A multi-sensor fusion acceleration prediction method and system for transport ships
Through the multi-sensor fusion acceleration prediction method, sensor data is used for effectiveness detection and processing, and combined with the RNN network model, the problem of low accuracy and reliability of ship state data is solved, and more accurate motion state prediction and decision support is achieved.
Patent Information
- Application Number
- CN202311407594.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-10-27
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2043-10-27
AI Technical Summary
In the prior art, the accuracy and reliability of ship status data are low, resulting in poor ship motion prediction results, and the inability to accurately understand the ship's operating status and provide optimized decisions.
The multi-sensor fusion acceleration prediction method is adopted, and multiple sets of sensors are arranged to collect data, perform effectiveness detection, filtering processing, data fusion and numerical differential processing, and the timing data set is constructed in combination with the RNN network model to output the ship's motion state prediction results.
Improve the reliability and accuracy of data, enable more accurate understanding of the ship's motion status, and provide effective decision-making support for ships' maritime navigation.
Smart Images

Figure CN117308962B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of ship acceleration, and in particular to a multi-sensor fusion acceleration prediction method and system for transport ships. Background Art
[0002] Ships and offshore platforms are the main tools at sea. Ships sailing on the sea will generate multiple status data that affect the motion state of the ship, such as linear acceleration, roll, pitch, bow pitch, sway, surge and heave. The abnormality of ship status data has an important impact on the stability and safety of the ship. Therefore, it is of great significance to monitor the motion state of the ship in real time and predict the motion behavior of the ship. The ship motion prediction can be used to evaluate the danger level of the ship operation and provide reasonable suggestions and decisions for ship motion control and offshore platform control.
[0003] In the existing technology, sensors are often installed on the ship to collect ship status data, and data deduction is performed based on the relevant data to evaluate whether the ship operation is dangerous and provide ship operation decisions. The collected ship status data is status data based on the sensor placement position, and the data does not contain data characteristics of all positions on the ship. In particular, when calculating the relevant data at the metacentric point, the data accuracy and reliability are low, which also affects the subsequent prediction results of the ship's motion, resulting in an inability to accurately understand the ship's operating status and provide better ship operation decisions.
[0004] Therefore, there is an urgent need for a multi-sensor fusion acceleration prediction method for transport ships to solve the problems of low accuracy and reliability of the above-mentioned state data and poor ship motion prediction results. Summary of the invention
[0005] In view of the deficiencies in the prior art, the present invention proposes a multi-sensor fusion acceleration prediction method and system for transport ships.
[0006] The first aspect of the present invention discloses a multi-sensor fusion acceleration prediction method for a transport ship, comprising:
[0007] S1: The first state data of each sampling point of the ship is collected by a plurality of sensor groups arranged on the ship, wherein the state data includes a first linear acceleration, a roll angle, a pitch angle, a bow angle, a sway angle, a pitch angle and a heave angle.
[0008] S2: pre-processing the state data based on the validity test, returning the first state data that failed the test, and obtaining the second state data that passed the test;
[0009] S3: subjecting the second state data to filtering, data fusion and numerical differentiation processing respectively, and outputting the second linear acceleration, ship angular velocity and ship angular acceleration corresponding to the sampling point;
[0010] S4: solving the preset observation point coordinates, the second linear acceleration, the ship angular velocity and the ship angular acceleration according to the acceleration calculation formula and data fusion processing to obtain the ship metacenter linear acceleration representing the ship linear acceleration and the observation point linear acceleration;
[0011] S5: construct a time series data set of the ship's motion state using the linear acceleration at the metacenter and the linear acceleration at the observation point within the prediction period, process the time series data set through the RNN network model, and output the prediction result of the ship's motion state.
[0012] In an optional embodiment, the sensor group includes a three-axis acceleration sensor and a two-axis angle sensor, and the sensor group is arranged in at least four groups. The arrangement rules of the sensors include that the sensor is arranged at one end away from the center of the ship, the sensor is arranged at one end away from the center of the ship, and the connection lines between the sensors are in a non-parallel state.
[0013] In an optional embodiment, the calculation formula of the acceleration in the prediction method is:
[0014] (x0,y0,z0)
[0015] (x1,y1,z1),(x2,y2,z2),(x3,y3,z3),(x4,y4,z4)
[0016]
[0017]
[0018]
[0019]
[0020]
[0021]
[0022] Among them, (x0, y0, z0) represents the coordinates of the ship's rotation center, (x i ,y i ,z i ) represents the location coordinates of the sampling points, i represents the number of sampling points, represents the radius vector of the linear acceleration at the ith sampling point relative to the center of the following rotation, (φ, θ, ψ) represent the roll angle, pitch angle and bow angle of the sampling point collected by the sensor, respectively. They represent the sway angular acceleration, pitch angular acceleration and heave angular acceleration collected by the sensor at the sampling point of the ship, is the angular velocity of the ship, Ship angular acceleration, Indicates the linear acceleration corresponding to the sampling point.
