Security situation awareness system for ship going out of sea

Abstract

The invention discloses a sailing ship safety situation awareness system, and belongs to the technical field of data analysis. The method is used for solving the technical problem that in an existing scheme, multi-dimensional data analysis and timely and reliable early warning cannot be carried out on complex sea conditions, and consequently the coping capacity of the complex sea conditions is weak. Through the whole process of biological anchoring anomaly suppression, data fusion and dynamic feature extraction, the limitation that traditional ship perception depends on a physical sensor can be broken through, high-precision perception of environmental risks and ship states under complex sea conditions is achieved, and compared with a traditional method, the abnormal sea conditions can be recognized in advance, and the method has the advantages of being high in practicability. The robustness of the data and the detection rate of small probability and high influence events can be effectively improved; cross-scale modeling is realized through deep coupling of a microscopic physical mechanism and macroscopic data features; the constructed multi-dimensional risk assessment matrix not only can output a risk value, but also can provide a risk propagation path, and can effectively improve the risk assessment precision of the ship under a complex sea condition.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data analysis, in particular to a sea-going ship safety situation awareness system. BACKGROUND

[0002] The sea-going ship safety situation awareness refers to using various technical means and information sources to monitor, analyze, evaluate and warn in real time the safety situation of the sea-going ship and the marine environment in which the ship is located, so as to form the overall cognition and judgment ability of the current and future safety risks, and the core goal is to improve the predictability, initiative and response ability of the marine ship safety risks, and to protect the safety of the ship, crew, cargo and marine environment.

[0003] The prior art scheme cannot perform multi-dimensional data analysis on complex sea conditions and provide timely and reliable early warning prompts in implementation, and has a low detection rate of identifying small probability high impact events in advance, resulting in weak response ability to complex sea conditions. SUMMARY

[0004] The purpose of the present application is to provide a sea-going ship safety situation awareness system to solve the technical problem that the prior art scheme cannot perform multi-dimensional data analysis on complex sea conditions and provide timely and reliable early warning prompts, resulting in weak response ability to complex sea conditions.

[0005] The purpose of the present application can be achieved by the following technical solutions:

[0006] A sea-going ship safety situation awareness system comprises:

[0007] A multi-source heterogeneous data dynamic perception enhancement module: real-time acquisition of multi-source heterogeneous data of the ship internal and external environment, extraction of time-varying sea condition feature vectors and ship dynamic state parameters through abnormal value dynamic suppression and adaptive weighted fusion;

[0008] A time-space correlation dynamic risk assessment prompt module: based on the obtained time-varying sea condition feature vectors and ship dynamic state parameters, a time-space correlation risk model is constructed by fusing physical constraints and data-driven, a multi-dimensional risk assessment matrix is output in real time, and multi-dimensional risk dynamic prompts of collision risk, capsizing risk and equipment failure risk are performed according to the multi-dimensional risk assessment matrix.

[0009] Preferably, the multi-source heterogeneous data of the ship internal and external environment is synchronously collected by a ship integrated perception system, including marine environment data, ship ontology data and biological acoustic data; the collected multi-source heterogeneous data is preprocessed to construct a unified time-space data set.

[0010] Preferably, the biological benchmark weighted correction is used to correct the qualified sensor data x(t): Wherein, x corr(t) is the modified sensor data; ω(t) is the first weight; is the environmental parameter estimate based on the biological benchmark.

[0011] Preferably, a three-modal knowledge graph is constructed, fuzzy rules are embedded into a graph neural network, node weights are dynamically adjusted through an attention mechanism, and a feature matrix is output.

