A Method and System for Monitoring the Condition of Ship Electric Propulsion Systems Based on Infrared Thermal Imaging

Abstract

This invention discloses a method and system for monitoring the condition of a ship's electric propulsion system based on infrared thermal imaging, belonging to the field of ship monitoring technology. This invention determines the electrical temperature rise characteristic points by using the connection edges from the electrical offset center to each of the high-temperature anomaly regions and the electrothermal variation coefficient of each connection edge. These characteristic points are then used to construct the structural element shape of the thermal distribution image. Furthermore, scaling the structural element shape constructs more accurate structural elements for each high-temperature anomaly region in the ship's electric propulsion system, achieving accurate morphological processing of the thermal distribution image and obtaining accurate thermal features. By fusing accurate thermal features and electrical parameters, comprehensive and accurate electrothermal fusion data is generated. This electrothermal fusion data enables the prediction of the state risk patterns of the ship's electric propulsion system, achieving accurate and comprehensive state monitoring of the ship's electric propulsion system using electrical parameters, thermal features, and state risk patterns.

Description

Technical Field

[0001] This invention belongs to the field of ship monitoring technology, and in particular relates to a method and system for monitoring the status of ship electric propulsion systems based on infrared thermal imaging. Background Technology

[0002] As a core component of modern ship propulsion systems, the normal and stable operation of marine electric propulsion systems is crucial to the ship's navigation. However, due to the significantly increased complexity of modern marine electric propulsion systems, which involve numerous electrical devices such as generators, transformers, frequency converters, and motors, these devices are susceptible to environmental humidity, vibration, and load fluctuations during long-term operation, leading to problems such as localized overheating and abnormal temperatures, which can cause equipment failure or even system paralysis. Therefore, real-time status monitoring of marine electric propulsion systems has become a key aspect of ensuring safe ship operation, extending equipment lifespan, and reducing maintenance costs.

[0003] Current condition monitoring solutions for marine electric propulsion systems mostly rely on infrared thermal imaging technology. This technology captures thermal distribution images of the system. To further mitigate the impact of the ship's operating environment, such as salt spray corrosion, mechanical vibration, electromagnetic interference, temperature and humidity variations, and surface contamination, morphological processing of the thermal distribution images is typically required. However, current morphological processing methods are based on fixed structural elements, resulting in inaccurate processing of the thermal distribution images and consequently affecting the accuracy of thermal distribution image analysis and, consequently, the accuracy of marine electric propulsion system condition monitoring. Furthermore, current marine electric propulsion system condition monitoring often focuses on real-time monitoring, lacking prediction of future system conditions, leading to a one-sided approach to condition monitoring. Summary of the Invention

[0004] This invention aims to provide a method and system for monitoring the condition of a ship's electric propulsion system based on infrared thermal imaging, in order to solve the above-mentioned technical problems. By constructing structural elements for each high-temperature abnormal region in the ship's electric propulsion system, accurate morphological processing of the thermal distribution image is achieved to obtain accurate thermal features. Thus, the state risk mode of the ship's electric propulsion system can be predicted by using accurate thermal features and electrical parameters, thereby realizing comprehensive and accurate condition monitoring of the ship's electric propulsion system.

[0005] To address the aforementioned technical problems, embodiments of the present invention provide a method for monitoring the condition of a ship's electric propulsion system based on infrared thermal imaging, comprising:

[0006] The thermal distribution images of various electrical devices in the ship's electric propulsion system are acquired using infrared thermal imaging technology, and several high-temperature abnormal areas are identified in each of the thermal distribution images.

[0007] Acquire several electrical parameters for each of the electrical devices, project the electrical parameters onto the corresponding thermal distribution image, and determine the electrical offset center of each thermal distribution image;

[0008] In each of the thermal distribution images, a connection edge is constructed from the electrical offset center to each of the high temperature anomaly regions; the electrothermal variation coefficient of each connection edge is calculated based on the thermal distribution image, and then several electrothermal temperature rise feature points are identified based on the electrothermal variation coefficient;

[0009] The shape of the structural element of each thermal distribution image is determined based on the aforementioned electrical temperature rise feature points; and the shape of the structural element is scaled based on the electrical-thermal variation coefficient to determine the structural element of each high-temperature anomaly region; and the thermal characteristics of each electrical device are determined based on the structural element and the high-temperature anomaly region.

[0010] The thermal characteristics and several electrical parameters of each electrical device are fused to determine the electrothermal fusion data of each electrical device, and the state risk mode of the ship's electric propulsion system is predicted based on the electrothermal fusion data;

[0011] Based on several electrical parameters and thermal characteristics of each of the electrical devices, and the state risk mode of the ship's electric propulsion system, the state monitoring results of the ship's electric propulsion system are determined.

[0012] Understandably, compared to existing technologies, this invention determines the electrical temperature rise feature points by using the connection edges from the electrical offset center to each of the aforementioned high-temperature anomaly regions and the electrothermal variation coefficient of each connection edge. Based on these feature points, the structural element shape of the thermal distribution image is constructed. Furthermore, by scaling the structural element shape, more accurate structural elements are constructed for each high-temperature anomaly region in the ship's electric propulsion system, achieving accurate morphological processing of the thermal distribution image and obtaining accurate thermal features. Next, by fusing accurate thermal features and electrical parameters, comprehensive and accurate electrothermal fusion data is generated. This data is then used to predict the state risk patterns of the ship's electric propulsion system. By combining electrical parameters, thermal features, and state risk patterns, accurate and comprehensive state monitoring of the ship's electric propulsion system is achieved, significantly improving the accuracy and comprehensiveness of state monitoring.

[0013] Accordingly, embodiments of the present invention provide a ship electric propulsion system status monitoring system based on infrared thermal imaging, including: a high temperature anomaly area identification module, an electrical parameter projection module, an electrical temperature rise feature point identification module, a structural element construction module, an electrothermal fusion prediction module, and a status monitoring result generation module;

[0014] The high-temperature anomaly area identification module is used to acquire thermal distribution images of various electrical devices in the ship's electric propulsion system based on infrared thermal imaging technology, and to identify several high-temperature anomaly areas in each thermal distribution image.

[0015] The electrical parameter projection module is used to acquire several electrical parameters of each electrical device, project the electrical parameters onto the corresponding thermal distribution image, and determine the electrical offset center of each thermal distribution image;

[0016] The electrical temperature rise feature point identification module is used to construct the connection edge from the electrical offset center to each of the high temperature anomaly regions in each of the thermal distribution images; calculate the electrical-thermal variation coefficient of each of the connection edges based on the thermal distribution images, and then identify several electrical temperature rise feature points based on the electrical-thermal variation coefficient;

[0017] The structural element construction module is used to determine the shape of the structural element of each thermal distribution image based on the plurality of electrical temperature rise feature points; and to scale the shape of the structural element based on the electrical temperature variation coefficient to determine the structural element of each high temperature anomaly region; and to perform morphological processing on the high temperature anomaly region based on the structural element to determine the thermal characteristics of each electrical device.

[0018] The electrothermal fusion prediction module is used to fuse the thermal characteristics and several electrical parameters of each electrical device to determine the electrothermal fusion data of each electrical device, and predict the state risk mode of the ship's electric propulsion system based on the electrothermal fusion data.

[0019] The condition monitoring result generation module is used to determine the condition monitoring result of the ship's electric propulsion system based on several electrical parameters and thermal characteristics of each electrical device, as well as the condition risk mode of the ship's electric propulsion system.

[0020] Understandably, compared to existing technologies, this system determines the electrical temperature rise feature points by using the connection edges from the electrical offset center to each of the aforementioned high-temperature anomaly regions and the electrothermal variation coefficient of each connection edge. Based on these feature points, it constructs the structural element shape of the thermal distribution image. Furthermore, by scaling the structural element shape, it constructs more accurate structural elements for each high-temperature anomaly region in the ship's electric propulsion system, achieving accurate morphological processing of the thermal distribution image and obtaining accurate thermal features. Next, by fusing accurate thermal features and electrical parameters, it generates comprehensive and accurate electrothermal fusion data. This data is then used to predict the state risk patterns of the ship's electric propulsion system. By combining electrical parameters, thermal features, and state risk patterns, it achieves accurate and comprehensive state monitoring of the ship's electric propulsion system, significantly improving the accuracy and comprehensiveness of state monitoring. Attached Figure Description

[0021] Figure 1 A flowchart illustrating the steps of a method for monitoring the condition of a ship's electric propulsion system based on infrared thermal imaging, as provided in an embodiment of the present invention.

