CN116702095A - Modularized marine ship motion attitude real-time forecasting method - Google Patents
Modularized marine ship motion attitude real-time forecasting method Download PDFInfo
- Publication number
- CN116702095A CN116702095A CN202310643493.XA CN202310643493A CN116702095A CN 116702095 A CN116702095 A CN 116702095A CN 202310643493 A CN202310643493 A CN 202310643493A CN 116702095 A CN116702095 A CN 116702095A
- Authority
- CN
- China
- Prior art keywords
- forecasting
- data
- output
- matrix
- model
- 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.)
- Granted
Links
- 230000033001 locomotion Effects 0.000 title claims abstract description 92
- 238000013277 forecasting method Methods 0.000 title claims abstract description 15
- 238000013528 artificial neural network Methods 0.000 claims abstract description 47
- 238000010238 partial least squares regression Methods 0.000 claims abstract description 10
- 239000011159 matrix material Substances 0.000 claims description 48
- 238000000034 method Methods 0.000 claims description 34
- 230000006870 function Effects 0.000 claims description 31
- 239000013598 vector Substances 0.000 claims description 20
- 238000012549 training Methods 0.000 claims description 12
- 238000000354 decomposition reaction Methods 0.000 claims description 8
- 238000004590 computer program Methods 0.000 claims description 6
- 238000013507 mapping Methods 0.000 claims description 6
- 230000004044 response Effects 0.000 claims description 6
- 230000008859 change Effects 0.000 claims description 3
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 238000012217 deletion Methods 0.000 claims description 3
- 230000037430 deletion Effects 0.000 claims description 3
- 210000002569 neuron Anatomy 0.000 claims description 3
- 238000000605 extraction Methods 0.000 claims description 2
- 238000004422 calculation algorithm Methods 0.000 description 4
- 238000004088 simulation Methods 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 238000012795 verification Methods 0.000 description 3
- 230000008878 coupling Effects 0.000 description 2
- 238000010168 coupling process Methods 0.000 description 2
- 238000005859 coupling reaction Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000007613 environmental effect Effects 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 230000002411 adverse Effects 0.000 description 1
- 238000009395 breeding Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005312 nonlinear dynamic Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000001556 precipitation Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000005096 rolling process Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The application provides a modularized marine ship motion attitude real-time forecasting method, which relates to the technical field of ship motion attitude forecasting and comprises the following steps: s1, acquiring ship motion six-degree-of-freedom motion element data of a target ship; s2, forecasting by using a recursive partial least squares regression model to obtain a first forecasting result; s3, obtaining an approximate component and a detail component; s4, carrying out time sequence forecasting on each approximate component and detail component by using a variable-structure radial basis function neural network; s5, establishing a ship motion prediction model based on a radial basis function neural network, and predicting by using the ship motion prediction model based on the radial basis function neural network to obtain a second prediction result; and S6, superposing the first forecasting result and the second forecasting result to obtain a final modularized ship motion forecasting result. The application combines the modularized forecasting strategy of the RPLS model and the VRBFN model, thereby improving the forecasting precision while guaranteeing the forecasting stability.
Description
Technical Field
The application relates to the technical field of ship motion attitude prediction, in particular to a modularized marine ship motion attitude real-time prediction method.
Background
The real-time prediction of the ship in the marine attitude plays an important role in ship steering in high storms, take-off and landing of an aircraft on a deck, retraction and extension of an underwater vehicle, offshore refuting, cable retraction and other offshore construction operations. However, when the ship sails on the sea, the ship is influenced by environmental factors such as wind, waves, currents and the like, and the swinging motion of the ship has the complex characteristics of nonlinearity, dynamic time variation and the like, so that the forecasting difficulty is increased.
The current general ship motion modularization forecasting is a method combining a linear model and a nonlinear model. Firstly, building and training a linear model, building a regression equation of ship motion sample data according to the linear model and obtaining parameters of the linear model, and building a ship motion forecast linear model; training nonlinear models such as a neural network and the like by utilizing the difference value between the measured value and the linear model, and determining the structure and parameters of the nonlinear model; and correspondingly adding the ship motion forecast value obtained by using the ship motion forecast nonlinear model and the regression error forecast value by using the linear model to obtain a final ship motion forecast value. The combination has the advantages that the linear model of the ship motion is established by utilizing the linear model, the online identification model of the ship motion is established by utilizing technologies such as a data-driven neural network and the like, and the ship motion prediction system with higher precision is established on the premise of ensuring the stability by combining the advantages of strong interpretation and strong stability of the mechanism model and the characteristics of strong nonlinear performance of the identification model such as the neural network and the like and being capable of reflecting the real-time dynamics of the ship motion.