[0012] Preferably, an abnormal wave generation probability P a , a current shear force gradient Δτ, a biological avoidance behavior index B a , and a ship body resonance frequency offset Δf are extracted from the feature matrix and combined to obtain a time-varying sea state feature vector S = [P a , Δτ, B a , Δf] T ;

[0013] The roll angular acceleration The spatial and temporal distribution of the water depth d(x, y) and the propeller power loss rate 77 are combined to obtain the ship dynamic state parameter

[0014] Preferably, when constructing a spatio-temporal correlation risk model that integrates physical constraints and data-driven methods, the overturning risk critical threshold is calculated based on the ship motion equation:

[0015] where GM is the initial stability height; φ0 is the initial value of the roll angle; d is the water depth;

[0016] The ship field boundary is calculated based on the AIS trajectory data, and the closest approach distance and the closest approach time are used to determine the collision danger zone.

[0017] Preferably, the closest approach distance, the closest approach time physical constraint, and the trajectory prediction error are integrated, and the collision risk R c is obtained by using the Logistic function normalization.

[0018] where k and m are different sensitivity coefficients; DCPA and TCPA are the closest approach distance and the closest approach time, respectively; DCPA lim and TCPA lim are the threshold values of the corresponding data items;

[0019] The overturning risk R o is calculated and obtained based on the roll angular acceleration and the overturning risk critical threshold φ lim :

[0020] where, for critical roll acceleration, g is the acceleration of gravity;

[0021] Based on the propulsion power loss rate η, the equipment failure risk R is calculated e :

[0022] wherein, sigma is a sigmoid function; omega t′ is the second weight, which is obtained by training the historical failure data by the XGBoost model; b is the bias; t' is the time step index.

[0023] Preferably, based on the collision risk R c , the capsizing risk R o , the equipment failure risk R e A multi-dimensional risk assessment matrix is constructed Wherein, the non-diagonal elements are the propagation probabilities between risk factors.

[0024] Preferably, according to the multi-dimensional risk assessment matrix, the risk value and the risk propagation path of the ship are obtained, and the multi-dimensional risk dynamic prompt of the collision risk, the capsizing risk and the equipment failure risk is carried out according to the risk value and the risk propagation path.

[0025] Preferably, the elements of the multi-dimensional risk assessment matrix are updated based on the newly collected time-varying sea state feature vector S and the ship dynamic state parameter P, and the high-frequency fluctuations are suppressed by the exponential moving average.

[0026] Compared with the prior art, the present application has the following beneficial effects:

[0027] Through the whole process of biological anchoring anomaly suppression, data fusion and dynamic feature extraction, the present application can break through the limitations of traditional ship perception relying on physical sensors, realize high-precision perception of environmental risk and ship state in complex sea conditions, and can identify abnormal sea conditions in advance and effectively improve the robustness of data and the detection rate of high-impact events with small probability.

[0028] The present application realizes cross-scale modeling through deep coupling of micro-physical mechanism and macro-data characteristics; the multi-dimensional risk assessment matrix constructed not only can output risk value, but also can provide risk propagation path, which provides a targeted intervention basis for subsequent response strategy generation; through the triple mechanism of physical constraint, data driving and causal reasoning, the risk assessment accuracy of the ship in complex sea conditions can be effectively improved. BRIEF DESCRIPTION OF DRAWINGS

[0029] The present application will be further described below with reference to the accompanying drawings.

[0030] Figure 1 The present application is a flow chart of the operation of a sea-going ship safety situation awareness system. DETAILED DESCRIPTION

[0031] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0032] like Figure 1 As shown, the present invention is a safety situation awareness system for ships at sea, comprising:

[0033] Multi-source heterogeneous data dynamic perception enhancement module: Real-time acquisition of multi-source heterogeneous data of the ship's internal and external environment; extraction of time-varying sea state feature vectors and ship dynamic state parameters through dynamic outlier suppression and adaptive weighted fusion; specific steps include:

[0034] The ship's integrated sensing system synchronously collects multi-source heterogeneous data on the ship's internal and external environment, including marine environmental data, ship body data, and bioacoustic data; specifically:

[0035] Marine environmental data: Shipborne weather radar collects wave height and wave direction data, Doppler current meter measures ocean current velocity profile, and satellite remote sensing receives sea surface temperature gradient data;

