CN-122020310-A - Ship electric propulsion efficiency evaluation method and system based on multi-mode data fusion

CN122020310ACN 122020310 ACN122020310 ACN 122020310ACN-122020310-A

Abstract

The invention discloses a ship electric propulsion efficiency evaluation method and a system based on multi-mode data fusion, wherein the method comprises the steps of acquiring a cavitation erosion image of a blade of a propeller, combining ship state parameters and electric propulsion parameters, and constructing a multi-mode propulsion state data set through space-time alignment; the method comprises the steps of extracting structural dynamic characteristics of a propeller and electric propulsion state characteristics, generating a real-time propulsion state characteristic vector in a fusion mode, obtaining a real-time propulsion efficiency estimated value and a corresponding propulsion efficiency risk level through a pre-built efficiency evaluation model based on the real-time propulsion state characteristic vector and a multi-mode propulsion state data set, carrying out risk correlation analysis by combining an efficiency risk mode library, determining an efficiency risk mode and correlation data, further obtaining data adjustment quantity through efficiency trend correlation analysis, and finally generating a ship evaluation result according to the data adjustment quantity and the efficiency risk mode to improve the integrity of ship evaluation.

Inventors

Zeng Peisheng
WEI JIALONG
XU YUANYUAN
LI YUNLU
An Liantong
YANG WENJIAO

Assignees

广东海洋大学

Dates

Publication Date: 20260512
Application Date: 20260202

Claims (10)