However, the current ship movement modularization forecasting method has certain defects. First, it is difficult to obtain an accurate linear model of ship motion with current forecasting methods. Because the ship motion online data are acquired in real time, serious multiple collinearity exists among partial variables, the serious uncertainty exists in the model parameter estimation, and the variance of the parameter estimation is close to infinity under serious conditions, so that the application of the parameter model in ship motion prediction is directly limited. Secondly, the non-parametric model in the current forecasting method often has difficulty in reflecting the complex influence of time-varying factors on the ship motion. For example, dynamic changes of time-varying factors such as hydrologic (wave, ocean current, tide, temperature, salinity and density) weather (wind force, wind direction, air temperature, air pressure and precipitation) of a sea area where a ship is located influence the time-varying state of the ship motion dynamics, and static nonlinear prediction models such as neural networks with fixed network scale, structure and weight are difficult to reflect the time-varying dynamics of the ship motion, so that the precision of ship motion prediction is difficult to ensure.
Disclosure of Invention
In view of the above, the present application is directed to a modular marine vessel motion gesture real-time prediction method, which updates a vessel motion Recursive Partial Least Squares (RPLS) prediction model in real time by adopting an iterative partial least squares learning strategy, and adjusts the structure and parameters of a variable structure radial basis function network (VRBFN) prediction model by using a sliding data window. Only iterative computation is needed to obtain the VRBFN fast when the VRBFN parameters are updated each time, and the partial least square regression method overcomes the multiple collinearity of the data; when VRBFN is updated, the network structure, parameters and orders are updated, and the adaptability to time-varying dynamics is ensured. By combining the modularized forecasting strategy of the RPLS model and the VRBFN model, the forecasting stability is ensured and the forecasting precision is improved.
The application adopts the following technical means:
a modularized marine ship motion attitude real-time forecasting method comprises the following steps:
s1, acquiring ship motion six-degree-of-freedom motion element data of a target ship;
s2, building and training a recursive partial least square model, and forecasting by using the recursive partial least square regression model to obtain a first forecasting result;
s3, performing difference between the measured data and the first forecasting result, and performing multi-scale wavelet decomposition on the difference between the measured data and the output value of the partial least square model to obtain an approximate component and a detail component;
s4, determining the order of time sequence forecasting of each approximate component and each detail component, and forecasting the time sequence of each approximate component and each detail component by using a radial basis function neural network of a variable structure;
s5, establishing a ship motion prediction model based on a radial basis function neural network, and predicting each approximate component and each detail component obtained by wavelet decomposition by using the ship motion prediction model based on the radial basis function neural network to obtain a second prediction result;
and S6, superposing the first forecasting result and the second forecasting result to obtain a final modularized ship motion forecasting result.
Further, the six-degree-of-freedom motion element data of the ship motion comprises ship six-degree-of-freedom attitude information and ship speed information, wherein the ship six-degree-of-freedom attitude information comprises a longitudinal position X, a transverse position Y, a vertical position Z and a roll angle of the shipThe ship speed information comprises a longitudinal speed mu, a transverse speed v and a head-shake angular speed r.
Further, S2 includes the following steps:
according to the variables needing d-step advanced prediction, respectively selecting related variables and orders from six-degree-of-freedom motion element data of the ship motion by using a partial correlation coefficient method as input of a recursive partial least square model, and establishing the recursive partial least square model;
let X (∈R) n×m ) For the input of the recursive partial least squares model, Y (∈R) n×p ) For the output of the recursive partial least squares model, the partial least squares method completes the following mapping:
wherein t is 1 And u 1 Is the score vector of the first partial least squares factor, p 1 And q 1 Is the load vector corresponding to the score vector, E 1 And F 1 Is the residual;
four vectors t 1 、u 1 、p 1 And q 1 By minimizing E 1 And F 1 Determining a score vector t 1 And u 1 The internal relationship between is obtained by a linear model:
u 1 ＝b 1 t 1 +ε 1
when t 1 And u 1 When the extraction information is insufficient, iteratively calculating further scores and loading vectors by extracting information from the residual matrix;
let the rank of X be r, after calculating r, Y be:
X＝TP T ,Y＝TBQ T +F r
due to the column of T and the output residual F r Orthogonal and mutually orthogonal, and therefore:
X T X＝PT T TP T
X T Y＝PT T TBQ T +F r ＝PBQ T
x and Y are represented by a parameter matrix of P, B and Q, and when a new data pair appears, the PLS regression formula is as follows:
the PLS regression equation for all data containing new data is as follows:
the recursive partial least squares model is implemented by updating the original obtained model and the new data.