[0036] Ship body data: Fiber optic gyroscope inertial measurement units collect the ship's roll angle φ and pitch angle θ, and shaft torque sensors record changes in propulsion system load;

[0037] Bioacoustic data: A towed single-beam sonar collects scattering signals from planktonic communities, extracts community density fluctuation characteristics through time-frequency analysis of the signals, and outputs a spectrum every 2 seconds;

[0038] Time alignment is achieved using GPS timing, and spatial registration is completed through a coordinate transformation matrix from the ship's coordinate system to the geodetic coordinate system, constructing a unified spatiotemporal dataset D(t)={D env (t), D hull (t), D bio (t)};where D env (t), D hull (t), D bio (t) represents marine environmental data, ship body data, and bioacoustic data, respectively; t is the time variable.

[0039] The LSTM network is used to predict the trajectory sequence of planktonic organisms and output the community movement trajectory for the next 2 seconds. Calculate the deviation between the measured trajectory y(t) and the predicted trajectory. Will As a biological benchmark for environmental stability;

[0040] Wherein, the LSTM (Long Short-Term Memory) network is a special recurrent neural network (RNN), which is specially designed to solve the problem that the standard RNN is difficult to learn long-term dependencies when processing long sequence data; by introducing the gating mechanism (forget gate, input gate, output gate) and cell state, the long-term dependence problem of the standard RNN is effectively solved, and information can be selectively remembered or forgotten, so as to perform well when processing long sequence data, and is one of the important tools for processing sequence tasks in deep learning;

[0041] In the embodiment of the application, the plankton community motion trajectory is taken as the environmental benchmark, the position of the biological group in the future 2 seconds is predicted by the LSTM network, and the plankton trajectory sequence is constructed.

[0042] For sensor data x(t), if e(t)>tau(t), tau(t) is a dynamic threshold, tau(t)=tau0(1+0.2L(t)), tau0 is a reference value, and L(t) is a sea state level, which is divided based on the international sea state level standard (1-5 levels); the biological benchmark is weighted and corrected:

[0043] Wherein, omega(t) is the first weight, omega(t)=exp(-0.8*e(t)), which dynamically reduces the weight of the measured value along with the increase of the deviation; It is an environmental parameter estimated value based on the biological benchmark, such as wave height inverted by plankton density;

[0044] It should be noted that the sensitivity of plankton to environmental changes is used to dynamically filter sensor noise and fault data, which can solve the problem that physical sensors are easily disturbed in high sea state, effectively improve the accuracy of abnormal value identification, data signal-to-noise ratio, and reduce the false alarm rate of sensor failure, and can provide high-reliability input data for subsequent fusion.

[0045] When adaptively weighting and fusing, a credibility factor lambda i (t) is assigned to each data source, and the sensor historical accuracy A i and the biological signal consistency coefficient C i are fused.

[0046] Wherein, i is different data sources, j is a summation index, which represents traversing all n data sources carried by the ship; when the wave height H>5m, the weight of the biological acoustic data is automatically increased, that is, C i x1.5, to compensate for the measurement error of the radar in high sea state;

[0047] A knowledge graph of environment-biology-hull three modalities is constructed, fuzzy rules are embedded into graph neural network (GNN), and fuzzy rules, such as the mutation rate Δp>0.3 / s of plankton density, are used to enhance the weight of ocean current data, and the node weight is dynamically adjusted through an attention mechanism:

[0048] wherein R i′,j′ is the association strength of nodes i' and j' in the knowledge graph; β is a rule influence coefficient, and the default value is 0.5; h i′ , h j′ are feature vectors of nodes i' and j' respectively; W Q , W K are query weight matrix and key weight matrix respectively; d is the output dimension of W Q , W K ; (h i′ W Q ) (h j′ W K ) T is the dot product of the query vector of node i' and the key vector of node j'; softmax() is a normalization function;

[0049] The fused output feature matrix F(t) ∈ R 6×T ; wherein T is the length of the time window, and the value is 50; 6 is the feature dimension;

[0050] It should be noted that by dynamically highlighting the weight of high-confidence data, the problem of insufficient robustness of traditional fixed-weight fusion in complex sea conditions can be solved, and the data reliability and robustness in high sea conditions can be effectively improved.