[0023] In an optional embodiment, the validity detection includes:
[0024] Range detection, used to perform data range detection on the first state data as a data source according to a preset basic data range, to determine whether the first state data falls within the basic data range;
[0025] Variance detection, used to perform variance data range detection on the variance of the first state data as a data source according to a preset variance data range, and determine whether the variance of the first state data falls within the variance data range;
[0026] Bad pixel detection is used to perform bad pixel data range detection on the bad pixels of the first state data as the data source according to a preset bad pixel data range, and determine whether the variance of the first state data falls within the bad pixel data range. The bad pixel detection adopts the sigma criterion based on the normal distribution for judgment.
[0027] In an optional embodiment, the numerical differentiation process is calculated by a three-point difference differential formula and a five-point difference differential formula in numerical analysis, wherein the three-point difference differential formula is calculated as follows:
[0028]
[0029] Among them, x0, x1, x2 represent the corresponding three known points calculated by the three-point interpolation differential formula, f() represents the Lagrange basis function corresponding to the known point, h represents the step size in the three-point interpolation differential formula, and P'2() represents the estimated value through the three-point interpolation differential formula;
[0030] The five-point differential formula is calculated as follows:
[0031]
[0032] Among them, x0, x1, x2, x3, and x4 represent the corresponding five known points calculated by the five-point interpolation differential formula, f() represents the Lagrange basis function corresponding to the known points, h represents the step size in the five-point interpolation differential formula, and m0, m1, m2, m3, and m4 represent the estimated values of the five known points calculated by the five-point interpolation differential formula.
[0033] In an optional embodiment, the second state data is subjected to filtering processing, data fusion processing and numerical differentiation processing respectively, and the second linear acceleration, ship angular velocity and ship angular acceleration corresponding to the output sampling point include:
[0034] S301: After filtering, the second state data is output as the linear acceleration and the deflection angle data of the sampling point respectively;
[0035] S302: performing data fusion processing on the deflection angle data of each sampling point and then performing numerical differentiation processing to obtain an angular velocity, and outputting the angular velocity of the ship after filtering processing;
[0036] S303: After the ship angular velocity is processed by numerical differentiation and filtering in sequence, the ship angular acceleration is output.
[0037] In an optional embodiment, the processing of the time series data set by using the RNN network model to output a prediction result of the ship motion state includes:
[0038] S501: Collecting ocean wave time series data within a prediction period, and adding the ocean wave time series data to the time series data set;
[0039] S502: Using the time series data set as the input of the RNN network model, iterating the first weight between the input layer and the middle layer, and iterating the second weight between the middle layer and the output layer, so as to complete the time series data processing process corresponding to a single moment;
[0040] S503: Iterating through the third weights of the middle layer at the previous moment and the middle layer at the next moment between adjacent moments, completing the iterative processing of the time series data set at adjacent moments in sequence, and then obtaining the prediction result after processing by the fully connected layer;
[0041] The calculation formula of the RNN network model is:
[0042] S t =f(W·S t-1 +U·X t )
[0043] O t =g(V·S t )
[0044] Among them, t represents a single moment corresponding to the time series, t-1 represents the previous moment corresponding to the single moment in the time series, and X t represents the input corresponding to the input layer at a single moment, S t Represents the value corresponding to the middle layer at a single moment, O t It represents the output value corresponding to the output layer at a single moment, U represents the weight from the input layer to the middle layer at a single moment, V represents the weight from the middle layer to the output layer at a single moment, and W represents the weight from the middle layer at a single moment to the middle layer at the previous moment.
[0045] The second aspect of the present invention discloses a multi-sensor fusion acceleration prediction system for a transport ship, the system comprising:
[0046] A collection module, used for collecting first state data of each sampling point of the ship through a plurality of sensor groups arranged on the ship, wherein the state data includes a first linear acceleration, a roll angle, a pitch angle, a bow angle, a sway angle, a pitch angle, and a heave angle;
[0047] A detection module, used for preprocessing the state data based on validity detection, returning the first state data that fails the detection, and obtaining the second state data that passes the detection;
[0048] a first calculation module, configured to subject the second state data to filtering, data fusion and numerical differentiation processing, and output a second linear acceleration, a ship angular velocity and a ship angular acceleration corresponding to a sampling point;
[0049] A second calculation module is used to solve the preset observation point coordinates, the second linear acceleration, the ship angular velocity and the ship angular acceleration according to the acceleration calculation formula and data fusion processing to obtain the ship metacenter linear acceleration representing the ship linear acceleration and the observation point linear acceleration;
[0050] The prediction module is used to construct a time series data set of the ship's motion state from the linear acceleration at the metacenter and the linear acceleration at the observation point within the prediction period, process the time series data set through the RNN network model, and output the prediction result of the ship's motion state.
[0051] The third aspect of the present invention discloses a multi-sensor fusion acceleration prediction device for a transport ship, comprising:
[0052] at least one processor, and
[0053] a memory communicatively connected to the at least one processor; wherein,
[0054] The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the multi-sensor fusion acceleration prediction method for a transport ship as disclosed in any one of the first aspects of the present invention.
[0055] The fourth aspect of the present invention discloses a computer-readable storage medium, which stores computer-executable instructions, and the computer-executable instructions are used to enable a computer to execute a multi-sensor fusion acceleration prediction method for a transport ship as described in any one of the first aspect of the present invention.