Further, S5 includes the following steps:
s51, establishing a sliding data window to observe the ship motion state, and dynamically adjusting a fitting model based on a radial basis function neural network by utilizing input and output data updated in real time;
the sliding window is a fixed-width first-in first-out data sample sequence, when a new group of input-output data is received, the new data group is added into the sliding window, and the earliest group of data is moved out of the sliding window; sliding window W at t moment SD Expressed as:
W SD ＝[(x t-L+1 ,x t-L+1 ),...,(y t ,y t )],
wherein L is the width of the sliding window; the real-time dynamics of the mapping relationship are represented by the input-output data sets in the sliding window, namely by the input matrix P and the corresponding output vector Q respectively:
Y＝[y t-L+1 ...y t ]
wherein n is p The dimension of the input matrix;
respectively using an input matrix P and a corresponding output Q as the input and the output of the radial basis function neural network, and training the neural network and dynamically adjusting hidden nodes of the network;
s52, in the aspect of increasing the strategy, the data which initially appears in the sliding window is used as an initial hidden node center; after each step receives new data samples, updating the sliding data window, adding the latest samples into the window, and deleting the earliest samples from the window; the structure adjustment of the variable structure neural network is divided into an adding strategy and a deleting strategy of hidden nodes;
in the hidden layer neuron addition strategy, when new sample data enters a sliding window, input sample data satisfying the following three conditions is selected as a hidden node center:
wherein the first condition is an input novelty criterion, the second condition is an output novelty criterion, and the third condition is a continuous novelty criterion;
wherein u is r (t) is the off-input x t E is closest to the center, e t Epsilon and e are the errors between the ideal value and the network output min Input and output novelty determination thresholds, e rms Is the root mean square error of the error between the selected ideal value and the network output, L is the width of the sliding data window, e max Is the iterative mean square error e rms Threshold value of (2), threshold value e max Is introduced to check whether the mean square error of the last L continuous outputs meets the requirement value, so that the change of the number of hidden programs is smoother, wherein f (x) t ) Is relative toInput x t The values of the neural network inputs of (1) are:
wherein I II the euclidean distance is represented as, sigma is the base function width; m is the number of hidden nodes;
s53, deleting hidden nodes with small contribution to output in continuous training steps on the deletion strategy of the hidden nodes, and calculating standardized contribution r of each hidden node in each step m (t)：
Wherein o is m (t)＝α m f(x t ),o max (t) is the largest |o among different hidden nodes max A value corresponding to (t) |;
if r of hidden node m The value of (t) is in the range of N w The step of calculation is smaller than a set threshold delta, and the hidden node is deleted from the hidden layer;
s54, after each step of hidden node determination, updating the connection weight from the hidden layer to the output layer;
the connection weight from the hidden layer matrix phi of the obtained radial basis function neural network to the output layer Q is obtained by a method of solving partial least square;
performing partial least square regression operation between the response matrix phi and the output matrix Y; after extracting the principal component matrix T, respectively projecting the response matrix phi and the output matrix Y onto the principal component matrix T to obtain a radial basis function neural network based on partial least squares regression, wherein the radial basis function neural network is as follows:
Q＝TR+F＝ΦWR+F
wherein T is the principal component matrix of phi; w is the conversion matrix of phi; r is regression coefficient matrix; f is the residual matrix.
The application also provides a storage medium comprising a stored program, wherein the program, when run, performs the modular marine vessel motion attitude real-time forecasting method of any one of the above.
The application also provides an electronic device, which comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, and is characterized in that the processor runs and executes the modularized marine ship motion attitude real-time forecasting method according to any one of the above steps through the computer program.
Compared with the prior art, the application has the following advantages:
compared with the traditional ship motion prediction based on a ship motion mathematical model and the nonlinear prediction based on a simple neural network, the method fully utilizes the advantages of stability and strong robustness of the linear prediction method and the self-adaption and nonlinearity of the nonlinear prediction method based on the neural network, combines the two methods, and can improve the prediction precision while guaranteeing the stability.
The method fully excavates the time sequence dynamic information of the actually measured ship motion state. The model established by the method reflects time-varying dynamics of ship motion, the adverse effect of multiple collinearity of data in real-time prediction is overcome by using a recursive partial least square method through information mining and time sequence prediction of the information, and more accurate marine ship motion prediction can be obtained by using the nonlinear dynamic fitting capability of a variable structure radial basis function neural network. In the traditional model-based prediction method, the established model cannot reflect complex influences of time-varying aperiodic hydrologic factors and the like, so that the situation that the prediction accuracy is not high occurs, and a large prediction error is easy to occur under the situation that the influence of external environmental factors is large.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described, and it is obvious that the drawings in the following description are some embodiments of the present application, and other drawings may be obtained according to the drawings without inventive effort to a person skilled in the art.
FIG. 1 is a flow chart of the method of the present application.
FIG. 2 is a schematic view of six degrees of freedom motion of the marine vessel according to the present application.
FIG. 3 is a flow chart of the modular model training of the present application.
FIG. 4 is a flow chart of modular model prediction according to the present application.