1. The ship electric propulsion efficiency evaluation method based on multi-mode data fusion is characterized by comprising the following steps of: Acquiring a blade image of a propeller in a ship, and identifying a cavitation region and a corrosion region in the blade image to obtain a blade cavitation corrosion image corresponding to the propeller; Acquiring ship state parameters and electric propulsion parameters of the ship, and performing space-time alignment on the cavitation erosion images of the blades, the ship state parameters and the electric propulsion parameters to obtain a multi-mode propulsion state data set of the ship; acquiring the structural dynamic characteristics of the propeller and the electric propulsion state characteristics of the ship according to the multi-mode propulsion state data set, and carrying out fusion processing on the structural dynamic characteristics and the electric propulsion state characteristics to obtain a real-time propulsion state characteristic vector of the ship; acquiring a real-time propulsion efficiency estimated value of the ship and a propulsion efficiency risk level corresponding to the real-time propulsion efficiency estimated value through a pre-constructed efficiency evaluation model according to the real-time propulsion state feature vector and the multi-mode propulsion state data set; Performing risk correlation analysis according to the real-time propulsion efficiency estimated value, the propulsion efficiency risk level and a pre-constructed efficiency risk mode library, obtaining an efficiency risk mode of the ship and obtaining efficiency risk correlation data from the multi-mode propulsion state dataset according to the efficiency risk mode; And performing performance trend correlation analysis on the performance risk correlation data and the real-time propulsion efficiency estimated value to obtain a data adjustment amount of the performance risk correlation data, so as to generate an evaluation result of the ship according to the data adjustment amount and the performance risk mode.
2. The method for evaluating the electric propulsion efficiency of the ship based on the multi-modal data fusion according to claim 1, wherein the steps of obtaining a blade image of a propeller in the ship, identifying a cavitation region and a corrosion region in the blade image, and obtaining a blade cavitation corrosion image corresponding to the propeller comprise: Acquiring an original image of the propeller, and inputting the original image into a pre-constructed instance segmentation model to acquire a plurality of blade images output by the instance segmentation model; for any one of the blade images, acquiring brightness values corresponding to each pixel in the blade image, so as to divide the blade image into a high brightness area and a low brightness area according to the brightness values and a preset brightness threshold; carrying out connected domain analysis and morphological analysis on the high-brightness regions to obtain morphological characteristics corresponding to each high-brightness region; Performing texture analysis and contour analysis on the low-brightness regions to obtain texture features and contour features corresponding to each low-brightness region respectively; Dividing the blade image according to the morphological characteristics and the pixel positions of the high-brightness area to obtain a cavitation image corresponding to the blade image; dividing the blade image according to the texture features, the contour features and the pixel positions of the low-brightness region to obtain a corrosion image corresponding to the blade image; And generating a blade cavitation erosion image corresponding to the propeller according to the cavitation image and the erosion image corresponding to each blade image.
3. The method for estimating electric propulsion efficiency of a vessel based on multi-modal data fusion according to claim 2, wherein the obtaining the vessel state parameter and the electric propulsion parameter of the vessel and performing space-time alignment on the blade cavitation erosion image, the vessel state parameter and the electric propulsion parameter to obtain the multi-modal propulsion state dataset of the vessel comprises: acquiring a multi-sensor time synchronization protocol of the ship so as to determine a time stamp of a unified time reference according to the multi-sensor time synchronization protocol; Acquiring cavitation image coordinates of the cavitation image and corrosion image coordinates of the corrosion image in the blade cavitation corrosion image, and performing coordinate conversion on the cavitation image coordinates and the corrosion image coordinates according to a pre-acquired coordinate conversion relation and a ship coordinate system to obtain a blade cavitation corrosion image after coordinate conversion; and taking the timestamp as an index, and taking the ship state parameter, the electric propulsion parameter and the blade cavitation erosion image after coordinate conversion at the same moment as multi-mode propulsion state data of the ship at each moment.
4. The method for estimating electric propulsion efficiency of a ship based on multi-modal data fusion according to claim 3, wherein the electric propulsion parameters include an output torque and an output rotation speed of each motor; The step of obtaining the structural dynamic characteristics of the propeller and the electric propulsion state characteristics of the ship according to the multi-mode propulsion state data set, and carrying out fusion processing on the structural dynamic characteristics and the electric propulsion state characteristics to obtain real-time propulsion state characteristic vectors of the ship, comprises the following steps: acquiring an operation mode of an electric propulsion system in the ship, and setting the window length according to the operation mode; acquiring a plurality of multi-mode propulsion state data in each time window according to the window length to obtain a plurality of time sequence data sets; For any time sequence data set, extracting cavitation area and corrosion area at each moment from a plurality of blade cavitation corrosion images of the time sequence data set to obtain cavitation area sequence and corrosion area sequence; extracting the output torque and the output rotating speed at each moment from a plurality of electric propulsion parameters in the time sequence data set to obtain a torque sequence and a rotating speed sequence; Acquiring structural dynamic characteristics corresponding to the time sequence data set according to the cavitation area sequence and the corrosion area sequence, and acquiring running state characteristics corresponding to the time sequence data set according to the torque sequence and the rotating speed sequence; fusing the structural dynamic characteristics and the running state characteristics through a cross attention mechanism to obtain real-time propulsion state characteristics of the ship in each time window; And obtaining the real-time propulsion state characteristic vector of the ship according to the real-time propulsion state characteristics and the window length corresponding to the multi-mode propulsion state data set.
5. The method for estimating electric propulsion efficiency of a ship based on multi-modal data fusion according to claim 4, wherein the steps of obtaining the structural dynamic characteristics corresponding to the time series data set according to the cavitation area sequence and the corrosion area sequence and obtaining the running state characteristics corresponding to the time series data set according to the torque sequence and the rotation speed sequence include: acquiring an area mean value, an area variance and an area change trend slope in each time window according to the cavitation area sequence and the corrosion area sequence, wherein the area mean value comprises a cavitation area mean value and a corrosion area mean value, the area variance comprises a cavitation area variance and a corrosion area variance, and the area change trend slope comprises a corrosion area change trend slope and a cavitation area change trend slope; Taking the area mean value, the area variance and the area change trend slope as the structural dynamic characteristics of the ship in each time window; Performing frequency domain conversion on the torque sequence to obtain a torsion spectrum, and calculating the energy duty ratio of the torsion spectrum in a pre-acquired frequency band; Performing linear fitting on the rotating speed sequence to obtain a rotating speed change rate; and taking the energy duty ratio and the rotating speed change rate as running state characteristics of the ship in each time window.
