Ship liquid level prediction method based on multi-sensor data fusion and dynamic correction

Abstract

The invention relates to a ship liquid level prediction method based on multi-sensor data fusion and dynamic correction, and the method comprises the following steps: S1, obtaining data collected by a plurality of sensors, and carrying out the quality scoring of the data, the quality scoring being based on the data loss rate and the data abnormality calculation; s2, according to a quality scoring result, calculating a weighting factor of each sensor, and carrying out weighted updating on data collected by each sensor; s3, fusing the weighted and updated sensor data by using an evidence theory method to obtain a preliminary ship liquid level prediction result; and S4, correcting the preliminary fusion result by adopting a differential reweighting strategy to obtain a final ship liquid level prediction result. Compared with the prior art, the measurement precision and reliability of each sensor are comprehensively considered, the data fusion result is dynamically corrected, and the precision of the liquid level prediction result can be further improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of artificial intelligence and intelligent ship data fusion, and in particular relates to a ship liquid level prediction method based on multi-sensor data fusion and dynamic correction. Background Art

[0002] With the rapid development of modern ship and ocean engineering technologies, the improvement of ship safety and operation efficiency increasingly depends on efficient monitoring and control systems. These systems can perform multiple tasks such as liquid level monitoring, environmental monitoring, fault diagnosis, and navigation state assessment through sensor networks. Especially in liquid level monitoring, by collecting and analyzing real-time data of ships and ocean equipment, it is possible to effectively prevent safety accidents caused by abnormal liquid levels and timely detect potential fault risks, thereby ensuring the safe navigation of ships and the normal operation of equipment.

[0003] However, in practical applications, how to efficiently and accurately fuse data from multiple sensors and effectively solve the incompleteness and inconsistency of data has become one of the research difficulties in the ship and ocean fields. In sensor networks, factors such as data quality differences, sensor failures, communication losses, and external interferences often lead to data loss, increased errors, or the appearance of abnormal values, thereby affecting the accuracy of liquid level prediction and the robustness of the system. Therefore, improving the accuracy and reliability of data fusion technology has become the key to promoting the intelligent development of intelligent ships and ocean engineering equipment. Currently, common multi-sensor data fusion methods include weighted average method, Kalman filter, particle filter, fuzzy logic control, etc. These methods can reduce the influence of single-sensor measurement errors on the liquid level prediction results to a certain extent by fusing data from multiple sensors. However, the existing fusion technologies still have several limitations: First, most methods fail to effectively address the problems of data loss and abnormal values, resulting in a possible significant decline in model performance in the case of incomplete or abnormal data; Second, many traditional fusion methods assume that the performance and data quality of sensors are stable, lacking a dynamic adaptation mechanism for changes in sensor states (such as failures, performance degradation, etc.), and it is difficult to maintain a high fusion accuracy in actual complex environments.

[0004] The Chinese patent application with the publication number CN119244799A discloses an intelligent regulation system for an adaptive multi - environment marine electric control valve actuator. Through closed - loop control and real - time feedback optimization, it ensures the precise operation of the electric control valve actuator and reduces the regulation error. However, this method has high requirements for hardware. At the same time, in the closed - loop control system, time delay and feedback delay may cause the system to be unstable or the performance to decline. Therefore, there is an urgent need to design a new multi - sensor data fusion method that can effectively handle data incompleteness, abnormality, and the dynamic changes of sensor states, thereby improving the accuracy of liquid level prediction and the robustness of the system, and contributing to the intelligent development of intelligent ships and ocean engineering equipment. Summary of the Invention

[0005] The purpose of the present invention is to overcome the defects of the above - mentioned existing technologies and provide a ship liquid level prediction method based on multi - sensor data fusion and dynamic correction, which comprehensively considers the measurement accuracy and reliability of each sensor and dynamically corrects the data fusion result, thereby further improving the accuracy of the liquid level prediction result.

