CN-121982863-A - Accurate positioning type ship methanol leakage real-time monitoring linkage early warning method and system

Abstract

The invention provides a linkage early warning method and system for monitoring methanol leakage of a precisely positioned ship in real time, which are characterized in that risk grades are divided through fluid dynamics simulation of a bilge water easily-leaked area, an integrated sensing unit is deployed to collect multidimensional water quality index and equipment running state data, a probe is used for compounding protection coating erosion resistance, data correction and abnormal elimination are carried out through combining self-adaptive calibration and wavelet transformation filtering, an edge computing unit is configured in each risk area, after receiving data through a private network, false signals are identified through multi-source data fusion analysis, a sliding window model pre-judging trend is combined, response delay is shortened, an early warning interval is set in a grading mode according to the explosion limit of methanol and related water quality threshold, a triple signal transmission mode is adopted, a data interface is opened based on a unified communication protocol, and various information is integrated through a related analysis model of a central control system to construct a full-link monitoring data link.

Inventors

• LIU CHEN
• CHEN RONGCHANG
• SUN SHUTING
• XING XINYING
• LIU CHUNLING

Assignees

• 交通运输部水运科学研究所

Dates

Publication Date

20260505

Application Date

20260202

Claims (10)

1. 1. The method for accurately positioning ship methanol leakage real-time monitoring linkage early warning is characterized by comprising the following steps of: The risk level is divided through the fluid dynamics simulation of the bilge water easy-leakage area, and an integrated sensing unit is deployed to collect multi-dimensional water quality index and equipment running state data; the probe is combined with the protection coating to resist erosion, and the self-adaptive calibration and wavelet transformation filtering are combined to correct data and reject anomalies; each risk zone is provided with an edge calculation unit, after receiving data through a private network, false signals are identified through multi-source data fusion analysis, and a sliding window model is combined to predict trend, so that response delay is shortened; Setting an early warning interval and associating a water quality threshold according to the methanol explosion limit in a grading manner, triggering multi-parameter cooperative early warning through a logic gating algorithm, and adopting a triple signal transmission mode; the data interface is opened based on a unified communication protocol, and multiple types of information are integrated through an association analysis model of the central control system to construct a full-link monitoring data chain; And training an LSTM model by utilizing historical monitoring data, outputting concentration change trend and leakage risk probability in real time, and triggering pretreatment measures in advance when the concentration change trend and leakage risk probability reach the standards.
2. 2. The method for real-time monitoring and linkage early warning of methanol leakage of a precisely positioned ship according to claim 1, wherein the hydrodynamic simulation of the bilge water easily-leaked area divides risk levels, and an integrated sensing unit is deployed to collect multidimensional water quality index and equipment operation state data, comprising: based on a ship design drawing, cabin layout parameters and methanol leakage characteristics, restoring a bilge water related structure by using simulation software, defining mixed fluid properties, setting boundary conditions, simulating methanol diffusion processes under different leakage intensities, ship postures and sea conditions, monitoring diffusion related characteristics and extracting data; The methanol concentration parameter, the diffusion coverage, the time to reach the equipment and the personnel safety threat degree in the simulation result are synthesized, the bilge water area is divided into a high-level risk area, a medium-level risk area and a low-level risk area, and the space boundary and the space range of each level area are defined; Adopting an anti-corrosion waterproof sealing shell, integrating a water quality monitoring module and an equipment state monitoring module, arranging sensing units in a different mode according to risk grades by a built-in data preprocessing chip and a wireless transmission module, arranging the sensing units densely in a high risk area, uniformly arranging the risk areas in the high risk area and sparsely arranging the low risk area, and connecting all the sensing units with a data acquisition gateway; The method comprises the steps of collecting water quality index and equipment running state related data through an integrated sensing unit, dynamically adjusting the collection frequency according to the risk level, preprocessing collected original data, transmitting the preprocessed data to a data collection gateway, primarily fusing and processing multi-source data through edge calculation by the gateway, and uploading the data to a central control system; And (3) carrying out field calibration on the sensing unit regularly, carrying out comprehensive inspection and aging part replacement by using the dock repairing period of the ship, and dynamically updating and optimizing the deployment of the sensing unit by combining the simulation model.
