CN113486574B - Sound velocity profile completion method and device based on historical data and machine learning - Google Patents
Sound velocity profile completion method and device based on historical data and machine learning Download PDFInfo
- Publication number
- CN113486574B CN113486574B CN202110688153.XA CN202110688153A CN113486574B CN 113486574 B CN113486574 B CN 113486574B CN 202110688153 A CN202110688153 A CN 202110688153A CN 113486574 B CN113486574 B CN 113486574B
- Authority
- CN
- China
- Prior art keywords
- data
- sound velocity
- sea area
- eof
- velocity profile
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/088—Non-supervised learning, e.g. competitive learning
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Testing Or Calibration Of Command Recording Devices (AREA)
Abstract
The invention relates to a sound velocity profile completion method based on historical data and machine learning, which comprises the following steps: obtaining sea area historical average data of a target sea area as first data and historical data collected by instrument equipment as second data; performing EOF analysis on the first data and the second data to obtain an EOF projection coefficient set; training the second data and the EOF projection coefficient set as training samples to obtain neurons representing different reference classification information; acquiring sea area field measured data of a target sea area, and calculating to obtain a reference neuron with the highest degree of association with the sea area field measured data; and reconstructing a sound velocity profile according to the reference neuron and the historical average data of the sea area to complete the completion of the sound velocity profile of the target sea area. The sound velocity profile reconstruction result established by the method accords with the disturbance rule, avoids the reconstruction result which does not accord with the reality, contains the characteristics of the fine structure of the time scale sound velocity profile in the target sea area, and can reflect the complex disturbance state in the practical situation.
Description
Technical Field
The invention relates to the technical field of sea area analysis, in particular to a sound velocity profile completion method and device based on historical data and machine learning.
Background
The sound velocity profile is a sound velocity structural feature in the vertical direction of seawater, reflects the hydrological element distribution of temperature and salt in local sea areas, and has decisive influence on sound propagation and underwater sound channel characteristics. The invention provides a sound velocity profile complementing method based on historical data and machine learning, which can estimate sound velocity distribution of the whole sea depth under the condition that only sound velocity values of partial depth are measured, and can provide reference for application of marine environment monitoring and a sonar system.
Therefore, mastering the time-space change characteristics of the full-depth sound velocity is important for applications such as ocean sound field modeling and sonar equipment performance estimation.
The most direct method for obtaining the sound velocity profile is to measure in the target sea area by using a sound velocity meter and other measuring equipment, but the method is time-consuming and labor-consuming, only the profile data which are discretely distributed in time and space can be obtained, and if the instant sound velocity distribution characteristics of a large area are to be obtained, the method is basically impossible.
With the development of marine observation means, marine remote sensing, underwater gliders, underwater unmanned underwater vehicles and other means make it possible to obtain marine environment parameters in a large area, but the marine environment data measured by the means are usually incomplete, only data of a certain depth point or depth range lack full-depth information, which brings difficulty to application of sound field prediction and the like, and complete sections need to be supplemented.
The existing completion method mainly adopts a method of reasonably combining historical average data and measured data to complete the sound velocity profile, which roughly comprises the following methods,
the method comprises the following steps:
(1) respectively calculating the interpolation of the temperature salinity value and the WOA2018 temperature salinity model value at the measurement depth;
(2) revising the temperature values of the water layers according to the difference obtained in the step (1) and the water depths of the water layers of the WOA2018 outside the measurement depth, so as to obtain the temperature and salt revision values of the water layers of the WOA2018 outside the effective measurement depth;
(3) and combining the temperature and salt revision values of the water layers to obtain a revised model temperature profile and salinity profile outside the effective measurement depth, and constructing a full-depth and high-precision temperature profile and salinity profile by combining the actually-measured temperature profile and salinity profile within the effective measurement depth.
(4) And calculating the sound velocity profile value through the constructed temperature salinity profile.
Reference patent: method for reconstructing full-depth sound velocity profile by combining WOA2018 model and actually measured temperature and salinity data (authorization notice number: 111523200)
However, the analysis can find that the first method has the following disadvantages:
the data models of the profiles such as WOA2018 are long-time average results, the profiles are subjected to smooth averaging and interpolation processing, long-time average characteristics are represented, and description of characteristics of fine structures of the small-time scale sound velocity profiles is lost.
