A ship running state multi-source data fusion evaluation method

Abstract

The application discloses a kind of ship running state multi-source data fusion evaluation method, belong to ship state monitoring technical field, comprising the following steps: S1, multi-source heterogeneous running state data acquisition and alignment;S2, multi-source data quality evaluation and pretreatment;S3, weighted fusion based on data value evaluation;S4, ship running state deduction;S5, ship running state comprehensive evaluation;S6, evaluation result and decision output;S7, multi-ship cooperative state evaluation.The application, by fusing ship-borne sensor real-time monitoring data (including six degrees of freedom motion posture, stress and strain, vibration, etc.), ship navigation state data, engine room equipment operating parameters (main engine power, rotating speed, exhaust temperature, etc.), ocean environment data (sea wave, ocean current, weather) and navigation situation data five kinds of multi-source heterogeneous information, constructs the all-around, multidimensional perception system of ship running state, provides data basis for accurate evaluation.

Description

Technical Field

[0001] This invention relates to the field of ship condition monitoring technology, and in particular to a method for evaluating the multi-source data fusion of ship operating status. Background Technology

[0002] With the rapid development of intelligent ships and autonomous navigation technology, real-time perception and accurate assessment of ship operating status have become key technical support for ensuring navigation safety, improving operational efficiency and reducing maintenance costs. As a complex mobile platform at sea, the operating status of a ship is affected by multiple factors such as its own equipment condition, navigation attitude, marine environment and navigation situation, exhibiting complex characteristics such as high nonlinearity, strong coupling and multi-scale.

[0003] In the existing technology, there are a variety of ship status monitoring and data fusion schemes. For example, the invention patent with announcement number CN119513796A discloses a method and device for ship trajectory data fusion. This method acquires shore-based base station AIS data, satellite-borne AIS data, and radar detection signals. After data preprocessing, it matches and merges the data based on static attributes. Then, it decomposes the data into subsets based on time and space. By analyzing the distance, time difference, and movement speed between adjacent data points, it judges the rationality of the trajectory and finally generates the ship trajectory. This invention has made a beneficial exploration in the multi-source fusion of AIS data and radar data.

[0004] However, the above application still has the following main shortcomings:

[0005] First, the data fused in this application is limited to shore-based AIS, spaceborne AIS and radar detection signals. The data dimensions are limited to external navigation information such as the ship's position, speed, heading and static attributes. It does not involve the ship's engine room equipment operating parameters, hull structure response data, and it is difficult to form a dynamic perception of the ship's operating status. Secondly, the application adopts a fusion method of merging records after static attribute matching, treating data from different sources equally, and failing to quantitatively evaluate the credibility, accuracy, timeliness, and consistency of each data source, thus affecting the accuracy and reliability of the fusion results; Finally, the speed change and heading angle change checks used in this application are only simple threshold judgment logic, lacking the ability to model ship kinematics and predict future states based on physical mechanisms. It cannot realize online dynamic simulation and trend prediction of ship operating status, and it is difficult to meet the actual needs of intelligent ships for early warning of operational risks.

[0006] Therefore, the aforementioned application has significant shortcomings in terms of data source coverage, fusion accuracy, dynamic extrapolation capability, and decision support capability, making it difficult to meet the comprehensive needs of intelligent ships in open sea areas for real-time perception, accurate assessment, and intelligent decision-making of operational status. Therefore, a multi-source data fusion assessment method for ship operational status is proposed to solve the above problems. Summary of the Invention

[0007] (a) Purpose of the invention To address the technical problems existing in the background technology, this invention proposes a multi-source data fusion evaluation method for ship operation status, which has the effects of comprehensive perception, accurate evaluation and intelligent decision-making integrated closed loop of ship operation status, effectively solving the defects of existing technologies such as single data source, low fusion accuracy and lack of online inference and diagnostic decision support.

[0008] (II) Technical Solution This invention provides a method for multi-source data fusion and evaluation of ship operating status, comprising the following steps: S1. Multi-source heterogeneous operation status data acquisition and alignment: Collect real-time monitoring data from shipborne sensors, ship navigation status data, engine room equipment operating parameters, marine environment data, and navigation situation data, uniformly calibrate to the UTC standard time base, and uniformly map data from different spatial reference systems to a relative coordinate system centered on the ship to construct a standardized multi-source dataset with spatiotemporal consistency. S2. Multi-source data quality assessment and preprocessing: The quality of each data source is quantitatively assessed from four dimensions: completeness, accuracy, timeliness, and consistency, and the comprehensive quality score Q of each data source is calculated. i =αI i +βA i +γT i +δC i , among which, I i A i ,T i C i The first i The quantified values ​​α, β, γ, and δ of each data source in terms of completeness, accuracy, timeliness, and consistency are the corresponding weight coefficients and satisfy α+β+γ+δ=1; a local outlier detection method is used to identify abnormal data, and a particle filtering method based on ship kinematics model constraints is used to reconstruct and correct the abnormal data; different types of data are filtered using appropriate noise reduction methods, and time-domain and frequency-domain features are extracted to construct a multi-dimensional operating state feature set; S3. Weighted fusion based on data value assessment: Construct a data value assessment function that comprehensively considers the overall quality score of the data source, the time decay factor, and the correlation coefficient with the assessment task, and calculates the dynamic value weight of each data source in real time; adopt a three-level fusion architecture of data layer, feature layer, and decision layer to deeply fuse multi-source data, and recalculate the weights at fixed time periods to achieve online adaptive adjustment of the fusion weights; S4. Ship Operation Status Simulation: Establish a six-degree-of-freedom kinematic model of the ship and a health status degradation model of the equipment. Use marine environmental data as external excitation input to simulate the evolution trajectory of the ship's motion status and the degradation trend of equipment performance within a future time window. Define a relative error index for state simulation. When the deviation between the simulated value and the actual observed value exceeds a preset threshold, use ensemble Kalman filtering to assimilate the real-time observed data into the digital twin model and update the model's state variables and parameters online. S5. Comprehensive assessment of ship operation status: Based on the fused multi-source data and digital twin simulation results, the ship operation status is assessed from four dimensions: navigation safety, equipment health status, energy efficiency and economy, and navigation situation risk. The weight of each dimension is determined by a combination of subjective and objective weighting method and entropy weight method. The comprehensive score and risk level of ship operation status are generated by weighted summation or fuzzy comprehensive evaluation. S6. Assessment Results and Decision Output: Construct a three-dimensional visualization situation map of the ship's operating status, and classify and dynamically display the assessment results; when the equipment health status assessment result is abnormal or severe, activate the root cause tracing engine to trace the root cause propagation path of the fault and output a semantic fault diagnosis report; automatically generate graded operation and maintenance decision suggestions based on the comprehensive assessment results and fault diagnosis conclusions, and trigger a new round of assessment process at fixed time intervals, compare the assessment results with subsequent actual operation data, optimize the assessment model parameters online, and form an assessment closed loop; S7. Multi-ship collaborative status assessment: Based on the multi-agent collaborative perception framework, each ship shares its own status assessment results through ship-to-ship communication links to construct a formation-level comprehensive situation map. By comparing the status assessment results of similar ships within the formation, abnormal ships are identified, thus realizing multi-ship collaborative assessment and information sharing.

