Intelligent monitoring system for liquid level of ship

Abstract

The invention provides a ship liquid level intelligent monitoring system. The ship liquid level intelligent monitoring system comprises a data acquisition layer, a data processing layer and an application layer, the data acquisition layer comprises an inertial measurement unit and a liquid level measurement unit, the inertial measurement unit is used for measuring ship attitude information, and the liquid level measurement unit is used for acquiring original liquid level information in a liquid tank; the data processing layer comprises a data preprocessing module, an intelligent liquid level correction module, a model training updating module and a local server, the data preprocessing module is used for preprocessing information, and the intelligent liquid level correction module can calculate static liquid level information according to the preprocessed information; the model training and updating module can utilize historical navigation data stored in the local database to train the liquid level dynamic correction model; the application layer is used for receiving and displaying the static liquid level information; according to the invention, the finally output liquid level value is close to the real static liquid level, and the measurement precision and reliability are greatly improved.

Description

Technical Field

[0001] This invention relates to the technical field of ship liquid level monitoring, and more specifically, to an intelligent ship liquid level monitoring system. Background Technology

[0002] In the shipping industry, accurate monitoring of liquid levels in tanks (such as fuel tanks, ballast water tanks, and freshwater tanks) is crucial, directly affecting ship stability, loading calculations, navigation safety, and economy. Currently, the liquid level monitoring devices commonly used in ships are mostly contact-type (such as float-type and pressure-type) or non-contact-type (such as radar and ultrasonic) sensors. Although these traditional technologies are mature, they have significant limitations in practical applications.

[0003] Ships inevitably experience rolling and pitching motions during navigation. This continuous dynamic swaying causes the liquid level in the tanks to tilt and slosh, resulting in a "false level" reading from a sensor at a single mounting point that is not the true vertical liquid level. This error leads to a significant discrepancy between the calculated tank capacity and the actual value, severely affecting the accuracy of loading calculations and posing a threat to the ship's safety and economic operation.

[0004] In existing technologies, while some solutions attempt to compensate for the interference of ship motion on liquid level measurement by introducing additional sensor data or algorithms, significant limitations remain. For example, some technologies rely solely on simple geometric models combined with static tilt angle data to correct liquid level errors. However, the motion of ships in complex sea conditions involves multi-degree-of-freedom coupled motions of roll, pitch, and vertical sway, and the sloshing of liquid within the tank exhibits significant nonlinear characteristics. Simple geometric models cannot effectively capture these complex dynamic disturbances, resulting in limited correction effects and failing to meet the requirements of high-precision monitoring.

[0005] Furthermore, there are shortcomings in the coordination between liquid level monitoring and other key systems (such as energy efficiency management and stability control) during ship operation. For example, Chinese invention patent CN118387260A discloses a method for predicting ship fuel consumption based on distributed energy efficiency calculation. Although it mentions collecting fuel quantity data through liquid level sensors to calculate the fuel consumption rate and constructing a fuel energy efficiency prediction model by combining buoyancy distribution and waterline stability, its core objective is to achieve fuel consumption prediction and speed optimization. Liquid level data is only used as an auxiliary input for fuel quantity statistics. It does not address the liquid level measurement error caused by ship motion, nor does it propose a dynamic correction scheme for the liquid level data itself, thus failing to provide technical support for improving the accuracy of liquid level monitoring.

[0006] Therefore, there is an urgent need for a monitoring solution that can deeply integrate the real-time motion attitude of ships and use intelligent algorithms to accurately analyze and correct dynamic liquid level data, so as to fundamentally improve measurement accuracy and reliability. Summary of the Invention

[0007] In view of this, the present invention aims to propose an intelligent monitoring system for ship liquid level to solve the problem of large liquid level measurement errors caused by the rolling and pitching motion of ships in the prior art.

[0008] To achieve the above objectives, the technical solution of the present invention is implemented as follows:

[0009] A ship liquid level intelligent monitoring system includes a data acquisition layer, a data processing layer, and an application layer;

[0010] The data acquisition layer includes an inertial measurement unit and a liquid level measurement unit installed in the liquid tank. The inertial measurement unit is used to measure the ship's attitude information in real time, including roll angle, pitch angle and sway acceleration. The liquid level measurement unit is used to collect the raw liquid level information in the liquid tank.