Detailed Description
In order that those skilled in the art will better understand the present application, a technical solution in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in which it is apparent that the described embodiments are only some embodiments of the present application, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the present application without making any inventive effort, shall fall within the scope of the present application.
It should be noted that the terms "first," "second," and the like in the description and the claims of the present application and the above figures are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that the embodiments of the application described herein may be implemented in sequences other than those illustrated or otherwise described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
As shown in fig. 1-4, the present application provides a modular marine vessel attitude real-time forecasting method. Firstly, building and training a Recursive Partial Least Squares (RPLS) model; according to the RPLS model, a partial least square regression equation of ship motion sample data is established, and a ship motion linear prediction model is established; establishing a variable structure radial basis function network (VRBFN) nonlinear prediction model; and correspondingly superposing a ship motion forecast value obtained by using the RPLS forecast model and a VRBFN model regression error forecast value to obtain a final ship motion forecast value. The application adopts an iterative partial least square learning strategy to update the ship motion RPLS forecasting model in real time, and utilizes a sliding data window to adjust the structure and parameters of the VRBFN forecasting model. Only iterative computation is needed to obtain the VRBFN fast when the VRBFN parameters are updated each time, and the partial least square regression method overcomes the multiple collinearity of the data; when VRBFN is updated, the network structure, parameters and orders are updated, and the adaptability to time-varying dynamics is ensured. By combining the modularized forecasting strategy of the RPLS model and the VRBFN model, the forecasting stability is ensured and the forecasting precision is improved.
The specific steps are as follows:
1. data acquisition
The six-degree-of-freedom motion element data of the ship motion is collected as standby information, and comprises (1) six-degree-of-freedom attitude information of the ship, including a longitudinal position X, a transverse position Y, a vertical position Z and a roll angle of the shipA pitch angle theta and a yaw angle phi; (2) the ship speed information comprises data of a longitudinal speed mu, a transverse speed v and a head-shake angular speed r.
2. Establishing partial least square model
According to the variables of the d-step advanced forecast, the related variables and orders are respectively selected in the ship motion state data by using a partial correlation coefficient method, such as the roll of ship motionThe selection may be:
as an input to the d-step lead forecast model,and (3) establishing a ship motion d-step advanced prediction model based on the partial least square method for output.
Let X (∈R) n×m ) And Y (∈R) n×p ) Is the model input and output, and the PLS method is to accomplish the following mapping:
wherein t1 and u1 are score vectors of the first PLS factor, p 1 And q 1 Is the corresponding load vector, E 1 And F 1 Is the residual. Four vectors t1, u1, p 1 And q 1 All by minimizing E 1 And F 1 To determine. The above equation is an external model of PLS, and the internal relationship between score vectors t1 and u1 is obtained by a linear model:
u 1 ＝b 1 t 1 +ε 1
if t 1 And u 1 Without extracting sufficient information, further score and load vectors are iteratively calculated by extracting information from the residual matrix. Assuming that the rank of X is r, after calculating r, Y can be written as:
X＝TP T ,Y＝TBQ T +F r
due to the column of T and the output residual F r Orthogonal and mutually orthogonal, it can therefore be deduced that:
X T X＝PT T TP T
X T Y＝PT T TBQ T +F r ＝PBQ T
so X and Y can be represented by a parameter matrix of P, B and Q. PLS regression when new pairs of data appear
PLS regression equivalent to all data containing new data:
therefore, the new PLS model is realized by updating the original obtained model and the new data, and the iterative calculation avoids repeated calculation of large-batch data, thereby effectively reducing the calculated amount.
3. Multi-scale wavelet decomposition and information screening
And carrying out multi-scale wavelet decomposition on the difference value output by the actual measurement data and the partial least square model.
The motions of the vessel at sea are affected by a number of factors, which are superimposed on each other, so that the vessel motions exhibit complex motion characteristics. By wavelet decomposition, the original ship motion time series x (t) can be divided into a plurality of sub-sequences including approximate components (A K ) And detail component (D) 1 ,D 2 ,…,D K ) This facilitates the input of the subsequence as a model. In the scheme, taking ship movement roll as an example, the roll angle of the ship movement is measuredAnd yaw angle +.>The difference r between them is subjected to wavelet decomposition to obtain an approximation component and a detail component.
And determining the order of the time sequence forecast of each component by calculating Lipschitz coefficients and the like, and after determining the input/output order, carrying out the time sequence forecast of each component by using a variable-structure radial basis function neural network.