[0051] Four core precursor features, i.e., abnormal wave generation probability P a , ocean current shear force gradient Δt, biological avoidance behavior index B a , and hull resonance frequency offset Δf, are extracted from the feature matrix F(t) and combined to obtain a time-varying sea condition feature vector S = [P a , Δτ, B a , Δf] T ;

[0052] wherein the abnormal wave generation probability P a is obtained by fitting the wave height distribution with a Gaussian mixture model:

[0053] wherein ω k is the weight of the kth Gaussian component; k is the Gaussian component index, k = 1, 2, 3; Φ() is a standard normal distribution cumulative function; H is an abnormal wave threshold; μ k is the mean of the kth Gaussian component; σ kStandard deviation of the k-th Gaussian component

[0054] Current shear gradient Δτ is calculated based on ADCP current velocity profile vertical gradient Combined with seawater density ρ = 1025 kg / m 3 And drag coefficient C d = 0.0025, we get:

[0055] Where |u| is the absolute value of the current velocity; vertical gradient Is the rate of change of the current velocity in the vertical direction, that is, the amount of change in the flow rate per unit depth; z is the depth;

[0056] Biological avoidance behavior index B a The index function is calculated by the prediction error of the plankton trajectory:

[0057] Where N is the prediction window size, which is 20; exp() is the exponential function; t is the time index; Is the predicted value of the plankton trajectory; y(t) is the measured value of the plankton trajectory; Is the mean absolute prediction error;

[0058] The ship body resonance frequency offset Δf is obtained by FFT analysis of the ship body vibration signal, and the measured main frequency f 实测 And the deviation of the design value f 设计 Is obtained:

[0059]

[0060] Get the roll angular acceleration calculated by the ship kinematics model The spatial and temporal distribution of the draft d(x, y), the propeller power loss rate η, and the combination of the ship dynamic state parameters

[0061] Where the roll angular acceleration Is obtained by twice differentiating the roll angle data collected by the fiber-optic gyroscope inertial measurement unit:

[0062] Where Δt is the sampling interval, which is 0.01;

[0063] The spatial and temporal distribution of the draft d(x, y) is obtained by processing the sensor array data by bicubic B-spline interpolation:

[0064] Where α′ i″j″ Is the B-spline control vertex coefficient; B i″ (u), B j″(v) are both B-spline basis functions; u, v are normalized parametric coordinates; i'', j'' are both indexes of B-spline basis functions, and each takes an integer value ranging from 0 to 3; a corresponds to the basis function index along the ship length direction, that is, the u parameter direction; b corresponds to the basis function index along the ship width direction, that is, the v parameter direction;

[0065] The propulsive power loss rate η is calculated based on shafting torque sensor data:

[0066] Wherein, T 实测 is the measured shafting torque; n is the measured rotation speed of the propeller shaft; P 额定 is the rated power of the propulsion system.

[0067] It should be noted that by converting the fusion data into a low-dimensional feature vector representing the environmental risk and the ship state, reliable data support can be provided for subsequent causal risk chain mining.

[0068] In the embodiment of the application, through the whole process of biological anchor anomaly suppression, data fusion and dynamic feature extraction, the limitations of traditional ship perception relying on physical sensors can be broken, high-precision perception of environmental risk and ship state in complex sea conditions can be realized, abnormal sea conditions can be identified in advance compared with traditional methods, and the robustness of data and the detection rate of high-impact events with low probability can be effectively improved.