[0011] The data processing layer includes a data preprocessing module, an intelligent liquid level correction module, a model training and update module, and a local server. The data preprocessing module is used to preprocess the collected information. The intelligent liquid level correction module is equipped with a liquid level dynamic correction model, which can calculate static liquid level information based on the preprocessed information. The local server is used to store the ship's historical navigation data in a local database. The model training and update module can use the historical navigation data stored in the local database to train the liquid level dynamic correction model and update the parameters of the liquid level dynamic correction model.

[0012] The application layer includes an output module, which is used to receive and display static liquid level information output by the data processing layer.

[0013] Furthermore, the data preprocessing module is used to filter and reduce noise on the collected raw liquid level information and ship attitude information, and to synchronize and align the raw liquid level information and ship attitude information based on timestamps.

[0014] Furthermore, the liquid level dynamic correction model is set as a gradient boosting tree regression model.

[0015] Furthermore, the liquid level dynamic correction model can calculate the static liquid level information for the current period in real time based on the static liquid level information calculated in the previous period and the currently collected original liquid level information and ship attitude information.

[0016] Furthermore, the historical navigation data includes raw liquid level information, ship attitude information, and verified static liquid level information under different sea conditions and loading states.

[0017] Furthermore, the inertial measurement unit includes a three-axis gyroscope and a three-axis accelerometer, and the inertial measurement unit is installed near the center of gravity of the ship or in the liquid tank that needs to be monitored.

[0018] Furthermore, the data acquisition layer and the data processing layer transmit data via a CAN bus or Ethernet.

[0019] Furthermore, the data preprocessing module uses a Kalman filter to smooth the ship's roll and pitch angle information, and a moving average filter to reduce noise in the original liquid level information.

[0020] Furthermore, when the planned berthing time of the vessel exceeds 24 hours, the model training and update module will use the newly added historical navigation data to incrementally train the liquid level dynamic correction model.

[0021] Furthermore, when the cumulative amount of newly added historical navigation data for the vessel reaches 100,000 records, the model training and update module will use all the historical navigation data to perform full training on the liquid level dynamic correction model.

[0022] Compared with existing technologies, the intelligent ship liquid level monitoring system of the present invention has the following advantages:

[0023] 1) It overcomes the problem of "false liquid level" caused by ship rolling, so that the final output liquid level value is infinitely close to the real static liquid level, and the measurement accuracy and reliability are greatly improved.

[0024] 2) The liquid level dynamic correction model can be continuously optimized with the accumulation of navigation data, and it shows good adaptability and robustness to different ships, different liquid tanks and different sea states. Attached Figure Description

[0025] Figure 1 This is a schematic diagram of the module connection of the intelligent ship liquid level monitoring system according to an embodiment of the present invention. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to embodiments. It should be understood that the described embodiments are only some, not all, of the embodiments of this invention. The specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0027] Example 1

[0028] like Figure 1 As shown, this embodiment provides a ship liquid level intelligent monitoring system, including a data acquisition layer, a data processing layer, and an application layer;

[0029] The data acquisition layer includes an inertial measurement unit and a liquid level measurement unit installed in the liquid tank. The inertial measurement unit is used to measure the ship's attitude information in real time, including roll angle, pitch angle and sway acceleration. The liquid level measurement unit is used to collect the raw liquid level information in the liquid tank.

[0030] The data processing layer includes a data preprocessing module, an intelligent liquid level correction module, a model training and update module, and a local server. The data preprocessing module is used to preprocess the collected information. The intelligent liquid level correction module is equipped with a liquid level dynamic correction model, which can calculate static liquid level information based on the preprocessed information. The local server is used to store the ship's historical navigation data in a local database. The model training and update module can use the historical navigation data stored in the local database to train the liquid level dynamic correction model and update the parameters of the liquid level dynamic correction model.

[0031] The application layer includes an output module, which is used to receive and display static liquid level information output by the data processing layer.