[0006] The purpose of the present invention can be achieved through the following technical solutions:

[0007] The present invention provides a ship liquid level prediction method based on multi - sensor data fusion and dynamic correction, including the following steps:

[0008] S1. Obtain the data collected by multiple sensors and perform quality scoring on them respectively. The quality scoring is calculated based on the data loss rate and data abnormality degree;

[0009] S2. According to the quality scoring results, calculate the weighting factors of each sensor and perform weighted update on the data collected by each sensor;

[0010] S3. Use the evidence theory method to fuse the weighted - updated sensor data to obtain a preliminary ship liquid level prediction result;

[0011] S4. Adopt a differential re - weighting strategy to correct the preliminary fusion result to obtain the final ship liquid level prediction result.

[0012] Furthermore, in step S1, after performing correlation processing and basic probability assignment determination on the obtained multiple - sensor data, perform quality scoring.

[0013] Furthermore, in step S1, the specific calculation formula of the quality scoring is as follows:

[0014] Q i (r j )=(1 - L i (t j ))·(1 - E i (tj ))

[0015] Among them, Q i (t j ) is the quality score of sensor i at time t j , L i (t j ) is the data loss rate of sensor i at time t j , and E i (t j ) is the data abnormality degree of sensor i at time t j .

[0016] Furthermore, the calculation formula of L i (t j ) is specifically as follows:

[0017]

[0018] Among them, N lost is the amount of lost data, and N total is the total amount of data.

[0019] Furthermore, the calculation formula of E i (t j ) is specifically as follows:

[0020]

[0021] Among them, x i (t j ) is the data of sensor i at time t j , μ i and σ i are respectively the mean and standard deviation of the data of sensor i within the time period m.

[0022] Furthermore, the specific process of step S2 is as follows:

[0023] S201. Calculate the weighting factor of each sensor according to the quality score:

[0024]

[0025] Among them, N is the number of sensors, w i (t j ) is the weighting factor of sensor i at time t j , the sum of the weighting factors of all sensors is 1, and Q i (t j ) is the quality score of sensor i at time t j ;

[0026] S202. Update the data collected by each sensor based on the weighting factor:

[0027] m j (t j )=w j (t j )·x j (t j )

[0028] Among them, x j (t j ) and m j (t j ) are sensor i at time t j Data before and after the update.

[0029] Furthermore, the specific calculation formula of step S3 is as follows:

[0030]

[0031] Among them, m i t i The basic probability distribution function of the time-independent evidence source, A i is m i The focal element, K is the conflict degree.

[0032] Furthermore, the calculation formula of the conflict degree K is as follows:

[0033]

[0034] Furthermore, the specific process of step S4 is as follows:

[0035] S401, calculate at t j The probability m of the liquid level prediction data fused at each moment fused (t j ) and the historical liquid level true value m of each sensor true (t j )’s internal consistency deviation ε j (t j ):

[0036] ε j (t j )=|m fused (t j )-m true (t j )|

[0037] Using ε j (t j ) Correct the weighting factors of each sensor calculated in step S2:

[0038] w′ j (t i )=wj (t i )·(1 - α·|ε j (t i )|)

[0039] where w j (t i ) and w' j (t i ) are the weighted factors before and after correction respectively, and α is the influencing factor;

[0040] S402. Repeat steps S2 and S3 based on the corrected weighted factor to obtain the final predicted result m corrected (t j ) of the ship liquid level:

[0041] m corrected (t j ) = m fused (t j ) + λ·ε j (t j )

[0042] where λ is the correction factor, which is calculated based on the trend of historical data or dynamic reference values and is used to control the correction amplitude.

[0043] Furthermore, the specific calculation formula of the correction factor λ is as follows:

[0044]

[0045] where k is the sensitivity parameter, which is used to control the response degree of λ to the deviation size.