[0069] The spatio-temporal correlation dynamic risk assessment prompt module: based on the obtained time-varying sea state feature vector and ship dynamic state parameters, a spatio-temporal correlation risk model integrating physical constraints and data driving is constructed, a multi-dimensional risk assessment matrix is output in real time, and multi-dimensional risk dynamic prompts of collision risk, capsizing risk and equipment failure risk are carried out according to the multi-dimensional risk assessment matrix; the specific steps include:

[0070] When constructing the spatio-temporal correlation risk model integrating physical constraints and data driving, the capsizing risk critical threshold is calculated based on the ship motion equation:

[0071] Wherein, GM is the initial stability height; φ0 is the initial value of the roll angle; d is the draft;

[0072] The ship field boundary is calculated based on the AIS trajectory data, and the closest approach distance and the closest approach time are used to determine the collision danger zone:

[0073] If the closest approach distance < 2L and the closest approach time < 300s, the collision risk warning is triggered; L is the ship length;

[0074] The obtained time-varying sea state feature vector S and ship dynamic state parameters P are input into a deep spatio-temporal network (DSTN) to realize data-driven feature enhancement;

[0075] The deep spatiotemporal network includes a temporal attention module and a spatial interaction module.

[0076] Time Attention Module: Assigns dynamic weights to the time-varying sea state feature vector S and ship dynamic state parameter P over the past 5 minutes, for example, recent data has higher weights;

[0077] Spatial Interaction Module: Employs a graph attention mechanism to capture the trajectory correlations of surrounding vessels. Node features include speed, heading angle, and the biological avoidance behavior index B. a ;

[0078] It should be noted that by integrating physical laws and implicit data features—physical laws such as ship stability and collision avoidance rules, and implicit data features such as biological precursors and sudden changes in sea state—we can provide reliable mechanistic data-driven input features for risk assessment. Compared with traditional pure data-driven models, physical constraints can effectively reduce the error in risk factor extraction.

[0079] When outputting the multidimensional risk assessment matrix R, the collision risk R is obtained by integrating the nearest encounter distance, the nearest encounter time physical constraint, and the trajectory prediction error, and then normalizing using the Logistic function. c :

[0080] Where k and m are different sensitivity coefficients, taking values ​​of 0.5 and 0.02 respectively; DCPA and TCPA are the nearest encounter distance and the nearest encounter time, respectively; DCPA lim TCPA lim These are the threshold values ​​for the corresponding data items, with values ​​of 2L and 300s respectively;

[0081] Combined with roll angle acceleration and the critical threshold for overturning risk φ lim Calculate and obtain the overturning risk R o :

[0082] in, The critical roll acceleration, g is the acceleration due to gravity;

[0083] The equipment failure risk R is calculated based on the propulsion power loss rate η. e :

[0084] Where σ is the Sigmoid function; ω t′ b is the second weight, obtained by training the XGBoost model on historical fault data; t′ is the bias; and t′ is the time step index.

[0085] Based on collision risk R c, capsizing risk R o , equipment failure risk R e Construct a multi-dimensional risk assessment matrix Where the off-diagonal elements are the propagation probabilities between risk factors.

[0086] According to the multi-dimensional risk assessment matrix, the risk value and risk propagation path of the ship are obtained, and the multi-dimensional risk dynamic prompt of the collision risk, the capsizing risk and the equipment failure risk is carried out according to the risk value and the risk propagation path.

[0087] Wherein, the risk value is the main diagonal element directly indicating the quantitative value of each type of risk; the risk propagation path, for example, R c→o Indicates the probability of collision risk triggering capsizing, forming path R c →R o .

[0088] Every 500ms, update the multi-dimensional risk assessment matrix elements based on the newly collected time-varying sea state feature vector S and ship dynamic state parameters P, and suppress high-frequency fluctuations through exponential moving average:

[0089] R(t) = δ·R(t-1) + (1-δ)R new (t); Wherein, δ is a smoothing coefficient, and the value is 0.2; R(t-1) is the smoothed risk assessment matrix at the last time t-1; R new (t) is the original risk assessment matrix at the current time t.

