CN114936429A

CN114936429A - Resistance prediction network training method, ship resistance prediction method and related device

Info

Publication number: CN114936429A
Application number: CN202210712961.XA
Authority: CN
Inventors: 蔡睿眸; 彭震; 黄琼; 王涛; 刘天夫; 徐超
Original assignee: Guangzhou Shipyard International Co Ltd
Current assignee: Guangzhou Shipyard International Co Ltd
Priority date: 2022-06-22
Filing date: 2022-06-22
Publication date: 2022-08-23

Abstract

The invention discloses a method for training a resistance prediction network and predicting resistance of a ship and a related device, wherein a plurality of first long-short term memory networks and second long-short term memory networks are trained by acquiring three-dimensional first simulation flow field data of a simulated ship in the sailing process, the first long-short term memory network outputs a plurality of predicted first resistances, the second long-short term memory network takes the spliced first resistance as input and outputs a second resistance predicted for the ship, the plurality of first long-short term memory networks and the second long-short term memory networks are updated through the second resistance, the plurality of first long-short term memory networks and the second long-short term memory networks which accord with preset evaluation standards are output as multi-scale resistance prediction networks, and due to the fact that the number of nodes among the plurality of first long-short term memory networks is different, the first long-short term memory networks with different node numbers learn time sequence characteristics of the first simulation flow field data in different scales through the first long-short term memory networks with different node numbers, the effect of sufficient learning is achieved, and the accuracy of resistance prediction is improved.

Description

Training of resistance prediction network, resistance prediction method of ship and related device

Technical Field

The invention relates to the technical field of ships, in particular to a resistance prediction network training method, a ship resistance prediction method and a related device.

Background

In the three-dimensional simulation of a ship, the resistance to which the ship is subjected during navigation is one of important factors to be considered in designing and optimizing the ship shape of the ship.

At present, in the three-dimensional simulation process of a ship, a prediction method of resistance borne by the ship mainly comprises two modes of CFD (Computational Fluid Dynamics) and long-short term memory network. When the resistance of the ship is predicted in a CFD mode, the resistance of the ship of a three-dimensional body is usually solved through a momentum conservation equation, an energy conservation equation, turbulent kinetic energy and a turbulent kinetic energy dissipation rate equation of fluid flow, the calculation process is long in time consumption and high in calculation cost, a high-dimensional large data set is often generated when the accuracy of the resistance calculated in the CFD mode is improved, the large data set contains a large number of degrees of freedom, and a large amount of calculation time and calculation resources are consumed for processing the degrees of freedom. When the resistance borne by the ship is predicted through the long-short term memory network, the long-short term memory network is often selected, a large amount of running data of the ship in operation is input into the network for training according to the characteristic that the long-short term memory network learns the time sequence characteristics of the data obtained from the data for prediction, then the obtained running data of the ship at the current moment is input into the trained long-short term memory network, and the resistance borne by the ship at the next moment is predicted.

Disclosure of Invention

The invention provides a resistance prediction network training and ship resistance prediction method and a related device, which are used for solving the problems of time consumption and high cost in calculation when the resistance borne by a three-dimensional ship body is solved by using computational fluid dynamics to carry out three-dimensional ship body numerical simulation.

According to an aspect of the present invention, there is provided a method for training a resistance prediction network, including: acquiring three-dimensional first simulation flow field data of a simulated ship in a sailing process by calculating fluid dynamics, wherein the first simulation flow field data comprise flow field data of a flow field where the ship is located and three-dimensional hull data of the ship;

training a plurality of first long-short term memory networks and a plurality of second long-short term memory networks, wherein the first long-short term memory network takes the three-dimensional first simulation flow field data as input and predicts first resistance suffered by the ship, and the second long-short term memory network takes the spliced plurality of first resistance as input and predicts second resistance suffered by the ship;

detecting whether a plurality of first long-short term memory networks and second long-short term memory networks meet preset evaluation criteria;

and if so, forming a plurality of the first long-short term memory networks and the second long-short term memory networks into a multi-scale resistance prediction network.

According to another aspect of the present invention, there is provided a resistance prediction method of a ship, including:

acquiring a multi-scale resistance prediction network consisting of a plurality of first long-short term memory networks and second long-short term memory networks trained by the method according to one aspect of the invention;

acquiring three-dimensional third simulation flow field data of the simulated ship at the first moment;

slicing the three-dimensional third simulation flow field data into two-dimensional third simulation flow field data;

inputting the two-dimensional third simulation flow field data into a plurality of first long-short term memory networks respectively to obtain a plurality of first target resistances;

inputting the spliced plurality of first target resistances into the second long-short term memory network to obtain a second target resistance, the second target resistance being a predicted resistance that the ship receives at a second time, the second time being adjacent to and after the first time.