[0056] Compared with the prior art, the present invention has the following advantages:
[0057] (1) The present invention can have corresponding acceleration monitoring and calculation algorithms through different acceleration and angle sensor layout schemes, and has high flexibility in sensor layout and installation.
[0058] (2) The present invention adopts a redundant sensor arrangement scheme, so that the same data can have multiple backups and can be compared with each other, thereby improving the reliability and accuracy of the data.
[0059] (3) The present invention uses the core algorithm of acceleration monitoring and calculation. In addition to acceleration calculation, it also has various data processing functions, including data validity monitoring function, filtering function, differential function and data fusion processing, etc. The data processing is relatively complete and the data of the motion status of any point of the ship can be obtained. The data has high reliability and accuracy.
[0060] (4) The present invention, by combining the currently widely used artificial neural network, can further predict the future state of the ship and provide effective decision-making for the ship's navigation at sea. BRIEF DESCRIPTION OF THE DRAWINGS
[0061] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0062] Figure 1 It is a flow chart of the multi-sensor fusion acceleration prediction method for transport ships of the present invention;
[0063] Figure 2 It is a schematic diagram of the structure of RNN, LSTM and GRU in the multi-sensor fusion acceleration prediction method for transport ships of the present invention;
[0064] Figure 3 It is a schematic diagram of a prediction model in a multi-sensor fusion acceleration prediction method for a transport ship according to the present invention;
[0065] Figure 4 It is a schematic diagram of ship motion state prediction in the multi-sensor fusion acceleration prediction method for transport ships of the present invention;
[0066] Figure 5 It is a schematic diagram of a multi-sensor fusion acceleration prediction system for transport ships according to the present invention. DETAILED DESCRIPTION
[0067] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0068] In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", "third", "fourth", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.
[0069] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.
[0070] In the embodiments of the present invention, it should be noted that:
[0071] Sigma criterion: The 3σ (Sigma) criterion is also called the Laida criterion. It first assumes that a set of test data contains only random errors, calculates and processes them to obtain the standard deviation, and determines an interval according to a certain probability. It is believed that any error exceeding this interval is not a random error but a gross error, and the data containing this error should be eliminated. The 3σ criterion is based on the normal distribution of equal-precision repeated measurements, and the interference or noise caused by singular data is difficult to satisfy the normal distribution. If the absolute value νi of the residual error of a certain measurement value in a set of measurement data is greater than 3σ, then the measurement value is a bad value and should be eliminated. Usually, the error equal to ±3σ is taken as the limit error. For a normally distributed random error, the probability of falling outside ±3σ is only 0.27%, and the possibility of it occurring in a finite number of measurements is very small, so there is a 3σ criterion. The 3σ criterion is the most commonly used error judgment criterion.
[0072] Acceleration sensor: A sensor that can measure acceleration. It is usually composed of a mass block, a damper, an elastic element, a sensitive element, and an adaptive circuit. During the acceleration process, the sensor measures the inertial force on the mass block and obtains the acceleration value using Newton's second law. In the present invention, it is used to monitor the linear acceleration of the X-axis, Y-axis, and Z-axis methods in the motion state of the ship.
[0073] Angle sensor: used to detect the deflection angle. In the present invention, it is used to measure the deflection angles of roll, pitch, bow roll, sway, surge and heave occurring in the motion state of a ship or an offshore platform.
[0074] Roll: Also known as roll angle, it refers to the rotational oscillation motion of an object immersed in water (i.e. ships and offshore platforms, etc.) around a horizontal axis in the longest extension direction or wave incidence direction; it also refers to the lateral inclination angle used to identify targets in navigation systems, and its value is equal to the angle between a line perpendicular to the bow and stern line on the plane where the target object is located and its projection on the horizontal plane. Ships are most likely to roll at sea and have the largest swaying amplitude, which has a serious impact on the use and safety of the ship, and the life and work of the crew. Therefore, the ship's roll must be studied first in the ship's seakeeping.
[0075] Pitch: The rotational oscillation of the hull around the transverse axis. Pitch and heave have many adverse effects on ships and are one of the main causes of stall, slamming and wave riding. Therefore, it is also necessary to study the pitch of ships when they are sailing at sea.
[0076] Bow roll: also known as vertical roll, refers to the rotational oscillation of an object immersed in water around the vertical axis of the hull. Generally, the bow roll of a ship can be regarded as the result of the following three factors in addition to poor steering: (1) unbalanced static pressure on the hull; (2) the circular motion of water in the waves; (3) gyroscopic effect. Therefore, the influence of bow roll on the ship also needs to be studied.
[0077] Sway: refers to the horizontal linear oscillating motion of an object immersed in water in a direction perpendicular to the longest extension direction. If the object (such as an axisymmetric object) does not have the longest horizontal extension direction, the sway motion direction is defined as the direction perpendicular to the incident wave direction. The sway motion can cause a variety of adverse phenomena on the ship, such as slamming, surfing, stalling, and propeller runaway, so the sway of the ship also needs to be closely monitored.