2. The sound velocity profile is calculated through a fixed sound velocity profile model relational expression, and because the sea is a complex power system which is difficult to describe by a formula, errors which are not consistent with the actual situation are easily generated by introducing an expression for calculation.
3. The sound velocity profile is reconstructed by means of pure mathematical processing such as variance, and reconstruction results which are not in accordance with reality are easily generated under the condition that the sound velocity profile is subjected to sudden change and the like.
The second method comprises the following steps:
the method is not specially used for obtaining a complete section, but is used for the modal analysis of the orthogonal empirical function of the complete section. The sound velocity profile compensation method presented therein is:
(1) carrying out interpolation processing on the measured Argo buoy data to obtain a smooth sound velocity distribution value of a known depth point;
(2) dividing the missing sound velocity profile into an upper layer and a lower layer for interpolation;
(3) for the sound velocity part of the upper ocean deletion, the sound velocity value of the shallowest point of the measured depth is taken as a reference, and the equal sound velocity of the shallowest point of the upper ocean deletion is set;
(4) for the sound velocity part of the lower ocean, extrapolation is carried out by adopting a polynomial fitting method;
(5) and combining the upper layer completed section, the measured section and the lower layer completed section to obtain a complete section.
Reference patent: a method for EOF analysis using modified Argo buoy data (application No.: 202011322448.7)
The method is rough, the ocean is directly divided into three layers on the basis of the measured profile, and a simple mathematical formula is used for interpolation. The method has low precision reliability, cannot reflect the real disturbance condition of the sound velocity profile, and has an effect inferior to that of the first method.
Disclosure of Invention
The present invention is directed to at least one of the deficiencies of the prior art, and provides a method and an apparatus for completing a sound velocity profile based on historical data and machine learning.
In order to achieve the purpose, the invention adopts the following technical scheme:
specifically, a sound velocity profile completion method based on historical data and machine learning is provided, which comprises the following steps:
acquiring sea area historical average data of a target sea area as first data and historical data acquired by instrument equipment as second data;
performing EOF analysis on the first data and the second data to obtain an EOF projection coefficient set;
training the second data and the EOF projection coefficient set as training samples through an unsupervised machine learning algorithm to obtain neurons representing different reference classification information;
obtaining sea area field measured data of a target sea area, calculating the correlation degree between the sea area field measured data and the neurons, and taking the neurons with the highest correlation degree as reference neurons;
and reconstructing a sound velocity profile according to the reference neuron and the historical average data of the sea area to obtain sound velocity profile data of the full sea depth, and completing the sound velocity profile completion of the target sea area.
Further, specifically, the sea area historical average data, that is, the first data, is historical smoothed average data of the WOA2018 database, and the historical data, that is, the second data, collected by the instrumentation equipment is Argo buoy data.
Further, specifically, the process of performing the EOF analysis on the first data and the second data to obtain the EOF projection coefficient set includes the following steps,
substituting the first data into a Del gross sound velocity empirical formula to obtain the distribution of the sound velocity value of the target sea area on the depth z, namely the steady-state background section c0(z); obtaining a temperature and salinity profile sample in the sea area with the target sea area coordinate as the center and the longitude and latitude of 1 degree through second data, and obtaining sound velocity step-by-step samples c (z, t) of the target sea area at different time t through a Del gross sound velocity empirical formula; the sound velocity step samples c (z, t) are subjected to EOF decomposition, and each sample can be represented as a superposition of a steady-state background profile and the EOF multiplied by a projection coefficient of each order, as follows:
wherein d is a sampling point on the depth, a is a corresponding projection coefficient, s represents an EOF order for reconstruction, and k is an EOF vector;
obtaining EOF projection coefficient sets [ A ] of sound velocity step samples c (z, t) at different times by performing EOF analysis on the first data and the second data1 A2…At]Each time corresponding sample vector At＝[a1t a2t…aNt]I.e. N projection coefficients at time t.