4. Ship motion prediction model based on radial basis function neural network
Establishing a sliding data window
The motion of the ship at sea has the characteristic of dynamic time variation, in order to reflect the latest ship motion state, a sliding data window is established to observe the ship motion state, and a fitting model based on a radial basis function neural network is dynamically adjusted by utilizing input and output data updated in real time;
the sliding window is a fixed-width first-in first-out data sample sequence, when a new input-output data set is received, the new data set is added into the sliding window, and the earliest data set is moved out of the sliding window. Sliding window W at t moment SD Expressed as:
W SD ＝[(x t-L+1 ,x t-L+1 ),...,(y t ,y t )],
wherein L is the width of the sliding window; the real-time dynamics of the mapping relationship are represented by the input-output data sets in the sliding window, namely by the input matrix P and the corresponding output vector Q respectively:
Y＝[y t-L+1 ...y t ]
wherein n is p The dimension of the input matrix;
the input matrix P and the corresponding output Q are used as the input and the output of the radial basis function neural network respectively, and training and dynamic adjustment of hidden nodes of the neural network are carried out.
In an incremental strategy, the data that initially appears within the sliding window serves as the initial hidden node center. After each step receives a new data sample, the sliding data window is updated, the latest sample is added to the window, and the earliest sample is deleted from the window. The structure adjustment of the variable structure neural network is divided into an adding strategy and a deleting strategy of hidden nodes.
In the hidden layer neuron addition strategy, when new sample data enters a sliding window, input sample data (x t ,y t ) Selected as the hidden node center:
wherein the first condition is to input a novelty criterion, the second condition is to output a novelty criterion, and the third condition is to continue the novelty criterion.
Wherein u is r (t) is the off-input x t E is closest to the center, e t Epsilon and e are the errors between the ideal value and the network output min Input and output novelty determination thresholds, e rms Is the root mean square error of the error between the selected ideal value and the network output, L is the width of the sliding data window, e max Is the iterative mean square error e rms Is set to a threshold value of (2). Threshold e max The introduction of (2) is to check whether the mean square error of the last L continuous outputs can meet the requirement value, so that the change of the number of hidden nodes is smoother. Wherein f (x) t ) For input x t The values of the neural network inputs of (1) are:
wherein I II the euclidean distance is represented as, sigma is the base function width; m is the number of hidden nodes.
On the deletion strategy of hidden nodes, those hidden nodes which have small contributions to the output in consecutive training steps are deleted. Calculating normalized contribution r of each hidden node at each step m (t)：
Wherein o is m (t)＝α m f(x t ),o max (t) is the largest |o among different hidden nodes max (t) | corresponding value.
If r of hidden node m The value of (t) is in the range of N w The step is calculated to be less than the set threshold delta, and the hidden node is deleted from the hidden layer.
After each step of hidden node determination, updating the connection weight value from the hidden layer to the output layer.
The connection weight of the hidden layer matrix phi of the obtained radial basis function neural network to the output layer Q is obtained by a method of solving partial least square.
Performing partial least square regression operation between the response matrix phi and the output matrix Y; after extracting a principal component matrix T, respectively projecting a response matrix phi and an output matrix Y onto the principal component matrix T to obtain a radial basis function neural network based on partial least squares regression, wherein the radial basis function neural network is as follows:
Q＝TR+F＝ΦWR+F
wherein T is the principal component matrix of phi; w is the conversion matrix of phi; r is regression coefficient matrix; f is the residual matrix.
5. Ship movement modularized forecast output
And after forecasting by using a recursive partial least square regression model and a variable structure radial basis function neural network, overlapping the two forecasting results to obtain a final modular ship motion forecasting result.
The application also provides a storage medium comprising a stored program, wherein the program, when running, executes the modular marine vessel motion attitude real-time forecasting method.
The application also provides an electronic device, which comprises a memory, a processor and a computer program stored in the memory and capable of running on the processor, and is characterized in that the processor runs and executes the modularized marine ship motion attitude real-time forecasting method through the computer program.
And (3) performing simulation verification on the online ship rolling by using the actual measurement data of the marine test of the spread-breeding wheel of the university of Dalian maritime. The ship coefficients of the 'yun kun' wheel are: the total length of the ship is 116m, the length between two columns is 105m, the model width is 18m, the total ton is 6106, the design navigational speed is 16.9kn, and the design draft is 5.4m.
The related variables of the ship motion attitude are ship roll phi, ship trim theta, ship bow phi and ship heave zeta. The input of the variable structure neural network is the previous ship motion state data phi (t-1), phi (t-2), theta (t-1), theta (t-2), phi (t-1), phi (t-2), zeta (t-1) and zeta (t-2); the output is the current vessel roll phi (t). The results of the simulation verification are shown in the following table.
| Algorithm | Identifying root mean square error (°) | Prediction root mean square error (°) |
| Combined prediction algorithm | 0.5128 | 0.5132 |
| Autoregressive linear method | 0.7139 | 0.6598 |
| BP neural network | 0.6411 | 0.5746 |
| Online sequential extreme learning machine | 0.5469 | 0.5361 |
The same tide data is utilized to carry out simulation verification by utilizing an autoregressive linear method, a BP neural network, an online sequential extreme learning machine and the like. Simulation results show that the prediction precision of the combined prediction algorithm based on the variable structure neural network is higher than that of other comparison prediction algorithms.