3. 3. The method for real-time monitoring and linkage early warning of methanol leakage of a precisely positioned ship according to claim 2, wherein the comprehensive simulation results include methanol concentration parameters, diffusion coverage, time to reach equipment and personnel safety threat level, dividing bilge water areas into high, medium and low three-level risk areas, defining space boundaries and ranges of all levels of areas, and comprising the following steps: Screening a methanol concentration peak value, an average concentration, a diffusion coverage area and a diffusion boundary coordinate of each bilge water area from a multi-scene simulation result, recording the shortest time for the methanol to reach different equipment, and quantifying the safety threat degree of personnel by combining the distance between the area and a personnel operation area and the living area and the ventilation condition to form a basic data set containing four indexes of concentration, diffusion, time and safety; Determining each index weight by adopting an analytic hierarchy process, wherein the methanol concentration parameter weight is 40%, the diffusion coverage weight is 25%, the arrival time weight of equipment is 20%, the personnel safety threat degree weight is 15%, and formulating each index quantization scoring standard, wherein the higher the concentration is, the larger the diffusion range is, the shorter the arrival time is, the more serious the personnel threat is, and the score is higher; assigning values to each index in the basic data set according to the scoring standard, and calculating the comprehensive risk value of each sub-region by a weighted summation formula, wherein the comprehensive score is = concentration score x 40% + diffusion range score x 25% + arrival time score x 20% + personnel safety score x 15%; Combining the methanol safety standard of the ship and the safety management requirement of a bilge water area, determining that the comprehensive risk score is more than or equal to 85 and is divided into a high risk area, 60-84 is divided into a medium risk area and <60 is divided into a low risk area, and determining the corresponding characteristics of all levels of risks; Based on a three-dimensional coordinate system of a simulation model, merging the continuous sub-areas with consistent risk levels by utilizing a spatial clustering algorithm according to the comprehensive risk levels of the sub-areas, extracting boundary coordinate points of the clustered areas to outline the spatial contour of each level of risk area, and defining the three-dimensional coordinate ranges, boundary trend and relative position relation with bilge structures of high, medium and low risk areas; And comparing risk level changes of the same area under different simulation scenes, eliminating misjudgment areas caused by scene fluctuation, adjusting boundary coordinates, and correcting boundary lines by combining physical structure limitation of the bilge water tank to finally form accurate and clear risk division results.
4. 4. The method for real-time monitoring and linkage early warning of methanol leakage of accurately positioned ships according to claim 1, wherein each risk zone is provided with an edge calculation unit, false signals are identified through multi-source data fusion analysis after data is received by a private network, and trend prediction and response delay shortening are combined with a sliding window model, and the method comprises the following steps: The high-risk area is configured with a high-performance multi-core edge gateway of the integrated GPU acceleration module, the risk area is configured with a standard edge computing unit, the low-risk area is configured with a lightweight edge node, all edge computing units are connected with corresponding area sensing units through a ship industry Ethernet private network, a UDP protocol is adopted to ensure low delay of data transmission, and a CRC32 check mechanism is arranged to ensure data integrity; After receiving multi-dimensional sensing data of methanol concentration, pH value, equipment vibration and temperature, an edge computing unit inputs an improved self-encoder fusion model, extracts feature vectors through an encoder, reconstructs data through a decoder and computes reconstruction errors, screens abnormal data by combining with a preset threshold value, adopts a DBSCAN clustering algorithm to perform clustering analysis on the abnormal data, distinguishes isolated false signals caused by sensor faults and environmental interference from continuous abnormal data caused by real leakage, and automatically marks and eliminates false signals; The method comprises the steps of setting a sliding window based on dynamic adjustment of the acquisition frequency of a risk area in an edge calculation unit, calculating the change slope of concentration and diffusion related indexes through linear regression analysis of data in the window, and pre-judging the leakage diffusion rate and the development trend; And establishing a cooperative communication mechanism among the edge computing units, realizing the real-time sharing of data processing results and trend prejudging information of edge nodes of high and medium risk areas, and guaranteeing the continuity and accuracy of trend prejudgment of leakage diffusion crossing risk areas.