[0032] This setup can simultaneously acquire ship attitude information (roll angle, pitch angle, and sway acceleration) and raw liquid level information, accurately addressing the core issue of "dynamic swaying interfering with liquid level measurement" during ship navigation. By using attitude data, it can directly capture the causes of liquid surface tilting and swaying, overcoming the "false liquid level" problem caused by ship swaying. This makes the final output liquid level value infinitely close to the true static liquid level, greatly improving measurement accuracy and reliability.

[0033] Preferably, the liquid level dynamic correction model is trained based on machine learning algorithms.

[0034] Preferably, the local server will continuously store the ship's navigation data (including raw liquid level information, ship attitude information and system-verified static liquid level information) into the local database, forming a continuously growing dataset that is exclusive to the ship.

[0035] As a preferred example of this application, the data preprocessing module is used to filter and reduce noise on the collected raw liquid level information and ship attitude information, and to synchronize and align the raw liquid level information and ship attitude information based on timestamps.

[0036] Specifically, the complex navigation environment of ships makes raw data susceptible to various types of interference. On the one hand, when a ship is continuously rolling or pitching, the raw liquid level information collected by the liquid level measurement unit may exhibit "spiking data" due to instantaneous swaying of the liquid surface, and the attitude information such as roll and pitch angles obtained by the inertial measurement unit may also experience random fluctuations due to hull vibration. On the other hand, electromagnetic interference may also cause outliers during data transmission. By filtering and noise reduction processing, these non-realistic interference data can be effectively eliminated, making the processed raw liquid level information closer to the actual trend of liquid surface changes, and the ship attitude information more accurately reflecting the true motion state of the hull, thus ensuring data quality from the source.

[0037] In the data acquisition layer, although the inertial measurement unit (IMU) and the liquid level measurement unit operate synchronously, slight differences in hardware response speed and data acquisition frequency can lead to minor time discrepancies between the attitude and liquid level information acquired at the same moment. For example, the IMU might acquire roll angle data at time t1, while the liquid level measurement unit might only acquire the corresponding liquid level data at t1+0.1 seconds. If data with this time difference is directly input into the liquid level dynamic correction model, the model will calculate the static liquid level based on the asynchronous attitude and liquid level data, resulting in a discrepancy between the calculated value and the actual situation. Synchronization and alignment based on timestamps ensure that each set of "ship attitude information and original liquid level information" involved in the calculation strictly corresponds to the same time node, enabling the model to accurately capture the real-time interference pattern of the ship's motion on the liquid surface at a given moment and avoiding correction errors caused by data time asynchrony.

[0038] As a preferred example of this application, the liquid level dynamic correction model is set as a gradient boosting tree regression model.

[0039] Specifically, liquid level disturbances during ship navigation exhibit significant nonlinear characteristics. On one hand, ship motion is a multi-degree-of-freedom coupled motion of roll, pitch, and vertical sway, rather than a linear oscillation in a single direction. On the other hand, the sloshing of liquid within the tanks exhibits different nonlinear patterns depending on the load, tank shape, and sea state. Traditional simple geometric or linear models cannot effectively characterize this complex relationship. The gradient boosting tree regression model, through iterative ensemble of multiple decision trees, can adaptively learn the nonlinear mapping relationship between input features (roll angle, pitch angle, sway acceleration, initial liquid level, etc.) and output (static liquid level). It can accurately capture the inherent laws of "ship attitude and liquid level disturbances" under different sea states and loading conditions, thereby avoiding correction errors caused by insufficient model fitting ability.

[0040] It should be noted here that the gradient boosting tree regression model is an existing technology and will not be elaborated on further.

[0041] As a preferred example of this application, the liquid level dynamic correction model can calculate the static liquid level information of the current period in real time based on the static liquid level information calculated in the previous period and the currently collected original liquid level information and ship attitude information.