Further, specifically, training the second data and the EOF projection coefficient set as training samples by an unsupervised machine learning algorithm to obtain neurons representing different pieces of reference classification information, including the following,
forming an input training matrix Q by the second data and the EOF projection coefficient set as:
the training matrix Q is a matrix of (M + N) x t, wherein t is the number of Argo buoy data samples; each training sample is composed of (M + N) elements and contains a sound speed value of the measured depth, c (z)Mt) Representing the sound velocity value of the t-th sample at the Mth depth, aNtRepresenting the projection coefficient corresponding to the EOF of the Nth order of the t sample; the number of neurons in the output layer needs to satisfy the rule of numerical values larger than the number of samples;
and training the self-organizing competitive neural network on the training matrix Q to obtain the neurons representing different reference classification information.
Further, the correlation between the sea area field measured data and the neuron is obtained by calculating the Euclidean distance, and specifically, the sea area field measured data is a measured limited depth sound velocity value X ═ c1 c2…cM]The finite depth acoustic velocity value and the neuron ref have been measuredpinter-Euclidean distance
Comprises the following steps:
wherein ref represents all neurons, p is the number of the neuron,
measuring a covariance matrix between a limited depth sound velocity value and a neuron, wherein an angle label avail represents measured information, missing represents unknown information of a measured vector relative to the neuron, and the part mainly refers to a projection coefficient part;
and selecting the neuron with the shortest distance as a reference neuron by calculating the Euclidean distance between the sea area field measured data and each neuron.
Further, specifically, reconstructing a sound velocity profile according to the reference neuron and the sea area historical average data to obtain sound velocity profile data of the whole sea depth, including the following steps,
wherein the projection coefficient asFor corresponding vectors of projection coefficients in reference neurons, crIs a reconstructed full-sea subsonic velocity profile.
The invention also provides a sound velocity profile complementing device based on historical data and machine learning, which comprises the following components:
the historical data acquisition module is used for acquiring sea area historical average data of a target sea area as first data and historical data acquired by instrument equipment as second data;
an EOF projection coefficient set making module, configured to perform EOF analysis on the first data and the second data to obtain an EOF projection coefficient set;
the training module is used for training the second data and the EOF projection coefficient set as training samples through an unsupervised machine learning algorithm to obtain neurons representing different reference classification information;
the correlation degree calculation module is used for acquiring sea area field measured data of a target sea area, calculating the correlation degree between the sea area field measured data and the neurons, and taking the neurons with the highest correlation degree as reference neurons;
and the sound velocity profile completion module is used for reconstructing a sound velocity profile according to the reference neuron and the historical average data of the sea area to obtain sound velocity profile data of the whole sea depth, and completing the sound velocity profile completion of the target sea area.
The invention has the beneficial effects that:
the invention aims to simultaneously adopt historical average data and real-time historical profile data measured by an Argo buoy, and finally realizes the reconstruction of a full-depth complete profile by establishing a self-organizing competitive neural network and introducing a reference profile.
Compared with the defects of the prior art, the method has the advantages that:
(1) while the WOA2018 historical average profile is adopted, Argo buoy data are introduced into the method, and the profiles measured in real time in the past contain the characteristics of the fine structure of the small-time-scale sound velocity profile of the target sea area;
(2) the characteristics of the sound velocity profile are not preset through a model or a relational expression, and the obtained sound velocity profile can reflect a complex disturbance state in an actual situation;
(3) the sound velocity profile is reconstructed by extracting the main component of the sound velocity disturbance, the reconstruction result accords with the disturbance rule, and the reconstruction result which does not accord with the actual result is avoided.
Drawings
The foregoing and other features of the present disclosure will become more apparent from the detailed description of the embodiments shown in conjunction with the drawings in which like reference characters designate the same or similar elements throughout the several views, and it is apparent that the drawings in the following description are merely some examples of the present disclosure and that other drawings may be derived therefrom by those skilled in the art without the benefit of any inventive faculty, and in which:
FIG. 1 is a flow chart of a method for complementing a sound velocity profile based on historical data and machine learning according to the present invention;
FIG. 2 is a block flow diagram illustrating an embodiment of a method for complementing a sound velocity profile based on historical data and machine learning according to the present invention;
fig. 3 is a schematic structural diagram of a sound velocity profile complementing device based on historical data and machine learning according to the present invention.