5. 5. The method for real-time monitoring and linkage early warning of methanol leakage of a precisely positioned ship according to claim 4, wherein after receiving the multi-dimensional sensing data of methanol concentration, pH value, equipment vibration and temperature, the edge calculation unit inputs an improved self-encoder fusion model, extracts feature vectors and decoder reconstruction data through an encoder and calculates reconstruction errors, screens the abnormal data by combining with a preset threshold value, adopts a DBSCAN clustering algorithm to perform clustering analysis on the abnormal data, distinguishes isolated false signals caused by sensor faults and environmental interference from continuous abnormal data caused by real leakage, and automatically marks and eliminates false signals, and comprises the following steps: processing the multi-dimensional sensing data of the methanol concentration, the pH value, the equipment vibration and the temperature received by the edge calculation unit, uniformly converting the original data of different types of sensors into a standardized format, eliminating dimension differences through Z-score normalization, removing impulse noise in the data by adopting a median filtering method, and filling the missing data by utilizing linear interpolation; The encoder of the model adopts a three-layer convolutional neural network structure, the first layer and the second layer are respectively provided with a corresponding convolutional kernel size, the third layer adopts a global average pooling layer to extract deep coupling feature vectors of multi-dimensional data, the decoder adopts a deconvolution structure to reconstruct the feature vectors into reconstruction data consistent with the input data dimension, an L1 regularization term is introduced into a model loss function, and the model is iteratively trained to be converged by adopting multi-source sensing data under a ship bilge water normal working condition and a simulated leakage working condition as a training data set through an Adam optimizer; Inputting the preprocessed multidimensional sensing data into a trained improved self-encoder fusion model in real time, calculating a reconstruction error of each data sample, setting and updating a dynamic threshold in real time by a moving average method in combination with a historical data statistical result under the actual running condition of the ship, and marking corresponding data as abnormal data when the reconstruction error exceeds the dynamic threshold; Constructing a data point set according to the time stamp of the abnormal data and the space position information of the deployment coordinates of the sensing units, adaptively adjusting the neighborhood radius epsilon based on the distance between the sensing units and the data acquisition frequency, setting the minimum clustering point MinPts by combining the number of the sensing units in the area, traversing the data point set to find a core point, a density reachable point and an isolated point, gathering the abnormal data with the density reachable to one type, and judging the isolated abnormal data as false signals; Automatically marking a real abnormal data cluster formed by clustering, reserving the real abnormal data cluster for subsequent analysis, generating a reject log for recording false signal sources, time and types, and synchronously uploading the reject log to a central control system.
6. 6. The method for real-time monitoring and linkage early warning of methanol leakage of a precisely positioned ship according to claim 5, wherein the encoder of the model adopts a three-layer convolutional neural network structure, the first layer and the second layer are both provided with corresponding convolutional kernel sizes, the third layer adopts a global average pooling layer to extract deep coupling feature vectors of multi-dimensional data, the decoder adopts a deconvolution structure to reconstruct the feature vectors into reconstructed data consistent with the input data dimension, an L1 regularization term is introduced into a model loss function, and the model is iteratively trained to be converged by adopting multi-source sensing data under normal conditions and simulated leakage conditions of ship bilge water as a training data set through an Adam optimizer, comprising: The encoder part constructs a three-layer convolution neural network architecture, wherein the input dimension of a first layer of convolution layer is matched with the preprocessed multidimensional sensing data dimension, the number and the dimension of convolution kernels are set, a ReLU activation function is adopted, the configuration step length is 1, the filling mode is the same parameter, the second layer of convolution layer takes the first layer output as the input, a larger number of convolution kernels with the same dimension are set, the ReLU activation function is adopted, the third layer adopts a global average pooling layer to reduce the dimension of the second layer output characteristic diagram, and the second layer of convolution layer is converted into a one-dimensional characteristic vector containing multidimensional data deep coupling characteristics; The decoder part constructs a deconvolution structure symmetrical to the encoder, the first deconvolution layer takes the output characteristic