[0022] Figure 2 This is a schematic diagram of the structure of a ship electric propulsion system condition monitoring system based on infrared thermal imaging, provided as an embodiment of the present invention. Detailed Implementation

[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] Example 1

[0025] Please refer to Figure 1 , Figure 1 The flowchart of the method for monitoring the status of a ship electric propulsion system based on infrared thermal imaging provided in the embodiments of the present invention includes steps S101 to S106.

[0026] Step S101: Acquire thermal distribution images of various electrical devices in the ship's electric propulsion system based on infrared thermal imaging technology, and identify several high-temperature abnormal areas in each thermal distribution image.

[0027] Step S102: Obtain several electrical parameters for each of the electrical devices, project the electrical parameters onto the corresponding thermal distribution image, and determine the electrical offset center of each thermal distribution image.

[0028] Step S103: In each of the thermal distribution images, construct the connection edge from the electrical offset center to each of the high temperature anomaly regions; calculate the electrothermal variation coefficient of each connection edge based on the thermal distribution image, and then identify several electrothermal temperature rise feature points based on the electrothermal variation coefficient.

[0029] Step S104: Determine the shape of the structural element of each thermal distribution image based on the several electrical temperature rise feature points; and scale the shape of the structural element based on the electrical-thermal variation coefficient to determine the structural element of each high-temperature anomaly region; perform morphological processing on the high-temperature anomaly region based on the structural element to determine the thermal characteristics of each electrical device.

[0030] Step S105: The thermal characteristics and several electrical parameters of each electrical device are fused to determine the electrothermal fusion data of each electrical device, and the state risk mode of the ship's electric propulsion system is predicted based on the electrothermal fusion data.

[0031] Step S106: Based on several electrical parameters and thermal characteristics of each electrical device, and the state risk mode of the ship's electric propulsion system, determine the state monitoring results of the ship's electric propulsion system.

[0032] This embodiment determines the electrical temperature rise feature points by using the connection edges from the electrical offset center to each of the high-temperature anomaly regions and the electrothermal variation coefficient of each connection edge. These feature points are then used to construct the structural element shape of the thermal distribution image. Furthermore, scaling the structural element shape allows for the construction of more accurate structural elements for each high-temperature anomaly region in the ship's electric propulsion system, achieving accurate morphological processing of the thermal distribution image and obtaining accurate thermal features. Next, by fusing accurate thermal features and electrical parameters, comprehensive and accurate electrothermal fusion data is generated. This data is then used to predict the state risk patterns of the ship's electric propulsion system. By combining electrical parameters, thermal features, and state risk patterns, accurate and comprehensive state monitoring of the ship's electric propulsion system is achieved, significantly improving the accuracy and comprehensiveness of state monitoring.

[0033] In this embodiment, thermal distribution images of various electrical devices in a ship's electric propulsion system are acquired based on infrared thermal imaging technology, and several high-temperature anomaly regions in each thermal distribution image are identified. This includes: acquiring thermal distribution images of various electrical devices in a ship's electric propulsion system using an infrared thermal imaging camera, wherein the electrical devices include generators, transformers, frequency converters, propulsion motors, etc.; after acquiring the thermal distribution images, the thermal distribution images are preprocessed, including: using Gaussian filtering to denoise the thermal distribution images, and then inputting the denoised thermal distribution images into a U-Net network, where the U-Net network performs image segmentation on the thermal distribution images and marks the high-temperature anomaly regions in the thermal distribution images.

[0034] It's important to note that a Gaussian filter is a linear smoothing filter. Its core principle is to use a Gaussian function (normal distribution function) as the weight kernel to perform a weighted average convolution operation on the image (or signal). Its primary goal is to effectively suppress noise (especially Gaussian noise) and smooth details while preserving the main structural features of the image. The U-Net network is an encoder-decoder architecture convolutional neural network (CNN) specifically designed for image segmentation tasks. Its core innovation lies in its unique symmetrical U-shaped structure and skip connections, enabling high-precision pixel-level segmentation even with small sample training.

[0035] In the above embodiments, thermal distribution images of various electrical devices are acquired using infrared thermal imaging technology. Gaussian filtering is then used to denoise the thermal distribution images, eliminating noise and improving their accuracy. Next, a U-Net network is used to accurately segment high-temperature anomaly regions from the thermal distribution images, providing an image basis for subsequent morphological processing and thus improving the accuracy of condition monitoring of the ship's electric propulsion system.

[0036] In this embodiment, the step of acquiring several electrical parameters of each electrical device, projecting the electrical parameters onto the corresponding thermal distribution image, and determining the electrical deviation center of each thermal distribution image includes: acquiring several electrical parameters of each electrical device and the rated electrical parameter corresponding to each electrical parameter; determining the electrical deviation degree of each electrical parameter based on the ratio of each electrical parameter to the corresponding rated electrical parameter; acquiring a device image of each electrical device, and registering the device image and thermal distribution image of each electrical device based on the SIFT algorithm to determine the electrothermal transformation matrix of each electrical device; and projecting several electrical parameters of each electrical device onto the corresponding thermal distribution image based on the transformation matrix and the measurement position of each electrical parameter at the electrical device to determine the electrical projection point of each electrical parameter on the corresponding thermal distribution image.

[0037] Based on the coordinates of the electrical projection points on the thermal distribution image and the corresponding electrical deviation, the electrical projection points on each thermal distribution image are clustered to determine the electrical deviation center of each thermal distribution image.

[0038] In one optional embodiment, different types of sensors are deployed at different measurement locations within each electrical device to measure the electrical parameters of the device. For example, sensors are deployed at locations such as the stator system, rotor system, and generator output bus of the generator to measure electrical parameters including: stator three-phase current, rotor winding resistance, generator output voltage, and generator output frequency. Simultaneously, the rated electrical parameters corresponding to each electrical parameter are obtained from the device's instruction manual. The ratio of the electrical parameter to the corresponding rated electrical parameter is then used as the electrical deviation of that parameter. Finally, an image of each electrical device is captured using a visible light camera, and the images are also processed to reduce their resolution. After noise reduction, the SIFT algorithm is used to register the equipment image and the thermal distribution image, generating an electrothermal transformation matrix between the equipment image and the thermal distribution image. Then, the electrothermal transformation matrix is ​​multiplied by the measurement position (i.e., coordinate form) of each electrical parameter on the electrical equipment to obtain the coordinates of the electrical parameter on the corresponding thermal distribution image. At the same time, a data point is generated at this coordinate, which is regarded as the electrical projection point of the electrical parameter on the thermal distribution image. Then, the K-Means clustering algorithm is set to 1 cluster, and the electrical deviation is used as the weight of the coordinates of the electrical projection point on the thermal distribution image. The electrical projection points on each thermal distribution image are clustered to determine the electrical deviation center of each thermal distribution image.

[0039] It should be noted that SIFT (Scale-Invariant Feature Transform) is an image registration method that extracts local feature points in an image that are invariant to scale, rotation, and brightness, and directly calculates the transformation matrix between two images based on the matching relationship of these feature points. K-Means clustering is a centroid-based iterative unsupervised learning algorithm. Its core objective is to divide n unlabeled observation data points into k clusters, such that each data point belongs to the cluster represented by its nearest centroid (mean point), and optimizes the clustering result by minimizing the Within-Cluster Sum of Squares (WCSS).

[0040] In the above embodiments, the ratio of electrical parameters to corresponding rated electrical parameters can quantify the degree of abnormality of electrical parameters. Then, an electrothermal transformation matrix is ​​constructed through SIFT registration. This allows the electrical parameters to be accurately projected onto the thermal distribution image using the electrothermal transformation matrix and the measurement location of the electrical equipment. Next, the electrical projection points are clustered using their coordinates on the thermal distribution image and corresponding electrical deviations. This clustering process considers not only the degree of abnormality of electrical parameters but also their position on the electrical equipment and the thermal distribution image. This ensures that electrical deviations from the center accurately and comprehensively characterize the degree of abnormality of electrical parameters, enabling the association of electrical data from the electrical equipment with image data from the thermal distribution image. Furthermore, the construction of structural elements comprehensively considers both electrical data and image data from the thermal distribution image, improving the accuracy of structural elements. This, in turn, enables accurate morphological processing of the thermal distribution image, obtaining accurate thermal features and improving the accuracy of monitoring the status of the ship's electric propulsion system.