Detailed Description
The conception, the specific structure and the technical effects of the present invention will be clearly and completely described in conjunction with the embodiments and the accompanying drawings to fully understand the objects, the schemes and the effects of the present invention. It should be noted that the embodiments and features of the embodiments in the present application may be combined with each other without conflict. The same reference numbers will be used throughout the drawings to refer to the same or like parts.
Referring to fig. 1, in embodiment 1, the present invention provides a sound velocity profile completion method based on historical data and machine learning, including the following steps:
step 110, obtaining sea area historical average data of a target sea area as first data and historical data collected by instrument equipment as second data;
step 120, performing EOF analysis on the first data and the second data to obtain an EOF projection coefficient set;
step 130, training the second data and the EOF projection coefficient set as training samples through an unsupervised machine learning algorithm to obtain neurons representing different reference classification information;
step 140, obtaining sea area field actual measurement data of a target sea area, calculating the correlation degree between the sea area field actual measurement data and the neurons, and taking the neurons with the highest correlation degree as reference neurons;
and 150, reconstructing a sound velocity profile according to the reference neuron and the historical average data of the sea area to obtain sound velocity profile data of the full sea depth, and completing the completion of the sound velocity profile of the target sea area.
In the embodiment, historical average data and real-time historical profile data measured by an Argo buoy are adopted at the same time, a reference profile is introduced by establishing a self-organizing competitive neural network, and finally the reconstruction of a full-depth complete profile is realized.
Compared with the defects of the prior art, the method has the advantages that:
(1) while the WOA2018 historical average profile is adopted, Argo buoy data are introduced into the method, and the profiles measured in real time in the past contain the characteristics of the fine structure of the small-time-scale sound velocity profile of the target sea area;
(2) the characteristics of the sound velocity profile are not preset through a model or a relational expression, and the obtained sound velocity profile can reflect a complex disturbance state in an actual situation;
(3) the sound velocity profile is reconstructed by extracting the main component of the sound velocity disturbance, the reconstruction result accords with the disturbance rule, and the reconstruction result which does not accord with the actual result is avoided.
As a preferred embodiment of the present invention, specifically, the first historical average data of the sea area, that is, the historical smoothed average data of the WOA2018 database, and the second historical data collected by the instrumentation equipment, that is, the second data, is Argo buoy data.
The historical smooth data of the WOA2018 database specifically refers to historical smooth data of temperature and salinity data in the database, and the Argo buoy data also corresponds to historical data of temperature and salinity collected by an Argo buoy.
In the preferred embodiment, while the WOA2018 historical average profile is adopted, the method also introduces Argo buoy data, and the profiles measured in real time in the past contain the characteristics of the fine structure of the small-time scale sound velocity profile in the target sea area.
The preferred embodiment adopts the historical smooth average data of the WOA2018 database, and can also be replaced by the historical average data of other database data in the sea area; the section sample mainly comprises Argo buoy data, and can also be historical data obtained by other instruments such as a thermohaline depth instrument, a sound velocimeter and the like.
With reference to fig. 2, as a preferred embodiment of the present invention, in detail, the process of performing the EOF analysis on the first data and the second data to obtain the EOF projection coefficient set includes the following steps,
substituting the first data into a Del gross sound velocity empirical formula to obtain the distribution of the sound velocity value of the target sea area on the depth z, namely the steady-state background section c0(z); obtaining a temperature and salinity profile sample in the sea area with the target sea area coordinate as the center and the longitude and latitude of 1 degree through the second data, and obtaining a target through a Del gross sound velocity empirical formulaStep-wise samples c (z, t) of the sound velocity of the sea area at different times t; the sound velocity step samples c (z, t) are subjected to EOF decomposition, and each sample can be represented as a superposition of a steady-state background profile and the EOF multiplied by a projection coefficient of each order, as follows:
wherein d is a sampling point on the depth, a is a corresponding projection coefficient, s represents an EOF order for reconstruction, and k is an EOF vector;
obtaining EOF projection coefficient sets [ A ] of sound velocity step samples c (z, t) at different times by performing EOF analysis on the first data and the second data1 A2…At]Each time corresponding sample vector At＝[a1t a2t…aNt]I.e. N projection coefficients at time t.