vector of the encoder as input, the deconvolution kernel size is matched with the second layer convolution kernel size of the encoder, the one-dimensional vector is restored to be a two-dimensional characteristic diagram, the second layer deconvolution kernel size is matched with the first layer convolution kernel size of the encoder, the characteristic dimension is gradually restored, the last deconvolution layer outputs reconstruction data consistent with the original input data dimension, and the activation function adopts Sigmoid; Introducing an L1 regularization term into the mean square error loss function, dynamically adjusting an L1 regularization coefficient according to a pre-training result, selecting an Adam optimizer, and setting learning rate, first-order moment estimation attenuation rate and second-order moment estimation attenuation rate parameters; Collecting multi-dimensional sensing data of methanol concentration, pH value, equipment vibration and temperature under normal working conditions and simulated leakage working conditions of ship bilge water, and dividing the sensing data into a training set, a verification set and a test set according to proportion after pretreatment; Inputting training set data into a model in batches, obtaining reconstruction data through forward propagation, calculating a mean square error loss value containing L1 regularization, updating model weight and bias through reverse propagation, evaluating reconstruction accuracy through a verification set after each training round, attenuating learning rate if the verification set loss value is continuously and repeatedly not reduced, stopping training when the training set loss value is reduced below a preset threshold value and the verification set loss value tends to be stable, storing model parameters, and verifying model reconstruction effects through a test set after training is completed.
7. 7. The method for real-time monitoring and linkage early warning of methanol leakage of a precisely positioned ship according to claim 1, wherein the steps of setting an early warning interval and associating water quality thresholds according to the explosion limit of methanol, triggering multi-parameter cooperative early warning through a logic gating algorithm, adopting a triple signal transmission mode, and comprising the following steps: Dividing a three-level early warning interval by combining the volume fraction range of the methanol explosion limit and the bilge water environment characteristic, wherein the first-level early warning interval corresponds to the lower limit of the methanol volume fraction to the lower limit of the explosion limit, the second-level early warning interval corresponds to the lower limit of the methanol explosion limit to the middle limit of the explosion limit, the third-level early warning interval corresponds to the middle limit of the methanol explosion limit to the upper limit of the explosion limit, and the three-level early warning interval corresponds to the middle limit of the methanol explosion limit to the upper limit of the explosion limit; The method comprises the steps of preprocessing sensing data of the concentration of methanol, the pH value and the dissolved oxygen content, inputting the sensing data into a logic gating algorithm, setting an AND gate logic judgment rule, outputting an effective early warning signal when the concentration of the methanol reaches a corresponding early warning interval threshold value and the pH value or the dissolved oxygen content meets an associated water quality threshold value, judging the sensing data as an interference signal when only a single parameter meets the standard, not triggering early warning, setting a parameter priority mechanism in the algorithm, enabling the weight of the concentration of the methanol to be higher than other water quality indexes, and directly improving the early warning grade when the concentration of the methanol is close to an explosion limit threshold value; The first is the marine industry Ethernet transmission, upload the early warning signal to the central control system in real time, the second is the wireless communication backup transmission of loRa, switch over to this radio link automatically when the Ethernet trouble, the third is the local audible and visual alarm transmission, control the desk, personnel's passageway to set up audible and visual alarm in the equipment near the area of leakage, start the local warning when the early warning triggers, each retransmission link adopts check mechanism and timestamp synchronization, triple link mutual redundancy, real-time supervision, when any one heavy link trouble, the system sends the unusual suggestion of link automatically.