6. The method for estimating electric propulsion efficiency of a ship based on multi-modal data fusion according to claim 3, wherein the obtaining, according to the real-time propulsion state feature vector and the multi-modal propulsion state data set, the real-time propulsion efficiency estimated value of the ship and the propulsion efficiency risk level corresponding to the real-time propulsion efficiency estimated value through a pre-constructed efficiency estimation model includes: Acquiring a ship state parameter sequence and an electric propulsion parameter sequence from the multi-mode propulsion state data set, and inputting the ship state parameter sequence and the electric propulsion parameter sequence into the efficiency evaluation model so as to extract a time sequence dependency feature vector through a long-period memory network preset in the efficiency evaluation model; selecting a key feature vector from the real-time propulsion state feature vector through a cross attention mechanism by taking the real-time propulsion state feature vector as a query vector and taking the time sequence dependency feature vector as a key vector and a value vector; And fusing the key feature vector and the time sequence dependent feature vector to obtain a comprehensive feature vector, and inputting the comprehensive feature vector into the efficiency evaluation model to output a real-time propulsion efficiency estimated value and a propulsion efficiency risk level of the ship through the efficiency evaluation model.
7. The method of claim 6, wherein the fusing the key feature vector and the time-dependent feature vector to obtain a comprehensive feature vector, and inputting the comprehensive feature vector into the performance evaluation model to output the real-time propulsion efficiency estimation value and the propulsion performance risk level of the ship through the performance evaluation model comprises: the comprehensive feature vectors are subjected to linear weighted summation through a weight matrix and a bias vector which are preset in the efficiency evaluation model, and transformation is carried out through a preset nonlinear activation function, so that high-dimensional nonlinear features are obtained; Mapping the high-dimensional nonlinear characteristics into one-dimensional scalar by the efficiency evaluation model, and applying a constraint function to the one-dimensional scalar to obtain a real-time propulsion efficiency estimated value of the ship; Classifying the comprehensive feature vectors through a classification function preset in the efficacy evaluation model to obtain initial probability distribution of each preset risk level; And carrying out logic judgment on the real-time propulsion efficiency estimated value according to a preset efficiency risk judgment rule base so as to correct the initial probability distribution according to the logic judgment result, thereby obtaining the propulsion efficiency risk level of the ship.
8. The method for assessing the electrical propulsion performance of a vessel based on multimodal data fusion according to any one of claims 6-7, wherein performing risk correlation analysis according to the real-time propulsion efficiency assessment value, the propulsion performance risk level and a pre-constructed performance risk pattern library, obtaining a performance risk pattern of the vessel and obtaining performance risk correlation data from the multimodal propulsion state dataset according to the performance risk pattern comprises: Acquiring a performance risk mode corresponding to the ship from a pre-constructed performance risk mode library according to the propulsion performance risk level of the ship, wherein each historical performance risk mode in the performance risk mode library is associated with a plurality of historical propulsion state feature vectors and a plurality of historical propulsion efficiency estimated values; acquiring an efficiency difference value of each historical propulsion efficiency estimated value and the real-time propulsion efficiency estimated value of the ship respectively; for any one of the historical propulsion state feature vectors, acquiring a feature difference vector of the historical propulsion state feature vector and the real-time propulsion state feature vector of the ship based on the same feature dimension; Determining a contribution rate of each of the feature dimensions to the efficiency difference value based on the feature difference vector and the efficiency difference value, to determine a number of key feature dimensions according to the contribution rate; acquiring data types corresponding to each key feature dimension according to the pre-acquired data mapping relation; And selecting a real-time performance risk data sequence from the multi-mode propulsion state data set of the ship and a historical performance risk data sequence from the performance risk mode library according to the data types to obtain performance risk associated data of the ship.
9. The method for estimating electric propulsion efficiency of a vessel based on multi-modal data fusion according to claim 8, wherein the performing efficiency trend correlation analysis on the efficiency risk correlation data and the real-time propulsion efficiency estimation value to obtain a data adjustment amount of the efficiency risk correlation data, so as to generate an estimation result of the vessel according to the data adjustment amount and the efficiency risk mode comprises: acquiring a historical propulsion efficiency estimated value sequence from a plurality of historical propulsion efficiency estimated values according to a preset time window length, and performing linear regression analysis on the historical propulsion efficiency estimated value sequence and the real-time propulsion efficiency estimated value to obtain a performance transformation trend; Acquiring a correlation model of the efficacy risk correlation data and the efficacy transformation trend through a preset multiple linear regression method; carrying out inverse solution on the correlation model based on the efficacy transformation trend to obtain data adjustment quantity corresponding to each class of efficacy risk correlation data; And matching the basic strategy type corresponding to the performance risk mode, generating a performance optimization strategy and an evaluation result corresponding to the ship according to the basic strategy type and the data adjustment amount, and executing the performance optimization strategy on the ship.
10. The ship electric propulsion efficiency evaluation system based on multi-mode data fusion is characterized by comprising a blade identification module, a multi-dimensional data alignment module, a multi-dimensional feature fusion module, an efficiency risk evaluation module, an efficiency risk analysis module and an efficiency evaluation module; The blade identification module is used for acquiring a blade image of a propeller in a ship, identifying a cavitation region and a corrosion region in the blade image, and obtaining a blade cavitation corrosion image corresponding to the propeller; The multi-dimensional data alignment module is used for acquiring ship state parameters and electric propulsion parameters of the ship, and carrying out space-time alignment on the cavitation erosion images of the blades, the ship state parameters and the electric propulsion parameters to obtain a multi-modal propulsion state data set of the ship; The multi-dimensional feature fusion module is used for acquiring the structural dynamic features of the propeller and the electric propulsion state features of the ship according to the multi-mode propulsion state data set, and carrying out fusion processing on the structural dynamic features and the electric propulsion state features to obtain real-time propulsion state feature vectors of the ship; The efficiency risk estimation module is used for acquiring a real-time propulsion efficiency estimation value of the ship and a propulsion efficiency risk grade corresponding to the real-time propulsion efficiency estimation value through a pre-constructed efficiency estimation model according to the real-time propulsion state feature vector and the multi-mode propulsion state data set; The performance risk analysis module is used for performing risk correlation analysis according to the real-time propulsion efficiency estimated value, the propulsion performance risk level and a pre-constructed performance risk mode library, obtaining a performance risk mode of the ship and obtaining performance risk correlation data from the multi-mode propulsion state data set according to the performance risk mode; The efficiency evaluation module is used for performing efficiency trend correlation analysis on the efficiency risk correlation data and the real-time propulsion efficiency estimated value to obtain a data adjustment amount of the efficiency risk correlation data, so as to generate an evaluation result of the ship according to the data adjustment amount and the efficiency risk mode.