[0046] Compared with the prior art, the present invention has the following beneficial effects:

[0047] The present invention proposes a ship liquid level prediction method based on multi - sensor data fusion and dynamic correction. On the one hand, the quality score is calculated based on the data loss rate and data abnormality degree, and then the weighted factor of each sensor is calculated based on the quality score result, and the data collected by each sensor is weighted and updated. By reasonably assigning a higher weight to reliable sensors and actively reducing the influence of data with poor quality, the interference of dirty data and outliers on the liquid level prediction result is effectively reduced. On the other hand, the differential re - weighting strategy is used to correct the preliminary fusion result, so as to further optimize the fusion result, reduce potential errors, and finally obtain a more accurate liquid level prediction output. This dynamic correction mechanism effectively improves the accuracy of prediction, enhances the adaptive ability and robustness of the method of the present invention in complex environments, and can further improve the accuracy of the liquid level prediction result. Description of the Drawings

[0048] Figure 1It is a flowchart of the method of the present invention;

[0049] Figure 2 It is an overall structural framework diagram of the ship liquid level prediction method in the present invention;

[0050] Figure 3 It is a comparison diagram of the correction effect of the differential reweighting strategy. Specific embodiments

[0051] The present invention will be described in detail below in conjunction with the accompanying drawings and specific embodiments. This embodiment is implemented on the premise of the technical solution of the present invention, and gives the detailed implementation manner and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0052] Embodiment:

[0053] This embodiment provides a ship liquid level prediction method based on multi-sensor data fusion and dynamic correction. As shown in Figure 1 and Figure 2 , it includes the following steps:

[0054] S1. Obtain the data collected by multiple sensors and perform quality scoring respectively. The quality scoring is calculated based on the data loss rate and data abnormality degree.

[0055] First, obtain the liquid level data from each sensor, then organize it into an evidence set, and finally align it to determine the basic probability assignment. The specific process is as follows:

[0056] S11. Collect data from N multiple types of sensors at different time points, and these data reflect the changes under different environmental conditions and measurement methods.

[0057] S12. Perform correlation processing on the original data collected by N sensors.

[0058] First, ensure that the data of each sensor can be effectively compared within the same time window. Subsequently, organize the various types of data collected by different sensors at the same moment into an evidence set, where x i (t j ) represents the liquid level data of the i-th sensor at t j moment.

[0059] S13. Determine the basic probability assignment of the sensor data, so as to provide a quantitative trust degree description m(A) for each sensor, which is used to represent its support degree for different hypotheses (target states, classifications, etc.), facilitating subsequent fusion, where A is the focal element of the basic probability assignment function of the evidence source.

[0060] S14. Calculate the quality score.

[0061] By analyzing the data loss rate and the results of outlier detection, evaluate the quality of each sensor and generate corresponding quality scores. The specific formula for calculating the quality score is as follows:

[0062] Q i (t j ) = (1 - L i (t j )) · (1 - E i (t j ))

[0063]

[0064] where Q i (t j ) is the quality score of sensor i at time t j , L i (t j ) is the data loss rate of sensor i at time t j , E i (t j ) is the data abnormality degree of sensor i at time t j ; N lost is the amount of lost data, N total is the total amount of data; x i (t j ) is the data of sensor i at time t j , μ i and σ i are the mean and standard deviation of the data of sensor i within the time period m, respectively.

[0065] S2. According to the quality score results, calculate the weighting factors of each sensor and perform weighted update on the data collected by each sensor.

[0066] Combined with the weighted voting strategy, perform weighted processing on the data output by the sensors, so as to achieve the goals of noise filtering, normalization processing, and reducing the impact of dirty data, etc., so as to ensure the consistency and reliability of the data. The specific process is as follows:

[0067] S21. Calculate the weighting factors of each sensor according to the quality score:

[0068]

[0069] where w i (t j ) is the weighting factor of sensor i at time t j , the sum of the weighting factors of all sensors is 1, Q i (t j ) is the quality score of sensor i at time t jQuality score on;

[0070] S22. Update the data collected by each sensor based on the weighting factor:

[0071] m j (t j ) = w j (t j )·x j (t j )

[0072] where x j (t j ) and m j (t j ) are the data of sensor i before and after update at time t j respectively.