8. 8. The method for real-time monitoring and linkage early warning of methanol leakage of a precisely positioned ship according to claim 1, wherein the method for constructing a full-link monitoring data chain by integrating multiple types of information through a correlation analysis model of a central control system based on a unified communication protocol is characterized by comprising the following steps: The method comprises the steps of (1) converting heterogeneous data of each system into standardized data frames in a unified format by adopting an industrial Ethernet protocol based on IEC 61162-450 standard as a unified communication standard according to the existing communication protocols of a carding ship central control system, a methanol leakage monitoring system, a fuel supply system, a ventilation system and an inerting device association system, defining field definitions of the data frames, and setting a check rule and an abnormal retransmission mechanism of interface data transmission; taking multidimensional sensing data, system running state data, risk area division data and leakage positioning data as inputs, adopting a correlation analysis algorithm based on machine learning, extracting characteristics of various data through characteristic engineering, constructing a multi-source data correlation matrix to mine potential correlation among different types of data, integrating influence factors of ship navigation states and environmental parameters, and optimizing model correlation analysis precision; The method comprises the steps of designing a full-link data flow architecture for data acquisition, transmission, processing, storage and application, acquiring data in real time through a multi-dimensional sensing array and system interfaces, transmitting the data to a central control system through a unified communication protocol, carrying out fusion association analysis on multi-source data through an association analysis model, filtering invalid data, marking abnormal information, generating a monitoring conclusion, adopting a storage scheme combining a distributed database and a time sequence database to store the data, establishing a data backup and disaster recovery mechanism, pushing the processed monitoring data to a visualization interface of the central control system, supporting early warning triggering, linkage treatment instruction generation and leakage treatment effect evaluation application scene, and forming a full-link monitoring data chain of closed loop flow.
9. 9. The method for real-time monitoring and linkage early warning of methanol leakage of a precisely positioned ship according to claim 8, wherein the method takes multidimensional sensing data, system running state data, risk area division data and leakage positioning data as input, adopts a correlation analysis algorithm based on machine learning, extracts characteristics of various data through characteristic engineering, constructs a multi-source data correlation matrix to mine potential correlation among different types of data, fuses influence factors of ship navigation states and environmental parameters, optimizes model correlation analysis precision, and comprises the following steps: uniformly converting multidimensional sensing data, system running state data, risk area division data, leakage positioning data, ship navigation state data and environmental parameters into a time-aligned structured data set, adopting Z-score normalization to eliminate dimension differences, filling missing data through sliding window interpolation, and removing impulse noise through median filtering to ensure consistency and integrity of input data; Extracting dynamic characteristics, state characteristics, space associated characteristics and interference characteristics from various input data, classifying and encoding the extracted characteristics to form a high-dimensional characteristic vector set, carrying out characteristic dimension reduction through mutual information calculation and principal component analysis, and retaining the characteristics strongly related to leakage events; constructing a multisource data association matrix by adopting an Apriori algorithm, taking feature vectors after dimension reduction as input, mining frequent item sets and association rules among different features, screening association relations with practical significance through confidence and support, introducing a random forest algorithm as an auxiliary verification module, inputting the association rules as features into the module, verifying the reliability of the association relations by calculating feature importance weights and dynamically adjusting the weight distribution of the association matrix, and integrating ship navigation states and environmental parameters as influencing factors to generate composite features through feature intersection so as to correct the confidence of the association rules; Training a model by adopting a ship historical leakage event data set and a simulated leakage scene data set, adjusting the minimum support degree, the confidence coefficient threshold value and the tree quantity parameters of a random forest of an association analysis algorithm through cross verification, continuously receiving new input data by the model in a real-time operation stage, dynamically updating an association matrix, iteratively optimizing an association rule through an online learning mechanism, and establishing a model effect evaluation mechanism taking the hit rate and the false alarm rate of the association rule as indexes to periodically evaluate the performance of the model.
10. 10. Accurate location formula boats and ships methyl alcohol leakage real-time supervision linkage early warning system, its characterized in that includes: A processor; a machine-readable storage medium storing machine-executable instructions for the processor; wherein the processor is configured to execute the precisely positioned ship methanol leak real-time monitoring linkage early warning method and system of any one of claims 1 to 9 via execution of the machine executable instructions.