[0009] Preferably, the multi-source heterogeneous data mentioned in step S1 specifically includes: 1) Includes six degrees of freedom motion attitude data of the ship, main engine speed and power, propeller torque and thrust, rudder angle and rudder status, liquid level and temperature of each compartment, stress and strain data of the hull structure, and vibration acceleration data. 2) Real-time data on the ship's latitude, longitude, speed, heading, bow direction, and draft collected through the AIS system; 3) Temperature, pressure, flow rate, speed, voltage, current, and power data of the main power system, power system, auxiliary system, and propulsion system; 4) Effective wave height, wave direction, ocean current velocity and direction, wind speed and direction, and tidal level data obtained through satellite remote sensing, weather forecasting, and tidal forecasting; 5) Data on the position, speed, course, and nearest encounter distance of surrounding vessels obtained through radar, AIS, and VHF.

[0010] Preferably, the multidimensional data quality assessment in step S2 specifically includes: The integrity index measures the data missing rate and sampling coverage; the accuracy index is calculated by combining the confidence of the data source and the sensor calibration error; the timeliness index evaluates the data transmission delay and sampling frequency; and the consistency index measures the degree of agreement between multiple sources of observation of the same physical quantity. The comprehensive quality score of each data source, Qi∈[0,1], is calculated by combining the above four dimensions. The abnormal data detection includes: identifying isolated outliers using a detection method based on local outlier factors, and judging continuous abnormal sequences using a method combining a sliding window and the 3σ criterion; The noise reduction method includes: using wavelet thresholding to denoise high-frequency signals, using Kalman filtering to smooth slowly varying signals, and using the Douglas-Peucker algorithm to compress tracks and extract feature points for sequence data.

[0011] Preferably, the data value evaluation function in step S3 is constructed as follows: V_i(t)=Q_i·e^(-λ·Δt)·ρ_i; Where, Q_i is the comprehensive quality score of the data source, e^(-λ·Δt) is the time decay factor, λ is the decay coefficient, Δt is the time difference between the data acquisition time and the current time, ρ_i is the correlation coefficient between the data source and the current evaluation task; dynamic value weight wi=V_i(t) / ΣV_j(t), N is the total number of data sources; In the three-level fusion architecture, the data layer fusion uses a confidence-based weighted average to perform primary fusion of sensor data of the same type and dimension; the feature layer fusion uses principal component analysis or autoencoder to perform feature space fusion and dimensionality reduction; and the decision layer fusion uses DS evidence theory or Bayesian inference to perform decision-level fusion and generate a comprehensive evaluation conclusion.

[0012] Preferably, the six-degree-of-freedom kinematic model of the ship in step S4 adopts the MMG ship maneuverability mathematical model, with the real-time collected rudder angle, main engine speed, and propeller thrust as input variables, and the ship position, speed, heading, roll angle, and pitch angle as state variables. The fourth-order Runge-Kutta method is used for numerical integration to simulate the evolution trajectory of the ship's motion state in the future time window online. The equipment health status degradation model uses an exponential degradation model HI(t)=HI_0·e^(-θ·t) to describe the long-term decay trend of the equipment health index. The degradation rate θ is identified online through real-time data and extended Kalman filtering. The remaining service life of the equipment is predicted based on the current health index and degradation rate. The formula for calculating the relative error of the state inference is ε_state=||X_predicted(t)-X_observed(t)|| / ||X_observed(t)||, where X_predicted(t) is the ship state vector inferred by digital twin, and X_observed(t) is the actual observation value of the sensor.

[0013] Preferably, the navigation safety assessment in step S5 includes seakeeping assessment, maneuverability assessment, and stability assessment. The seakeeping assessment compares the real-time roll angle, pitch angle, and heave acceleration with the safety threshold to calculate the margin ratio. The maneuverability assessment calculates the ratio of the actual turning diameter to the design turning diameter and the heading hold error. The stability assessment calculates the ship's stability height based on real-time draft and compartment liquid level data and compares it with the minimum stability requirement. The equipment health status assessment adopts a multi-source information fusion algorithm based on joint Kalman filtering, which integrates vibration characteristics, temperature characteristics, and power characteristics to construct the equipment health index HI∈[0,1], and divides the health level into four levels: normal, attention, abnormal, and severe. The energy efficiency and economic assessment calculates the ship's real-time energy efficiency operation index and predicts fuel consumption and carbon emissions for future voyages. The navigation situation risk assessment calculates the nearest encounter distance and the nearest encounter time between the vessel and other vessels. When the CPA is less than 2 nautical miles and the TCPA is less than 12 minutes, it is determined to be a high-risk encounter.