[0041] In this embodiment, in each of the thermal distribution images, a connection edge is constructed from the electrical offset center to each of the high-temperature anomaly regions; the electrothermal variation coefficient of each connection edge is calculated based on the thermal distribution image, and then several electrothermal temperature rise feature points are identified based on the electrothermal variation coefficient, including:

[0042] In each of the thermal distribution images, the centroid of each of the high-temperature anomaly regions is obtained;

[0043] Connect the electrical offset center to the centroid of each of the high-temperature anomaly regions to construct a connection edge from the electrical offset center to each of the high-temperature anomaly regions;

[0044] The temperature value of each pixel on each of the connecting edges is determined based on the thermal distribution image.

[0045] Based on the temperature value of each pixel on each of the connecting edges, calculate the standard deviation and average value of the temperature values ​​of all pixels on each of the connecting edges;

[0046] The electrothermal variation coefficient of each connection edge is determined based on the ratio of the standard deviation of the temperature value to the average temperature value of the pixels on each connection edge.

[0047] Based on the electrothermal variation coefficient of each of the connecting edges, the centroid of the high-temperature anomaly region is screened to determine several characteristic points of electrothermal temperature rise.

[0048] In an optional embodiment, the centroid of each high-temperature anomaly region is obtained in each thermal distribution image. Since there are mature mathematical calculation methods for determining the centroid of an image, this embodiment will not elaborate further. Then, the electrical offset center is connected to the centroid of each high-temperature anomaly region, forming a connection edge from the electrical offset center to each high-temperature anomaly region on the thermal distribution image. This establishes a spatial correlation between the electrical parameters, represented by the electrical offset center, and the high-temperature anomaly region on the thermal distribution image. Since the essence of the thermal distribution image is temperature data, the temperature value of each pixel on the thermal distribution image can be directly read from the image. Next, for each connection edge, the standard deviation and average temperature value of all pixels on the connection edge are calculated. Then, the standard deviation is divided by the average temperature value, and the ratio of the standard deviation to the average temperature value is used as the electrothermal variation coefficient for each connection edge. The electrothermal variation coefficient reflects the degree of temperature change of the pixels on the connection edge; a larger electrothermal variation coefficient indicates a more drastic temperature change, and a greater impact of the deviation of the surface electrical parameters on the spatial correlation of the high-temperature anomaly region. Finally, by filtering the centroid of the high-temperature anomaly region using the electrothermal variation coefficient of the connecting edges, we can obtain electrothermal and temperature rise feature points that are more related to electrical and temperature anomalies. Furthermore, to simplify the calculation, this embodiment provides a set of examples where all temperature values ​​are integers. Assuming the temperature values ​​of pixels on a connecting edge are 45, 64, 67, 79, 93, 94, and 97 respectively, then the standard deviation of the temperature values ​​is approximately 17.90 (the overall standard deviation is used here, and two decimal places are retained), and the average temperature value is 77. Therefore, the electrothermal variation coefficient is (17.8965 ÷ 77) × 100% ≈ 0.23242 × 100% ≈ 23.24%.

[0049] In the above embodiments, the spatial correlation between electrical parameters and the high-temperature anomaly region on the thermal distribution image can be established through the connecting edge between the centroid of the high-temperature anomaly region and the electrical offset center. Then, the electrothermal variation coefficient of the connecting edge is calculated using the temperature value of each pixel. This electrothermal variation coefficient can characterize the intensity of the spatial correlation between electrical parameters and the high-temperature anomaly region. Consequently, electrothermal temperature rise feature points more relevant to electrical and temperature anomalies can be selected using the electrothermal variation coefficient. This makes the structural elements constructed from these electrothermal temperature rise feature points more accurate, enabling accurate morphological processing of the thermal distribution image, obtaining accurate thermal features, and improving the accuracy of monitoring the status of ship electric propulsion systems.

[0050] In this embodiment, the process of filtering the centroid of the high-temperature anomaly region based on the electrothermal variation coefficient of each of the connecting edges to determine several electrothermal temperature rise characteristic points includes:

[0051] The average value of the electrothermal variation coefficient of the heat distribution image is determined by summing and averaging the electrothermal variation coefficients of each of the connecting edges.

[0052] The first connecting edge with the largest electrothermal variation coefficient is selected from the connecting edges, and the centroid of the high temperature anomaly region corresponding to the first connecting edge is taken as the maximum point of electrothermal temperature rise in the heat distribution image.

[0053] The second connecting edge with the smallest electrothermal variation coefficient is selected from the connecting edges, and the centroid of the high temperature anomaly region corresponding to the second connecting edge is taken as the minimum point of electrothermal temperature rise in the heat distribution image.

[0054] The absolute deviation of the electrothermal variation coefficient of each of the connecting edges is determined based on the electrothermal variation coefficient of each edge and the average value of the electrothermal variation coefficient.

[0055] The third connecting edge with the smallest absolute deviation of the electrothermal variation coefficient is selected from the connecting edges, and the centroid of the high temperature anomaly region corresponding to the third connecting edge is taken as the average point of the electrothermal temperature rise of the heat distribution image.

[0056] Based on the maximum, minimum, and average points of the electrical temperature rise, several characteristic points of electrical temperature rise are determined.

[0057] In one optional embodiment, the first connecting edge with the largest electrothermal variation coefficient is selected from the connecting edges. Since the first connecting edge has the largest electrothermal variation coefficient, it directly reflects the most severe situation in the spatial correlation influence intensity of the heat distribution image. Then, the centroid of the high-temperature anomaly region corresponding to the first connecting edge is marked as the maximum electrothermal temperature rise point. Similarly, the second connecting edge with the largest electrothermal variation coefficient is selected from the connecting edges. Since the second connecting edge has the smallest electrothermal variation coefficient, it directly reflects the least severe situation in the spatial correlation influence intensity of the heat distribution image. Then, the centroid of the high-temperature anomaly region corresponding to the second connecting edge is marked as the minimum electrothermal temperature rise point. After capturing the two electrothermal temperature rise feature points (maximum and minimum electrothermal temperature rise points) representing the most and least severe spatial correlation influence intensity, it is also necessary to consider the overall intensity of spatial correlation influence in the heat distribution image. To determine the average electrothermal coefficient of variation of the thermal distribution image, the electrothermal coefficient of variation of each connecting edge is first accumulated and averaged. Then, the electrothermal coefficient of variation of each connecting edge is subtracted from the average electrothermal coefficient of variation, and the absolute value is taken to obtain the absolute deviation of the electrothermal coefficient of variation, which can measure the difference from the average level of the spatial correlation influence in the thermal distribution image. Therefore, the third connecting edge with the smallest absolute deviation of the electrothermal coefficient of variation is selected, that is, the third connecting edge that is closest to the average level and has the most universal temperature anomaly is selected. The centroid of the high temperature anomaly region corresponding to the third connecting edge is taken as the average point of the electrothermal temperature rise of the thermal distribution image. Thus, the structural elements used in traditional morphological operations are mostly fixed rectangles and circles. Although these shapes can also achieve image processing, their fixed shapes make it impossible to achieve accurate morphological operations for different images during use. In this embodiment, by filtering the first, second, and third connecting edges of each thermal distribution image, the electrical temperature rise feature points not only consider the two extreme cases of the intensity of spatial correlation influence in the thermal distribution image, but also take into account the average level of the intensity of spatial correlation influence. This allows the electrical temperature rise feature points to accurately and comprehensively reflect the temperature anomalies of the corresponding thermal distribution image. As a result, the shape of the subsequently constructed structural elements can be adaptively adjusted according to the characteristics of each thermal distribution image, thereby improving the accuracy of morphological processing.