Description

Accurate positioning type ship methanol leakage real-time monitoring linkage early warning method and system Technical Field The invention relates to the technical field of monitoring linkage early warning methods, in particular to a precise positioning type ship methanol leakage real-time monitoring linkage early warning method and system. Background The ship methanol fuel has the advantages of cleanness, environmental protection and high energy density, is increasingly widely applied to the environment-friendly transformation of the shipping industry, but methanol has the characteristics of strong volatility and wide explosion limit range (volume fraction is 6% -36.5%), once leakage occurs in a bilge water area, the bilge water area is easy to mix with seawater to form a flammable and explosive environment, and the leakage and diffusion are influenced by factors such as ship postures, sea conditions, cabin layout and the like, so that safety accidents are extremely easy to cause. The current ship methanol leakage monitoring is mostly dependent on a single concentration sensor, and has the problems of lack of pertinence in deployment, easy environmental interference of data, high false alarm rate, difficult accurate capturing of dynamic process and potential risk of leakage diffusion and the like. The existing monitoring system has the defects of serious data island, non-uniform communication protocol, longer response delay and the like, the data of each related system (fuel supply, ventilation, inerting devices and the like) cannot be effectively integrated, and most of the monitoring systems are used for passive early warning after leakage occurs, and the prediction and advanced treatment capability of leakage trend is lacking. Meanwhile, the severe environment of high salt, high humidity and strong corrosion of bilge water is easy to cause the performance attenuation and data distortion of a sensor, further influences the monitoring precision and the system reliability, and is difficult to meet the severe requirements of real-time prevention and control of methanol leakage in the ship navigation process. Disclosure of Invention In view of the above-mentioned problems, in combination with the first aspect of the present invention, an embodiment of the present invention provides a method and a system for precisely positioning a ship for real-time monitoring and linkage early warning of methanol leakage, where the method includes: The risk level is divided through the fluid dynamics simulation of the bilge water easy-leakage area, and an integrated sensing unit is deployed to collect multi-dimensional water quality index and equipment running state data; the probe is combined with the protection coating to resist erosion, and the self-adaptive calibration and wavelet transformation filtering are combined to correct data and reject anomalies; each risk zone is provided with an edge calculation unit, after receiving data through a private network, false signals are identified through multi-source data fusion analysis, and a sliding window model is combined to predict trend, so that response delay is shortened; Setting an early warning interval and associating a water quality threshold according to the methanol explosion limit in a grading manner, triggering multi-parameter cooperative early warning through a logic gating algorithm, and adopting a triple signal transmission mode; the data interface is opened based on a unified communication protocol, and multiple types of information are integrated through an association analysis model of the central control system to construct a full-link monitoring data chain; And training an LSTM model by utilizing historical monitoring data, outputting concentration change trend and leakage risk probability in real time, and triggering pretreatment measures in advance when the concentration change trend and leakage risk probability reach the standards. In still another aspect, an embodiment of the present invention further provides a precise positioning type ship methanol leakage real-time monitoring and linkage early warning system, which is characterized by comprising: The system comprises a processor, a machine-readable storage medium, a machine-executable instruction, a computer-readable storage medium and a computer-readable program, wherein the processor is configured to execute the accurate positioning type ship methanol leakage real-time monitoring linkage early warning method and system by executing the machine-executable instruction. In still another aspect, an embodiment of the present invention further provides a computer program product, where the computer program product includes machine executable instructions, where the machine executable instructions are stored in a computer readable storage medium, and a processor of a computer device reads the machine executable instructions from the computer readable storage medium, and the processor executes the machi