According to another aspect of the present invention, there is provided a training device for a resistance prediction network, comprising:

the first simulation flow field data acquisition module is used for acquiring three-dimensional first simulation flow field data of a simulated ship in a navigation process, wherein the first simulation flow field data comprises flow field data of a flow field where the ship is located and three-dimensional hull data of the ship;

the network training module is used for training a plurality of first long-short term memory networks and second long-short term memory networks, the first long-short term memory networks predict first resistance suffered by the ship by taking three-dimensional first simulation flow field data as input, and the second long-short term memory networks predict second resistance suffered by the ship by taking a plurality of spliced first resistances as input;

the network detection module is used for detecting whether the plurality of first long-short term memory networks and the plurality of second long-short term memory networks meet preset evaluation standards or not;

and the network composition module is used for forming the plurality of first long-short term memory networks and the second long-short term memory networks into a multi-scale resistance prediction network if the first long-short term memory networks and the second long-short term memory networks are matched.

According to another aspect of the present invention, there is provided a resistance prediction apparatus of a ship, including:

a network obtaining module, configured to obtain a multi-scale resistance prediction network composed of a plurality of first long-short term memory networks and second long-short term memory networks trained by the apparatus according to an aspect of the present invention;

the third simulation flow field data acquisition module is used for acquiring three-dimensional third simulation flow field data of the simulated ship at the first moment;

the third simulation flow field data slicing module is used for slicing the three-dimensional third simulation flow field data into two-dimensional third simulation flow field data;

the first target resistance acquisition module is used for respectively inputting the two-dimensional third simulation flow field data into a plurality of first long-short term memory networks so as to acquire a plurality of first target resistances;

and the second target resistance prediction module is used for inputting the spliced plurality of first target resistances into the second long-short term memory network to obtain a second target resistance, wherein the second target resistance is predicted resistance which the ship receives at a second moment, and the second moment is adjacent to and behind the first moment.

According to another aspect of the present invention, there is provided an electronic apparatus including:

at least one processor; and

a memory communicatively coupled to the at least one processor; wherein the content of the first and second substances,

the memory stores a computer program executable by the at least one processor, the computer program being executable by the at least one processor to enable the at least one processor to perform a method of training a resistance prediction network, a resistance prediction method for a marine vessel, as described in any of the embodiments of the invention.

According to another aspect of the present invention, there is provided a computer-readable storage medium storing computer instructions for causing a processor to implement a method for training a resistance prediction network and predicting resistance of a ship according to any one of the embodiments of the present invention when the computer instructions are executed.

The technical scheme provided by the invention includes that a plurality of first long-short term memory networks and second long-short term memory networks are trained by acquiring three-dimensional first simulation flow field data of a simulated ship in a sailing process, the first long-short term memory networks predict first resistance suffered by the ship and the second long-short term memory networks predict second resistance suffered by the ship by using a plurality of spliced first resistances as input by using the three-dimensional first simulation flow field data as input, the first long-short term memory networks and the second long-short term memory networks are detected whether to accord with preset evaluation standards after training, if so, the first long-short term memory networks and the second long-short term memory networks are combined into a multi-scale resistance prediction network, wherein due to the fact that the number of nodes among the plurality of first long-short term memory networks is different, the first long-short term memory networks with different node numbers learn time sequence characteristics of the same three-dimensional first simulation flow field data in different scales through the first long-short term memory networks with different node numbers, the effect of sufficient learning is achieved, and therefore when the resistance of the ship is predicted according to the multi-scale resistance prediction network which is obtained by training and consists of the first long-short term memory network and the second long-short term memory network, the accuracy of resistance prediction is improved. It should be understood that the statements in this section do not necessarily identify key or critical features of the embodiments of the present invention, nor do they necessarily limit the scope of the invention. Other features of the present invention will become apparent from the following description.

Drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.

FIG. 1 is a flow chart of a training method for a resistance prediction network according to an embodiment of the present invention;

FIG. 2 is a schematic diagram of a multi-scale resistance prediction network according to an embodiment of the present invention;

FIG. 3 is a flow chart of a method for predicting the resistance of a ship according to a second embodiment of the present invention;

FIG. 4 is a schematic structural diagram of a training device of a resistance prediction network according to a third embodiment of the present invention;

fig. 5 is a schematic structural diagram of a resistance prediction device of a ship according to a fourth embodiment of the present invention;

fig. 6 is a schematic structural diagram of an electronic device for implementing the method for training the resistance prediction network and predicting the resistance of the ship according to the embodiment of the present invention.

Detailed Description

In order to make the technical solutions of the present invention better understood, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

It should be noted that the terms "first," "second," and the like in the description and claims of the present invention and in the drawings described above are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the invention described herein are capable of operation in sequences other than those illustrated or described herein. Moreover, 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.

Example one

Fig. 1 is a flowchart of a method for training a resistance prediction network according to an embodiment of the present invention, which is applicable to a situation where resistance experienced by a ship is predicted according to three-dimensional flow field data of a simulated ship, and the method may be performed by a training device of the resistance prediction network, which may be implemented in hardware and/or software, and the training device of the resistance prediction network may be configured in a computer device. As shown in fig. 1, the method includes:

s110, acquiring three-dimensional first simulation flow field data of the simulated ship in the sailing process through computational fluid dynamics, wherein the first simulation flow field data comprise flow field data of a flow field where the ship is located and three-dimensional hull data of the ship.