[0042] Specifically, the complex navigation environment of ships means that the raw liquid level information or ship attitude information collected in the current cycle may contain outliers due to sudden interference (such as violent fluctuations in the liquid surface caused by sudden strong waves or brief electromagnetic interference from sensors). If the model relies solely on the data from the current cycle, these outliers will cause "jumps" in the static liquid level, affecting the reliability of subsequent functions such as tank capacity calculation and over-limit alarms. However, by incorporating the static liquid level information from the previous cycle, the model can verify the current data using historical data, identifying and mitigating the impact of outliers. For example, if the raw liquid level in the current cycle is displayed as 4.8m due to interference (the static liquid level in the previous cycle was 5.2m, and there was no significant liquid level consumption), the model, combined with the ship attitude information, determines that this change does not conform to the continuity of the liquid level and will refer to the data from the previous cycle to correct the current result, avoiding calculation errors caused by outliers. This "dual-cycle data cross-validation" mechanism is equivalent to adding a fault-tolerant barrier to the static liquid level calculation, which can significantly improve the system's anti-interference capability and data stability under complex sea conditions.

[0043] Preferably, one cycle is 1 second.

[0044] As a preferred example of this application, the historical navigation data includes raw liquid level information, ship attitude information, and verified static liquid level information under different sea conditions and loading states.

[0045] Specifically, during ship navigation, liquid level measurement is affected by both sea state and loading status. The more complex the sea state, the more severe the interference of hull rolling on the liquid surface. Different loading statuses also significantly change the nonlinear characteristics of liquid sloshing within the tanks. If historical data only covers a single sea state or a single loading status, the model can only learn the "attitude-liquid level interference pattern" of a local scenario. When switching scenarios (such as from calm sea states to rough sea states), the correction accuracy will drop significantly. However, data that includes multiple sea states and loading statuses allows the model to cover most scenarios in actual ship operation, ensuring that the model can output accurate static liquid levels in any scenario and avoiding measurement errors caused by insufficient scenario adaptation.

[0046] As a preferred example of this application, the inertial measurement unit includes a three-axis gyroscope and a three-axis accelerometer, and the inertial measurement unit is installed near the center of gravity of the ship or in the liquid tank that needs to be monitored.

[0047] Specifically, a three-axis gyroscope can accurately measure the angular velocity of a ship around its X-axis (roll), Y-axis (pitch), and Z-axis (bow), reflecting the rate of change of the ship's attitude in real time and avoiding the "attitude change lag" problem caused by only measuring angles and not rates. A three-axis accelerometer can simultaneously collect the linear acceleration of the ship along the X, Y, and Z axes, especially capturing the acceleration changes of vertical swaying (such as the ship's up-and-down rolling). Although this vertical motion does not directly change the horizontal tilt angle of the liquid surface, it will cause the liquid surface to "fluctuate up and down." Relying solely on angle data will miss this type of interference. The combination of the three-axis gyroscope and the three-axis accelerometer forms a full-dimensional attitude monitoring system of "angular velocity + linear acceleration," which can completely cover all dynamic interference sources that affect liquid level measurement, providing accurate information for subsequent data preprocessing and liquid level correction models.

[0048] The ship's center of gravity is the "reference point" for its motion. Whether it's rolling, pitching, or vertical swaying, the attitude change at the center of gravity best represents the ship's overall motion state and is least affected by local structural deformation. Installing an inertial measurement unit (IMU) at this location provides global attitude information, suitable for scenarios requiring simultaneous monitoring of multiple liquid tanks (such as fuel tanks and ballast water tanks). This eliminates the need for a separate IMU for each tank, reducing system cost and data redundancy.

[0049] When high-precision monitoring is required only for a specific liquid tank (such as a fuel tank), the inertial measurement unit can be directly installed inside the liquid tank to directly measure the local attitude changes of the liquid tank, avoiding the problem that "global attitude data cannot reflect local differences".

[0050] Preferably, the liquid level measurement unit is configured as a non-contact radar level gauge and installed at the center of the top of the liquid tank. The environment inside ship liquid tanks (such as fuel tanks and ballast water tanks) is harsh, with problems such as liquid sloshing, media corrosion (fuel oil, seawater), and temperature and humidity fluctuations. Traditional contact level gauges (such as float-type and pressure-type) are easily affected by these factors, leading to measurement failure. Non-contact radar level gauges, however, measure the distance to the liquid surface by emitting high-frequency electromagnetic waves (such as microwaves), without direct contact with the liquid. This avoids the failures caused by wear and jamming of mechanical parts in float-type level gauges, and eliminates concerns about the accuracy reduction caused by media corrosion and blockage in pressure-type level gauges, significantly reducing equipment maintenance costs and failure rates.