In conjunction with fig. 2, as a preferred embodiment of the present invention, specifically, training the second data and the EOF projection coefficient set as training samples by an unsupervised machine learning algorithm to obtain neurons representing different pieces of reference classification information includes the following,
forming an input training matrix Q by the second data and the EOF projection coefficient set as:
the training matrix Q is a matrix of (M + N) x t, wherein t is the number of Argo buoy data samples; each training sample is composed of (M + N) elements and contains a sound speed value of the measured depth, c (z)Mt) Representing the sound velocity value of the t-th sample at the Mth depth, aNtRepresenting the projection coefficient corresponding to the EOF of the Nth order of the t sample; the number of neurons in the output layer needs to satisfy the rule of numerical values larger than the number of samples;
in the preferred embodiment, neurons representing different pieces of reference classification information are obtained by training the training matrix Q with a self-organizing competitive neural network. Of course, it is possible to use some other unsupervised machine learning algorithm to complete the classification process, if reasonable.
With reference to fig. 2, as a preferred embodiment of the present invention, the correlation between the sea area field measurement data and the neuron is obtained by calculating the Euclidean distance, and specifically, the sea area field measurement data is the measured finite depth sound velocity value X ═ c1 c2…cM]The finite depth acoustic velocity value and the neuron ref have been measuredpinter-Euclidean distance
Comprises the following steps:
wherein ref represents all neurons, p is the number of the neuron,
measuring a covariance matrix between a limited depth sound velocity value and a neuron, wherein an angle label avail represents measured information, missing represents unknown information of a measured vector relative to the neuron, and the part mainly refers to a projection coefficient part;
in the preferred embodiment, the Euclidean distance between the sea area field measurement data and each neuron is calculated, and the neuron with the shortest distance is selected as the reference neuron.
With reference to fig. 2, as a preferred embodiment of the present invention, specifically, reconstructing a sound velocity profile according to the reference neuron and the historical average data of the sea area to obtain sound velocity profile data of the whole sea depth, including the following,
wherein the projection coefficient asFor corresponding projection coefficient vectors in reference neurons，crIs a reconstructed full-sea subsonic velocity profile.
Referring to fig. 3, in embodiment 2, the present invention further provides a sound velocity profile completion apparatus based on historical data and machine learning, including the following:
a historical data acquisition module 100, configured to acquire sea area historical average data of a target sea area as first data and historical data acquired by an instrument as second data;
an EOF projection coefficient set creating module 200, configured to perform EOF analysis on the first data and the second data to obtain an EOF projection coefficient set;
a training module 300, configured to train the second data and the EOF projection coefficient set as training samples through an unsupervised machine learning algorithm to obtain neurons representing different pieces of reference classification information;
the correlation calculation module 400 is configured to obtain sea area field actual measurement data of a target sea area, calculate a correlation between the sea area field actual measurement data and the neuron, and use the neuron with the highest correlation as a reference neuron;
and an acoustic velocity profile completion module 500, configured to reconstruct an acoustic velocity profile according to the reference neuron and the sea area historical average data, to obtain acoustic velocity profile data of a full sea depth, and complete acoustic velocity profile completion of the target sea area.
In the embodiment, a device applying the method is provided, and the method of the invention is supported on hardware, so that the following advantages can be embodied on hardware,
(1) while the WOA2018 historical average profile is adopted, Argo buoy data is introduced into the method, and the profiles measured in real time in the past contain the characteristics of the fine structure of the small-time scale sound velocity profile in the target sea area.
(2) The characteristics of the sound velocity profile are not preset through a model or a relational expression, and the obtained sound velocity profile can reflect a complex disturbance state in an actual situation.
(3) The sound velocity profile is reconstructed by extracting the main component of the sound velocity disturbance, the reconstruction result accords with the disturbance rule, and the reconstruction result which does not accord with the actual result is avoided.
The modules described as separate parts may or may not be physically separate, and parts displayed as modules may or may not be physical modules, may be located in one place, or may be distributed on a plurality of network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution in this embodiment.