[0073] S3. Use the evidence theory method to fuse the weighted and updated sensor data to obtain a preliminary ship liquid level prediction result.

[0074] The Dempster rule in the evidence theory is used to achieve effective fusion of multi-source data. Its main formula is:

[0075]

[0076] where m i is the basic probability assignment function of the independent evidence source at time t i , A i is the focal element of m i , and K is the conflict degree, defined as follows:

[0077]

[0078] S4. Adopt a differential re-weighting strategy to correct the preliminary fusion result to obtain the final ship liquid level prediction result. The specific process is as follows:

[0079] S41. Calculate the internal consistency deviation ε j between the probability m fused (t j ) of the liquid level prediction data fused at time t true and the historical true value m j (t j ) of each sensor: j :

[0080] ε j (t j ) = |m fused (t j ) - m true (t j )|

[0081] Then, use ε j (t j ) to correct the weighting factors of each sensor calculated in step S2:

[0082] w′ j (t i ) = w j (t i ) · (1 - α · |ε j (t i )|)

[0083] where, w j (t i ) and w′ j (t i ) are the weighting factors before and after correction respectively, and α is the influencing factor;

[0084] S402. Repeat steps S2 and S3 based on the corrected weighting factors to obtain the final ship liquid level prediction result m corrected (t j ) :

[0085] m corrected (t j ) = m fused (t j ) + λ · ε j (t j )

[0086] where, λ is the correction factor, which is calculated based on the trend of historical data or dynamic reference values and is used to control the correction amplitude. Its specific calculation formula is as follows:

[0087]

[0088] where, k is the sensitivity parameter, which is used to control the response degree of λ to the deviation size. Usually, its value range is between [0.1, 1], and it can be adjusted according to actual needs. <>

[0089] As Figure 3 shown, in the case of no dynamic correction of the liquid level, the fused liquid level data shows a large fluctuation range, resulting in a significant deviation in the final prediction result; however, after dynamic correction, the fluctuation range of the liquid level data is significantly reduced, and the prediction accuracy is significantly improved, making the final result more accurate and reliable.

[0090] The above method predicts the ship liquid level aiming at the feature heterogeneity and distribution difference of the intelligent ship perception data in complex dynamic ocean scenarios. Compared with the existing technology, it has the following advantages:

[0091] (1) A ship liquid level prediction method based on multi-sensor data fusion and dynamic correction is proposed. The output weights of sensors are dynamically adjusted through a weighted voting strategy. Combining data loss rate and outlier filtering significantly reduces the interference of dirty data and improves the liquid level prediction accuracy and system robustness.

[0092] (2) A differential re-weighting method is designed. Based on the deviation between the fused liquid level data and the historical liquid level data of each sensor, the sensor weights are dynamically corrected, and the historical performance of the sensors is quantitatively analyzed to optimize the current prediction output, further improving the accuracy and reliability of the liquid level prediction.

[0093] (3) A dynamic correction mechanism is constructed. Combining the historical data of sensors and real-time deviation information enhances the adaptive ability of the prediction model in complex environments, ensuring that the ship liquid level prediction system remains efficient and stable under changing conditions.

[0094] If the above method is implemented in the form of software functional units and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.

[0095] The above description of the embodiments is for the convenience of those of ordinary skill in the art to understand and use the invention. Persons familiar with the art can obviously make various modifications to these embodiments easily and apply the general principles described herein to other embodiments without creative labor. Therefore, the present invention is not limited to the above embodiments, and the improvements and modifications made by those skilled in the art without departing from the scope of the present invention should be within the protection scope of the present invention.