[0014] Preferably, in the fusion of comprehensive evaluation indicators in step S5, the analytic hierarchy process (AHP) constructs a judgment matrix based on expert experience to calculate subjective weights, and the entropy weight method calculates objective weights based on the degree of variation of historical data of each dimension's evaluation indicators. The final weight is the weighted average of the subjective weights and the objective weights, with a subjective weight coefficient of 0.4 and an objective weight coefficient of 0.6. The comprehensive score S_total = Σw_j·S_j, where S_j is the evaluation score of each dimension and w_j is the corresponding weight. Map S_total to a comprehensive risk level: S_total≥85 is low risk, 70≤S_total<85 is medium risk, 50≤S_total<70 is high risk, and S_total<50 is extremely high risk.

[0015] Preferably, the hierarchical operation and maintenance decision-making recommendations in step S6 include: Normal operation outputs routine inspection suggestions; Note the recommendations to strengthen status output monitoring and status tracking; In abnormal situations, a maintenance work order is automatically generated, which includes a fault description, suggested maintenance plan, a list of required spare parts and estimated working hours, and triggers the resource scheduling module. In critical situations, emergency shutdown recommendations and emergency response procedures are output, and shore-based support centers are automatically notified. In the dynamic update of the evaluation results, a new round of evaluation process is automatically triggered every minute, and the evaluation accuracy is back-calculated every 24 hours. When the evaluation accuracy is lower than the preset threshold, the evaluation model parameters are automatically adjusted using the Bayesian optimization method.

[0016] Preferably, the root cause tracing engine in step S6 constructs a logical reasoning system based on the ship equipment expert knowledge base, combines the temporal correlation analysis of multi-source data, and uses Granger causality test and transfer entropy method to automatically trace the root cause propagation path of the fault, outputs the most likely root cause of the fault, propagation path, scope of impact and urgency according to probability, and forms a semantic fault diagnosis report.

[0017] Preferably, in the multi-ship collaborative status assessment described in step S7, each ship broadcasts its own status assessment results, including a comprehensive score, risk level, and equipment health summary, once per second via a ship-to-ship communication link. After receiving the assessment results from other ships, each ship in the formation constructs a comprehensive situational map of the formation. By comparing the status assessment results of similar ships in the same sea area and at the same time, ships whose status deviates from the formation average by more than 2 standard deviations are identified as abnormal ships. The marine environmental data measured simultaneously by multiple ships are integrated, and a high-resolution environmental field is generated through spatial interpolation to realize a collective intelligent assessment mode of individual ship perception, formation sharing, and collaborative assessment.

[0018] Compared with the prior art, the above-mentioned technical solution of the present invention has the following beneficial technical effects: 1. This multi-source data fusion assessment method for ship operation status integrates five types of heterogeneous information: real-time monitoring data from shipborne sensors (including six-degree-of-freedom motion attitude, stress and strain, vibration, etc.), ship navigation status data, engine room equipment operating parameters (main engine power, speed, exhaust temperature, etc.), marine environmental data (waves, ocean currents, weather), and navigation situation data. This constructs a comprehensive and multi-dimensional perception system for ship operation status, providing a data foundation for accurate assessment.

[0019] 2. The multi-source data fusion assessment method for ship operation status constructs a data quality quantitative scoring system from four dimensions: completeness, accuracy, timeliness, and consistency. It establishes a dynamic evaluation function of data value based on fusion time decay factor and task relevance, enabling online adaptive adjustment of fusion weights. At the same time, it adopts a three-level fusion architecture of data layer, feature layer, and decision layer to effectively handle the spatiotemporal inconsistency of heterogeneous data, ensuring that the fusion results always prioritize the use of high-confidence data sources, thereby improving the accuracy and reliability of the assessment.

[0020] 3. The multi-source data fusion assessment method for ship operation status establishes a six-degree-of-freedom kinematic model of the ship and a health status degradation model of the equipment. Using real-time control commands and environmental loads as inputs, it online extrapolates the ship's motion evolution trajectory and equipment performance degradation trend within future time windows. It introduces a relative error index for state extrapolation and an ensemble Kalman filter closed-loop correction mechanism to ensure that the digital twin is synchronized with the actual ship in the long term, thereby achieving early warning and trend judgment of the operation status.

[0021] 4. The multi-source data fusion assessment method for ship operation status establishes a comprehensive assessment system covering four dimensions: navigation safety, equipment health, energy efficiency and economy, and navigation risks. It automatically generates comprehensive scores and risk levels. When an anomaly occurs, the root cause tracing engine is activated to trace the fault propagation path and outputs semantic diagnostic reports. Based on the assessment level, it outputs operation and maintenance decision suggestions (routine inspection / enhanced monitoring / maintenance work orders / emergency response), opening up a complete link from status perception to intelligent decision-making. Furthermore, it cyclically assesses and optimizes model parameters online at fixed intervals, forming a continuously evolving closed-loop mechanism.

[0022] 5. The multi-source data fusion assessment method for ship operation status supports the expansion scheme of multi-ship status assessment based on multi-agent collaborative perception. It realizes intra-formation status sharing, abnormal ship identification and environmental field fusion enhancement through ship-to-ship communication links, and can flexibly adapt to the group assessment and collaborative decision-making needs of future intelligent ship formation operation. Attached Figure Description

[0023] Figure 1 This is a schematic diagram of the method flow of the present invention. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments and the accompanying drawings. It should be understood that these descriptions are merely exemplary and not intended to limit the scope of the invention. Furthermore, descriptions of well-known structures and techniques are omitted in the following description to avoid unnecessarily obscuring the concept of the invention.