[0051] In addition, contact level gauges are prone to "data jumps" due to liquid level fluctuations, while radar level gauges can smooth instantaneous fluctuations through signal filtering algorithms. Furthermore, electromagnetic wave propagation is far less affected by liquid level disturbances than mechanical contact, enabling more stable acquisition of raw liquid level information and reducing measurement noise in dynamic environments.

[0052] As a preferred example of this application, the data acquisition layer and the data processing layer transmit data via a CAN bus or Ethernet.

[0053] Specifically, the ship's navigation environment presents challenges such as strong electromagnetic interference (e.g., electromagnetic radiation from engines and radar equipment), vibration and shock (e.g., hull rolling and wave impact), and drastic temperature and humidity changes (e.g., high temperatures in the engine room and high humidity at sea). Ordinary data transmission methods (such as RS485) are easily affected by these factors, leading to data transmission errors. The CAN bus, using differential signal transmission, effectively suppresses electromagnetic interference, maintaining a very low data transmission error rate even in strong electromagnetic environments. Furthermore, the CAN bus hardware interface features short-circuit and overvoltage protection, withstands ship vibration and temperature and humidity fluctuations, and is suitable for installation in harsh areas such as engine rooms and liquid tanks. Ethernet, using shielded twisted-pair cables or fiber optic transmission, further enhances electromagnetic interference resistance; fiber optic transmission can even completely eliminate electromagnetic interference. This strong environmental adaptability ensures stable and reliable data transmission links throughout the ship's entire navigation cycle, preventing data interruptions or errors due to environmental interference and guaranteeing system stability.

[0054] As a preferred example of this application, the data preprocessing module uses a Kalman filter to smooth the ship's roll and pitch angle information, and uses a moving average filter to reduce noise in the original liquid level information.

[0055] Specifically, the roll and pitch angles exhibit "continuous dynamic changes" with the ship's motion. The interference is mostly transient and random high-frequency noise, and the changes in attitude information follow clear physical laws (such as the roll angle fluctuating periodically with the wave cycle). The core advantage of the Kalman filter is that through the iterative logic of "prediction-update," combined with the statistical characteristics of the data and the physical model, it can smooth noise while preserving the dynamic trend of the data. It can eliminate high-frequency interference caused by instantaneous vibrations without causing attitude changes to lag due to excessive smoothing, accurately restoring the true attitude motion law of the hull and avoiding the impact of attitude data distortion on subsequent liquid level correction. For attitude information, the Kalman filter can effectively filter out instantaneous outliers caused by mechanical vibration and electromagnetic interference, making the processed attitude data closer to the actual motion state of the hull and avoiding model misjudgment caused by such outliers.

[0056] The raw liquid level information is affected by surface sloshing, resulting in relatively smooth fluctuations. The interference is mostly continuous, low-amplitude random noise (such as fluctuations in liquid level data caused by tiny ripples on the liquid surface). The moving average filter can efficiently smooth out this type of low-amplitude noise by calculating the average value of the liquid level data within a certain window. Moreover, the algorithm logic is simple and the calculation time is short, which can quickly complete the processing within the monitoring period and avoid data delays caused by complex calculations. At the same time, the smooth filtering effect can preserve the overall trend of liquid level changes (such as the slow drop in liquid level caused by fuel consumption) and will not destroy the continuity of liquid level data, which meets the requirement of continuous and stable input for the liquid level dynamic correction model. For the raw liquid level information, the moving average filter can smooth the continuous fluctuations caused by local surface sloshing and small sensor errors, making the processed liquid level data closer to the true average liquid level height, avoiding the model from misjudging local fluctuations as real changes in liquid level, and reducing correction errors.