Claims

1. A ship liquid level prediction method based on multi-sensor data fusion and dynamic correction, characterized in that, It includes the following steps: S1. Obtain the data collected by multiple sensors and perform quality scoring on them respectively. The quality scoring is calculated based on the data loss rate and data abnormality degree; S2. According to the quality scoring results, calculate the weighting factors of each sensor and perform weighted update on the data collected by each sensor; S3. Use the evidence theory method to fuse the weighted updated sensor data to obtain a preliminary ship liquid level prediction result; S4. Adopt a differential reweighting strategy to correct the preliminary fusion result to obtain the final ship liquid level prediction result.

2. A method for predicting the liquid level of a ship based on multi-sensor data fusion and dynamic correction according to claim 1, characterized in that, In step S1, after performing correlation processing and basic probability assignment determination on the obtained multiple sensor data, quality scoring is carried out.

3. A method for predicting the liquid level of a ship based on multi-sensor data fusion and dynamic correction according to claim 1, characterized in that, In step S1, the specific calculation formula of the quality scoring is as follows: Q i (r j )=(1-L i (t j ))·(1-E i (t j )) Among them, Q i (t j ) is the quality score of sensor i at time t j , L i (t j ) is the data loss rate of sensor i at time t j , and E i (t j ) is the data abnormality degree of sensor i at time t j .

4. A ship liquid level prediction method based on multi-sensor data fusion and dynamic correction according to claim 3, characterized in that, L i (t j ) is calculated as follows: Among them, N lost is the amount of lost data, and N total is the total amount of data.

5. The ship liquid level prediction method based on multi-sensor data fusion and dynamic correction according to claim 3, wherein E i (t j ) is calculated as follows: where x i (t j ) is the data of sensor i at time t j , μ i and σ i are the mean and standard deviation of the data of sensor i within the time period m, respectively.

6. A method for predicting the liquid level of a ship based on multi-sensor data fusion and dynamic correction according to claim 1, characterized in that, The specific process of step S2 is as follows: S201. Calculate the weighting factors of each sensor according to the quality scoring; where N is the number of sensors, and w i (t j ) is the weighting factor of sensor i at time t j , and the sum of the weighting factors of all sensors is 1. Q i (t j ) is the quality score of sensor i at time t j . S202. Update the data collected by each sensor based on the weighting factors; m j (t j ) = w j (t j )·x j (t j ) where x j (t j ) and m j (t j ) are the data of sensor i before and after the update at time t j respectively.

7. A ship liquid level prediction method based on multi-sensor data fusion and dynamic correction according to claim 1, characterized in that, The specific calculation formula of step S3 is as follows: Among them, m i is the basic probability assignment function of the evidence sources independent at time t i , A i is the focal element of m i , and K is the conflict degree.

8. A method for predicting the liquid level of a ship based on multi-sensor data fusion and dynamic correction according to claim 7, characterized in that, The specific calculation formula of the conflict degree K is as follows:

9. A ship liquid level prediction method based on multi-sensor data fusion and dynamic correction according to claim 1, characterized in that, The specific process of step S4 is as follows: S401. Calculate the internal consistency deviation ε j at time t fused for the probability m j of the predicted liquid level data fused at time t true and the true historical liquid level m j of each sensor at time t j : j ​ ε j (t j )=|m fused (t j )-m true (t j )| Using ε j (t j ) to correct the weighting factors of each sensor calculated in step S2: w′ j (t i ) = w j (t i )·(1 - α·|ε j (t i )|) where, w j (t i ) and w′ j (t i ) are the weighted factors before and after correction respectively, and α is the influencing factor; S402. Repeat steps S2 and S3 based on the corrected weighting factor to obtain the final predicted result m of the ship liquid level corrected (t j ): m corrected (t j ) = m fused (t j ) + λ·ε j (t j ) Among them, λ is a correction factor, which is calculated according to the trend of historical data or dynamic reference value and is used to control the correction amplitude.

10. A method for predicting the liquid level of a ship based on multi-sensor data fusion and dynamic correction according to claim 9, characterized in that, The specific calculation formula of the correction factor λ is as follows: Among them, k is a sensitivity parameter, which is used to control the response degree of λ to the deviation size.

Images