[0025] In the description of the invention, it should be noted that the terms "upper," "lower," "inner," "outer," "front end," "rear end," "both ends," "one end," and "the other end," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0026] In the description of the invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installed," "equipped with," and "connected," etc., should be interpreted broadly. For example, "connected" can be a fixed connection, such as welding, riveting, or bonding; it can also be a detachable connection, such as threaded connection, keyed connection, or pin connection; or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium; or it can be a connection within two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0027] like Figure 1 As shown, the present invention proposes a multi-source data fusion assessment method for ship operating status, which includes the following steps: S1. Multi-source heterogeneous operating status data acquisition and spatiotemporal alignment First, comprehensive, multi-source data collection is conducted, and a distributed sensor network is deployed in key locations such as the ship's engine room, bridge, and hull structure to collect real-time data on the ship's operational status. Real-time data acquisition from shipborne sensors: Six-degree-of-freedom motion attitude (puff, sway, heave, roll, pitch, and bow) of the ship are acquired via fiber optic gyroscopes and accelerometers; main engine speed, power, torque, exhaust temperature, and cooling water temperature and pressure are acquired via the main engine monitoring system; propeller thrust and torque are acquired via shaft power meter; real-time rudder angle is acquired via rudder angle sensor; liquid levels in each oil and water tank are acquired via the compartment liquid level monitoring system; stress, strain, and vibration data of key structural parts of the hull are acquired via strain gauges and accelerometers. Ship navigation status data acquisition: The system collects navigation parameters such as ship latitude and longitude, speed, heading, bow, and draft through the AIS system; it acquires high-precision positioning data through the GPS / BeiDou system; and it obtains the ship's speed over water through a speed log. Engine room equipment operating parameter acquisition: The operating parameters of key equipment such as the main power system, power system, auxiliary machine system, and propulsion system are collected through distributed sensors, including cylinder exhaust temperature, turbocharger speed, fuel injection pressure, lubricating oil temperature and pressure, cooling water temperature, generator voltage and current, and switch cabinet power. Marine environmental data acquisition: Real-time acquisition of wind speed and direction, significant wave height, wave direction, ocean current speed and direction in the sea area where the ship is located through shipborne anemometers, wave height meters, and current meters; receiving satellite remote sensing sea state data and weather forecast data through satellite communication links; Navigation situation data acquisition: Dynamic information such as the position, speed, heading, name, and MMSI of surrounding vessels is obtained through shipborne radar, AIS receiver, and VHF communication system; All the above multi-source data are unified with a time coordinate and calibrated to the UTC international standard time base. Data from different spatial reference systems are uniformly mapped to a relative coordinate system with the ship's center of mass as the origin, thus constructing a standardized multi-source dataset that is consistent in time and space.

[0028] S2. Multi-source data quality assessment and preprocessing Implementation of multidimensional data quality assessment: For each data source, a comprehensive quality score Q_i∈[0,1] is calculated. The completeness index is obtained by the missing rate and sampling coverage of statistical data. The accuracy index is calculated by combining sensor calibration error, historical confidence of the data source, and cross-validation results with other redundant data sources. The timeliness index is obtained by measuring the deviation of data transmission delay and sampling frequency from the ideal value. The consistency index is obtained by calculating the correlation coefficient and standard deviation between multiple observations of the same physical quantity. The comprehensive quality score of each data source is calculated by combining the four-dimensional indexes and using a weighted geometric mean. Anomaly detection and correction implementation: A sliding time window (window length is 30 seconds) is used to monitor each data stream in real time. For data points within the window, the Local Outlier Factor (LOF) is calculated. When the LOF value exceeds the preset threshold, it is determined to be an isolated outlier. If more than 3 consecutive data points deviate from the 3σ range, it is determined to be a continuous anomaly sequence. For the identified anomaly data, a particle filtering method based on ship kinematics model constraints is used for data reconstruction. The ship kinematics equation is used as the state transition model, and normal sensor observations are used as the measurement update. The true state value at the time of the anomaly is estimated by particle filtering. Data denoising and feature extraction implementation: For high-frequency signals such as vibration and stress (sampling frequency ≥ 1kHz), 4-level wavelet decomposition and threshold denoising are performed using the db4 wavelet basis; for slowly varying signals such as temperature and pressure (sampling frequency ≤ 10Hz), Kalman filtering is used for smoothing; for sequence data such as AIS trajectories, the Douglas-Peucker algorithm is used with a compression threshold of 1% of the nautical mileage for track compression and feature point extraction. Based on the denoised data, time-domain statistical features (mean, variance, peak value, kurtosis, root mean square value, etc.) and frequency-domain features (FFT spectrum, power spectral density, characteristic frequency amplitude, etc.) are extracted to construct a multi-dimensional operational status feature set.

[0029] S3. Adaptive weighted fusion based on data value assessment Implementation of dynamic data value assessment: Construct a data value assessment function V_i(t)=Q_i·e^(-λ·Δt)·ρ_i, where Q_i is the comprehensive quality score of the data source, e^(-λ·Δt) is the time decay factor (λ is 0.02, Δt is the time difference between the data collection time and the current time), and ρ_i is the correlation coefficient between the data source and the current assessment task (the value range is 0~1, and it is pre-calibrated by expert experience). Calculate the dynamic value weight w_i=V_i(t) / ΣV_j(t) of each data source in real time. Multi-level converged architecture implementation: Data layer fusion: For redundant sensor data of the same type and dimension (such as temperature and pressure sensors installed at the bow, midship, and stern), a weighted average based on confidence level is used for primary fusion. The confidence level is determined by the sensor calibration accuracy and the current data quality score. After removing abnormal sensor data with a confidence level below 0.3, the weighted average is calculated. Feature layer fusion involves concatenating feature vectors extracted from different types of data to form high-dimensional feature vectors, then using principal component analysis (PCA) for dimensionality reduction, retaining principal components with a cumulative variance contribution rate of over 95%; or using a stacked autoencoder for nonlinear feature fusion and dimensionality reduction. The decision-level fusion process involves using Dempster evidence theory to integrate preliminary evaluation results from different evaluation subsystems. This process maps the evaluation conclusions of each subsystem to a basic probability assignment function (BPA) and performs evidence fusion according to Dempster's synthesis rules to generate a fused comprehensive evaluation conclusion and confidence level. Dynamic weight update mechanism implementation: Set an update cycle of 5 minutes. When the cycle is reached, the quality score and value weight of each data source are recalculated. When the quality score of a data source is lower than 0.3 for 3 consecutive cycles, its weight is automatically reduced to 0 and sensor calibration or maintenance reminder is triggered. When a new data source is connected, its quality score is automatically initialized and incorporated into the fusion framework.