[0058] In the above embodiments, by selecting the first connecting edge with the largest electrothermal variation coefficient and the second connecting edge with the smallest electrothermal variation coefficient, the two most intense and least intense electrothermal temperature rise feature points in the intensity of spatial correlation influence can be captured. This allows the structural elements constructed based on the electrothermal temperature rise feature points to reflect the two extreme temperature anomalies in the thermal distribution image. Subsequently, by calculating the absolute deviation of the electrothermal variation coefficient and selecting the third connecting edge with the smallest absolute deviation, the average electrothermal temperature rise point closest to the average level in the intensity of spatial correlation influence can be captured. This allows the subsequent structural elements constructed based on the electrothermal temperature rise feature points to also take into account the temperature anomalies closest to the average level and the most common temperature anomalies, achieving comprehensive and accurate subsequent structural element construction. This enables the structural elements to comprehensively and accurately perform morphological processing on the thermal distribution image, thereby obtaining accurate thermal features and improving the accuracy of monitoring the status of the ship's electric propulsion system.

[0059] In this embodiment, the step of determining the shape of the structural element of each thermal distribution image based on the plurality of electrothermal temperature rise feature points; scaling the shape of the structural element based on the electrothermal variation coefficient to determine the structural element of each high-temperature anomaly region; and performing morphological processing on the high-temperature anomaly region based on the structural element to determine the thermal characteristics of each electrical device includes:

[0060] Using the maximum, minimum, and average points of electrical and thermal temperature rise in each thermal distribution image as vertices, construct a geometric triangle for each thermal distribution image;

[0061] The geometric triangle of each heat distribution image is used as the shape of the structural element of each heat distribution image, and the area of ​​the geometric triangle of each heat distribution image is used as the reference area of ​​the structural element of each heat distribution image.

[0062] The area of ​​each high-temperature anomaly region in each of the thermal distribution images is obtained, and the area change rate of each high-temperature anomaly region is determined based on the ratio of the area of ​​each high-temperature anomaly region to the reference area of ​​the structural element of the thermal distribution image to which it belongs.

[0063] Based on the electrothermal variation coefficient and the area change rate of the connecting edge corresponding to each of the high-temperature anomaly regions, the scaling factor of each of the high-temperature anomaly regions is determined.

[0064] The scaling factor of each of the high temperature anomaly regions is multiplied by the reference area of ​​the structural element of the corresponding thermal distribution image to scale the shape of the structural element of the corresponding thermal distribution image, thereby determining the area of ​​the structural element of each of the high temperature anomaly regions.

[0065] Based on the area of ​​the structural element of each high-temperature anomaly region and the shape of the structural element of the corresponding thermal distribution image, the structural element of each high-temperature anomaly region is constructed.

[0066] Morphological operations are performed on the high-temperature anomaly regions based on the structural elements of each region to determine the hot spot area and temperature rise gradient of each electrical device. Then, based on the hot spot area and temperature rise gradient, the thermal characteristics of each electrical device are determined.

[0067] In an optional embodiment, since the electrical temperature rise feature points not only consider the two extreme cases of the intensity of spatial correlation influence in the thermal distribution image, but also take into account the average level of the intensity of spatial correlation influence, the maximum, minimum, and average electrical temperature rise points are used as vertices, and the maximum, minimum, and average electrical temperature rise points are connected sequentially to construct the geometric triangle of the thermal distribution image. Each geometric triangle of the thermal distribution image is then used as the shape of the structural element of each thermal distribution image. Such geometric triangles can also consider the two extreme cases and the average level of the intensity of spatial correlation influence in the thermal distribution image. Because different thermal distribution images have different electrical temperature rise feature points, the geometric triangle of each thermal distribution image can also adapt to changes, improving the accuracy of the structural element shape and avoiding inaccuracies in morphological calculations caused by the fixed shape of conventional structural elements. Next, the geometric triangles of each heat distribution image are used as the structural element shape of each heat distribution image. Since the geometric triangles are constructed on the heat distribution image, the area of ​​the geometric triangle can be determined by counting the number of pixels in the geometric triangle and combining it with the area of ​​a single pixel (since this process is a conventional image calculation, it will not be elaborated on in this embodiment). Thus, the area of ​​the geometric triangles of the heat distribution image is used as the structural element reference area of ​​the heat distribution image. Similarly, the area of ​​each high-temperature anomaly region in the heat distribution image can also be obtained in this way. Then, the ratio of the area of ​​the high-temperature anomaly region to the structural element reference area of ​​the corresponding heat distribution image is used as the area change rate of the high-temperature anomaly region.

[0068] Next, the electrothermal variation coefficient of the connecting edge corresponding to the high-temperature anomaly region is multiplied by the area change rate to obtain the scaling factor of the high-temperature anomaly region. This scaling factor not only considers the temperature influence caused by the area of ​​the high-temperature anomaly region itself, but also the intensity of the spatial correlation between the high-temperature anomaly region and electrical parameters. Then, the scaling factor of the high-temperature anomaly region is multiplied by the reference area of ​​the structural element of the corresponding thermal distribution image to obtain the area of ​​the structural element of the high-temperature anomaly region. By multiplying the scaling factor by the reference area of ​​the structural element, the area of ​​the structural element of the high-temperature anomaly region can be adapted to the high-temperature anomaly region, improving the accuracy of the structural element of the high-temperature anomaly region. After obtaining the structural element of each high-temperature anomaly region, morphological operations are performed on the high-temperature anomaly region using the structural element to determine the hot spot area and temperature rise gradient of each electrical device, and then the thermal characteristics of each electrical device are determined based on the hot spot area and temperature rise gradient. Thus, in traditional morphological operations, the structural element is mostly a 3×3 or 5×5 area. This fixed area structural element has poor scale adaptability when dealing with different sized areas to be processed, resulting in loss of detail or overprocessing. In this embodiment, the scaling factor is used to scale the shape of the structural elements so that the structural elements of each high-temperature anomaly region can be adapted to the area of ​​the high-temperature region itself, thus ensuring accurate morphological operations on the high-temperature anomaly region.

[0069] In the above embodiments, geometric triangles are constructed using the maximum, minimum, and average points of electrical and thermal temperature rise as vertices to form the shape of the structural element. This allows the structural element shape to comprehensively and accurately represent temperature anomalies on the thermal distribution image. Then, by calculating the area change rate of the high-temperature anomaly region, a scaling factor is constructed using the area change rate and the electrothermal variation coefficient. This scaling factor, applied to the structural element shape, ensures that the area of ​​the structural element in the high-temperature anomaly region not only considers the temperature influence caused by the area of ​​the high-temperature anomaly region itself but also the intensity of the spatial correlation between the high-temperature anomaly region and electrical parameters. This allows the structural element area to adapt to the high-temperature anomaly region, improving its accuracy. Next, by considering the structural element shape of the overall temperature anomaly in the thermal distribution image and the structural element area considering the local factors of the high-temperature anomaly region itself, the structural element of the high-temperature anomaly region is constructed. This improves the accuracy and comprehensiveness of the structural element, enabling it to comprehensively and accurately perform morphological processing on the thermal distribution image, thereby obtaining accurate thermal features and improving the accuracy of monitoring the status of the ship's electric propulsion system.

[0070] In this embodiment, the step of performing morphological operations on the high-temperature anomaly region based on the structural elements of each high-temperature anomaly region to determine the hot spot area and temperature rise gradient of each electrical device, and then determining the thermal characteristics of each electrical device based on the hot spot area and temperature rise gradient, includes:

[0071] Based on the structural elements of each high-temperature anomaly region, morphological opening and closing operations are performed on the high-temperature anomaly regions to obtain the first high-temperature anomaly region corresponding to each high-temperature anomaly region.

[0072] Connectivity analysis is performed on several first high-temperature anomaly regions in each of the thermal distribution images to determine the hot spot area of ​​each thermal distribution image; then, the hot spot area of ​​each electrical device is determined based on the hot spot area of ​​each thermal distribution image.

[0073] The temperature rise gradient of each of the first high-temperature anomaly regions is calculated based on a preset gradient operator;

[0074] Based on the area of ​​each first high temperature anomaly region in each of the thermal distribution images, the temperature rise gradient weight of each first high temperature anomaly region is determined.

[0075] Based on the temperature rise gradient weight, the temperature rise gradient of the first high temperature anomaly region in each of the thermal distribution images is weighted and summed to determine the temperature rise gradient of each of the thermal distribution images, and then the temperature rise gradient of each of the electrical devices is determined.

[0076] The thermal characteristics of each electrical device are determined based on the hot spot area and temperature rise gradient of each device.