In addition, functional modules in the embodiments of the present invention may be integrated into one processing module, or each of the modules may exist alone physically, or two or more modules are integrated into one module. The integrated module can be realized in a hardware mode, and can also be realized in a software functional module mode.
The integrated module, if implemented in the form of a software functional module and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, all or part of the flow of the method according to the embodiments of the present invention may also be implemented by a computer program, which may be stored in a computer-readable storage medium and can implement the steps of the above-described method embodiments when the computer program is executed by a processor. Wherein the computer program comprises computer program code, which may be in the form of source code, object code, an executable file or some intermediate form, etc. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U.S. disk, removable hard disk, magnetic diskette, optical disk, computer Memory, Read-Only Memory (ROM), Random Access Memory (RAM), electrical carrier wave signal, telecommunications signal, and software distribution medium, etc. It should be noted that the computer readable medium includes content that can be suitably increased or decreased according to the requirements of legislation and patent practice in jurisdictions, for example, in some jurisdictions, computer readable media does not include electrical carrier signals and telecommunication signals according to legislation and patent practice.
While the present invention has been described in considerable detail and with particular reference to a few illustrative embodiments thereof, it is not intended to be limited to any such details or embodiments or any particular embodiments, but it is to be construed as effectively covering the intended scope of the invention by providing a broad, potential interpretation of such claims in view of the prior art with reference to the appended claims. Furthermore, the foregoing describes the invention in terms of embodiments foreseen by the inventor for which an enabling description was available, notwithstanding that insubstantial modifications of the invention, not presently foreseen, may nonetheless represent equivalent modifications thereto.
The above description is only a preferred embodiment of the present invention, and the present invention is not limited to the above embodiment, and the present invention shall fall within the protection scope of the present invention as long as the technical effects of the present invention are achieved by the same means. The invention is capable of other modifications and variations in its technical solution and/or its implementation, within the scope of protection of the invention.
Claims (2)
1. The sound velocity profile completion method based on historical data and machine learning is characterized by comprising the following steps of:
acquiring sea area historical average data of a target sea area as first data and historical data acquired by instrument equipment as second data;
performing EOF analysis on the first data and the second data to obtain an EOF projection coefficient set;
training the second data and the EOF projection coefficient set as training samples through an unsupervised machine learning algorithm to obtain neurons representing different reference classification information;
obtaining sea area field measured data of a target sea area, calculating the correlation degree between the sea area field measured data and the neurons, and taking the neurons with the highest correlation degree as reference neurons;
reconstructing a sound velocity profile according to the reference neuron and the historical average data of the sea area to obtain sound velocity profile data of the full sea depth, and completing the completion of the sound velocity profile of the target sea area;
specifically, the sea area historical average data, i.e., the first data, is historical smooth average data of the WOA2018 database, and the historical data, i.e., the second data, collected by the instrument equipment is Argo buoy data;
specifically, the process of performing the EOF analysis on the first data and the second data to obtain the EOF projection coefficient set includes the following steps,
substituting the first data into a Del gross sound velocity empirical formula to obtain the distribution of the sound velocity value of the target sea area on the depth z, namely the steady-state background section c0(z); obtaining temperature and salinity profile samples in the sea area with the target sea area coordinate as the center and the longitude and latitude of 1 degree through second data, and obtaining sound velocity step-by-step samples c (z, t) of the target sea area at different time t through a Del gross sound velocity empirical formula; performing EOF decomposition on the sound velocity step samples c (z, t), where each sample can be expressed as a superposition of the steady-state background profile and the EOF multiplied by the projection coefficient of each order, as follows:
wherein d is a sampling point on the depth, a is a corresponding projection coefficient, s represents an EOF order for reconstruction, and k is an EOF vector;