[0030] S4. Dynamic simulation of ship operation status based on digital twin. Ship kinematics and dynamics modeling implementation: Establish the six-degree-of-freedom rigid body motion equations of the ship, adopt the MMG (Mathematical Manning Model) ship maneuvering mathematical model as the core of kinematics, use real-time collected control commands such as rudder angle δ, main engine speed n, and propeller thrust T as input variables, and use the ship's current position (x,y), speed V, heading ψ, roll angle φ, and pitch angle θ as state variables. Use the fourth-order Runge-Kutta method to perform numerical integration at a time step Δt=0.1s, and simulate the evolution trajectory of the ship's motion state in the next 120 seconds online; Equipment health status degradation modeling implementation: For key equipment such as ship's main propulsion system and propulsion system, a health status degradation model based on a combination of physical mechanism and data-driven approach is established. The equipment's factory performance curve is used as the benchmark model, and real-time monitoring data (exhaust temperature, vibration amplitude, power output, etc.) is used as the correction basis. The exponential degradation model HI(t)=HI_0·e^(-θ·t) is used to describe the long-term decay trend of the equipment health index. The degradation rate θ is identified online through real-time data and extended Kalman filter (EKF). Based on the current health index and degradation rate, the remaining service life of the equipment is predicted as RUL=-ln(HI_threshold / HI_current) / θ. Multi-field coupling modeling implementation: Real-time collected data on the effective wave height and direction of ocean waves, the velocity and direction of ocean currents, and the speed and direction of sea winds are used as external environmental excitation inputs to the digital twin model. The environmental load is calculated using the empirical transfer function, the wave excitation force is calculated using the ITTC recommended spectrum and the wave drift force formula, the ocean current force is calculated using the drag coefficient method, and the wind force is calculated using the Isherwood formula. The environmental load is superimposed as an external force term on the right side of the ship's six-degree-of-freedom motion equations to achieve high-fidelity simulation of the ship's operating state under environmental coupling. Digital twin closed-loop correction implementation: Define the relative error of state projection ε_state=||X_predicted(t)-X_observed(t)|| / ||X_observed(t)||, where X_predicted(t) is the ship state vector projected by the digital twin, and X_observed(t) is the actual sensor observation value. Set the error threshold ε_threshold=0.05 (5%). When ε_state>ε_threshold, the ensemble Kalman filter (EnKF) algorithm is enabled: Generate a state set containing 50 ensemble members, each member is superimposed with a Gaussian perturbation that conforms to the observation noise distribution; use real-time observation data as the measurement update, calculate the Kalman gain and update the state value of each ensemble member; take the ensemble mean as the corrected digital twin state output to realize the continuous self-correction of the digital twin.

[0031] S5. Comprehensive Assessment of Multi-Dimensional Ship Operation Status Navigation safety assessment implementation: Navigation safety is assessed from three dimensions: seakeeping, maneuverability, and stability. Seakeeping assessment—real-time roll angle φ, pitch angle θ, and heave acceleration a_z are compared with the safety thresholds specified by the IMO (roll angle <30°, pitch angle <10°, heave acceleration <0.3g), and the margin ratio of each indicator is calculated. Maneuverability assessment—the ratio of the actual turning diameter to the design turning diameter and the root mean square value of the heading hold error are calculated to assess whether the ship's maneuvering response meets the design requirements. Stability assessment—based on real-time draft and compartment level data, the ship's current stability height GM is calculated and compared with the minimum stability requirement GM_min. Combining the above three sub-dimensions, a weighted summation is used to calculate the navigation safety score S_safety∈[0,100]. Equipment health status assessment implementation: Health status assessments are conducted on key equipment such as the ship's main engine, auxiliary engines, propulsion system, and power system. Taking the main engine as an example, multi-dimensional features such as vibration acceleration characteristics (root mean square value, peak factor), exhaust temperature characteristics (temperature difference between cylinders, average temperature deviation), and power characteristics (ratio of output power to rated power) are integrated. A multi-source information fusion algorithm based on joint Kalman filtering is used to integrate the above multi-dimensional features to construct an equipment health index HI∈[0,1]. Combined with the RUL predicted by the equipment health degradation model in step four, the equipment health levels are divided as follows: HI≥0.85 is normal (green), 0.70≤HI<0.85 is alert (yellow), 0.50≤HI<0.70 is abnormal (orange), and HI<0.50 is serious (red). Energy efficiency and economic assessment implementation: Based on real-time main engine power P, fuel consumption rate SFC, ship speed V, and sailing resistance R, the real-time energy efficiency operation index EEOI is calculated as EEOI = FC / (D·C), where FC is the fuel consumption of the voyage, D is the sailing distance, and C is the cargo capacity. The real-time EEOI is compared with the benchmark EEOI of the same type of ship to calculate the energy efficiency deviation rate. Combined with the ship's future sailing status and sea state forecast data derived from digital twin simulation, the fuel consumption and carbon emissions in the next hour are predicted to assess the economic efficiency of the voyage. Navigation situation risk assessment implementation: Integrating AIS and radar data, a navigation situation map within a 3-nautical-mile radius around the vessel is constructed in real time. For each other vessel, the nearest encounter distance (CPA) and nearest encounter time (TCPA) are calculated. A CPA threshold of 2 nautical miles and a TCPA threshold of 12 minutes are set. When CPA < 2 nautical miles and TCPA < 12 minutes, it is judged as a high-risk encounter. Combined with the future trajectory of the vessel in the digital twin simulation, an early warning is given for potential navigation conflicts. The warning time window is set at 3 minutes. The comprehensive evaluation indicators are integrated and implemented as follows: The weights of each dimension are determined by a combination of subjective and objective weighting methods, which combines the Analytic Hierarchy Process (AHP) and the entropy weight method. The AHP method constructs a judgment matrix based on expert experience to calculate subjective weights; the entropy weight method calculates objective weights based on the degree of variation of historical data of each dimension's evaluation indicators. The final weight is the weighted average of the two (subjective weight coefficient 0.4, objective weight coefficient 0.6). The comprehensive score of the ship's operating status, S_total = Σw_j·S_j (j = four dimensions: safety, health, energy efficiency, and navigation risk), is calculated by weighted summation. S_total is then mapped to a comprehensive risk level: S_total ≥ 85 is low risk (green), 70 ≤ S_total < 85 is medium risk (yellow), 50 ≤ S_total < 70 is high risk (orange), and S_total < 50 is extremely high risk (red).