The foregoing embodiment numbers of the present application are merely for the purpose of description, and do not represent the advantages or disadvantages of the embodiments.
In the foregoing embodiments of the present application, the descriptions of the embodiments are emphasized, and for a portion of this disclosure that is not described in detail in this embodiment, reference is made to the related descriptions of other embodiments.
In the several embodiments provided in the present application, it should be understood that the disclosed technology may be implemented in other manners. The above-described embodiments of the apparatus are merely exemplary, and the division of the units, for example, may be a logic function division, and may be implemented in another manner, for example, a plurality of units or components may be combined or may be integrated into another system, or some features may be omitted, or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be through some interfaces, units or modules, or may be in electrical or other forms.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
In addition, each functional unit in the embodiments of the present application may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit. The integrated units may be implemented in hardware or in software functional units.
The integrated units, if implemented in the form of software functional units and sold or used as stand-alone products, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present application may be embodied essentially or in part or all of the technical solution or in part in the form of a software product stored in a storage medium, including instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a Read-only memory (ROM), a random access memory (RAM, random Access Memory), a removable hard disk, a magnetic disk, or an optical disk, or other various media capable of storing program codes.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present application, and not for limiting the same; although the application has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some or all of the technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit of the application.
Claims (6)
1. The real-time forecasting method for the motion gesture of the modularized marine ship is characterized by comprising the following steps of:
s1, acquiring ship motion six-degree-of-freedom motion element data of a target ship;
s2, building and training a recursive partial least square model, and forecasting by using the recursive partial least square regression model to obtain a first forecasting result;
s3, performing difference between the measured data and the first forecasting result, and performing multi-scale wavelet decomposition on the difference between the measured data and the output value of the partial least square model to obtain an approximate component and a detail component;
s4, determining the order of time sequence forecasting of each approximate component and each detail component, and forecasting the time sequence of each approximate component and each detail component by using a radial basis function neural network of a variable structure;
s5, establishing a ship motion prediction model based on a radial basis function neural network, and predicting each approximate component and each detail component obtained by wavelet decomposition by using the ship motion prediction model based on the radial basis function neural network to obtain a second prediction result;
and S6, superposing the first forecasting result and the second forecasting result to obtain a final modularized ship motion forecasting result.
2. The method for real-time forecasting of a motion attitude of a modular marine vessel according to claim 1, wherein the six degrees of freedom motion element data of the vessel motion comprises six degrees of freedom attitude information of the vessel and vessel speed information, and the six degrees of freedom attitude information of the vessel comprises a longitudinal position X, a transverse position Y, a vertical position Z and a roll angle of the vesselThe ship speed information comprises a longitudinal speed mu, a transverse speed v and a head-shake angular speed r.
3. The method for real-time forecasting of the motion attitude of a modular marine vessel according to claim 1, wherein S2 comprises the steps of:
according to the variables needing d-step advanced prediction, respectively selecting related variables and orders from six-degree-of-freedom motion element data of the ship motion by using a partial correlation coefficient method as input of a recursive partial least square model, and establishing the recursive partial least square model;
let X (∈R) n×m ) For the input of the recursive partial least squares model, Y (∈R) n×p ) For the output of the recursive partial least squares model, the partial least squares method completes the following mapping:
wherein t is 1 And u 1 Is the score vector of the first partial least squares factor, p 1 And q 1 Is the load vector corresponding to the score vector, E 1 And F 1 Is the residual;
four vectors t 1 、u 1 、p 1 And q 1 By minimizing E 1 And F 1 Determining a score vector t 1 And u 1 The internal relationship between is obtained by a linear model:
u 1 ＝b 1 t 1 +ε 1
when t 1 And u 1 When the extraction information is insufficient, iteratively calculating further scores and loading vectors by extracting information from the residual matrix;
let the rank of X be r, after calculating r, Y be:
X＝TP T ,Y＝TBQ T +F r
due to the column of T and the output residual F r Orthogonal and mutually orthogonal, and therefore:
X T X＝PT T TP T
X T Y＝PT T TBQ T +F r ＝PBQ T
x and Y are represented by a parameter matrix of P, B and Q, and when a new data pair appears, the PLS regression formula is as follows:
the PLS regression equation for all data containing new data is as follows:
the recursive partial least squares model is implemented by updating the original obtained model and the new data.