[0077] It should be noted that morphological operations are based on mathematical morphology. By sliding structuring elements across an image, they perform logical operations (such as AND, OR, NOT) on pixel neighborhoods, thereby altering the image's geometry. Their combinational operations include morphological opening and closing operations. Connected Component Analysis (CCA) is a core technique in digital image processing used to identify, label, and statistically analyze independent connected regions in images. It is widely applied in object detection, image segmentation, and character recognition. Gradient operators are core tools in digital image processing for detecting image edges and texture changes. They highlight salient features in an image by calculating the spatial rate of change (i.e., gradient) of pixel grayscale values.

[0078] In one optional embodiment, morphological opening operations are performed on the high-temperature anomaly region using its structural elements to smooth its boundaries. Morphological closing operations are then performed to fill small holes in the high-temperature anomaly region, resulting in a first high-temperature anomaly region. Connected Component Analysis (CCA) is then performed on the first high-temperature anomaly region to obtain its hotspot area. The hotspot areas of all first high-temperature anomaly regions in each heat distribution image are summed to obtain the hotspot area of ​​the heat distribution image. This hotspot area is then used as the hotspot area of ​​the corresponding electrical equipment. The gradient operator is set to the Sobel operator, and the temperature rise gradient of each of the first high temperature anomaly regions is calculated using the Sobel operator. Then, the area of ​​the first high temperature anomaly region is divided by the sum of the areas of all the first high temperature anomaly regions in this heat distribution image to obtain the temperature rise gradient weight. The temperature rise gradient of the first high temperature anomaly region in each heat distribution image is weighted and summed using the temperature rise gradient weight to determine the temperature rise gradient of each heat distribution image. Then, the temperature rise gradient of the heat distribution image is used as the temperature rise gradient of the corresponding electrical equipment. Finally, the hot spot area and temperature rise gradient of the electrical equipment are used as the thermal characteristics of the electrical equipment.

[0079] In the above embodiments, by performing morphological processing on the high-temperature anomaly region using accurate and comprehensive structural elements, noise points in the high-temperature anomaly region can be accurately removed and the boundaries of the high-temperature anomaly region can be smoothed, resulting in a more accurate and realistic first high-temperature anomaly region. Next, by analyzing the connected regions of the first high-temperature anomaly region, the hotspot area of ​​the thermal distribution image can be quantified, enabling accurate acquisition of the hotspot area of ​​the electrical equipment. Simultaneously, the gradient operator accurately determines the temperature rise gradient of the first high-temperature anomaly region, and using the area of ​​the first high-temperature anomaly region, the influence of the temperature rise gradient on the thermal distribution image can be accurately quantified, resulting in a more accurate temperature rise gradient weight. Furthermore, by weighted summing of the temperature rise gradient weights, a more accurate temperature rise gradient of the electrical equipment can be obtained. Finally, by determining the accurate thermal characteristics using the hotspot area and temperature rise gradient, accurate condition monitoring of the ship's electric propulsion system can be achieved.

[0080] In this embodiment, the process of fusing the thermal characteristics and several electrical parameters of each electrical device to determine the electrothermal fusion data of each electrical device, and predicting the state risk mode of the ship's electric propulsion system based on the electrothermal fusion data, includes:

[0081] Obtain the electrothermal knowledge graph of the ship's electric propulsion system;

[0082] Based on the electrothermal knowledge graph, the thermal characteristics and several electrical parameters of each electrical device are fused to determine the electrothermal fusion data of each electrical device;

[0083] The electrothermal fusion data of each electrical device is input into the pre-trained overheating prediction model to determine the overheating prediction result of each electrical device.

[0084] Keyword extraction is performed on the overheating prediction results of each electrical device to determine the state risk similarity between any two electrical devices, and the state risk pattern of the ship's electric propulsion system is inferred based on the state risk similarity.

[0085] In one optional embodiment, an electrothermal knowledge graph of the ship's electric propulsion system is obtained. This electrothermal knowledge graph is constructed from historical monitoring data of various electrical devices within the ship's electric propulsion system. Specifically, the historical monitoring data includes: several historical electrical parameters, historical hotspot areas, and historical temperature rise gradients obtained from each state monitoring of each electrical device at a historical moment. Each electrical device is assigned a device ID, and each historical electrical parameter, device ID, and the corresponding state monitoring timestamp are used as node attributes to construct electrical parameter nodes. Historical temperature rise gradients, device IDs, and the corresponding state monitoring timestamps are used as node attributes to construct temperature rise gradient nodes. Historical hotspot areas, device IDs, and the corresponding state monitoring timestamps are used as node attributes to construct hotspot area nodes. Simultaneously, undirected edges are formed between temperature rise gradient nodes, electrical parameter nodes, and hotspot area nodes with the same device ID and the same state monitoring timestamp to obtain the electrothermal knowledge graph of the ship's electric propulsion system.

[0086] Then, the thermal characteristics and several electrical parameters of each electrical device are fused using an electrothermal knowledge graph to determine the electrothermal fusion data for each device. This electrothermal fusion data is then input into a pre-trained overheating prediction model. The overheating prediction model performs time-series analysis on the electrothermal fusion data to determine the overheating prediction result for each electrical device. The overheating prediction model is trained on a neural network model using historical electrothermal fusion data and the corresponding overheating results of the electrical devices. Next, keywords are extracted from the overheating prediction results of each electrical device to determine the state risk similarity between any two devices. Based on this state risk similarity, the state risk pattern of the ship's electric propulsion system is inferred. Furthermore, in this embodiment, the overheating prediction result refers to text data, which shows which components of the electrical device are overheating. For example, taking a generator as an example, one possible overheating prediction result is: localized overheating of the stator winding and overheating of the diodes in the excitation rectifier cabinet located at the air outlet side of the duct.

[0087] It should be noted that neural network models include, but are not limited to, LSTM models, CNN-LSTM models, Transformer models, Transformer-LSTM models, and BP neural network models. Since there are already relatively mature technologies for obtaining prediction data through time series analysis using neural network models, and for training neural network models using corresponding data, this embodiment will not elaborate further.

[0088] In the above embodiments, the fusion of accurate thermal characteristics and electrical parameters through the electrothermal knowledge graph of the ship's electric propulsion system improves the quality of the fusion process, thereby enhancing the accuracy of the electrothermal fusion data. Consequently, the overheating prediction results of electrical equipment based on the electrothermal fusion data are more accurate. Furthermore, keyword extraction improves the accuracy of state risk similarity calculation, enabling accurate reasoning about the state risk patterns of the ship's electric propulsion system based on state risk similarity. This results in more comprehensive and accurate state monitoring results based on electrical parameters, thermal characteristics, and state risk patterns, achieving accurate and comprehensive state monitoring of the ship's electric propulsion system and significantly improving the accuracy and comprehensiveness of state monitoring.

[0089] In this embodiment, the step of fusing the thermal characteristics and several electrical parameters of each electrical device based on the electrothermal knowledge graph to determine the electrothermal fusion data of each electrical device includes:

[0090] Based on the hot spot area and temperature rise gradient of each electrical device, the electrothermal knowledge graph is identified to determine a number of hot spot area nodes corresponding to the hot spot area and a number of temperature rise gradient nodes corresponding to the temperature rise gradient.

[0091] Based on the hot spot area node and the temperature rise gradient node, the starting point and ending point of the electric heating path are determined, and the electric heating knowledge graph is searched for connected paths to determine several electric heating paths for each electrical device.

[0092] The electrical parameter nodes on each of the electrothermal paths are statistically analyzed to generate a list of electrical parameter nodes for each of the electrical devices;

[0093] Based on the list of electrical parameter nodes, several electrical parameters of each electrical device are filtered to determine the associated electrical parameters of each electrical device;

[0094] The associated electrical parameters, hot spot area, and temperature rise gradient of each electrical device are spliced ​​together to generate electrothermal fusion data for each electrical device.