[0078] Surge: refers to the horizontal linear oscillatory motion of an object immersed in water along the longest extension direction. If the object (such as an axisymmetric object) does not have the longest horizontal extension direction, the direction of the surge motion is defined as the direction of the incident wave.
[0079] Heave: Heave is usually caused by the periodic change of buoyancy when waves pass through the hull. The hull is subject to greater damping during heave, so the free heave decays rapidly. Therefore, it is necessary to add time series data of wave history to the research process.
[0080] Acceleration monitoring and calculation method: It is often used to monitor the motion status of ships and offshore platforms, and to assess the degree of danger of ships and offshore platforms. It can usually be used in ship monitoring and decision-making systems. At present, many companies abroad have related products for shipborne monitoring and decision-making systems, such as OCTOPUS developed by ABB, MO4 by Mocean Offshore and Siri Marine.
[0081] OCTOPUS system: A central node that can connect to onboard sensors, collect data from these sensors, and make the data available for display. Servers can connect directly to it and upload configurations. The OCTOPUS system includes navigation, load, meteorology, dynamic positioning, fuel, motion and other related monitoring. It is a highly integrated ship motion monitoring and decision support system for ships and other marine floating structures.
[0082] The MO4 system includes motion monitoring module, cable laying module, route planning module, dynamic positioning module, etc. The system can realize related functions through the operation of different modules based on external input data, such as weather forecast data and sensor data, and present them in the system in real time in the form of images.
[0083] The development of shipborne motion monitoring and decision-making systems is still in a blank stage in China, but many companies abroad currently have related products. Therefore, how to design a set of ship acceleration monitoring and calculation methods with complete functions, reliable data and convenient deployment is the focus of the present invention. The present invention is based on general acceleration and angle sensor monitoring, and derives a set of acceleration monitoring and calculation methods, which can realize the acceleration and angle data measured from fixed points on the ship, and calculate the acceleration information of any point. It has the advantages of simple layout, clear logic, complete process, etc. The technical solution of the present invention will be described in detail below.
[0084] Example 1
[0085] See also Figure 1-Figure 4 The embodiment of the present invention discloses a multi-sensor fusion acceleration prediction method for a transport ship, comprising:
[0086] S1: The first state data of each sampling point of the ship is collected by a plurality of sensor groups arranged on the ship, wherein the state data includes a first linear acceleration, a roll angle, a pitch angle, a bow angle, a sway angle, a pitch angle and a heave angle.
[0087] S2: pre-processing the state data based on the validity test, returning the first state data that failed the test, and obtaining the second state data that passed the test;
[0088] S3: subjecting the second state data to filtering, data fusion and numerical differentiation processing respectively, and outputting the second linear acceleration, ship angular velocity and ship angular acceleration corresponding to the sampling point;
[0089] S4: solving the preset observation point coordinates, the second linear acceleration, the ship angular velocity and the ship angular acceleration according to the acceleration calculation formula and data fusion processing to obtain the ship metacenter linear acceleration representing the ship linear acceleration and the observation point linear acceleration;
[0090] S5: construct a time series data set of the ship's motion state using the linear acceleration at the metacenter and the linear acceleration at the observation point within the prediction period, process the time series data set through the RNN network model, and output the prediction result of the ship's motion state.
[0091] In an optional embodiment, the sensor group includes a three-axis acceleration sensor and a two-axis angle sensor, and the sensor group is arranged in at least four groups. The arrangement rules of the sensors include that the sensor is arranged at one end away from the center of the ship, the sensor is arranged at one end away from the center of the ship, and the connection lines between the sensors are in a non-parallel state.
[0092] It should be noted that the installation of the sensors can be arranged in the front and rear arrangement of the bottom of the ship and the pipe gallery in the middle of the ship, and can be arranged in the empty warehouses on both sides of the ship's side. At the same time, the sensor positions arranged should not be connected in a straight line, so as to obtain the data of linear acceleration and angle change in the X-axis direction, Y-axis direction and Z-axis direction at the arrangement position. It can be known by those skilled in the art that the arrangement of the sensor group can be more than 4 sensor group arrangement schemes. In this embodiment, the arrangement scheme of at least 4 sensor groups is used to support the calculation of the acceleration monitoring method. Therefore, the acceleration at any position of the ship can be further calculated based on the collected data.
[0093] In an optional embodiment, the calculation formula of the acceleration in the prediction method is:
[0094] (x0,y0,z0)
[0095] (x1,y1,z1),(x2,y2,z2),(x3,y3,z3),(x4,y4,z4)
[0096]
[0097]
[0098]
[0099]
[0100]
[0101]
[0102] Among them, (x0, y0, z0) represents the coordinates of the ship's rotation center, (x i ,y i ,z i ) represents the location coordinates of the sampling points, i represents the number of sampling points, represents the radius vector of the linear acceleration at the ith sampling point relative to the center of the following rotation, (φ, θ, ψ) represent the roll angle, pitch angle and bow angle of the sampling point collected by the sensor, respectively. They represent the sway angular acceleration, pitch angular acceleration and heave angular acceleration collected by the sensor at the sampling point of the ship, is the angular velocity of the ship, Ship angular acceleration, Indicates the linear acceleration corresponding to the sampling point.