[0032] S6. Visualization of evaluation results and intelligent decision output Implementation of multidimensional assessment results visualization: Construct a three-dimensional visualization situation map of the ship's operating status. With a high-precision three-dimensional ship model as the core, the ship's six degrees of freedom attitude (rendered by OpenGL or Unity3D engine), equipment health status (marked on the equipment model with red-yellow-green colors), surrounding navigation situation (represented by three-dimensional arrows and trajectory lines to indicate the movement of other ships), and marine environmental field (represented by color cloud maps to indicate wave height and wind speed distribution) are displayed in real time in the three-dimensional scene. The bridge human-machine interface adopts a layered display strategy. The top layer displays the comprehensive risk level and key alarm information, the middle layer displays the assessment details of each dimension, and the bottom layer displays the raw data and trend curves. Fault Root Cause Tracing and Diagnosis Implementation: When the equipment health status assessment result is abnormal (HI<0.70) or severe (HI<0.50), the root cause tracing engine is automatically activated. Based on the logical reasoning system built on the ship equipment expert knowledge base, combined with the time-series correlation analysis of multi-source data (using Granger causality test and transfer entropy method), the system automatically traces the propagation path of the fault root cause. For example, when the main engine exhaust temperature rises abnormally, the system automatically traces to possible causes such as the fuel system, turbocharging system, and cooling system, and outputs the top 3 most likely fault root causes according to probability. Finally, it outputs a semantic fault diagnosis report containing the fault root cause, propagation path, scope of impact, and urgency level. Intelligent O&M Decision Recommendation Implementation: Based on the comprehensive evaluation results and fault diagnosis conclusions, hierarchical O&M decision recommendations are automatically generated. Normal status (green): Output routine inspection recommendations and the time window for the next comprehensive assessment; Warning status (yellow): An enhanced monitoring command has been issued, suggesting that the monitoring cycle for this device / system be shortened to 50% of the original cycle. Abnormal status (orange): Automatically generate a maintenance work order, including a fault description, suggested maintenance plan, required spare parts list and estimated man-hours; at the same time, trigger the resource scheduling module to dynamically coordinate personnel and spare parts according to the navigation criticality; Critical situation (red): Outputs emergency shutdown recommendations and emergency response procedures, automatically notifying shore support centers and ship management personnel; All decision recommendations are linked to specific equipment numbers, system names, and standard operating procedure numbers to ensure their feasibility. Implementation of dynamic updates and closed-loop feedback for assessment results: A new round of assessment process is automatically triggered every minute to achieve continuous dynamic updates of the ship's operating status. The assessment accuracy is back-calculated every 24 hours, comparing the assessment prediction results of the previous 24 hours with the subsequent actual operating data to calculate the accuracy and false alarm rate of each dimension of the assessment. When the assessment accuracy is lower than a preset threshold (e.g., 85%), online optimization of model parameters is automatically triggered: The AHP judgment matrix and fusion weight coefficients are adjusted using the Bayesian optimization method to enable the assessment model to continuously adapt to changes in the actual operating conditions of the ship, forming a complete closed loop of data collection, fusion, assessment, decision-making, and feedback.

[0033] S7, Multi-ship Coordination Status Assessment Based on a multi-agent collaborative perception framework, collaborative assessment and information sharing of the operational status of vessels in a formation are achieved. Each vessel broadcasts its own status assessment results (comprehensive score, risk level, and equipment health summary) once per second via a ship-to-ship communication link (VDE-SAT or LTE Mesh network). After receiving the assessment results from other vessels, each vessel in the formation constructs a formation-level comprehensive situation map. By comparing the status assessment results of similar vessels in the same sea area and at the same time, abnormal vessels whose status deviates significantly from the formation average (deviation exceeding 2 standard deviations) are identified. The fusion of multi-vehicle sensor data can improve the perception accuracy of the regional marine environment—for example, wave data measured synchronously by multiple vessels can generate a higher resolution wave field through spatial interpolation. Ultimately, a collective intelligent assessment mode of individual vessel perception, formation sharing, and collaborative assessment is realized.