[0095] In one optional embodiment, based on the device ID and hotspot area of ​​the electrical equipment, the node attributes of the hotspot area nodes in the electrothermal knowledge graph are matched, where matching means having the same device ID and hotspot area; this yields several hotspot area nodes corresponding to the hotspot area. Based on the device ID and temperature rise gradient of the electrical equipment, the node attributes of the temperature rise gradient nodes in the electrothermal knowledge graph are matched, yielding several temperature rise gradient nodes corresponding to the temperature rise gradient; this matching means having the same device ID and temperature rise gradient. Then, using any hotspot area node as the starting point of the electrothermal path and any temperature rise gradient node as the ending point, breadth-first search (BFS) is used to search for connected paths in the electrothermal knowledge graph. Because the constructed electrothermal knowledge graph restricts undirected edges to only nodes with the same device ID and the same status monitoring timestamp (temperature rise gradient, electrical parameter, and hot spot area), each electrothermal path obtained in this way will only contain electrical parameter nodes. Furthermore, electrical parameter nodes, hot spot area nodes, and temperature rise gradient nodes all share the same device ID, meaning they belong to the same electrical device. Therefore, compared to traditional knowledge graphs which include a large amount of redundant data, this embodiment, by restricting undirected edges to nodes with the same device ID and the same status monitoring timestamp, can further improve the efficiency of path search, thereby improving data fusion efficiency.

[0096] Then, by statistically analyzing the electrical parameter nodes along the electrothermal path, an electrical parameter node list is generated for each electrical device. This list includes several historical electrical parameters of the electrical device in the electrothermal knowledge graph. The electrical parameters of the electrical device are then filtered using this list; if a parameter is not present in the list, it is removed, leaving the associated electrical parameters. The associated electrical parameters, hotspot area, and temperature rise gradient of the electrical device are then concatenated as vectors to generate electrothermal fusion data. Traditional data fusion schemes often use vector concatenation, which still retains irrelevant data, and excessive data can reduce the accuracy of the fused data. This embodiment, however, uses path search and electrical parameters from the electrothermal knowledge graph to filter associated electrical parameters closely related to hotspot area and temperature rise gradient, while removing electrical parameters unrelated to high-temperature anomalies. This ensures that, even with a large number of electrical parameters for each electrical device in a ship's electric propulsion system, the electrothermal fusion data not only avoids excessive data length but also accurately reflects the electrical and temperature states of the electrical devices.

[0097] In the above embodiments, by mapping the hotspot area and temperature rise gradient to hotspot area nodes and temperature rise gradient nodes in the electrothermal knowledge graph, and then using these nodes as the start and end points of the electrothermal path, a connected path search is performed on the electrothermal knowledge graph. This allows the electrothermal path to be associated with electrical parameter nodes closely related to the hotspot area and temperature rise gradient, thereby achieving effective screening of several electrical parameters of the electrical equipment. Noise from electrical parameters unrelated to the high-temperature anomaly of the electrical equipment is eliminated, making the electrothermal fusion data, which is composed of associated electrical parameters, hotspot area, and temperature rise gradient, accurately reflect the electrical and temperature states of the electrical equipment. This improves the accuracy of the electrothermal fusion data, making the overheating prediction results and state risk models of the electrical equipment based on the electrothermal fusion data more accurate. Consequently, the state monitoring results based on electrical parameters, thermal characteristics, and state risk models are more comprehensive and accurate, achieving accurate and comprehensive state monitoring of the ship's electric propulsion system, and significantly improving the accuracy and comprehensiveness of state monitoring of the ship's electric propulsion system.

[0098] In this embodiment, the step of extracting keywords from the overheat prediction results of each electrical device, determining the state risk similarity between any two electrical devices, and inferring the state risk pattern of the ship's electric propulsion system based on the state risk similarity includes:

[0099] Based on a preset text extraction algorithm, keywords are extracted from the overheating prediction results of each electrical device to determine the keyword set for each electrical device.

[0100] The correlation score between each two sets of keywords is calculated based on the Jaccard similarity coefficient algorithm to determine the similarity of the state risk of each two electrical devices.

[0101] Based on the similarity of the state risk of every two electrical devices, a similarity matrix of the ship's electric propulsion system is constructed.

[0102] The key electrical equipment of the ship's electric propulsion system is identified by extracting features from the similarity matrix using the PCA algorithm.

[0103] The keyword set of the ship's electric propulsion system is determined based on the union of the keyword sets of all the key electrical equipment.

[0104] The keyword set of the ship electric propulsion system is arranged based on a preset semantic model to determine the state risk tendency of the ship electric propulsion system.

[0105] Based on the state risk tendency query of the ship electric propulsion system, the pre-acquired risk tendency operation and maintenance database is used to determine the load adjustment strategy and early warning strategy corresponding to the ship electric propulsion system.

[0106] Based on the state risk tendency, load regulation strategy, and early warning strategy of the ship's electric propulsion system, the state risk mode of the ship's electric propulsion system is determined.

[0107] In one optional embodiment, the text extraction algorithm is set to the TF-IDF algorithm. Keyword extraction is performed based on the overheating prediction results of electrical equipment using the TF-IDF algorithm to determine the keyword set for each electrical equipment. For example, taking a generator as an example, one overheating prediction result for a generator is local overheating of the stator winding and overheating of the diode located on the air outlet side of the excitation rectifier cabinet. Therefore, its keyword set includes overheating, stator winding, excitation rectifier cabinet, diode, and air outlet of the air duct. Taking a transformer as an example, one overheating prediction result for a transformer is overheating of the winding and overheating of the tap changer contact. Therefore, its keyword set includes winding, tap changer, contact, and overheating. Then, the Jaccard similarity coefficient algorithm is used to calculate the relevance score between each pair of keyword sets, and this relevance score is used as the state risk similarity between the two electrical devices. After obtaining the state risk similarity between each pair of electrical devices, a similarity matrix of the ship's electric propulsion system is constructed. Features are extracted from the similarity matrix using the PCA algorithm to identify principal components, and the electrical devices corresponding to the principal components are identified as the key electrical devices of the ship's electric propulsion system. Then, a semantic model (e.g., BERT model, GPT model) is used to arrange the keyword set of the ship's electric propulsion system to determine the state risk tendency of the ship's electric propulsion system. For example, assuming the keyword set of the ship's electric propulsion system includes: overheating, stator winding, excitation rectifier cabinet, diode, duct outlet, winding, tap changer, and contact. After arrangement using the semantic model, the resulting state risk tendencies include: localized overheating of the generator's stator winding and overheating of the diode in the excitation rectifier cabinet located on the duct outlet side, overheating of the transformer winding, and overheating of the tap changer contact. Then, by querying the pre-acquired risk tendency maintenance database through the state risk tendency query, the corresponding load adjustment strategy and early warning strategy for the ship's electric propulsion system are determined. The risk tendency maintenance database stores the load adjustment strategy and early warning strategy for each electrical device in the ship's electric propulsion system. For example, if the state risk tendency is local overheating of the generator stator winding and overheating of the diode in the excitation rectifier cabinet located on the air outlet side of the air duct, overheating of the transformer winding and overheating of the tap changer contact, then the load adjustment strategy obtained by querying the risk tendency maintenance database at this time is to immediately implement three-phase load balance optimization to reduce the reactive power output of the generator; the early warning strategy is to issue an overheating warning for the transformer tap changer through the ship's alarm bell.

[0108] In traditional ship electric propulsion system state risk pattern identification, most methods rely on threshold processing to filter overheating prediction results of electrical equipment to obtain state risk patterns. However, this threshold setting method ignores the risk correlation between different electrical equipment. This embodiment extracts a set of keywords from the overheating prediction results and constructs a similarity matrix using the keyword set. This allows for the measurement of the fault risk correlation of different electrical equipment in overheating prediction, thereby cleverly identifying key electrical equipment in the ship electric propulsion system through the PCA algorithm, improving the comprehensiveness and accuracy of the state risk patterns of the ship electric propulsion system.

[0109] It's important to note that TF-IDF is a classic statistical method used in information retrieval and text mining to assess the importance of a word to a specific document within a document set. Its core idea is that the higher the frequency of a word in the current document (high TF) and the lower its frequency across all documents (high IDF), the stronger its representativeness (i.e., discriminative power) for that document. The Jaccard Similarity Coefficient is a statistical indicator used to measure the similarity between two finite sets. It is defined as the ratio of the size of the intersection to the size of the union of two sets, with values ​​ranging from [0,1]. A higher value indicates greater similarity between the two sets. Principal Component Analysis (PCA) is a linear dimensionality reduction method that projects high-dimensional data into a low-dimensional space through orthogonal transformation, preserving the direction of maximum variance (principal components) to achieve data compression and feature extraction. Mathematically, it performs eigenvalue decomposition on the covariance matrix or similarity matrix, selecting the eigenvector corresponding to the largest eigenvalue as the principal component direction.