In this embodiment, when optimizing and designing the ship shape of the ship, the resistance of the ship during sailing is one of the main considerations that affect the design, and it is generally preferable to reduce the resistance of the ship during the design and optimization of the ship shape, so as to reduce the energy consumption of the ship. Furthermore, by means of computational fluid dynamics, the ship's sailing process is simulated under different conditions, for example, the ship sailing in turbulent flow or the ship sailing in a real water area, so that the optimized or designed ship shape can be more suitable for operation under the corresponding scene. In this embodiment, the navigation behavior of the ship in the turbulent flow can be simulated to simulate the ship, and after the simulation, three-dimensional first simulation flow field data about the ship can be obtained, where the first simulation flow field data may include various three-dimensional numerical hull data of the simulated ship and various three-dimensional flow field data simulating a flow field in which the ship is located, such as a river flow field of the turbulent flow.

And S120, training a plurality of first long-short term memory networks and second long-short term memory networks, wherein the first long-short term memory networks predict the first resistance of the ship by taking the first simulation flow field data as input, and the second long-short term memory networks predict the second resistance of the ship by taking the spliced plurality of first resistances as input.

In this embodiment, after the three-dimensional first simulation flow field data is acquired, the resistance prediction network of the ship may be trained according to the three-dimensional first simulation flow field data, where the resistance prediction network may be a neural network formed by a plurality of long-term and short-term memory networks. In traditional calculation, resistance prediction of a ship usually adopts CFD to simulate a three-dimensional ship, a mode of calculating the resistance of the ship is solved by using a momentum conservation equation, an energy conservation equation and a turbulent kinetic energy and turbulent kinetic energy dissipation rate equation of fluid flow, a differential equation set for calculating the resistance is obtained in the calculation mode, and in order to determine the resistance of the ship in a specific area in the sailing process, the differential equation is subjected to discrete processing and is converted into an algebraic equation, so that the discrete distribution of the resistance of the ship in a flow field is obtained. However, the traditional method for calculating the resistance has the disadvantages of time consumption and high calculation cost, when the accuracy of the calculation method is improved, namely a high-fidelity CFD technology is adopted, a large data set with extremely high dimension is often generated, a large number of degrees of freedom are included in the data set, and a new technical problem is formed how to effectively process and analyze the degrees of freedom. However, in the face of these large high-dimensional data sets, the deep learning method provides a new idea, and the process of deep learning learns the similarity relation between data based on a data-driven method, so that errors generated by grid division and parameter setting during processing of the data sets are reduced, and the accuracy of calculating the resistance borne by the ship by using CFD is improved. In this embodiment, a long-short term memory network is selected as a neural network model for resistance calculation, and includes a plurality of first long-short term memory networks and a single second long-short term memory network, and the plurality of first long-short term memory networks and the plurality of second long-short term memory networks are trained, that is, deep learning, through first simulation flow field data. A Long Short-Term Memory Network (LSTM) is a special variant of RNN (Recurrent Neural Network), and overcomes stability bottleneck encountered in the traditional RNN, so that the RNN can be practically applied. The long-short term memory network can learn from the data and take advantage of the time dependencies, and also take advantage of its internal memory to predict the most recent context in the input sequence as input, rather than the current state network as current input. The long-short term memory network can see one observation at a time from a sequence and can understand which observations it saw before are relevant and how to use them for prediction. In this embodiment, the obtained data of the turbulence in the river flow field has a time sequence characteristic, and therefore, by using the long-term and short-term memory network as the neural network for calculating the resistance in this embodiment, the resistance of the ship at the next moment can be predicted after the turbulence data of the ship at the current moment is obtained according to the learned time sequence characteristic of the turbulence data.

However, the accuracy of the resistance prediction is also reduced if the time sequence characteristics of the turbulent flow data learned through the single long-short term memory network are actually insufficient, so the resistance prediction network proposed in this embodiment may be composed of a plurality of first long-short term memory networks and a second long-short term memory network, the number of nodes in the plurality of first long-short term memory networks is different, and different time sequence characteristics may be learned when deep learning is performed according to the first long-short term memory networks with different numbers of nodes. The number of the first long-short term memory networks in the embodiment can be set according to the ship size in the established three-dimensional model of the ship, so that the prediction of the resistance borne by the ships with different sizes is realized.

In this embodiment, the first long-short term memory network takes the first simulation flow field data as input, and specifically, because the format of the input data that can be accepted by the first long-short term memory network is a two-dimensional mode, when the first simulation flow field data is input to the first long-short term memory network in this embodiment, the three-dimensional first simulation flow field data may be sliced first.

The slicing process in this application can be expressed as down-sampling the three-dimensional first simulation flow field data to obtain the two-dimensional first simulation flow field data, and the slicing process for the three-dimensional first simulation flow field data in this embodiment can be specifically expressed as:

in this embodiment, the number of nodes of different first long-short term memory networks is different, so that different first long-short term memory networks can learn the timing characteristics of the turbulence data in different scales according to different numbers of nodes.