[0103] It should be noted that the acceleration calculation formula in this embodiment can be used to solve a total of 12 unknowns: Therefore, the linear acceleration value, angular velocity value and angular acceleration value corresponding to any point of the ship, as well as the coordinates of the ship's rotation center can be obtained.
[0104] Furthermore, it is easy for a person skilled in the art to know that, assuming that in addition to the linear acceleration sensor, an angle sensor is additionally arranged, then The data measured by the angle sensor can be directly obtained through data processing operations such as differential filtering; then, the acceleration at the metacenter of each single-line acceleration sensor can be derived in the following way.
[0105] If we consider the rotation in one direction alone, taking pitch as an example, we can Decompose into So
[0106]
[0107]
[0108] but
[0109]
[0110] Can write
[0111]
[0112] but
[0113]
[0114] The above are the relevant principles and theories of sensor arrangement in this embodiment. The acceleration at the metacenter can be derived by using sensors arranged in this way, and the numbers 1-4 corresponding to the subscripts of the relevant parameters are data collected by four sensor groups at four positions. In practical applications, several data collected by several sensor groups may be included. This embodiment only illustrates the technical solution for calculating the acceleration of a ship using data at four positions. It is easy for those skilled in the art to know that the method of calculating the acceleration of a ship using data at more than four positions is still applicable even when additional parameters are added.
[0115] In an optional embodiment, the validity detection includes:
[0116] Range detection, used to perform data range detection on the first state data as a data source according to a preset basic data range, to determine whether the first state data falls within the basic data range;
[0117] Variance detection, used to perform variance data range detection on the variance of the first state data as a data source according to a preset variance data range, and determine whether the variance of the first state data falls within the variance data range;
[0118] Bad pixel detection, used to perform bad pixel data range detection on the bad pixels of the first state data as the data source according to a preset bad pixel data range, and determine whether the variance of the first state data falls within the bad pixel data range, wherein the bad pixel detection is determined using a sigma criterion based on a normal distribution;
[0119] In an optional embodiment, the numerical differentiation process is calculated by a three-point difference differential formula and a five-point difference differential formula in numerical analysis, wherein the three-point difference differential formula is calculated as follows:
[0120]
[0121] Among them, x0, x1, x2 represent the corresponding three known points calculated by the three-point interpolation differential formula, f() represents the Lagrange basis function corresponding to the known point, h represents the step size in the three-point interpolation differential formula, and P'2() represents the estimated value through the three-point interpolation differential formula;
[0122] The five-point differential formula is calculated as follows:
[0123]
[0124] Among them, x0, x1, x2, x3, and x4 represent the corresponding five known points calculated by the five-point interpolation differential formula, f() represents the Lagrange basis function corresponding to the known points, h represents the step size in the five-point interpolation differential formula, and m0, m1, m2, m3, and m4 represent the estimated values of the five known points calculated by the five-point interpolation differential formula.
[0125] In an optional embodiment, the second state data is subjected to filtering processing, data fusion processing and numerical differentiation processing respectively, and the second linear acceleration, ship angular velocity and ship angular acceleration corresponding to the output sampling point include:
[0126] S301: After filtering, the second state data is output as the linear acceleration and the deflection angle data of the sampling point respectively;
[0127] S302: performing data fusion processing on the deflection angle data of each sampling point and then performing numerical differentiation processing to obtain an angular velocity, and outputting the angular velocity of the ship after filtering processing;
[0128] S303: After the ship angular velocity is processed by numerical differentiation and filtering in sequence, the ship angular acceleration is output.
[0129] It should be noted that in the acceleration monitoring and calculation method of this embodiment, 4 groups of sensors are provided, each group includes 2 triaxial acceleration sensors and 1 biaxial inclination sensor, so the sensor data will be tested for validity when input, and the sensor data that fails the test will be eliminated. Usually, the failure of the validity test may be due to problems with the sensor itself, or errors in the installation of the sensor. Redundant acceleration sensors and angle sensors can still provide acceleration and angle data at corresponding positions when the sensors in the same group fail, and can fuse similar data to make the data more accurate and reliable. After the validity test passes, these data are filtered in the method, and the processed data of the same type will be fused, such as acceleration data at the same point, and angle data of the whole ship. After preliminary calculation of these data, some output data can be obtained, such as the angular velocity of the whole ship, angular acceleration, acceleration of each point line, etc. The above data are further calculated, mainly through the acceleration calculation formula derived in the previous text, to obtain the acceleration value at the metacentric position. Since there are multiple acceleration sensors, the acceleration value at the metacenter can be calculated based on the data of each sensor. After data fusion of these values, the final acceleration value at the metacenter can be obtained as the acceleration data of the whole ship. And through the acceleration data of the whole ship and the coordinates of the observation point of interest, the acceleration of any point on the whole ship can be obtained.