[0110] In the above embodiments, the text extraction algorithm can accurately extract the keyword set from the overheating prediction results, making the state risk similarity of each pair of electrical devices calculated using the keyword set more accurate; and the state risk similarity can accurately quantify the fault risk correlation between different electrical devices. Then, by using the PCA algorithm to extract features from the similarity matrix, based on considering the fault risk correlation between different electrical devices, key electrical devices with a more closely related impact on the state of the ship's electric propulsion system can be further identified, improving the accuracy of the keyword set for the state representation of the ship's electric propulsion system. Next, by arranging the keyword set of the ship's electric propulsion system using a semantic model, a more accurate state risk tendency of the ship's electric propulsion system can be obtained. The state risk tendency is then converted into corresponding load adjustment strategies and early warning strategies using a risk tendency operation and maintenance library. From the multi-angle data of state risk tendency, load adjustment strategies, and early warning strategies, the comprehensiveness and accuracy of the state risk pattern of the ship's electric propulsion system are improved. This results in more comprehensive and accurate state monitoring results based on electrical parameters, thermal characteristics, and state risk patterns, achieving accurate and comprehensive state monitoring of the ship's electric propulsion system and significantly improving the accuracy and comprehensiveness of state monitoring of the ship's electric propulsion system.

[0111] Example 2

[0112] Please refer to Figure 2 , Figure 2 A schematic diagram of the structure of a ship electric propulsion system status monitoring system based on infrared thermal imaging provided in an embodiment of the present invention includes: a high temperature anomaly area identification module 201, an electrical parameter projection module 202, an electrical temperature rise feature point identification module 203, a structural element construction module 204, an electrothermal fusion prediction module 205, and a status monitoring result generation module 206.

[0113] The high-temperature anomaly area identification module 201 is used to acquire thermal distribution images of various electrical devices in the ship's electric propulsion system based on infrared thermal imaging technology, and to identify several high-temperature anomaly areas in each thermal distribution image.

[0114] The electrical parameter projection module 202 is used to acquire several electrical parameters of each electrical device, project the electrical parameters onto the corresponding thermal distribution image, and determine the electrical offset center of each thermal distribution image;

[0115] The electrical temperature rise feature point identification module 203 is used to construct the connection edge from the electrical offset center to each of the high temperature anomaly regions in each of the thermal distribution images; calculate the electrical-thermal variation coefficient of each of the connection edges based on the thermal distribution images; and then identify several electrical temperature rise feature points based on the electrical-thermal variation coefficient.

[0116] The structural element construction module 204 is used to determine the shape of the structural element of each thermal distribution image based on the plurality of electrical temperature rise feature points; and to scale the shape of the structural element based on the electrical temperature variation coefficient to determine the structural element of each high temperature anomaly region; and to perform morphological processing on the high temperature anomaly region based on the structural element to determine the thermal characteristics of each electrical device.

[0117] The electrothermal fusion prediction module 205 is used to fuse the thermal characteristics and several electrical parameters of each electrical device to determine the electrothermal fusion data of each electrical device, and predict the state risk mode of the ship electric propulsion system based on the electrothermal fusion data.

[0118] The condition monitoring result generation module 206 is used to determine the condition monitoring result of the ship electric propulsion system based on several electrical parameters and thermal characteristics of each electrical device, as well as the condition risk mode of the ship electric propulsion system.

[0119] In this embodiment, the electrical parameter projection module 202 includes: an electrical parameter projection unit;

[0120] The electrical parameter projection unit is used to acquire several electrical parameters of each electrical device and the rated electrical parameters corresponding to each electrical parameter; determine the electrical deviation of each electrical parameter based on the ratio of each electrical parameter to the corresponding rated electrical parameter; acquire the device image of each electrical device, and register the device image and thermal distribution image of each electrical device based on the SIFT algorithm to determine the electrothermal transformation matrix of each electrical device; project several electrical parameters of each electrical device onto the corresponding thermal distribution image based on the transformation matrix and the measurement position of each electrical parameter at the electrical device, and determine the electrical projection point of each electrical parameter on the corresponding thermal distribution image; and cluster the electrical projection points on each thermal distribution image based on the coordinates of the electrical projection points on the thermal distribution image and the corresponding electrical deviation to determine the electrical deviation center of each thermal distribution image.

[0121] In this embodiment, the electrical temperature rise feature point recognition module 203 includes: an electrical temperature rise feature point recognition unit;

[0122] The electrical temperature rise feature point recognition unit is used to obtain the centroid of each of the high temperature anomaly regions in each of the thermal distribution images;

[0123] Connect the electrical offset center to the centroid of each of the high-temperature anomaly regions to construct a connection edge from the electrical offset center to each of the high-temperature anomaly regions;

[0124] The temperature value of each pixel on each of the connecting edges is determined based on the thermal distribution image.

[0125] Based on the temperature value of each pixel on each of the connecting edges, calculate the standard deviation and average value of the temperature values ​​of all pixels on each of the connecting edges;

[0126] The electrothermal variation coefficient of each connection edge is determined based on the ratio of the standard deviation of the temperature value to the average temperature value of the pixels on each connection edge.

[0127] Based on the electrothermal variation coefficient of each of the connecting edges, the centroid of the high-temperature anomaly region is screened to determine several characteristic points of electrothermal temperature rise.

[0128] In this embodiment, the electrical temperature rise feature point identification unit includes: an electrical temperature rise feature point identification subunit;

[0129] The electrothermal temperature rise feature point identification subunit is used to accumulate and average the electrothermal variation coefficients of each of the connecting edges to determine the average electrothermal variation coefficient of the thermal distribution image.

[0130] The first connecting edge with the largest electrothermal variation coefficient is selected from the connecting edges, and the centroid of the high temperature anomaly region corresponding to the first connecting edge is taken as the maximum point of electrothermal temperature rise in the heat distribution image.

[0131] The second connecting edge with the smallest electrothermal variation coefficient is selected from the connecting edges, and the centroid of the high temperature anomaly region corresponding to the second connecting edge is taken as the minimum point of electrothermal temperature rise in the heat distribution image.

[0132] The absolute deviation of the electrothermal variation coefficient of each of the connecting edges is determined based on the electrothermal variation coefficient of each edge and the average value of the electrothermal variation coefficient.

[0133] The third connecting edge with the smallest absolute deviation of the electrothermal variation coefficient is selected from the connecting edges, and the centroid of the high temperature anomaly region corresponding to the third connecting edge is taken as the average point of the electrothermal temperature rise of the heat distribution image.

[0134] Based on the maximum, minimum, and average points of the electrical temperature rise, several characteristic points of electrical temperature rise are determined.

[0135] In this embodiment, the structural element construction module 204 includes: a structural element construction unit;

[0136] The structural element building unit is used to construct the geometric triangle of each heat distribution image, with the maximum, minimum, and average points of electrical and thermal temperature rise of each heat distribution image as vertices.

[0137] The geometric triangle of each heat distribution image is used as the shape of the structural element of each heat distribution image, and the area of ​​the geometric triangle of each heat distribution image is used as the reference area of ​​the structural element of each heat distribution image.

[0138] The area of ​​each high-temperature anomaly region in each of the thermal distribution images is obtained, and the area change rate of each high-temperature anomaly region is determined based on the ratio of the area of ​​each high-temperature anomaly region to the reference area of ​​the structural element of the thermal distribution image to which it belongs.

[0139] Based on the electrothermal variation coefficient and the area change rate of the connecting edge corresponding to each of the high-temperature anomaly regions, the scaling factor of each of the high-temperature anomaly regions is determined.

[0140] The scaling factor of each of the high temperature anomaly regions is multiplied by the reference area of ​​the structural element of the corresponding thermal distribution image to scale the shape of the structural element of the corresponding thermal distribution image, thereby determining the area of ​​the structural element of each of the high temperature anomaly regions.

[0141] Based on the area of ​​the structural element of each high-temperature anomaly region and the shape of the structural element of the corresponding thermal distribution image, the structural element of each high-temperature anomaly region is constructed.

[0142] Morphological operations are performed on the high-temperature anomaly regions based on the structural elements of each region to determine the hot spot area and temperature rise gradient of each electrical device. Then, based on the hot spot area and temperature rise gradient, the thermal characteristics of each electrical device are determined.