obtaining EOF projection coefficient sets [ A ] of sound velocity step samples c (z, t) at different times by performing EOF analysis on the first data and the second data1 A2…At]Each time corresponding sample vector At＝[a1t a2t…aNt]N projection coefficients at time t;
specifically, the second data and the EOF projection coefficient set are used as training samples to be trained through an unsupervised machine learning algorithm to obtain neurons representing different reference classification information, and the method comprises the following steps,
forming an input training matrix Q by the second data and the EOF projection coefficient set as:
the training matrix Q is a matrix of (M + N) x t, wherein t is the number of Argo buoy data samples; each training sample consists of (M + N) elements, and contains the sound velocity value of the measured depth, c (z)Mt) Representing the sound velocity value of the t-th sample at the Mth depth, aNtRepresenting the projection coefficient corresponding to the EOF of the Nth order of the t sample; the number of neurons in the output layer needs to satisfy the rule of numerical values larger than the number of samples;
training a self-organizing competitive neural network on the training matrix Q to obtain neurons representing different reference classification information;
the correlation degree between the sea area field measured data and the neuron is obtained by calculating the Euclidean distance, and the sea area field measured data is a measured limited depth sound velocity value X ═ c1 c2…cM]The finite depth acoustic velocity value and the neuron ref have been measuredpinter-Euclidean distance
Comprises the following steps:
wherein ref represents all neurons, p is the number of the neuron,
measuring a covariance matrix between a limited depth sound velocity value and a neuron, wherein an angle label avail represents measured information, and missing represents unknown information of a measured vector relative to the neuron;
selecting the neuron with the shortest distance as a reference neuron by calculating the Euclidean distance between the field measured data of the sea area and each neuron;
specifically, reconstructing a sound velocity profile according to the reference neuron and the sea area historical average data to obtain sound velocity profile data of the whole sea depth, including the following steps,
wherein the projection coefficient asFor corresponding vectors of projection coefficients in reference neurons, crIs a reconstructed full-sea subsonic velocity profile.
2. The sound velocity profile complementing device based on historical data and machine learning, characterized in that the sound velocity profile complementing method based on historical data and machine learning described in claim 1 is applied, and comprises the following steps:
the historical data acquisition module is used for acquiring sea area historical average data of a target sea area as first data and historical data acquired by instrument equipment as second data;
an EOF projection coefficient set making module, configured to perform EOF analysis on the first data and the second data to obtain an EOF projection coefficient set;
the training module is used for training the second data and the EOF projection coefficient set as training samples through an unsupervised machine learning algorithm to obtain neurons representing different reference classification information;
the correlation degree calculation module is used for acquiring sea area field measured data of a target sea area, calculating the correlation degree between the sea area field measured data and the neurons, and taking the neurons with the highest correlation degree as reference neurons;
and the sound velocity profile completion module is used for reconstructing a sound velocity profile according to the reference neuron and the historical average data of the sea area to obtain sound velocity profile data of the full sea depth, and completing the sound velocity profile completion of the target sea area.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110688153.XA CN113486574B (en) | 2021-06-21 | 2021-06-21 | Sound velocity profile completion method and device based on historical data and machine learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110688153.XA CN113486574B (en) | 2021-06-21 | 2021-06-21 | Sound velocity profile completion method and device based on historical data and machine learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113486574A CN113486574A (en) | 2021-10-08 |
| CN113486574B true CN113486574B (en) | 2022-07-01 |
Family
ID=77933940
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110688153.XA Active CN113486574B (en) | 2021-06-21 | 2021-06-21 | Sound velocity profile completion method and device based on historical data and machine learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113486574B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113916347A (en) * | 2021-11-05 | 2022-01-11 | 哈尔滨工程大学 | Seawater sound velocity profile continuation method and device |
| CN114781576B (en) * | 2022-04-19 | 2023-04-07 | 广东海洋大学 | Sound velocity profile estimation method and device based on random forest algorithm |