[0057] As a preferred example of this application, when the planned berthing time of the vessel exceeds 24 hours, the model training and update module will use the newly added historical navigation data to incrementally train the liquid level dynamic correction model.

[0058] Specifically, when a ship is scheduled to be docked for more than 24 hours (such as during port unloading, replenishment, or maintenance), the real-time requirements of the liquid level monitoring system are significantly reduced (liquid tank level changes are gradual and do not require high-frequency correction), and the server has idle computing power. At this time, incremental training can be triggered to complete the model update without affecting the key navigation functions, perfectly adapting to the ship operation rules of "navigation priority and docking optimization", balancing the needs of real-time system monitoring and model iteration.

[0059] As a preferred example of this application, when the cumulative amount of newly added historical navigation data of the ship reaches 100,000 records, the model training and update module will use all the historical navigation data to perform full training on the liquid level dynamic correction model.

[0060] Specifically, while incremental training can quickly adapt to new scenarios, its long-term reliance on "local data iterative parameter updates" may lead to "local drift" in model parameters. That is, the parameters only adapt to recently added scenarios and gradually deviate from the optimal fit to early multi-scenario data. In contrast, full training can "globally recalibrate" the model parameters by calling all historical data.

[0061] For example, in incremental training, the tree structure and feature weights of a gradient boosting tree regression model may be overweighted for sea states 5-6 due to frequent recent encounters with such sea states. During full training, the model will redistribute the weights based on all sea state data to ensure that the feature weights for all sea states, such as 0-2, 3-4, and 7-8, are in the optimal state, avoiding biased "specialization".

[0062] This global calibration can completely eliminate the local errors accumulated during incremental training, enabling the liquid level dynamic correction model to be continuously optimized as navigation data accumulates, and showing good adaptability and robustness to different ships, different liquid tanks and different sea states.

[0063] As a preferred example of this application, the model training and update module can update the trained liquid level dynamic correction model to the intelligent liquid level correction module online.

[0064] Specifically, the online update mode can complete the model replacement while the intelligent liquid level correction module is running. During the update process, the old model can still process data and output static liquid level normally. After the new model is loaded and verified in the background, it can seamlessly take over the work of the old model. This update method can ensure continuous and stable liquid level monitoring throughout the entire cycle, avoid functional stagnation caused by model updates, and improve the stability and safety of ship operation.

[0065] Preferably, the liquid level dynamic correction model after incremental training or full training, after verification, will seamlessly replace the old model by hot deployment or planned restart, thereby optimizing the system. All data and algorithms are retained on the ship's side without the need to connect to an external network.

[0066] Preferably, the static liquid level information received by the application layer is displayed on the monitoring screens of the bridge and engine room, and is used to automatically calculate the liquid tank capacity, update the ship's loading system and energy efficiency management system, and the system can determine whether the liquid level exceeds the limit. If it does, an alarm will be triggered to issue an audible and visual alarm signal.

[0067] The measurement errors of the intelligent ship liquid level monitoring system in this embodiment are compared with those of a traditional liquid level monitoring system, and the results are shown in Table 1:

[0068] Table 1 - Comparison of Measurement Errors between This Embodiment and Traditional Measurement Systems

[0069]

[0070] As can be seen from Table 1, the intelligent ship liquid level monitoring system in this embodiment has a smaller measurement error compared to the traditional liquid level monitoring system.

[0071] In this embodiment, when the cumulative amount of historical navigation data added to the local database reaches 100,000 records, the model training and update module is used to perform full training on the liquid level dynamic correction model, and the measurement errors of the liquid level dynamic correction model before and after training are compared. The measurement errors are shown in Table 2.

[0072] Table 2 - Comparison of Measurement Errors Before and After Full Training of the Model in this Embodiment

[0073]

[0074] As can be seen from Table 2, after training the liquid level dynamic correction model in this embodiment based on the ship's historical navigation data, its measurement accuracy has been greatly improved compared with the liquid level dynamic correction model before training. Moreover, with the continuous accumulation of data, its measurement accuracy will be further improved.