[0130] In an optional embodiment, the processing of the time series data set by using the RNN network model to output a prediction result of the ship motion state includes:
[0131] S501: Collecting ocean wave time series data within a prediction period, and adding the ocean wave time series data to the time series data set;
[0132] S502: Using the time series data set as the input of the RNN network model, iterating the first weight between the input layer and the middle layer, and iterating the second weight between the middle layer and the output layer, so as to complete the time series data processing process corresponding to a single moment;
[0133] S503: Iterating through the third weights of the middle layer at the previous moment and the middle layer at the next moment between adjacent moments, completing the iterative processing of the time series data set at adjacent moments in sequence, and then obtaining the prediction result after processing by the fully connected layer;
[0134] The calculation formula of the RNN network model is:
[0135] S t =f(W·S t-1 +U·X t )
[0136] O t=g(V·S t )
[0137] Among them, t represents a single moment corresponding to the time series, t-1 represents the previous moment corresponding to the single moment in the time series, and X t represents the input corresponding to the input layer at a single moment, S t Represents the value corresponding to the middle layer at a single moment, O t It represents the output value corresponding to the output layer at a single moment, U represents the weight from the input layer to the middle layer at a single moment, V represents the weight from the middle layer to the output layer at a single moment, and W represents the weight from the middle layer at a single moment to the middle layer at the previous moment.
[0138] It should be noted that by establishing a deep learning prediction model, the ship's motion in several cycles is obtained according to the ship's motion history data and marine environmental conditions. According to the characteristics of the ship and marine platform motion prediction problem, a recurrent neural network model suitable for processing time series data is selected as the basis of the prediction model. Recurrent Neural Network (RNN) is a special type of neural network algorithm, which has great advantages in processing time series data. The usual neural network model has a one-to-one correspondence between input and output, and there is no connection between the two adjacent inputs and outputs. For some data with actual sequence relationships, this information will be ignored and lost during processing. Therefore, it is necessary to add weights between each network layer and weights between adjacent moments (that is, if a single moment is t, the previous moment of the single moment is represented as t-1) to associate the input time series data and make the prediction results more accurate.
[0139] Furthermore, in predicting the movement of ships and offshore platforms, the expected input is the wave history and motion history, both of which are time series data, and in actual situations, the motion response at a certain moment must be the result of the combined action of the previous moment's motion and waves, and has a certain continuity. Therefore, it is reasonable and accurate to use RNN as an algorithm to establish a forecast model, and it has a good performance in processing relevant time series data. The input of the forecast model uses environmental condition data (such as wave history data) and motion history data (such as longitudinal and vertical motion), and the dimension of the input layer is 2. The type of RNN layer can be selected from LSTM or GRU, and finally the fully connected layer is used as the final output.
[0140] Furthermore, for example, the most widely used long short-term memory network (LSTM) and gated recurrent unit (GRU) in RNN, the differences and connections between LSTM and GRU are mainly:
[0141] The performance of LSTM and GRU is comparable in many tasks;
[0142] GRU has fewer parameters, 1 / 3 fewer parameters, so it is easier to converge, but in the case of large data sets, LSTM performs better;
[0143] GRU has only two gates (update, reset), LSTM has three gates (forget, input, output). GRU directly passes the hidden state to the next unit, while LSTM wraps the hidden state with a memory cell.
[0144] Both can be used as choices for model building, and existing mature third-party libraries contain their original models. Considering that modeling accuracy and prediction accuracy are the first criteria for measurement, it is appropriate to use the LSTM model for prediction.
[0145] The present invention can have corresponding acceleration monitoring and calculation algorithms through different acceleration and angle sensor layout schemes, and has high flexibility in sensor layout and installation. By adopting a redundant sensor layout scheme, the same type of data can have multiple backups and can be compared with each other, thereby improving the reliability and accuracy of the data. Through the core algorithm of acceleration monitoring and calculation, in addition to acceleration calculation, there are also various data processing functions, including data validity monitoring functions, filtering functions, differential functions, and data fusion processing, etc. The data processing is relatively complete, and the data of the motion state of any point of the ship can be obtained, and the data has high reliability and accuracy. By combining the currently more widely used artificial neural network, the future state of the ship can be further predicted, providing effective decision-making for ships sailing at sea.
[0146] like Figure 5 As shown, the second aspect of the present invention discloses a multi-sensor fusion acceleration prediction system for a transport ship, the system comprising:
[0147] A collection module, used for collecting first state data of each sampling point of the ship through a plurality of sensor groups arranged on the ship, wherein the state data includes a first linear acceleration, a roll angle, a pitch angle, a bow angle, a sway angle, a pitch angle, and a heave angle;
[0148] A detection module, used for preprocessing the state data based on validity detection, returning the first state data that fails the detection, and obtaining the second state data that passes the detection;
[0149] a first calculation module, configured to subject the second state data to filtering, data fusion and numerical differentiation processing, and output a second linear acceleration, a ship angular velocity and a ship angular acceleration corresponding to a sampling point;
[0150] A second calculation module is used to solve the preset observation point coordinates, the second linear acceleration, the ship angular velocity and the ship angular acceleration according to the acceleration calculation formula and data fusion processing to obtain the ship metacenter linear acceleration representing the ship linear acceleration and the observation point linear acceleration;
[0151] The prediction module is used to construct a time series data set of the ship's motion state from the linear acceleration at the metacenter and the linear acceleration at the observation point within the prediction period, process the time series data set through the RNN network model, and output the prediction result of the ship's motion state.