| CN115238569B (en) * | 2022-06-21 | 2023-04-28 | 中国人民解放军61540部队 | Warm salt reconstruction method and system based on information embedding depth FFNN model |
| CN115307714B (en) * | 2022-10-12 | 2023-02-03 | 中国海洋大学 | Rapid sound velocity distribution estimation method based on cross-space-time sound velocity profile clustering |
| CN116202611B (en) * | 2023-05-06 | 2023-07-14 | 中国海洋大学 | Small sample sound velocity profile inversion method based on meta-learning |
| CN117151198B (en) * | 2023-09-06 | 2024-04-09 | 广东海洋大学 | Underwater sound passive positioning method and device based on self-organizing competitive neural network |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106950568A (en) * | 2017-02-23 | 2017-07-14 | 山东科技大学 | A kind of construction method of self adaptation multinode equivalent sound velocity profile |
| CN108802735A (en) * | 2018-06-15 | 2018-11-13 | 华南理工大学 | A kind of submarine target positioning and speed-measuring method and device for unknown velocity of sound environment |
| CN112598113A (en) * | 2020-12-15 | 2021-04-02 | 广东海洋大学 | Ocean sound velocity profile acquisition method based on self-organizing competitive neural network |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2006084187A (en) * | 2004-09-14 | 2006-03-30 | Oki Electric Ind Co Ltd | Spike noise removing device of ocean data profile, and spike noise removing method of ocean data profile |
| CN103591942B (en) * | 2013-11-07 | 2014-07-23 | 国家海洋局第二海洋研究所 | Method for quickly simplifying and automatically optimizing sound velocity profile based on maximum offset of sound velocity |
| CN105388457B (en) * | 2015-12-03 | 2017-10-27 | 山东科技大学 | A kind of Long baselines hydrolocation method based on equivalent sound velocity gradient |
-
2021
- 2021-06-21 CN CN202110688153.XA patent/CN113486574B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106950568A (en) * | 2017-02-23 | 2017-07-14 | 山东科技大学 | A kind of construction method of self adaptation multinode equivalent sound velocity profile |
| CN108802735A (en) * | 2018-06-15 | 2018-11-13 | 华南理工大学 | A kind of submarine target positioning and speed-measuring method and device for unknown velocity of sound environment |
| CN112598113A (en) * | 2020-12-15 | 2021-04-02 | 广东海洋大学 | Ocean sound velocity profile acquisition method based on self-organizing competitive neural network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113486574A (en) | 2021-10-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113486574B (en) | Sound velocity profile completion method and device based on historical data and machine learning | |
| CN108489498A (en) | A kind of AUV collaborative navigation methods without mark particle filter based on maximum cross-correlation entropy | |
| CN111307266B (en) | Sound velocity obtaining method and global ocean sound velocity field construction method based on same | |
| CN116450711B (en) | GNSS coordinate time sequence data stream matching method | |
| CN108981957A (en) | Submarine temperatures field reconstructing method based on self organizing neural network and Empirical Orthogonal Function | |
| CN113218493B (en) | Acoustic velocity profile inversion method based on empirical orthogonal function method | |
| CN114782745B (en) | Ocean sound velocity profile classification method and device based on machine learning | |
| Thomson et al. | AUV localization in an underwater acoustic positioning system | |
| CN112598113A (en) | Ocean sound velocity profile acquisition method based on self-organizing competitive neural network | |
| CN113901927B (en) | Underwater object shape recognition method based on flow field pressure time course | |
| CN114781576B (en) | Sound velocity profile estimation method and device based on random forest algorithm | |
| CN110837791A (en) | Sound velocity profile inversion method based on over-complete dictionary | |
| CN115859116A (en) | Marine environment field reconstruction method based on radial basis function regression interpolation method | |
| CN113916347A (en) | Seawater sound velocity profile continuation method and device | |
| CN115049026A (en) | Regression analysis method of space non-stationarity relation based on GSNNR | |
| White et al. | Localisation of sperm whales using bottom-mounted sensors | |
| CN115952472B (en) | Sound velocity field estimation method, sound velocity field estimation device and storage medium | |
| CN111142134B (en) | Coordinate time series processing method and device | |
| Apolinario et al. | Estimation of 2D velocity model using acoustic signals and convolutional neural networks | |
| Taroudakis et al. | Tomographic and bottom geoacoustic inversions using genetic algorithms and a statistical characterization of the acoustic signal | |
| CN113657416B (en) | Deep sea sound source ranging method and system based on improved deep neural network | |
| CN117763953A (en) | Reconstruction method, device and equipment of ocean sound velocity field and storage medium | |
| CN117494573B (en) | Wind speed prediction method and system and electronic equipment | |
| CN114580615B (en) | Distributed small platform underwater pure orientation positioning method based on neural network | |
| CN117195958A (en) | Surface sea current prediction method based on attention mechanism TCN-LSTM model |
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 |