[0075] The intelligent ship liquid level monitoring system described in this application has the following advantages:

[0076] 1) High measurement accuracy: It overcomes the problem of "false liquid level" caused by ship rolling, so that the final output liquid level value is infinitely close to the true static liquid level, and the measurement accuracy and reliability are greatly improved;

[0077] 2) Strong adaptability: The liquid level dynamic correction model can be continuously optimized with the accumulation of navigation data, and it shows good adaptability and robustness to different ships, different liquid tanks and different sea states.

[0078] 3) High safety and economy: Accurate liquid level information can enhance the reliability of ship stability calculation and loading scheme formulation, and significantly reduce the safety risks caused by improper loading; at the same time, accurate fuel tank monitoring helps to optimize energy efficiency management, reduce energy waste, and provide more accurate measurement basis for trade transactions, thereby improving economic efficiency.

[0079] 4) High degree of automation: It reduces human intervention and misjudgment, and improves the level of automation and decision-making efficiency of ship operation management.

[0080] While the present invention has been disclosed above, it is not limited thereto. Any person skilled in the art can make various modifications and alterations without departing from the spirit and scope of the invention; therefore, the scope of protection of the present invention should be determined by the scope defined in the claims.

Claims

1. A ship liquid level intelligent monitoring system, characterized in that, It includes a data acquisition layer, a data processing layer, and an application layer; The data acquisition layer includes an inertial measurement unit and a liquid level measurement unit installed in the liquid tank. The inertial measurement unit is used to measure the ship's attitude information in real time, including roll angle, pitch angle and sway acceleration. The liquid level measurement unit is used to collect the raw liquid level information in the liquid tank. The data processing layer includes a data preprocessing module, an intelligent liquid level correction module, a model training and update module, and a local server. The data preprocessing module is used to preprocess the collected information. The intelligent liquid level correction module is equipped with a liquid level dynamic correction model, which can calculate static liquid level information based on the preprocessed information. The local server is used to store the ship's historical navigation data in a local database. The model training and update module can use the historical navigation data stored in the local database to train the liquid level dynamic correction model and update the parameters of the liquid level dynamic correction model. The application layer includes an output module, which is used to receive and display static liquid level information output by the data processing layer.

2. The intelligent ship liquid level monitoring system according to claim 1, characterized in that, The data preprocessing module is used to filter and reduce noise on the collected raw liquid level information and ship attitude information, and to synchronize and align the raw liquid level information and ship attitude information based on timestamps.

3. The intelligent ship liquid level monitoring system according to claim 1, characterized in that, The liquid level dynamic correction model is set as a gradient boosting tree regression model.

4. The intelligent ship liquid level monitoring system according to claim 1, characterized in that, The liquid level dynamic correction model can calculate the static liquid level information of the current period in real time based on the static liquid level information calculated in the previous period and the currently collected original liquid level information and ship attitude information.

5. The intelligent ship liquid level monitoring system according to claim 1, characterized in that, The historical navigation data includes raw liquid level information, ship attitude information, and verified static liquid level information under different sea conditions and loading conditions.

6. The intelligent ship liquid level monitoring system according to claim 1, characterized in that, The inertial measurement unit includes a three-axis gyroscope and a three-axis accelerometer, and the inertial measurement unit is installed near the center of gravity of the ship or in the liquid tank that needs to be monitored.

7. The intelligent ship liquid level monitoring system according to claim 1, characterized in that, The data acquisition layer and the data processing layer transmit data via CAN bus or Ethernet.

8. The intelligent ship liquid level monitoring system according to claim 1, characterized in that, The data preprocessing module uses a Kalman filter to smooth the ship's roll and pitch angle information, and a moving average filter to reduce noise in the raw liquid level information.

9. The intelligent ship liquid level monitoring system according to claim 1, characterized in that, When the planned berthing time of the vessel exceeds 24 hours, the model training and update module will use the newly added historical navigation data to incrementally train the liquid level dynamic correction model.

10. The intelligent ship liquid level monitoring system according to claim 9, characterized in that, When the cumulative amount of newly added historical navigation data for the vessel reaches 100,000 records, the model training and update module will use all the historical navigation data to perform full training on the liquid level dynamic correction model.

Images