The target value is determined, the target value is equal to the number of the queried nodes, the target value is determined according to the number of the nodes of the first long-short term memory network for the slicing times of the three-dimensional first simulation flow field data, and the target value can be equal to the number of the nodes of the first long-short term memory network, so that the size of the first simulation flow field data input into the first long-short term memory network is matched with the number of the nodes of the first long-short term memory network.

When slicing, slicing operation with the three-dimensional first simulation flow field data as a target value can be executed along a preset slicing direction, so as to obtain two-dimensional first simulation flow field data, that is, in the embodiment, each time the first simulation flow field data is sliced along the preset slicing direction, corresponding two-dimensional first simulation flow field data can be obtained,

and then inputting two-dimensional first simulation flow field data into first long and short term memory networks with different numbers of nodes to detect first resistance applied to the ship, wherein the number of the two-dimensional first simulation flow field data corresponds to the number of the nodes of the input first long and short term memory networks.

In this embodiment, after the first resistances output by the plurality of first long-short term memory networks are obtained, the plurality of first resistances may be expanded, spliced and fused to obtain a fused first resistance, and the fused first resistance includes timing characteristics of different scales learned by the first long-short term memory networks with different numbers of nodes, so that more accurate resistance prediction may be achieved.

In this embodiment, the fusion of the first resistances may be completed through the second long-term and short-term memory network, the multiple first resistances are first spliced before the fusion, and the splicing process is represented as slicing the first resistances to obtain one-dimensional first resistances, and then the multiple one-dimensional first resistances are connected end to obtain the spliced first resistances. And inputting the spliced first resistance into a second long-short term memory network, and performing feature fusion learning on the input spliced first resistance by the second long-short term memory network so as to detect the second resistance on the ship. In this embodiment, to facilitate the fusion of the first resistances, a plurality of first long-short term memory networks may be set to have the same preset number of network layers.

In the training process of the resistance prediction network in this embodiment, when the second resistance output by the second long-short term memory network is obtained, the first long-short term memory network and the second long-short term memory network in the resistance prediction network may be updated according to a difference between the predicted second resistance and the actual resistance received by the ship, and a specific process may be represented as:

acquiring a first real resistance received by a simulated ship in a sailing process, calculating a difference between a second resistance and the first real resistance to serve as a loss value, wherein the smaller the loss value calculated in network training is, the better the network training performance is, therefore, in the embodiment, after the loss value is calculated, whether the loss value is smaller than or equal to a preset threshold value or not can be judged, if so, the performance of a resistance prediction network obtained by training at the moment can meet requirements, the training of the resistance prediction network can be determined to be completed, if not, the resistance prediction network obtained by training at the moment can update a first long-short term memory network and a second long-short term memory network according to the loss value, and the training is returned to perform the training of a plurality of first long-short term memory networks and a plurality of second long-short term memory networks until a difference value between a second resistance output by the second long-term memory network obtained by retraining and the first real resistance meets the preset threshold value, the training of the resistance prediction network is ended.

S130, detecting whether the first long-short term memory network and the second long-short term memory network meet a predetermined evaluation criterion, and if so, executing step S140.

In this embodiment, after completing the training of the first long-short term memory network and the second long-short term memory network in the resistance prediction network, flow field data different from that used in the training is input into the first long-short term memory network and the second long-short term memory network, and it is verified whether the flow field data meets a preset evaluation criterion, and this detection process in this embodiment may be specifically expressed as:

in this embodiment, the three-dimensional second simulation flow field data may be obtained at a different time period in the same navigation process of the ship as the three-dimensional first simulation flow field data, and the second simulation flow field data may also include flow field data of a flow field where the simulated ship is located and three-dimensional hull data of the ship.

The three-dimensional second simulation flow field data is sliced to obtain two-dimensional second simulation flow field data, and based on the reason that the three-dimensional first simulation flow field data is sliced to be the same, in the embodiment, after the three-dimensional second simulation flow field data is obtained, the three-dimensional second simulation flow field data is sliced to obtain two-dimensional second simulation flow field data, and then the two-dimensional second simulation flow field data is input into the first long-short term memory network with different numbers of the plurality of nodes to detect third resistance suffered by the ship, wherein the third resistance is the same as the first resistance in the embodiment.

And then splicing a plurality of third resistances output by the first long-short term memory network, and inputting the spliced third resistances into a second long-short term memory network to detect a fourth resistance applied to the ship, wherein the fourth resistance has the same property as the second resistance in the embodiment.

After the fourth resistance borne by the ship is obtained, the second real resistance borne by the simulated ship in the sailing process can be obtained, whether the first long-short term memory network and the second long-short term memory network meet the preset evaluation standard or not is judged according to the average absolute error between the fourth resistance and the second real resistance, for example, in the embodiment, the evaluation standard can be set to be that the average absolute percentage error is smaller than 3%, and if the average absolute percentage error obtained through calculation is smaller than the preset error threshold value, it can be determined that the plurality of first long-short term memory networks and the second long-short term memory networks meet the preset evaluation standard.