[0090] It should be noted that by fusing physical constraints, data-driven prediction and causal propagation law into a structured risk matrix, all-round assessment of collision, capsizing and equipment failure can be realized. In addition, multi-dimensional risk coupling assessment can effectively reduce the false alarm rate, and the matrix dynamic updating mechanism can reduce the response delay of sudden sea conditions.

[0091] In the embodiment of the application, by deeply coupling micro physical mechanism and macro data characteristics, cross-scale modeling is realized; the constructed multi-dimensional risk assessment matrix can not only output risk value, but also provide risk propagation path, which provides targeted intervention basis for subsequent response strategy generation. Through the triple mechanism of physical constraints, data-driven and causal reasoning, the risk assessment accuracy of the ship in complex sea conditions can be effectively improved.

[0092] Adaptive response strategy generation module: according to the obtained multi-dimensional risk assessment matrix R, combined with the ship maneuvering performance parameters, a cooperative response strategy set containing route correction, speed adjustment and emergency operation is dynamically generated. The specific steps include:

[0093] Obtain the ship static maneuvering performance parameter library, and combine the ship dynamic state parameters The upper limit of the ship speed and the minimum turning radius in the static maneuvering performance parameter library are corrected in real time.

[0094] wherein, when g is the gravity acceleration, the upper limit of the ship speed is reduced by the following formula:

[0095] wherein, κ is the roll influence coefficient, and the value is 0.5; V max is the preset basic upper limit of the ship speed; V′ max (t) is the corrected upper limit of the ship speed;

[0096] Based on the space-time distribution d(x, y) of the draft, the minimum safe turning radius in the shallow water area is calculated by bicubic interpolation:

[0097] wherein, d min (t) is the real-time minimum draft; d design is the design draft; R min is the preset minimum safe turning radius in the shallow water area; R′ min (t) is the corrected minimum safe turning radius in the shallow water area;

[0098] The corrected upper limit of the ship speed and the corrected minimum safe turning radius in the shallow water area are sorted and combined to obtain the corrected ship maneuvering performance parameters;

[0099] According to the generated multi-dimensional risk assessment matrix R, the causal propagation chain between the risk factors is extracted, for example: main path 1: corresponding to abnormal waves→roll aggravation→risk of capsizing;

[0100] Main path 2: Δτ→R c →R e , corresponding to current shear→collision risk→equipment overload failure;

[0101] For the source risk factor of each propagation path, a targeted intervention strategy is matched from a preset strategy library:

[0102] If P a > 0.6, corresponding to high abnormal wave risk, a combination strategy of route deviation and speed reduction is triggered, and the adjustment parameters corresponding to the route deviation and the speed reduction are determined according to the existing strategy, and the specific content is not limited;

[0103] If R c > 0.7, corresponding to high collision risk, a heading angle correction amount Δθ is generated by AIS trajectory prediction, and the expression involved is: wherein, V(t) is the current speed;

[0104] Generate atomic strategies satisfying physical feasibility based on constraints; Atomic strategy types include speed adjustment and course correction;

[0105] Where constraints are: speed V(t)∈[0.3V′ max , V′ max ], rudder angle θ(t)∈[-θ max , θ max ];

[0106] Speed adjustment: generate ΔV candidate set, satisfying speed constraint V(t)∈[0.3V′ max , V′ max ];

[0107] Course correction: generate Δθ candidate set, satisfying rudder angle constraint θ(t)∈[-θ max , θ max ];

[0108] Based on the adapted ship maneuvering performance correction parameters, detect whether there are conflicting items in the atomic strategy; For example:

[0109] "Speed up to avoid" (ΔV = +2 knots) and "speed down" (ΔV = -3 knots) speed instruction conflict;

[0110] "Left turn" (Δθ = +15°) and "right turn" (Δθ = -10°) course instruction conflict;

[0111] Eliminate conflicting items, retain non-conflicting strategies and combine to get composite strategy set;

[0112] For each strategy in the composite strategy set, calculate its risk reduction rate J1 and execution cost J2 to quantify the performance of the strategy:

