Banazadeh et al. Frequency domain identification of the Nomoto model to facilitate Kalman filter estimation and PID heading control of a patrol vessel
CN104992002B (en) A kind of strain transducer layout method towards smart skins antenna
CN114218875A (en) Acceleration method and device for flow field prediction
Shafiei et al. Application of neural network and genetic algorithm in identification of a model of a variable mass underwater vehicle
CN110837680A (en) Underwater towing cable steady-state motion multi-objective optimization method and system
CN115496006A (en) High-precision numerical simulation method suitable for hypersonic aircraft
CN107292015B (en) Neural network algorithm-based underwater vehicle equilibrium submerged model evaluation method
Njaka et al. Method for improving existing maneuvering models to accommodate large drift angles
CN116992577B (en) Simulation method, system, equipment and storage medium of cross-medium aircraft
CN112414668A (en) Wind tunnel test data static bomb correction method, device, equipment and medium
CN114611420A (en) Unsteady aerodynamic force calculation precision evaluation and correction method
Zhou et al. A study of hybrid prediction method for ship parametric rolling
Kim et al. A study on the sensitivity analysis of the hydrodynamic derivatives on the maneuverability of kvlcc2 in shallow water
CN110937082A (en) Ship overturning risk testing method based on random wind field and sea waves
CN113761645A (en) Method, device and equipment for designing underwater vehicle shell
Peric et al. Trends in industry applications of computational fluid dynamics for maritime flows
CN111783957A (en) Model quantitative training method and device, machine-readable storage medium and electronic equipment
CN108711203B (en) Damaged hull wave load rapid forecasting method based on proxy model
Zhuoyi et al. Structure design of an autonomous underwater vehicle made of composite material
Koskinen Numerical simulation of ship motion due to waves and manoeuvring
CN111177855A (en) Pneumatic structure solving method and system in global aeroelasticity optimization
Lin et al. Study on resistance of multi-function small surface boat design
CN110926496B (en) Method, device and system for detecting motion abnormity of underwater vehicle
CN116702571B (en) Numerical simulation method and device based on multiple smoothness measurement factors

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant