Ship methanol leakage dynamic diffusion prediction and intervention system and method

Abstract

The invention discloses a ship methanol leakage dynamic diffusion prediction and intervention system and method. The system comprises a multi-source sensing layer, a dynamic modeling layer, a prediction and control layer and a data interaction layer. The multi-source sensing layer captures a methanol leakage signal through a sensor deployed on a ship, and summarizes real-time monitoring data to the dynamic modeling layer; the dynamic modeling layer comprises a data preprocessing module, a cabin digital twin module and a CFD dynamic diffusion engine; after the dynamic modeling layer receives real-time data, a data preprocessing module preprocesses the data, a cabin digital twin module loads a three-dimensional model of a corresponding ship, and a CFD dynamic diffusion engine starts real-time calculation based on the preprocessed data and updates a diffusion cloud picture; and the prediction and control layer receives the dynamic cloud picture output by the CFD, and triggers multi-level response when a prediction result shows that the high-risk area relates to the key cabin. Through dynamic fusion of ship motion parameters, diffusion path prediction errors are reduced, a high-risk area is predicted in advance, and key time is won for emergency response.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of ship safety monitoring, in particular to a ship methanol leakage dynamic diffusion prediction and intervention system and method. BACKGROUND

[0002] With the global shipping industry towards low-carbon transformation, methanol as a ship alternative fuel application is accelerating popularization. In the application of ship, methanol is usually stored in liquid state in independent tank, but there is a risk of leakage in the process of filling and transfer. Methanol vapor diffusion is significantly affected by environmental wind speed and air pressure, and is easy to accumulate in enclosed space, forming regional toxic and explosive risk.

[0003] Nowadays, the methanol leakage diffusion model has the following problems: the model defaults the ship to be static, does not consider the airflow disturbance in the cabin caused by rolling and pitching during sailing, the actual diffusion path has a large deviation from the prediction, and the real-time concentration monitoring alarm lacks forward-looking prediction and cannot avoid high-risk areas in advance. SUMMARY

[0004] The purpose of the present application is to provide a ship methanol leakage dynamic diffusion prediction and intervention system and method, which combines a ship motion methanol leakage dynamic diffusion prediction system to realize high-precision prediction of diffusion path, early warning and linkage prevention and control.

[0005] Technical scheme: the present application includes a multi-source perception layer, a dynamic modeling layer, a prediction and control layer, and a data interaction layer; after the sensor array deployed at the key position of the ship captures the methanol leakage signal, the multi-source perception layer collects the real-time monitoring data to the dynamic modeling layer; the dynamic modeling layer includes a data preprocessing module, a cabin digital twin module and a CFD dynamic diffusion engine; after receiving the real-time data, the data preprocessing module first preprocesses the data, at the same time, the cabin digital twin module loads the three-dimensional model of the corresponding ship type, and the CFD dynamic diffusion engine starts real-time calculation based on the preprocessed data, and updates the three-dimensional diffusion cloud picture once every period of time; the prediction and control layer receives the dynamic cloud picture output by the CFD engine, and when the prediction result shows that the high-risk area involves the key cabin, it triggers multi-level response.

[0006] The multi-source perception layer includes a ship attitude perception unit, an environment perception unit and a data aggregation module; the ship attitude perception unit includes a three-axis MEMS gyroscope and a Beidou / GPS dual-mode positioning module, the environment perception unit includes a methanol sensor deployed at the leakage risk point and a wind speed sensor at the cabin vent, and the data aggregation module aggregates sensor data and synchronizes with timestamp.

[0007] The data preprocessing module is used for filtering and normalizing the original data, and extracting ship motion characteristic parameters, including rolling period and pitch angular velocity.

[0008] The cabin digital twin module is used to pre-construct a three-dimensional digital model containing cabin structure, ventilation network, supports dynamic loading of cabin parameters of different ship types; the cabin structure includes cabin wall, equipment layout, and the ventilation network includes air outlet position, air duct size.

[0009] The CFD dynamic diffusion engine is based on the pre-processed roll angle, pitch angle, speed, ventilation rate, initial leakage concentration and digital twin model, and solves the diffusion equation in real time. When solving the diffusion equation, a ship attitude correction factor is introduced, the gravity field component is dynamically adjusted, the pressure difference between the cabin and the outside is calculated combined with the speed, the ventilation port boundary condition is corrected, the Navier-Stokes equation is solved by using the finite volume method, and the three-dimensional diffusion cloud picture is updated once every period of time.

[0010] When solving the diffusion equation, a ship attitude correction factor is introduced, the roll angle θ is converted into the x-axis gravity component (g·sinθ), and the pitch angle is converted into the y-axis gravity component to correct the flow field driving force.

[0011] The prediction and control layer includes a short-term prediction module and a linkage control module. The short-term prediction module is based on the diffusion cloud picture output by the CFD engine, and uses a long short-term memory network to predict the vapor diffusion trajectory in the future period of time. The linkage control module is used to generate multi-level control instructions when the predicted high-risk area involves a key area.

[0012] The multi-level control instructions include first-level instructions, second-level instructions and third-level instructions. The first-level instructions are used to control the electric louvers to change the ventilation direction and guide the vapor to diffuse to the deck exhaust port. The second-level instructions are used to start the regional explosion-proof fan to enhance the air exchange. The third-level instructions are used to trigger the audible and light alarms and display the dynamic evacuation path on the central control screen.

[0013] The data interaction layer synchronously stores real-time monitoring data, CFD model parameters and control instruction records.

[0014] A ship methanol leakage dynamic diffusion prediction and intervention method, comprising the following steps:

[0015] S1: The multi-source perception layer collects ship motion parameters and environmental data, and transmits them to the dynamic modeling layer after time stamp synchronization;

[0016] S2: The dynamic modeling layer pre-processes the data, based on the cabin digital twin model and the CFD engine, and updates the three-dimensional diffusion cloud picture once every period of time;

[0017] S3: The prediction and control layer uses an LSTM network to predict high-risk areas in the future. If critical areas are involved, multi-level control commands are triggered. At the same time, the data interaction layer stores in real time the raw monitoring data collected by the multi-source sensing layer, the diffusion model parameters generated by the dynamic modeling layer, and the control commands output by the prediction and control layer.

[0018] Beneficial effects: This invention reduces the error in diffusion path prediction by dynamically integrating ship motion parameters, predicts high-risk areas in advance, buys critical time for emergency response, and links the ventilation system to change the diffusion direction, reducing the concentration in high-risk areas; it supports ships of different tonnages and various cabin types, ensuring the safety of personnel in the current environment. Attached Figure Description

[0019] Figure 1 This is a system architecture diagram of the present invention;

[0020] Figure 2 This is a flowchart of the CFD dynamic diffusion model parameter correction process of this invention;

[0021] Figure 3 This is a comparison chart of the diffusion prediction and control effects under ship roll conditions according to the present invention. Detailed Implementation

[0022] The invention will now be further described with reference to the accompanying drawings.

[0023] Example 1

[0024] like Figure 1 As shown, the ship methanol leakage dynamic diffusion prediction and intervention system of this embodiment includes a multi-source sensing layer, a dynamic modeling layer, a prediction and control layer, and a data interaction layer. The multi-source sensing layer includes a ship attitude sensing unit, an environmental sensing unit, and a data aggregation module. The ship attitude sensing unit includes a three-axis MEMS gyroscope and a Beidou / GPS dual-mode positioning module. The three-axis MEMS gyroscope and the Beidou / GPS dual-mode positioning module are used to collect roll angle, pitch angle, speed, and heading angle in real time. The environmental sensing unit includes a non-dispersive infrared methanol sensor deployed at leakage risk points and a wind speed sensor at the cabin ventilation opening. The data aggregation module aggregates sensor data via industrial Ethernet and uses timestamp synchronization.

[0025] The dynamic modeling layer includes a data preprocessing module, a cabin digital twin module, and a CFD dynamic diffusion engine. The data preprocessing module is used to filter (adaptive Kalman filter to remove ship body vibration noise) and normalize the original data, extract ship motion characteristic parameters, and the ship motion characteristic parameters include roll period and pitch angular velocity. The cabin digital twin module is used to pre-construct a three-dimensional digital model containing cabin structure and ventilation network, and supports dynamic loading of cabin parameters of different ship types. The cabin structure includes bulkhead and equipment layout, and the ventilation network includes air outlet position and air duct size. The CFD dynamic diffusion engine solves the diffusion equation in real time based on the pre-processed roll angle, pitch angle, speed, ventilation rate, initial leakage concentration, and digital twin model. When solving the diffusion equation, a ship attitude correction factor is introduced, the gravity field component is dynamically adjusted, the pressure difference between the cabin and the outside is calculated combined with the speed, the ventilation port boundary condition is corrected, the Navier-Stokes equation is solved by using the finite volume method, and the three-dimensional diffusion cloud picture is updated every 0.5 seconds. In the dynamic modeling layer, the relationship between the speed and the pressure difference between the cabin and the outside is calculated based on the Bernoulli equation: ΔP = 0.5ρv 2 , where ρ is the air density (1.225 kg / m 3 ), v is the speed (unit: m / s); the ventilation port wind speed is corrected as the superposition value of the reference wind speed v0 and the associated increment of the speed: v1 = v0 + 0.05v (0.05 is an empirical coefficient).

[0026] The prediction and control layer includes a short-term prediction module and a linkage control module. When the predicted high-risk area involves a key area, the linkage control module generates multi-level control instructions. The short-term prediction module uses a long short-term memory network (LSTM) to predict the vapor diffusion trajectory within 5 seconds in the future based on the diffusion cloud picture output by the CFD engine, and identifies high-risk areas (such as living area passages and power distribution cabins) with a concentration ≥200ppm. The LSTM network of the prediction and control layer uses 3 layers of hidden layers (64 neurons per layer), the input is the spatial coordinates (x, y, z) and the time sequence (the first 10 seconds of data) of the diffusion cloud picture, and the training data includes 1000 groups of real ship leakage simulation data and 200 groups of historical accident data. The linkage control module is used to generate multi-level control instructions when the predicted high-risk area involves a key area. The multi-level control instructions include primary instructions, secondary instructions, and tertiary instructions. The primary instructions are used to control the electric louvers to change the ventilation direction and guide the vapor to diffuse towards the deck exhaust port. The secondary instructions are used to start the regional explosion-proof fan to enhance ventilation. The tertiary instructions are used to trigger the sound and light alarm and display the dynamic evacuation path on the central control screen.

[0027] The data interaction layer synchronously stores real-time monitoring data, CFD model parameters and control instruction records, realizes data intercommunication with the ship energy efficiency management system through an Ethernet interface, and simultaneously provides a 5G remote access channel, supports the real-time retrieval of a diffusion cloud map, a predicted trajectory and a control state by a shore-based monitoring center, and forms a full closed-loop management from leakage monitoring, dynamic prediction to linkage prevention and control.

[0028] After the multi-source perception layer captures the methanol leakage signal through the sensor array deployed at key positions of the ship, real-time monitoring data is aggregated to the dynamic modeling layer; the real-time monitoring data covers ship motion data and environmental state data, the ship motion data includes roll angle, pitch angle collected by a three-axis MEMS gyroscope and speed, heading angle obtained by a Beidou / GPS module, and the environmental state data includes concentration readings of a non-dispersive infrared methanol sensor at risk points such as a storage tank interface and a pipeline flange, and real-time air volume of a cabin ventilation port air speed sensor. The data aggregation module realizes time stamp synchronization of sensor data through a redundant industrial Ethernet, and ensures the spatio-temporal matching of motion parameters and environmental parameters.

[0029] As shown in Figure 2 , after the dynamic modeling layer receives real-time data, the data preprocessing module first removes high-frequency noise introduced by ship vibration by using adaptive Kalman filtering, and then performs normalization processing on the data to extract key motion characteristic parameters such as roll period, pitch angular velocity and speed; at the same time, the cabin digital twin module loads a three-dimensional model corresponding to the type of the ship, the model includes cabin wall coordinates, equipment layout and ventilation network parameters. The CFD dynamic diffusion engine starts real-time calculation based on the preprocessed data: introduce the ship attitude correction factor, convert the roll angle θ into the x-axis gravity component (g sin θ), convert the pitch angle into the y-axis gravity component to correct the flow field driving force; calculate the cabin pressure difference by combining the speed, and correct the ventilation port air speed as the superposition value of the reference air speed and the speed-related increment; use the finite volume method to solve the Navier-Stokes equation to divide the calculation domain, and update the three-dimensional diffusion cloud map once every 0.5 seconds to intuitively present the spatial distribution of methanol vapor.

[0030] After the prediction and control layer receives the dynamic cloud map output by the CFD engine, the short-term prediction module deduces the diffusion trajectory for the next 5 seconds through an LSTM network, and automatically identifies high-risk areas with a concentration ≥200 ppm. When the prediction result shows that the high-risk area involves a key cabin, the linkage control module triggers multi-level response: the first-level instruction adjusts the angle of the electric louver through a servo motor to guide the vapor to diffuse to the deck explosion-proof exhaust port; the second-level instruction starts the regional explosion-proof fan to enhance the ventilation efficiency; the third-level instruction triggers an audible and visual alarm, and generates a dynamic evacuation path on the central control screen.

[0031] As shown in Figure 3Left side (no control): roll angle -8°~+8°, 5 seconds later, the high concentration area (>200ppm) spreads to the living area, and the concentration is 250ppm; right side (with control): after adjusting the direction of the ventilation port, 5 seconds later, the concentration of the high concentration area is reduced to 180ppm, and it does not spread to the living area; the legend contains airflow direction, low / medium / high concentration area (50-100ppm, 100-200ppm, >200ppm) annotation.

[0032] Example 2

[0033] The ship methanol leakage dynamic diffusion prediction and intervention method of the embodiment comprises the following steps:

[0034] S1: The multi-source perception layer collects ship motion parameters (roll angle, pitch angle, etc.) and environmental data (methanol concentration, ventilation rate), and transmits the data to the dynamic modeling layer after time stamp synchronization;

[0035] S2: The dynamic modeling layer preprocesses the data through adaptive Kalman filtering, and updates the three-dimensional diffusion cloud map every 0.5 seconds based on the cabin digital twin model and the CFD engine;

[0036] S3: The prediction and control layer predicts the high-risk area in the next 5 seconds through the LSTM network, and triggers multi-level control instructions if the key area is involved; at the same time, the data interaction layer stores the original monitoring data collected by the multi-source perception layer, the diffusion model parameters generated by the dynamic modeling layer, and the control instruction records output by the prediction and control layer, and supports the docking with the ship energy efficiency management system (SEEMP) based on the stored data, and provides Ethernet / 5G interface to realize remote access and data sharing, forming a linkage mechanism of parallel intervention action and data storage.

[0037] The specific working process and principle of the application are as follows: the multi-source perception layer collects motion parameters such as roll angle, pitch angle, and speed through the ship attitude perception unit (three-axis MEMS gyroscope, Beidou / GPS module), and collects cabin concentration and ventilation data through the environmental perception unit (methanol sensor, wind speed sensor), and transmits the data to the dynamic modeling layer after time stamp synchronization through the data aggregation module; the dynamic modeling layer first preprocesses the data through adaptive Kalman filtering to remove noise, and then introduces the ship attitude correction factor (roll angle θ corresponds to x-axis gravity component gx=g*sinθ, pitch angle corresponding to y-axis component The gravity field is adjusted, and the pressure difference between the inside and outside of the cabin is calculated based on the Bernoulli equation to correct the boundary conditions of the ventilation port. The finite volume method is used to update the three-dimensional diffusion cloud picture every 0.5 seconds; the prediction and control layer predicts the high-risk area with a concentration of ≥200ppm 5 seconds in advance through an LSTM network based on the spatiotemporal data of the diffusion cloud picture, and if the key area is involved, multi-level control instructions (adjusting the direction of the electric louver, starting the explosion-proof fan, triggering the sound-light alarm, etc.) are triggered to block the diffusion; the data interaction layer synchronously stores data, supports docking with the ship energy efficiency management system and remote access, forms a full closed-loop management from monitoring, dynamic modeling, prediction to intervention, solves the prediction deviation and early warning lag caused by the neglect of ship motion in traditional static models, and realizes accurate prediction and active prevention and control.

Claims

1. A methanol leak dynamic diffusion prediction and intervention system for a marine vessel, characterized in that, The system comprises a multi-source perception layer, a dynamic modeling layer, a prediction and control layer, and a data interaction layer; after the sensor array deployed at the key position of the ship captures the methanol leakage signal, the multi-source perception layer collects real-time monitoring data to the dynamic modeling layer; the dynamic modeling layer comprises a data preprocessing module, a cabin digital twin module, and a CFD dynamic diffusion engine; after receiving real-time data, the data preprocessing module first preprocesses the data, at the same time, the cabin digital twin module loads the three-dimensional model of the corresponding ship type, and the CFD dynamic diffusion engine starts real-time calculation based on the preprocessed data, and updates the three-dimensional diffusion cloud picture once every certain period of time; the prediction and control layer receives the dynamic cloud picture output by the CFD engine, and when the prediction result shows that the high-risk area involves the key cabin, a multi-level response is triggered.