[0113]

[0114] J2 = ΔV 2 + Δθ 2 + δ 应急 ;

[0115] Where R target is the safety threshold matrix; δ 应急 is the emergency operation penalty term, taking the value of 10; trace is the matrix trace, i.e. the sum of the main diagonal elements, which comprehensively reflects the overall risk reduction proportion; ΔV is the speed adjustment;

[0116] Select Pareto optimal strategies through multi-objective optimization algorithm to form a set of collaborative strategies C;

[0117] Among them, the multi-objective optimization algorithm, such as NSGA-III, NSGA-III is an advanced evolutionary algorithm for solving multi-objective optimization problems (Multi-Objective Optimization Problems, MOOPs), which is an improved version of the NSGA-II algorithm, and can effectively solve the problem of uneven distribution of solution set in high-dimensional multi-objective optimization problem by introducing a reference point mechanism instead of traditional congestion calculation; it can significantly improve the diversity and distribution of the population while maintaining the convergence of the solution, and is an important tool for solving complex multi-objective optimization problems, and the specific implementation steps of NSGA-III are not described here, which is the existing conventional technical solution.

[0118] The Pareto optimal solution is: if the strategy c satisfies "there is no other strategy c such that J1(c) >= J1(c*) and J2(c) <= J2(c*)", then c* is the Pareto optimal solution, that is, it is impossible to reduce the cost without reducing the risk reduction rate, or to increase the risk reduction rate without increasing the cost;

[0119] Select the final collaborative strategy set C from the Pareto optimal solution to ensure that the strategy has no conflict and covers different optimization directions;

[0120] Among them, the strategy selection principle includes the diversity principle and the non-redundancy principle;

[0121] The diversity principle: retain strategies with different J1 and J2 trade-off relationships, such as "high J1-high J2", "medium J1-medium J2", and "low J1-low J2";

[0122] The non-redundancy principle: eliminate strategies with highly similar effects, such as the risk reduction rate difference between ΔV=-2 sections and ΔV=-1.8 sections is less than 5%; The implementation of the strategy selection principle can be determined based on the application requirements of the actual application scene.

[0123] It should be noted that by resolving strategy conflicts, a Pareto optimal strategy set that takes into account risk reduction and execution cost can be generated, and by using the Pareto optimal strategy, the strategy conflict rate can be effectively reduced.

[0124] In the embodiment of the application, by targeting the risk transmission source, the risk reduction efficiency in complex sea conditions can be significantly improved; at the same time, the dynamic adaptation strategy is combined with the real-time maneuvering performance of the ship to ensure conflict-free execution, and the collaborative optimization of ship safety and energy efficiency is realized.

[0125] In several embodiments provided by the present application, it should be understood that the disclosed system can be implemented in other ways. For example, the above-described embodiments of the application are only illustrative, and for example, the division of modules is only a logical functional division, and actual implementation can have another division way.

[0126] The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical modules, and may be located in one place or distributed over multiple network modules. Part or all of the modules can be selected according to actual needs to achieve the purpose of the embodiment.

[0127] In addition, each functional module in each embodiment of the present application can be integrated into one processing module, or each module can exist physically alone, or two or more modules can be integrated into one module. The integrated module can be realized in the form of hardware or in the form of hardware plus software functional module.

[0128] It is obvious to those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and the present application can be implemented in other specific forms without departing from the essential characteristics of the present application.

[0129] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and are not limiting. Although the present application has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solutions of the present application can be modified or replaced equivalently without departing from the spirit and scope of the technical solutions of the present application.

Claims

1. A safety situation awareness system for ships at sea, characterized in that, include: Multi-source heterogeneous data dynamic perception enhancement module: Real-time collection of multi-source heterogeneous data of the ship's internal and external environment, and extraction of time-varying sea state feature vectors and ship dynamic state parameters through outlier dynamic suppression and adaptive weighted fusion; Spatiotemporal correlation dynamic risk assessment and prompting module: Based on the acquired time-varying sea state feature vector and ship dynamic state parameters, a spatiotemporal correlation risk model integrating physical constraints and data-driven approaches is constructed, and a multidimensional risk assessment matrix is ​​output in real time. Based on the multidimensional risk assessment matrix, multidimensional risk dynamic prompts for collision risk, capsizing risk, and equipment failure risk are provided.