And S140, forming a multi-scale resistance prediction network by the plurality of first long-short term memory networks and the plurality of second long-short term memory networks.

In this embodiment, when it is determined that the plurality of first long-short term memory networks and the plurality of second long-short term memory networks meet the preset evaluation criteria, the plurality of first long-short term memory networks and the plurality of second long-short term memory networks may be combined into a multi-scale long-short term memory network, that is, a resistance prediction network. Assuming that the number of the first long-short term memory networks is 2 in this embodiment, the connection and composition relationship between the resistance prediction network 210 and the first long-short term memory network 220 and the second long-short term memory network 230 can be as shown in fig. 2.

The technical scheme provided by the invention includes that three-dimensional first simulation flow field data of a simulated ship in a sailing process are obtained, a plurality of first long-short term memory networks and second long-short term memory networks are trained, the three-dimensional first simulation flow field data are input by the first long-short term memory networks, first resistance borne by the ship is predicted, a plurality of spliced first resistances are input by the second long-short term memory networks, second resistance borne by the ship is predicted, the first long-short term memory networks and the second long-short term memory networks are detected whether to meet preset evaluation standards or not after training, if yes, the first long-short term memory networks and the second long-short term memory networks are combined into a multi-scale resistance prediction network, wherein due to the fact that the number of nodes among the plurality of first long-short term memory networks is different, time sequence characteristics of the same three-dimensional first simulation flow field data in different scales are learned through the first long-short term memory networks with different numbers of nodes, the effect of sufficient learning is achieved, and therefore when the resistance of the ship is predicted according to the multi-scale resistance prediction network which is obtained by training and consists of the first long-short term memory network and the second long-short term memory network, the accuracy of resistance prediction is improved. It should be understood that the statements in this section do not necessarily identify key or critical features of the embodiments of the present invention, nor do they necessarily limit the scope of the invention. Other features of the present invention will become readily apparent from the following description

Example two

Fig. 3 is a flowchart of a resistance prediction method for a ship according to a second embodiment of the present invention, where the present embodiment predicts the resistance of the ship by using a multi-scale resistance prediction network obtained through training in the foregoing embodiment. The method may be performed by a resistance prediction device of the vessel, which may be implemented in the form of hardware and/or software, which may be configured in a computer apparatus. As shown in fig. 3, the method includes:

step 301, obtaining a multi-scale resistance prediction network composed of a plurality of first long-short term memory networks and second long-short term memory networks.

After the multi-scale resistance prediction network is obtained through training in the first embodiment, the multi-scale resistance prediction network can be applied to optimization and improvement of the ship type of the ship, and the resistance of the ship in the flow field can be predicted.

And 302, acquiring three-dimensional third simulation flow field data of the simulated ship at the first moment.

In this embodiment, when resistance of the simulated ship is predicted in the preset flow field, third simulated flow field data of a three-dimensional mode of the simulated ship at a first time before the time when the simulated ship receives the predicted resistance is also acquired, where the third simulated flow field data includes flow field data of a flow field where the simulated ship is located and three-dimensional ship body data of the ship, so that the resistance of the ship at a later time of the first time, that is, at a second time, can be predicted according to the third simulated flow field data.

And 303, slicing the three-dimensional third simulation flow field data into two-dimensional third simulation flow field data.

In this embodiment, the three-dimensional third simulated flow field data obtained first is not suitable for being used as an input of the first long-term and short-term memory network in the multi-scale resistance prediction network, so that in this embodiment, the three-dimensional third simulated flow field data may be sliced before resistance is predicted according to the third simulated flow field data, so as to obtain the two-dimensional third simulated flow field data.

And step 304, inputting the two-dimensional third simulation flow field data into a plurality of first long-short term memory networks respectively to obtain a plurality of first target resistances.

In this embodiment, after the two-dimensional third simulated flow field data is obtained, the two-dimensional third simulated flow field data may be input into the plurality of first long-short term memory networks obtained in this embodiment and trained according to the step of inputting the two-dimensional first simulated flow field data into the first long-short term memory network in the first embodiment, so as to obtain the plurality of first target resistances.

And 305, inputting the spliced plurality of first target resistances into a second long-short term memory network to obtain a second target resistance, wherein the second target resistance is predicted resistance which the ship receives at a second moment, and the second moment is adjacent to and behind the first moment.

In this embodiment, after the plurality of first target resistances are obtained, the plurality of first target resistances may be fused according to the second long-short term memory network to obtain a second target resistance, where the second target resistance may be a resistance that is predicted for the simulated ship in this embodiment and is received at a second time in the preset flow field scene, and the second time is a time adjacent to and after the first time from an input third simulation flow field data acquisition time — the first time. In this embodiment, the calculated first target resistances may be subjected to splicing operation before being input into the second long-term and short-term memory network, and a specific splicing method may refer to the method for splicing the first resistances in the embodiment.