[0152] The third aspect of the present invention discloses a multi-sensor fusion acceleration prediction device for a transport ship, comprising:
[0153] at least one processor, and
[0154] a memory communicatively connected to the at least one processor; wherein,
[0155] The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the multi-sensor fusion acceleration prediction method for a transport ship as disclosed in any one of the first aspects of the present invention.
[0156] The computer device may be a terminal, and the computer device includes a processor, a memory, a network interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a multi-sensor fusion acceleration prediction method for transport ships is implemented. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covered on the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.
[0157] The fourth aspect of the present invention discloses a computer-readable storage medium, which stores computer-executable instructions, and the computer-executable instructions are used to enable a computer to execute a multi-sensor fusion acceleration prediction method for a transport ship as described in any one of the first aspect of the present invention.
[0158] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the multi-sensor fusion acceleration prediction method for transport ships as described above. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.
[0159] Alternatively, if the above modules of the present invention are implemented in the form of software function modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of the present invention can be essentially or partly reflected in the form of a software product that contributes to the relevant technology. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, terminal, or network device, etc.) to execute all or part of the methods of each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as mobile storage devices, RAM, ROM, magnetic disks or optical disks.
[0160] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims
1. A multi-sensor fusion acceleration prediction method for transport ships, characterized in that: The method comprises: S1: collecting first state data of each sampling point of the ship through a plurality of sensor groups arranged on the ship, wherein the state data includes a first linear acceleration, a roll angle, a pitch angle, a bow angle, a sway angle, a pitch angle and a heave angle; S2: pre-processing the state data based on the validity test, returning the first state data that failed the test, and obtaining the second state data that passed the test; The effectiveness detection includes: Range detection, used to perform data range detection on the first state data as a data source according to a preset basic data range, to determine whether the first state data falls within the basic data range; Variance detection, used to perform variance data range detection on the variance of the first state data as a data source according to a preset variance data range, and determine whether the variance of the first state data falls within the variance data range; Bad pixel detection, used to perform bad pixel data range detection on the bad pixels of the first state data as the data source according to a preset bad pixel data range, and determine whether the variance of the first state data falls within the bad pixel data range, wherein the bad pixel detection is determined using a sigma criterion based on a normal distribution; S3: subjecting the second state data to filtering, data fusion and numerical differentiation processing respectively, and outputting the second linear acceleration, ship angular velocity and ship angular acceleration corresponding to the sampling point; S4: solving the preset observation point coordinates, the second linear acceleration, the ship angular velocity and the ship angular acceleration according to the acceleration calculation formula and data fusion processing to obtain the ship metacenter linear acceleration representing the ship linear acceleration and the observation point linear acceleration; S5: constructing a time series data set of the ship's motion state by using the linear acceleration at the metacenter and the linear acceleration at the observation point within the prediction period, processing the time series data set through the RNN network model, and outputting a prediction result of the ship's motion state; The processing of the time series data set by the RNN network model and outputting the ship motion state prediction result includes: S501: Collecting ocean wave time series data within a prediction period, and adding the ocean wave time series data to the time series data set; S502: Using the time series data set as the input of the RNN network model, iterating the first weight between the input layer and the middle layer, and iterating the second weight between the middle layer and the output layer, so as to complete the time series data processing process corresponding to a single moment; S503: Iterating through the third weights of the middle layer at the previous moment and the middle layer at the next moment between adjacent moments, completing the iterative processing of the time series data set at adjacent moments in sequence, and then obtaining the prediction result after processing by the fully connected layer; The calculation formula of the RNN network model is: S t =f(W·S t-1 +U·X t ) O t =g(V·S t ) Among them, t represents a single moment corresponding to the time series, t-1 represents the previous moment corresponding to the single moment of the time series, and X t represents the input corresponding to the input layer at a single moment, S t Represents the value corresponding to the middle layer at a single moment, O t It represents the output value corresponding to the output layer at a single moment, U represents the weight from the input layer to the middle layer at a single moment, V represents the weight from the middle layer to the output layer at a single moment, and W represents the weight from the middle layer at a single moment to the middle layer at the previous moment.
2. The multi-sensor fusion acceleration prediction method for transport ships according to claim 1, characterized in that: The sensor group includes a three-axis acceleration sensor and a two-axis angle sensor. The sensor group is arranged in at least four groups. The arrangement rules of the sensors include that the sensor is arranged at one end away from the center of the ship, the sensor is arranged at one end away from the metacenter of the ship, and the connection lines between the sensors are in a non-parallel state.
3. The multi-sensor fusion acceleration prediction method for transport ships according to claim 2, characterized in that: The calculation formula of the acceleration in the prediction method is: (x0,y0,z0) (x1,y1,z1),(x2,y2,z2),(x3,y3,z3),(x4,y4,z4) Among them, (x0, y0, z0) represents the coordinates of the ship's rotation center, (x i ,y i ,z i ) represents the location coordinates of the sampling points, i represents the number of sampling points, represents the radius vector of the linear acceleration at the ith sampling point relative to the center of the following rotation, (φ, θ, ψ) represent the roll angle, pitch angle and bow angle of the sampling point collected by the sensor, respectively. They represent the sway angular acceleration, pitch angular acceleration and heave angular acceleration collected by the sensor at the sampling point of the ship, is the angular velocity of the ship, Ship angular acceleration, Indicates the linear acceleration corresponding to the sampling point.