4. The method for real-time forecasting of the motion attitude of a modular marine vessel according to claim 1, wherein S5 comprises the steps of:
s51, establishing a sliding data window to observe the ship motion state, and dynamically adjusting a fitting model based on a radial basis function neural network by utilizing input and output data updated in real time;
the sliding window is a fixed-width first-in first-out data sample sequence, when a new group of input-output data is received, the new data group is added into the sliding window, and the earliest group of data is moved out of the sliding window; sliding window W at t moment SD Expressed as:
W SD ＝[(x t-L+1 ,x t-L+1 ),...,(y t ,y t )],
wherein L is the width of the sliding window; the real-time dynamics of the mapping relationship are represented by the input-output data sets in the sliding window, namely by the input matrix P and the corresponding output vector Q respectively:
Y＝[y t-L+1 ...y t ]
wherein n is p The dimension of the input matrix;
respectively using an input matrix P and a corresponding output Q as the input and the output of the radial basis function neural network, and training the neural network and dynamically adjusting hidden nodes of the network;
s52, in the aspect of increasing the strategy, the data which initially appears in the sliding window is used as an initial hidden node center; after each step receives new data samples, updating the sliding data window, adding the latest samples into the window, and deleting the earliest samples from the window; the structure adjustment of the variable structure neural network is divided into an adding strategy and a deleting strategy of hidden nodes;
in the hidden layer neuron addition strategy, when new sample data enters a sliding window, input sample data satisfying the following three conditions is selected as a hidden node center:
wherein the first condition is an input novelty criterion, the second condition is an output novelty criterion, and the third condition is a continuous novelty criterion;
wherein u is r (t) is the off-input x t E is closest to the center, e t Epsilon and e are the errors between the ideal value and the network output min Input and output novelty determination thresholds, e rms Is the root mean square error of the error between the selected ideal value and the network output, L is the width of the sliding data window, e max Is the iterative mean square error e rms Threshold value of (2), threshold value e max Is introduced to check whether the mean square error of the last L continuous outputs meets the requirement value, so that the change of the number of hidden programs is smoother, wherein f (x) t ) For input x t The values of the neural network inputs of (1) are:
wherein I II the euclidean distance is represented as, sigma is the base function width; m is the number of hidden nodes;
s53, deleting hidden nodes with small contribution to output in continuous training steps on the deletion strategy of the hidden nodes, and calculating standardized contribution r of each hidden node in each step m (t)：
Wherein o is m (t)＝α m f(x t ),o max (t) is the largest |o among different hidden nodes max A value corresponding to (t) |;
if r of hidden node m The value of (t) is in the range of N w The steps are all smaller than the setA threshold delta, the hidden node is to be deleted from the hidden layer;
s54, after each step of hidden node determination, updating the connection weight from the hidden layer to the output layer;
the connection weight from the hidden layer matrix phi of the obtained radial basis function neural network to the output layer Q is obtained by a method of solving partial least square;
performing partial least square regression operation between the response matrix phi and the output matrix Y; after extracting the principal component matrix T, respectively projecting the response matrix phi and the output matrix Y onto the principal component matrix T to obtain a radial basis function neural network based on partial least squares regression, wherein the radial basis function neural network is as follows:
Q＝TR+F＝ΦWR+F
wherein T is the principal component matrix of phi; w is the conversion matrix of phi; r is regression coefficient matrix; f is the residual matrix.
5. A storage medium comprising a stored program, wherein the program, when run, performs the modular marine vessel motion profile real-time forecasting method of any one of claims 1-4.
6. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor is operative with the computer program to perform the modular marine vessel motion profile real-time forecasting method of any one of claims 1-4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310643493.XA CN116702095B (en) | 2023-06-01 | 2023-06-01 | Modularized marine ship motion attitude real-time forecasting method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310643493.XA CN116702095B (en) | 2023-06-01 | 2023-06-01 | Modularized marine ship motion attitude real-time forecasting method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116702095A true CN116702095A (en) | 2023-09-05 |
| CN116702095B CN116702095B (en) | 2024-04-05 |
Family
ID=87828567
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310643493.XA Active CN116702095B (en) | 2023-06-01 | 2023-06-01 | Modularized marine ship motion attitude real-time forecasting method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116702095B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117494057A (en) * | 2023-12-29 | 2024-02-02 | 大连海事大学 | Ship motion prediction method based on hierarchical multi-scale decomposition |