According to the technical scheme provided by the embodiment, the multi-scale resistance prediction network trained by the method in the first embodiment is adopted, the three-dimensional third simulation flow field data of the simulated ship at the first moment are obtained, the one-dimensional third simulation flow field data formed by slicing the three-dimensional third simulation flow field data are input into the plurality of first long and short term memory networks in the resistance prediction network to predict the first target resistance, the spliced first target resistance is input into the second long and short term memory network to obtain the second target resistance predicted for the ship, so that when the resistance of the ship is predicted by the multi-scale resistance prediction network consisting of the first long and short term memory network and the second long and short term memory network obtained by training, the accuracy of resistance prediction is improved, and compared with the resistance prediction by a method for calculating fluid dynamics, the analysis time of a high-dimensional large data set is saved, the efficiency of resistance prediction is carried out to boats and ships has been promoted.

EXAMPLE III

Fig. 4 is a schematic structural diagram of a training device of a resistance prediction network according to a third embodiment of the present invention. As shown in fig. 4, the apparatus includes:

a first simulation flow field data obtaining module 410, configured to obtain, through computational fluid dynamics, three-dimensional first simulation flow field data of a simulated ship during a sailing process, where the first simulation flow field data includes flow field data of a flow field in which the ship is located and three-dimensional hull data of the ship;

a network training module 420, configured to train a plurality of the first long-short term memory networks and the second long-short term memory networks, where the first long-short term memory network predicts a first resistance experienced by the ship by using the three-dimensional first simulated flow field data as an input, and the second long-short term memory network predicts a second resistance experienced by the ship by using the spliced plurality of the first resistances as an input;

a network detection module 430, configured to detect whether the plurality of first long-short term memory networks and the plurality of second long-short term memory networks meet a preset evaluation criterion;

a network composition module 440, configured to, if yes, compose a plurality of the first long-short term memory networks and the second long-short term memory networks into a multi-scale long-short term memory network.

Optionally, the network training module 420 includes:

the first slicing module is used for slicing the three-dimensional first simulation flow field data to obtain two-dimensional first simulation flow field data;

a first resistance prediction module, configured to input the two-dimensional flow field data into the first long-short term memory network with different numbers of the nodes, so as to detect a first resistance experienced by the ship;

the first resistance splicing module is used for splicing a plurality of first resistances;

the second resistance prediction module is used for inputting the spliced first resistance into the second long-short term memory network so as to detect second resistance on the ship;

and the network updating module is used for updating the plurality of first long-short term memory networks and the second long-short term memory networks according to the second resistance.

Optionally, the first dicing module includes:

a node number query module, configured to query the number of nodes of the first long-short term memory network as a first node number;

a target value determination module for determining a target value, the target value being equal to the first node number;

and the two-dimensional data acquisition module is used for executing slicing operation with the three-dimensional first simulation flow field data as the target value along a preset slicing direction to obtain two-dimensional first simulation flow field data.

Optionally, the first resistance splice module comprises:

the second slicing module is used for slicing the first resistance to obtain one-dimensional first resistance;

the one-dimensional first resistance splicing module is used for connecting the first resistances of the multiple one-dimensional models end to obtain spliced first resistance.

Optionally, the network update module includes:

the first real resistance acquisition module is used for acquiring first real resistance of the simulated ship in the sailing process;

a loss value calculation module for calculating a difference between the second resistance and the first true resistance as a loss value;

the loss value judging module is used for judging whether the loss value is smaller than or equal to a preset threshold value, if so, the training completion determining module is called, and if not, the loss value updating module is called;

a training completion determining module for determining completion of training;

and the loss value updating module is used for updating the first long-short term memory network and the second long-short term memory network according to the loss value and returning to call the loss value calculating module.

Optionally, the network detecting module 430 includes:

the second simulation flow field data acquisition module is used for acquiring three-dimensional second simulation flow field data of the simulated ship in the sailing process;

the third slicing module is used for slicing the three-dimensional first simulation flow field data to obtain two-dimensional first simulation flow field data;

a third resistance prediction module, configured to input the two-dimensional flow field data into the first long-short term memory network with different numbers of the plurality of nodes, so as to detect a third resistance experienced by the ship;

the third resistance splicing module is used for splicing a plurality of third resistances;

the fourth resistance calculation module is used for inputting the spliced third resistance into the second long-short term memory network so as to detect the fourth resistance on the ship;

the second real resistance acquisition module is used for acquiring a second real resistance of the simulated ship in the sailing process;

a mean absolute error calculation module for calculating a mean absolute error between the fourth resistance and the second true resistance;

and the evaluation standard conformity determining module is used for determining that the plurality of first long-short term memory networks and the plurality of second long-short term memory networks conform to the preset evaluation standard if the average absolute error is smaller than a preset error threshold.

The training device of the resistance prediction network provided by the embodiment of the invention can execute the training method of the resistance prediction network provided by any embodiment of the invention, and has corresponding functional modules and beneficial effects of the execution method.