[0034] The working principle of the above embodiments is as follows: This invention breaks through the perception limitations of traditional single-source data monitoring by comprehensively collecting and spatiotemporally aligning multi-source heterogeneous data; it ensures high accuracy and reliability of the fusion results through multi-dimensional data quality assessment and adaptive weighted fusion; it enables forward-looking prediction of ship operating status through dynamic inference driven by digital twins; it comprehensively depicts ship operating status from four dimensions: navigation safety, equipment health, energy efficiency and economy, and navigation risks through a multi-dimensional comprehensive evaluation system; it achieves a complete closed loop from status perception to intelligent decision-making through an integrated link of evaluation, diagnosis, and decision-making; and it ensures continuous optimization and long-term accuracy of the evaluation model through dynamic updates and closed-loop feedback. It can be widely applied to scenarios such as intelligent ships, autonomous navigation ships, and intelligent fleet management.

[0035] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for evaluating the multi-source data fusion of ship operating status, characterized in that, Includes the following steps: S1. Multi-source heterogeneous operation status data acquisition and alignment: Collect real-time monitoring data from shipborne sensors, ship navigation status data, engine room equipment operating parameters, marine environment data, and navigation situation data, uniformly calibrate to the UTC standard time base, and uniformly map data from different spatial reference systems to a relative coordinate system centered on the ship to construct a standardized multi-source dataset with spatiotemporal consistency. S2. Multi-source data quality assessment and preprocessing: The quality of each data source is quantitatively assessed from four dimensions: completeness, accuracy, timeliness, and consistency, and the comprehensive quality score Q of each data source is calculated. i =αI i +βA i +γT i +δC i , among which, I i A i ,T i C i The first i The quantified values ​​α, β, γ, and δ of each data source in terms of completeness, accuracy, timeliness, and consistency are the corresponding weight coefficients and satisfy α+β+γ+δ=1; a local outlier detection method is used to identify abnormal data, and a particle filtering method based on ship kinematics model constraints is used to reconstruct and correct the abnormal data; different types of data are filtered using appropriate noise reduction methods, and time-domain and frequency-domain features are extracted to construct a multi-dimensional operating state feature set; S3. Weighted fusion based on data value assessment: Construct a data value assessment function that comprehensively considers the overall quality score of the data source, the time decay factor, and the correlation coefficient with the assessment task, and calculates the dynamic value weight of each data source in real time; adopt a three-level fusion architecture of data layer, feature layer, and decision layer to deeply fuse multi-source data, and recalculate the weights at fixed time periods to achieve online adaptive adjustment of the fusion weights; S4. Ship Operation Status Simulation: Establish a six-degree-of-freedom kinematic model of the ship and a health status degradation model of the equipment. Use marine environmental data as external excitation input to simulate the evolution trajectory of the ship's motion status and the degradation trend of equipment performance within a future time window. Define a relative error index for state simulation. When the deviation between the simulated value and the actual observed value exceeds a preset threshold, use ensemble Kalman filtering to assimilate the real-time observed data into the digital twin model and update the model's state variables and parameters online. S5. Comprehensive assessment of ship operation status: Based on the fused multi-source data and digital twin simulation results, the ship operation status is assessed from four dimensions: navigation safety, equipment health status, energy efficiency and economy, and navigation situation risk. The weight of each dimension is determined by a combination of subjective and objective weighting method and entropy weight method. The comprehensive score and risk level of ship operation status are generated by weighted summation or fuzzy comprehensive evaluation. S6. Assessment Results and Decision Output: Construct a three-dimensional visualization situation map of the ship's operating status, and classify and dynamically display the assessment results; when the equipment health status assessment result is abnormal or severe, activate the root cause tracing engine to trace the root cause propagation path of the fault and output a semantic fault diagnosis report; automatically generate graded operation and maintenance decision suggestions based on the comprehensive assessment results and fault diagnosis conclusions, and trigger a new round of assessment process at fixed time intervals, compare the assessment results with subsequent actual operation data, optimize the assessment model parameters online, and form an assessment closed loop; S7. Multi-ship collaborative status assessment: Based on the multi-agent collaborative perception framework, each ship shares its own status assessment results through ship-to-ship communication links to construct a formation-level comprehensive situation map. By comparing the status assessment results of similar ships within the formation, abnormal ships are identified, thus realizing multi-ship collaborative assessment and information sharing.

2. The method for multi-source data fusion assessment of ship operating status according to claim 1, characterized in that, The multi-source heterogeneous data mentioned in step S1 specifically includes: 1) Includes six degrees of freedom motion attitude data of the ship, main engine speed and power, propeller torque and thrust, rudder angle and rudder status, liquid level and temperature of each compartment, stress and strain data of the hull structure, and vibration acceleration data. 2) Real-time data on the ship's latitude, longitude, speed, heading, bow direction, and draft collected through the AIS system; 3) Temperature, pressure, flow rate, speed, voltage, current, and power data of the main power system, power system, auxiliary system, and propulsion system; 4) Effective wave height, wave direction, ocean current velocity and direction, wind speed and direction, and tidal level data obtained through satellite remote sensing, weather forecasting, and tidal forecasting; 5) Data on the position, speed, course, and nearest encounter distance of surrounding vessels obtained through radar, AIS, and VHF.

3. The method for multi-source data fusion evaluation of ship operating status according to claim 1, characterized in that, The multidimensional data quality assessment described in step S2 specifically includes: The integrity index measures the data missing rate and sampling coverage; the accuracy index is calculated by combining the confidence of the data source and the sensor calibration error; the timeliness index evaluates the data transmission delay and sampling frequency; and the consistency index measures the degree of agreement between multiple sources of observation of the same physical quantity. The comprehensive quality score of each data source, Qi∈[0,1], is calculated by combining the above four dimensions. The abnormal data detection includes: identifying isolated outliers using a detection method based on local outlier factors, and judging continuous abnormal sequences using a method combining a sliding window and the 3σ criterion; The noise reduction method includes: using wavelet thresholding to denoise high-frequency signals, using Kalman filtering to smooth slowly varying signals, and using the Douglas-Peucker algorithm to compress tracks and extract feature points for sequence data.