| CN117669396A (en) * | 2024-02-02 | 2024-03-08 | 陕西欧卡电子智能科技有限公司 | Unmanned ship movement detection method and device, unmanned ship and storage medium |
| CN117893575A (en) * | 2024-03-15 | 2024-04-16 | 青岛哈尔滨工程大学创新发展中心 | Ship motion prediction method and system with self-attention mechanism integrated by graph neural network |
Family Cites Families (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR100584074B1 (en) * | 2004-07-28 | 2006-05-29 | 학교법인 울산공업학원 | non-linear data classification method using partial least square and residual compensator |
| CN101446828A (en) * | 2008-12-30 | 2009-06-03 | 东北大学 | Nonlinear process quality prediction method |
| CN101872163B (en) * | 2010-05-21 | 2012-02-01 | 杭州电子科技大学 | Process monitoring method based on recursive non-linear partial least squares |
| CN101872182A (en) * | 2010-05-21 | 2010-10-27 | 杭州电子科技大学 | Batch process monitoring method based on recursive non-linear partial least square |
| RU2690001C1 (en) * | 2017-12-29 | 2019-05-30 | Федеральное государственное бюджетное образовательное учреждение высшего образования "Московский государственный университет имени М.В. Ломоносова" (МГУ) | Method of processing vector signals for pattern recognition based on wavelet analysis |
| CN110609476B (en) * | 2019-09-19 | 2022-04-12 | 太原理工大学 | Multivariate nonlinear dynamic system model prediction control method based on Gaussian process model |
| CN112906955B (en) * | 2021-02-04 | 2022-06-21 | 广东海洋大学 | Tidal data fitting method based on multi-source data driving |
| CN113156815B (en) * | 2021-03-10 | 2022-05-20 | 广东海洋大学 | Data-driven marine ship motion attitude real-time forecasting method |
| CN114298382A (en) * | 2021-12-13 | 2022-04-08 | 燕山大学 | Method for forecasting motion pose of ship body in extremely short term and real time |
-
2023
- 2023-06-01 CN CN202310643493.XA patent/CN116702095B/en active Active
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117494057A (en) * | 2023-12-29 | 2024-02-02 | 大连海事大学 | Ship motion prediction method based on hierarchical multi-scale decomposition |
| CN117669396A (en) * | 2024-02-02 | 2024-03-08 | 陕西欧卡电子智能科技有限公司 | Unmanned ship movement detection method and device, unmanned ship and storage medium |
| CN117669396B (en) * | 2024-02-02 | 2024-05-24 | 陕西欧卡电子智能科技有限公司 | Unmanned ship movement detection method and device, unmanned ship and storage medium |
| CN117893575A (en) * | 2024-03-15 | 2024-04-16 | 青岛哈尔滨工程大学创新发展中心 | Ship motion prediction method and system with self-attention mechanism integrated by graph neural network |
| CN117893575B (en) * | 2024-03-15 | 2024-05-31 | 青岛哈尔滨工程大学创新发展中心 | Ship motion prediction method and system with self-attention mechanism integrated by graph neural network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116702095B (en) | 2024-04-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN116702095B (en) | Modularized marine ship motion attitude real-time forecasting method | |
| Yin et al. | A real-time ship roll motion prediction using wavelet transform and variable RBF network | |
| Yin et al. | On-line prediction of ship roll motion during maneuvering using sequential learning RBF neuralnetworks | |
| CN113156815B (en) | Data-driven marine ship motion attitude real-time forecasting method | |
| CN112684701A (en) | Ship motion forecasting method based on long-time and short-time memory network and Gaussian process regression | |
| US5649064A (en) | System and method for modeling the flow performance features of an object | |
| CN107944648B (en) | Large ship speed and oil consumption rate prediction method | |
| CN114021441A (en) | Ship motion attitude prediction method based on CNN-BilSTM | |
| CN110806692A (en) | Wave compensation prediction method based on CNN-LATM combined model | |
| CN114692501A (en) | Pneumatic data fusion method and device based on multi-precision deep neural network | |
| CN103400158B (en) | Based on the level set tracking of dynamic shape code book study | |
| CN112802011A (en) | Fan blade defect detection method based on VGG-BLS | |
| CN117494057A (en) | Ship motion prediction method based on hierarchical multi-scale decomposition | |
| CN106971189A (en) | A kind of noisy method for recognising star map of low resolution | |
| CN116861202A (en) | Ship motion envelope forecasting method and system based on long-term and short-term memory neural network | |
| CN114298382A (en) | Method for forecasting motion pose of ship body in extremely short term and real time | |
| CN116822336B (en) | Multi-model combined tide forecasting method | |
| CN116150934A (en) | Ship maneuvering Gaussian process regression online non-parameter identification modeling method | |
| EP3956733A1 (en) | Method for computer-implemented determination of a drag coefficient of a wind turbine | |
| CN112800075A (en) | Ship operation forecast database updating method based on real ship six-degree-of-freedom attitude data | |
| CN117109580A (en) | AUV navigation method based on transducer model | |
| CN114936669B (en) | Mixed ship rolling prediction method based on data fusion | |
| CN116026325A (en) | Navigation method and related device based on neural process and Kalman filtering | |
| CN115994481A (en) | Ship motion attitude prediction method based on multiple combinations | |
| Bisinotto et al. | Sea state estimation with neural networks based on the motion of a moored fpso subjected to campos basin metocean conditions |
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 |