2. The maritime vessel safety situation awareness system according to claim 1, characterized in that, The ship integrated sensing system synchronously collects multi-source heterogeneous data of the ship's internal and external environment, including marine environmental data, ship body data, and bioacoustic data; the collected multi-source heterogeneous data is preprocessed to construct a unified spatiotemporal dataset.

3. The maritime vessel safety situation awareness system according to claim 2, characterized in that, The eligible sensor data x(t) is corrected using a biological benchmark weighting method: Where, x corr (t) represents the corrected sensor data; ω(t) represents the first weight; These are estimates of environmental parameters based on biological benchmarks.

4. The maritime vessel safety situation awareness system according to claim 3, characterized in that, A trimodal knowledge graph is constructed, fuzzy rules are embedded into a graph neural network, node weights are dynamically adjusted through an attention mechanism, and a feature matrix is ​​output.

5. A maritime vessel safety situation awareness system according to claim 4, characterized in that, Extract the anomalous wave generation probability P from the feature matrix. a Ocean current shear gradient Δτ, biological avoidance behavior index B a The four core precursor features of the ship's resonant frequency shift Δf are combined to obtain the time-varying sea state feature vector S = [P]. a ,Δτ,B a ,Δf] T ; Obtain the roll acceleration calculated from the hull kinematic model. By combining the spatiotemporal distribution of draft d(x, y) and propulsion power loss rate η, the dynamic state parameters of the ship can be obtained.

6. The maritime vessel safety situation awareness system according to claim 1, characterized in that, When constructing a spatiotemporal correlation risk model that integrates physical constraints and data-driven approaches, the critical threshold for capsizing risk is calculated based on the ship's motion equations: Where GM is the initial stability height; φ0 is the initial roll angle; and d is the draft. The boundary of the ship's domain is calculated based on AIS trajectory data, and the collision hazard zone is determined by the nearest encounter distance and the nearest encounter time.

7. A maritime vessel safety situation awareness system according to claim 6, characterized in that, By integrating the physical constraints of nearest encounter distance and nearest encounter time with trajectory prediction error, and normalizing using the Logistic function, the collision risk R is obtained. c : Where k and m are different sensitivity coefficients; DCPA and TCPA are the nearest encounter distance and nearest encounter time, respectively; DCPA lim TCPA lim These are all threshold values ​​for the corresponding data items; Combined with roll angle acceleration and the critical threshold for overturning risk φ lim Calculate and obtain the overturning risk R o : in, The critical roll acceleration, g is the acceleration due to gravity; The equipment failure risk R is calculated based on the propulsion power loss rate η. e : Where σ is the Sigmoid function; ω t′ is the second weight, obtained by training the XGBoost model on historical fault data; b is the bias; t′ is the time step index.

8. A maritime vessel safety situation awareness system according to claim 7, characterized in that, Based on collision risk R c Overturning risk R o Equipment failure risk R e Constructing a multidimensional risk assessment matrix Among them, the off-diagonal elements represent the propagation probability between risk factors.

9. A maritime vessel safety situation awareness system according to claim 8, characterized in that, The risk value and risk propagation path of the ship are obtained by using a multi-dimensional risk assessment matrix, and multi-dimensional risk dynamic prompts are made based on the risk value and risk propagation path for collision risk, capsizing risk and equipment failure risk.

10. A maritime vessel safety situation awareness system according to claim 9, characterized in that, The elements of the multidimensional risk assessment matrix are updated based on the newly acquired time-varying sea state feature vector S and ship dynamic state parameters P, and high-frequency fluctuations are suppressed by exponential moving average.

Images