Example four

Fig. 5 is a schematic structural diagram of a resistance prediction apparatus for a ship according to a fourth embodiment of the present invention. As shown in fig. 5, the apparatus includes:

a network obtaining module 510, configured to obtain a multi-scale resistance prediction network composed of a plurality of first long-short term memory networks and second long-short term memory networks, trained by the apparatus according to the third embodiment;

a third simulation flow field data obtaining module 520, configured to obtain three-dimensional third simulation flow field data of the simulated ship at the first time;

a third simulation flow field data slicing module 530, configured to slice the three-dimensional third simulation flow field data into two-dimensional third simulation flow field data;

a first target resistance obtaining module 540, configured to input the two-dimensional third simulation flow field data into a plurality of first long-term and short-term memory networks, respectively, so as to obtain a plurality of first target resistances;

a second target resistance obtaining module 550, configured to input the spliced plurality of first target resistances into the second long-short term memory network to obtain a second target resistance, where the second target resistance is a predicted resistance that the ship receives at a second time, and the second time is adjacent to and after the first time.

The ship resistance prediction device provided by the embodiment of the invention can execute the ship resistance prediction method provided by any embodiment of the invention, and has corresponding functional modules and beneficial effects of the execution method.

EXAMPLE five

FIG. 6 illustrates a schematic structural diagram of an electronic device 10 that may be used to implement an embodiment of the present invention. Electronic devices are intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. The electronic device may also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smart phones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions, are meant to be exemplary only, and are not meant to limit implementations of the inventions described and/or claimed herein.

As shown in fig. 6, the electronic device 10 includes at least one processor 11, and a memory communicatively connected to the at least one processor 11, such as a Read Only Memory (ROM)12, a Random Access Memory (RAM)13, and the like, wherein the memory stores a computer program executable by the at least one processor, and the processor 11 may perform various suitable actions and processes according to the computer program stored in the Read Only Memory (ROM)12 or the computer program loaded from the storage unit 18 into the Random Access Memory (RAM) 13. In the RAM 13, various programs and data necessary for the operation of the electronic apparatus 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other via a bus 14. An input/output (I/O) interface 15 is also connected to bus 14.

A number of components in the electronic device 10 are connected to the I/O interface 15, including: an input unit 16 such as a keyboard, a mouse, or the like; an output unit 17 such as various types of displays, speakers, and the like; a storage unit 18 such as a magnetic disk, an optical disk, or the like; and a communication unit 19 such as a network card, modem, wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information/data with other devices via a computer network such as the internet and/or various telecommunication networks.

The processor 11 may be a variety of general and/or special purpose processing components having processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a Central Processing Unit (CPU), a Graphics Processing Unit (GPU), various specialized Artificial Intelligence (AI) computing chips, various processors running machine learning model algorithms, a Digital Signal Processor (DSP), and any suitable processor, controller, microcontroller, or the like. The processor 11 performs the various methods and processes described above, such as training of a resistance prediction network, a resistance prediction method for a ship.

In some embodiments, the training of the resistance prediction network, the resistance prediction method of the vessel, may be implemented as a computer program tangibly embodied in a computer-readable storage medium, such as the storage unit 18. In some embodiments, part or all of the computer program may be loaded and/or installed onto the electronic device 10 via the ROM 12 and/or the communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the method of resistance prediction of a vessel, the training of the resistance prediction network described above, may be performed. Alternatively, in other embodiments, the processor 11 may be configured by any other suitable means (e.g., by means of firmware) to perform the training of the resistance prediction network, the resistance prediction method of the vessel.

Various implementations of the systems and techniques described here above may be implemented in digital electronic circuitry, integrated circuitry, Field Programmable Gate Arrays (FPGAs), Application Specific Integrated Circuits (ASICs), Application Specific Standard Products (ASSPs), system on a chip (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and/or combinations thereof. These various embodiments may include: implemented in one or more computer programs that are executable and/or interpretable on a programmable system including at least one programmable processor, which may be special or general purpose, receiving data and instructions from, and transmitting data and instructions to, a storage system, at least one input device, and at least one output device.

A computer program for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus, such that the computer programs, when executed by the processor, cause the functions/acts specified in the flowchart and/or block diagram block or blocks to be performed. A computer program can execute entirely on a machine, partly on a machine, as a stand-alone software package partly on a machine and partly on a remote machine or entirely on a remote machine or server.

In the context of the present invention, a computer-readable storage medium may be a tangible medium that can contain, or store a computer program for use by or in connection with an instruction execution system, apparatus, or device. A computer readable storage medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. Alternatively, the computer readable storage medium may be a machine readable signal medium. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

To provide for interaction with a user, the systems and techniques described here can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to a user; and a keyboard and a pointing device (e.g., a mouse or a trackball) by which a user can provide input to the electronic device. Other kinds of devices may also be used to provide for interaction with a user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form, including acoustic, speech, or tactile input.

The systems and techniques described here can be implemented in a computing system that includes a back-end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front-end component (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such back-end, middleware, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: local Area Networks (LANs), Wide Area Networks (WANs), blockchain networks, and the internet.