4. The multi-sensor fusion acceleration prediction method for transport ships according to claim 3, characterized in that: The numerical differential processing is calculated by the three-point differential formula and the five-point differential formula in numerical analysis, wherein the three-point differential formula is calculated as follows: Among them, x0, x1, x2 represent the corresponding three known points calculated by the three-point interpolation differential formula, f() represents the Lagrange basis function corresponding to the known point, h represents the step size in the three-point interpolation differential formula, and P'2() represents the estimated value through the three-point interpolation differential formula; The five-point differential formula is calculated as follows: Among them, x0, x1, x2, x3, and x4 represent the corresponding five known points calculated by the five-point interpolation differential formula, f() represents the Lagrange basis function corresponding to the known points, h represents the step size in the five-point interpolation differential formula, and m0, m1, m2, m3, and m4 represent the estimated values of the five known points calculated by the five-point interpolation differential formula.
5. The multi-sensor fusion acceleration prediction method for transport ships according to claim 4, characterized in that: The second state data is subjected to filtering processing, data fusion processing and numerical differentiation processing respectively, and the second linear acceleration, ship angular velocity and ship angular acceleration corresponding to the output sampling point are outputted, including: S301: After filtering, the second state data is output as the linear acceleration and the deflection angle data of the sampling point respectively; S302: performing data fusion processing on the deflection angle data of each sampling point and then performing numerical differentiation processing to obtain an angular velocity, and outputting the angular velocity of the ship after filtering processing; S303: After the ship angular velocity is processed by numerical differentiation and filtering in sequence, the ship angular acceleration is output.
6. A multi-sensor fusion acceleration prediction system for transport ships, characterized in that: The system comprises: A collection module, used for collecting first state data of each sampling point of the ship through a plurality of sensor groups arranged on the ship, wherein the state data includes a first linear acceleration, a roll angle, a pitch angle, a bow angle, a sway angle, a pitch angle, and a heave angle; A detection module, used for preprocessing the state data based on validity detection, returning the first state data that fails the detection, and obtaining the second state data that passes the detection; The effectiveness detection includes: Range detection, used to perform data range detection on the first state data as a data source according to a preset basic data range, to determine whether the first state data falls within the basic data range; Variance detection, used to perform variance data range detection on the variance of the first state data as a data source according to a preset variance data range, and determine whether the variance of the first state data falls within the variance data range; Bad pixel detection, used to perform bad pixel data range detection on the bad pixels of the first state data as the data source according to a preset bad pixel data range, and determine whether the variance of the first state data falls within the bad pixel data range, wherein the bad pixel detection is determined using a sigma criterion based on a normal distribution; a first calculation module, configured to subject the second state data to filtering, data fusion and numerical differentiation processing, and output a second linear acceleration, a ship angular velocity and a ship angular acceleration corresponding to a sampling point; A second calculation module is used to solve the preset observation point coordinates, the second linear acceleration, the ship angular velocity and the ship angular acceleration according to the acceleration calculation formula and data fusion processing to obtain the ship metacenter linear acceleration representing the ship linear acceleration and the observation point linear acceleration; A prediction module, used to construct a time series data set of the ship's motion state from the linear acceleration at the metacenter and the linear acceleration at the observation point within a prediction period, process the time series data set through an RNN network model, and output a prediction result of the ship's motion state; The method of processing the time series data set through the RNN network model and outputting the prediction result of the ship motion state includes: collecting the wave time series data within the prediction period and adding the wave time series data to the time series data set; using the time series data set as the input of the RNN network model, iterating the first weight between the input layer and the middle layer, and iterating the second weight between the middle layer and the output layer to complete the time series data processing process corresponding to a single moment; iterating the third weight of the middle layer of the previous moment and the middle layer of the next moment between adjacent moments, completing the iterative processing of the time series data set between adjacent moments in sequence, and then obtaining the prediction result after processing by the fully connected layer; The calculation formula of the RNN network model is: S t =f(W·S t-1 +U·X t ) O t =g(V·S t ) Among them, t represents a single moment corresponding to the time series, t-1 represents the previous moment corresponding to the single moment of the time series, and X t represents the input corresponding to the input layer at a single moment, S t Represents the value corresponding to the middle layer at a single moment, O t It represents the output value corresponding to the output layer at a single moment, U represents the weight from the input layer to the middle layer at a single moment, V represents the weight from the middle layer to the output layer at a single moment, and W represents the weight from the middle layer at a single moment to the middle layer at the previous moment.
7. A multi-sensor fusion acceleration prediction device for transport ships, characterized in that: include: at least one processor, and a memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the multi-sensor fusion acceleration prediction method for a transport ship as described in any one of claims 1 to 5.
8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are used to enable a computer to execute the multi-sensor fusion acceleration prediction method for a transport ship according to any one of claims 1 to 5.