4. The method for multi-source data fusion evaluation of ship operating status according to claim 1, characterized in that, The data value evaluation function described in step S3 is constructed as follows: V_i(t) = Q_i·e^(-λ·Δt)·ρ_i; Where, Q_i is the comprehensive quality score of the data source, e^(-λ·Δt) is the time decay factor, λ is the decay coefficient, Δt is the time difference between the data acquisition time and the current time, ρ_i is the correlation coefficient between the data source and the current evaluation task; dynamic value weight wi=V_i(t) / ΣV_j(t), N is the total number of data sources; In the three-level fusion architecture, the data layer fusion uses a confidence-based weighted average to perform primary fusion of sensor data of the same type and dimension; the feature layer fusion uses principal component analysis or autoencoder to perform feature space fusion and dimensionality reduction; and the decision layer fusion uses DS evidence theory or Bayesian inference to perform decision-level fusion and generate a comprehensive evaluation conclusion.

5. The method for multi-source data fusion evaluation of ship operating status according to claim 1, characterized in that, The six-degree-of-freedom kinematic model of the ship described in step S4 adopts the MMG ship maneuverability mathematical model, with real-time collected rudder angle, main engine speed, and propeller thrust as input variables, and ship position, speed, heading, roll angle, and pitch angle as state variables. The fourth-order Runge-Kutta method is used for numerical integration to online deduce the evolution trajectory of the ship's motion state within the future time window. The equipment health status degradation model uses an exponential degradation model HI(t)=HI_0·e^(-θ·t) to describe the long-term decay trend of the equipment health index. The degradation rate θ is identified online through real-time data and extended Kalman filtering. The remaining service life of the equipment is predicted based on the current health index and degradation rate. The formula for calculating the relative error of the state inference is ε_state=||X_predicted(t)-X_observed(t)|| / ||X_observed(t)||, where X_predicted(t) is the ship state vector inferred by digital twin, and X_observed(t) is the actual observation value of the sensor.

6. The method for multi-source data fusion assessment of ship operating status according to claim 1, characterized in that, The navigation safety assessment described in step S5 includes seakeeping assessment, maneuverability assessment, and stability assessment. The seakeeping assessment compares the real-time roll angle, pitch angle, and heave acceleration with the safety threshold to calculate the margin ratio. The maneuverability assessment calculates the ratio of the actual turning diameter to the design turning diameter and the heading hold error. The stability assessment calculates the ship's stability height based on real-time draft and compartment liquid level data and compares it with the minimum stability requirement. The equipment health status assessment adopts a multi-source information fusion algorithm based on joint Kalman filtering, which integrates vibration characteristics, temperature characteristics, and power characteristics to construct the equipment health index HI∈[0,1], and divides the health level into four levels: normal, attention, abnormal, and severe. The energy efficiency and economic assessment calculates the ship's real-time energy efficiency operation index and predicts fuel consumption and carbon emissions for future voyages. The navigation situation risk assessment calculates the nearest encounter distance and the nearest encounter time between the vessel and other vessels. When the CPA is less than 2 nautical miles and the TCPA is less than 12 minutes, it is determined to be a high-risk encounter.

7. The method for multi-source data fusion evaluation of ship operating status according to claim 1, characterized in that, In the integration of comprehensive evaluation indicators described in step S5, the analytic hierarchy process (AHP) constructs a judgment matrix based on expert experience to calculate subjective weights, while the entropy weight method calculates objective weights based on the degree of variation in historical data of each dimension's evaluation indicators. The final weight is the weighted average of the subjective and objective weights, with a subjective weight coefficient of 0.4 and an objective weight coefficient of 0.

6. The comprehensive score S_total = Σw_j·S_j, where S_j is the evaluation score for each dimension and w_j is the corresponding weight. Map S_total to a comprehensive risk level: S_total≥85 is low risk, 70≤S_total<85 is medium risk, 50≤S_total<70 is high risk, and S_total<50 is extremely high risk.

8. The method for multi-source data fusion assessment of ship operating status according to claim 1, characterized in that, The hierarchical operation and maintenance decision recommendations mentioned in step S6 include: Normal operation outputs routine inspection suggestions; Note the recommendations to strengthen status output monitoring and status tracking; In abnormal situations, a maintenance work order is automatically generated, which includes a fault description, suggested maintenance plan, a list of required spare parts and estimated working hours, and triggers the resource scheduling module. In critical situations, emergency shutdown recommendations and emergency response procedures are output, and shore-based support centers are automatically notified. In the dynamic update of the evaluation results, a new round of evaluation process is automatically triggered every minute, and the evaluation accuracy is back-calculated every 24 hours. When the evaluation accuracy is lower than the preset threshold, the evaluation model parameters are automatically adjusted using the Bayesian optimization method.

9. The method for multi-source data fusion assessment of ship operating status according to claim 1, characterized in that, The root cause tracing engine described in step S6 is based on a logical reasoning system built on a ship equipment expert knowledge base. It combines time-series correlation analysis of multi-source data and uses Granger causality test and transfer entropy method to automatically trace the root cause propagation path of the fault. It outputs the most likely root cause of the fault, propagation path, scope of impact and urgency in probability sorting, forming a semantic fault diagnosis report.

10. The method for multi-source data fusion evaluation of ship operating status according to claim 1, characterized in that, In the multi-ship collaborative status assessment described in step S7, each ship broadcasts its own status assessment results, including a comprehensive score, risk level, and equipment health summary, once per second via a ship-to-ship communication link. After receiving the assessment results from other ships, each ship in the formation constructs a comprehensive situational map of the formation. By comparing the status assessment results of similar ships in the same sea area and at the same time, ships whose status deviates from the formation average by more than 2 standard deviations are identified as abnormal ships. The marine environmental data measured simultaneously by multiple ships are integrated, and a high-resolution environmental field is generated through spatial interpolation to realize a collective intelligent assessment mode of individual ship perception, formation sharing, and collaborative assessment.

Images