The computing system may include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. The server can be a cloud server, also called a cloud computing server or a cloud host, and is a host product in a cloud computing service system, so that the defects of high management difficulty and weak service expansibility in the traditional physical host and VPS service are overcome.

It should be understood that various forms of the flows shown above may be used, with steps reordered, added, or deleted. For example, the steps described in the present invention may be executed in parallel, sequentially, or in different orders, and are not limited herein as long as the desired results of the technical solution of the present invention can be achieved.

The above-described embodiments should not be construed as limiting the scope of the invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may be made in accordance with design requirements and other factors. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method of training a resistance prediction network, the resistance prediction network including a first long-short term memory network and a second long-short term memory network having a different number of nodes, the method comprising:

acquiring three-dimensional first simulation flow field data of a simulated ship in a sailing process through computational fluid dynamics, wherein the first simulation flow field data comprise flow field data of a flow field where the ship is located and three-dimensional hull data of the ship;

2. The method of claim 1, wherein training the plurality of first long-short term memory networks, the second long-short term memory network, comprises:

slicing the three-dimensional first simulation flow field data to obtain two-dimensional first simulation flow field data;

inputting the two-dimensional first simulation flow field data into a plurality of first long-short term memory networks with different numbers of nodes to detect first resistance on the ship;

splicing a plurality of the first resistance forces;

inputting the spliced first resistance into the second long-short term memory network to detect a second resistance suffered by the ship;

and updating the plurality of first long-short term memory networks and the second long-short term memory network according to the second resistance.

3. The method of claim 2, wherein said slicing the three-dimensional first simulated flow field data to obtain two-dimensional first simulated flow field data comprises:

querying the number of nodes of a plurality of the first long-short term memory networks;

determining a target value, wherein the target value is equal to the number of the nodes;

and performing slicing operation with the three-dimensional first simulation flow field data as the target value along a preset slicing direction to obtain two-dimensional first simulation flow field data.

4. The method of claim 2, wherein said splicing a plurality of said first resistances comprises:

slicing the first resistance to obtain a one-dimensional first resistance;

and connecting the first resistances of a plurality of one dimensions end to obtain the spliced first resistance.

5. The method of claim 2, wherein said updating a plurality of said first long-short term memory networks, said second long-short term memory network based on said second resistance comprises:

acquiring a first real resistance of the simulated ship in the sailing process;

calculating a difference between the second resistance and the first true resistance as a loss value;

judging whether the loss value is less than or equal to a preset threshold value or not;

if so, determining that the training is finished;

and if not, updating the first long-short term memory network and the second long-short term memory network according to the loss value, and returning to calculate the difference between the second resistance and the first real resistance to be used as the loss value.

6. The method according to claim 1, wherein the detecting whether the plurality of first long-short term memory networks and the second long-short term memory networks meet the preset evaluation criteria comprises:

acquiring three-dimensional second simulation flow field data of the simulated ship in the sailing process;

slicing the three-dimensional second simulation flow field data to obtain two-dimensional second simulation flow field data;

inputting the two-dimensional second simulation flow field data into the first long-short term memory networks with different numbers of the nodes to detect third resistance suffered by the ship;

splicing a plurality of the third resistance forces;

inputting the spliced third resistance into the second long-short term memory network to detect a fourth resistance suffered by the ship;

acquiring a second real resistance of the simulated ship in the sailing process;

calculating an average absolute percentage error of the fourth resistance from the second true resistance;

and if the average absolute percentage error is smaller than a preset error threshold, determining that the plurality of first long-short term memory networks and the plurality of second long-short term memory networks meet a preset evaluation standard.

7. The method of claim 1, wherein the number of preset network layers of the first long-short term memory network is the same.

8. A resistance prediction method for a ship, applied to the ship, the method comprising:

obtaining a multi-scale resistance prediction network consisting of a plurality of first long-short term memory networks and second long-short term memory networks trained by the method of any one of claims 1-7;

9. An apparatus for training a resistance prediction network, comprising:

the first simulation flow field data acquisition module is used for acquiring three-dimensional first simulation flow field data of a simulated ship in a sailing process through computational fluid dynamics, wherein the first simulation flow field data comprise flow field data of a flow field where the ship is located and three-dimensional hull data of the ship;

10. A resistance prediction device for a ship, comprising:

the first target resistance acquisition module is used for respectively inputting the two-dimensional third simulation flow field data into a plurality of first long-short term memory networks to acquire a plurality of first target resistances;

11. An electronic device, characterized in that the electronic device comprises:

one or more processors;

a memory for storing one or more programs,

when executed by the one or more processors, cause the one or more processors to implement a method of training a resistance prediction network according to any one of claims 1-7 or a method of resistance prediction for a vessel according to claim 8.

12. A computer-readable storage medium, characterized in that the computer-readable storage medium stores computer instructions for causing a processor, when executed, to implement the method of training a resistance prediction network according to any one of claims 1-7 or the method of resistance prediction of a vessel according to claim 8.