2. The marine vessel methanol leak dynamic dispersion prediction and intervention system of claim 1, wherein, The multi-source perception layer comprises a ship attitude perception unit, an environment perception unit, and a data aggregation module; the ship attitude perception unit comprises a three-axis MEMS gyroscope and a Beidou / GPS dual-mode positioning module, the environment perception unit comprises a methanol sensor deployed at the leakage risk point and a wind speed sensor at the cabin vent, and the data aggregation module aggregates sensor data and synchronizes with timestamps.

3. The marine vessel methanol leak dynamic dispersion prediction and intervention system of claim 1, wherein, The data preprocessing module is used for filtering and normalizing the original data, and extracting ship motion characteristic parameters, wherein the ship motion characteristic parameters comprise roll period and pitch angular velocity.

4. The marine vessel methanol leak dynamic dispersion prediction and intervention system of claim 1, wherein, The cabin digital twin module is used for pre-building a three-dimensional digital model comprising a cabin structure and a ventilation network, and supporting dynamic loading of cabin parameters of different ship types; the cabin structure comprises a bulkhead and equipment layout, and the ventilation network comprises a vent position and a duct size.

5. The marine vessel methanol leak dynamic dispersion prediction and intervention system of claim 1, wherein, The CFD dynamic diffusion engine solves the diffusion equation in real time based on the preprocessed roll angle, pitch angle, speed, ventilation rate, initial leakage concentration, and digital twin model, introduces a ship attitude correction factor when solving the diffusion equation, dynamically adjusts the gravity field component, calculates the cabin pressure difference combined with the speed, corrects the boundary conditions of the ventilation port, solves the Navier-Stokes equation by using the finite volume method, and updates the three-dimensional diffusion cloud picture once every certain period of time.

6. The ship methanol leak dynamic dispersion prediction and intervention system of claim 5, wherein, When solving the diffusion equation, the ship attitude correction factor is introduced, the roll angle θ is converted into the x-axis gravity component (g·sinθ), and the pitch angle φ is converted into the y-axis gravity component (g·sinφ) to correct the flow field driving force.

7. The marine vessel methanol leak dynamic dispersion prediction and intervention system of claim 1, wherein, The prediction and control layer comprises a short-term prediction module and a linkage control module; the short-term prediction module predicts the vapor diffusion trajectory in a future period of time based on the diffusion cloud picture output by the CFD engine by using a long short-term memory network; and the linkage control module is used for generating multi-level control instructions when predicting that the high-risk area involves the key area.

8. The marine vessel methanol leak dynamic dispersion prediction and intervention system of claim 7, wherein, The multi-level control instructions comprise first-level instructions, second-level instructions, and third-level instructions; the first-level instructions are used for controlling the electric louvers to change the ventilation direction and guide the vapor to diffuse to the deck exhaust port; the second-level instructions are used for starting the regional explosion-proof fan to enhance ventilation; and the third-level instructions are used for triggering the sound and light alarm and displaying the dynamic evacuation path on the central control screen.

9. The marine vessel methanol leak dynamic dispersion prediction and intervention system of claim 1, wherein, The data interaction layer synchronously stores real-time monitoring data, CFD model parameters, and control instruction records.

10. A method of predicting and intervening in dynamic diffusion of methanol leakage in a marine vessel, characterized by, The method is based on the ship methanol leakage dynamic diffusion prediction and intervention system according to any one of claims 1-9, and specifically comprises the following steps: S1: The multi-source perception layer collects ship motion parameters and environmental data, and transmits the data to the dynamic modeling layer after time stamp synchronization; S2: The dynamic modeling layer pre-processes the data, and updates the three-dimensional diffusion cloud map once every period of time based on the cabin digital twin model and the CFD engine; S3: The prediction and control layer predicts the high-risk area in the future period of time through the LSTM network, and triggers multi-level control instructions if the key area is involved; at the same time, the data interaction layer stores the original monitoring data collected by the multi-source perception layer, the diffusion model parameters generated by the dynamic modeling layer, and the control instruction records output by the prediction and control layer in real time.

Images