Description

Ship electric propulsion efficiency evaluation method and system based on multi-mode data fusion Technical Field The invention relates to the technical field of ship control, in particular to a ship electric propulsion efficiency evaluation method and system based on multi-mode data fusion. Background In marine electric propulsion systems, a main engine or generator generates electrical energy, which is then fed to an electric motor, which drives a propeller or other propulsion device to navigate the marine vessel. The electric propulsion system of the ship has become the main stream of the power device of the ship due to the characteristics of high efficiency, environmental protection and energy conservation. However, due to the complexity of the operating environment of the ship, the propulsion efficiency of the electric propulsion system of the ship often changes, which affects not only the normal operation of the ship but also the life safety of the crew. Therefore, the method and the device can timely and accurately identify the propulsion efficiency of the ship electric propulsion system and improve the propulsion efficiency of the ship electric propulsion system to generate the evaluation result of the efficiency, and are important to ensure the normal operation of the ship and improve the work efficiency of the crew. In the prior art, when the propulsion efficiency of the ship electric propulsion system is estimated, the propulsion efficiency is estimated by collecting the ratio of the electric energy generated in the ship electric propulsion system to the electric energy transmitted to the propeller, but the estimation method does not consider the influence of the changes of the ship body and the propeller on the propulsion efficiency, so that the obtained estimation result is disjointed from the real running state of the ship, the propulsion efficiency cannot be accurately estimated, and the technology can only identify the change trend of the efficiency, but cannot accurately position the reason causing the change of the propulsion efficiency, and further cannot comprehensively estimate the propulsion efficiency of the electric propulsion system in the ship. Disclosure of Invention In order to solve the technical problems, the invention discloses a ship electric propulsion efficiency evaluation method and a system based on multi-mode data fusion, which are used for improving the comprehensiveness of ship electric propulsion efficiency evaluation. In order to achieve the above object, the present invention discloses a ship electric propulsion efficiency evaluation method based on multi-modal data fusion, comprising: Acquiring a blade image of a propeller in a ship, and identifying a cavitation region and a corrosion region in the blade image to obtain a blade cavitation corrosion image corresponding to the propeller; Acquiring ship state parameters and electric propulsion parameters of the ship, and performing space-time alignment on the cavitation erosion images of the blades, the ship state parameters and the electric propulsion parameters to obtain a multi-mode propulsion state data set of the ship; acquiring structural dynamic characteristics of the propeller and electric propulsion state characteristics of the ship according to the multi-mode propulsion state data set so as to fuse the structural dynamic characteristics and the electric propulsion state characteristics and obtain a real-time propulsion state characteristic vector of the ship; acquiring a real-time propulsion efficiency estimated value of the ship and a propulsion efficiency risk level corresponding to the real-time propulsion efficiency estimated value through a pre-constructed efficiency evaluation model according to the real-time propulsion state feature vector and the multi-mode propulsion state data set; Performing risk association analysis according to the real-time propulsion efficiency estimated value, the propulsion efficiency risk level and a pre-constructed efficiency risk mode library to obtain an efficiency risk mode of the ship and select efficiency risk association data from the multi-mode propulsion state data set; And performing performance trend correlation analysis on the performance risk correlation data and the real-time propulsion efficiency estimated value to obtain a data adjustment amount of the performance risk correlation data, so as to generate an evaluation result of the ship according to the data adjustment amount and the performance risk mode. The invention discloses a multi-modal data fusion-based ship electric propulsion efficiency evaluation method, which comprises the steps of firstly carrying out space-time alignment on an acquired cavitation erosion image of a blade of a propeller in a ship, a ship state parameter of the ship and an electric propulsion parameter to acquire a multi-modal propulsion state data set representing the overall operation of the ship, and then extracting multi-modal