[0143] In this embodiment, the structural element construction unit includes: a morphological operation subunit;

[0144] The morphological operation subunit is used to perform morphological opening and closing operations on the high temperature anomaly region based on the structuring element of each high temperature anomaly region, to obtain the first high temperature anomaly region corresponding to each high temperature anomaly region.

[0145] Connectivity analysis is performed on several first high-temperature anomaly regions in each of the thermal distribution images to determine the hot spot area of ​​each thermal distribution image; then, the hot spot area of ​​each electrical device is determined based on the hot spot area of ​​each thermal distribution image.

[0146] The temperature rise gradient of each of the first high-temperature anomaly regions is calculated based on a preset gradient operator;

[0147] Based on the area of ​​each first high temperature anomaly region in each of the thermal distribution images, the temperature rise gradient weight of each first high temperature anomaly region is determined.

[0148] Based on the temperature rise gradient weight, the temperature rise gradient of the first high temperature anomaly region in each of the thermal distribution images is weighted and summed to determine the temperature rise gradient of each of the thermal distribution images, and then the temperature rise gradient of each of the electrical devices is determined.

[0149] The thermal characteristics of each electrical device are determined based on the hot spot area and temperature rise gradient of each device.

[0150] In this embodiment, the electrothermal fusion prediction module 205 includes: an electrothermal fusion prediction unit;

[0151] The electrothermal fusion prediction unit is used to obtain the electrothermal knowledge graph of the ship's electric propulsion system.

[0152] Based on the electrothermal knowledge graph, the thermal characteristics and several electrical parameters of each electrical device are fused to determine the electrothermal fusion data of each electrical device;

[0153] The electrothermal fusion data of each electrical device is input into the pre-trained overheating prediction model to determine the overheating prediction result of each electrical device.

[0154] Keyword extraction is performed on the overheating prediction results of each electrical device to determine the state risk similarity between any two electrical devices, and the state risk pattern of the ship's electric propulsion system is inferred based on the state risk similarity.

[0155] In this embodiment, the electrothermal fusion prediction unit includes: an electrothermal fusion data acquisition subunit;

[0156] The electrothermal fusion data acquisition subunit is used to identify nodes in the electrothermal knowledge graph based on the hot spot area and temperature rise gradient of each electrical device, and to determine a number of hot spot area nodes corresponding to the hot spot area and a number of temperature rise gradient nodes corresponding to the temperature rise gradient.

[0157] Based on the hot spot area node and the temperature rise gradient node, the starting point and ending point of the electric heating path are determined, and the electric heating knowledge graph is searched for connected paths to determine several electric heating paths for each electrical device.

[0158] The electrical parameter nodes on each of the electrothermal paths are statistically analyzed to generate a list of electrical parameter nodes for each of the electrical devices;

[0159] Based on the list of electrical parameter nodes, several electrical parameters of each electrical device are filtered to determine the associated electrical parameters of each electrical device;

[0160] The associated electrical parameters, hot spot area, and temperature rise gradient of each electrical device are spliced ​​together to generate electrothermal fusion data for each electrical device.

[0161] In this embodiment, the electrothermal fusion prediction unit includes: a state risk mode reasoning subunit;

[0162] The state risk pattern reasoning subunit is used to extract keywords from the overheating prediction results of each electrical device according to a preset text extraction algorithm, and determine the keyword set of each electrical device.

[0163] The correlation score between each two sets of keywords is calculated based on the Jaccard similarity coefficient algorithm to determine the similarity of the state risk of each two electrical devices.

[0164] Based on the similarity of the state risk of every two electrical devices, a similarity matrix of the ship's electric propulsion system is constructed.

[0165] The key electrical equipment of the ship's electric propulsion system is identified by extracting features from the similarity matrix using the PCA algorithm.

[0166] The keyword set of the ship's electric propulsion system is determined based on the union of the keyword sets of all the key electrical equipment.

[0167] The keyword set of the ship electric propulsion system is arranged based on a preset semantic model to determine the state risk tendency of the ship electric propulsion system.

[0168] Based on the state risk tendency query of the ship electric propulsion system, the pre-acquired risk tendency operation and maintenance database is used to determine the load adjustment strategy and early warning strategy corresponding to the ship electric propulsion system.

[0169] Based on the state risk tendency, load regulation strategy, and early warning strategy of the ship's electric propulsion system, the state risk mode of the ship's electric propulsion system is determined.

[0170] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A method for monitoring the condition of a ship's electric propulsion system based on infrared thermal imaging, characterized in that, include: The thermal distribution images of various electrical devices in the ship's electric propulsion system are acquired using infrared thermal imaging technology, and several high-temperature abnormal areas are identified in each of the thermal distribution images. Acquire several electrical parameters for each of the electrical devices, project the electrical parameters onto the corresponding thermal distribution image, and determine the electrical offset center of each thermal distribution image; In each of the thermal distribution images, a connection edge is constructed from the electrical off-center to each of the high-temperature anomaly regions; The electrothermal variation coefficient of each connecting edge is calculated based on the thermal distribution image, and then several electrothermal temperature rise feature points are identified based on the electrothermal variation coefficient. Specifically, in each thermal distribution image, the centroid of each high-temperature anomaly region is obtained; the electrical offset center is connected to the centroid of each high-temperature anomaly region to construct a connecting edge from the electrical offset center to each high-temperature anomaly region; the temperature value of each pixel on each connecting edge is determined based on the thermal distribution image; based on the temperature value of each pixel on each connecting edge, the standard deviation and average temperature value of all pixels on each connecting edge are calculated; the electrothermal variation coefficient of each connecting edge is determined based on the ratio of the standard deviation and average temperature value of the pixels on each connecting edge; the centroid of the high-temperature anomaly region is screened based on the electrothermal variation coefficient of each connecting edge to determine several electrothermal temperature rise feature points, including: the electrothermal variation coefficient of each connecting edge... The coefficients of variation are accumulated and averaged to determine the average electrothermal coefficient of variation of the heat distribution image. The first connecting edge with the largest electrothermal coefficient of variation is selected from the connecting edges, and the centroid of the high-temperature anomaly region corresponding to the first connecting edge is taken as the maximum point of electrothermal temperature rise in the heat distribution image. The second connecting edge with the smallest electrothermal coefficient of variation is selected from the connecting edges, and the centroid of the high-temperature anomaly region corresponding to the second connecting edge is taken as the minimum point of electrothermal temperature rise in the heat distribution image. Based on the electrothermal coefficient of variation of each connecting edge and the average electrothermal coefficient of variation, the absolute deviation of the electrothermal coefficient of variation of each connecting edge is determined. The third connecting edge with the smallest absolute deviation of the electrothermal coefficient of variation is selected from the connecting edges, and the centroid of the high-temperature anomaly region corresponding to the third connecting edge is taken as the average point of electrothermal temperature rise in the heat distribution image. Based on the maximum point, minimum point, and average point of electrothermal temperature rise, several electrothermal temperature rise feature points are determined. The shape of the structural element of each heat distribution image is determined based on the aforementioned electrical temperature rise feature points; and the shape of the structural element is scaled based on the electrical-thermal variation coefficient to determine the structural element of each high-temperature anomaly region; morphological processing is performed on the high-temperature anomaly region based on the structural element to determine the thermal characteristics of each electrical device; wherein, the maximum, minimum, and average electrical temperature rise points of each heat distribution image are used as vertices to construct a geometric triangle of each heat distribution image; the geometric triangle of each heat distribution image is used as the shape of the structural element of each heat distribution image, and the area of ​​the geometric triangle of each heat distribution image is used as the reference area of ​​the structural element of each heat distribution image; the area of ​​each high-temperature anomaly region in each heat distribution image is obtained, and the area of ​​each high-temperature anomaly region is compared with the structural element of its respective heat distribution image. The ratio of the reference area of ​​each structural element is used to determine the area change rate of each high-temperature anomaly region; based on the electrothermal variation coefficient of the connecting edge corresponding to each high-temperature anomaly region and the area change rate, a scaling factor for each high-temperature anomaly region is determined; the scaling factor of each high-temperature anomaly region is multiplied by the reference area of ​​the structural element of the corresponding thermal distribution image to scale the shape of the structural element of the corresponding thermal distribution image, thereby determining the area of ​​the structural element of each high-temperature anomaly region; based on the area of ​​the structural element of each high-temperature anomaly region and the shape of the structural element of the corresponding thermal distribution image, a structural element for each high-temperature anomaly region is constructed; based on the structural element of each high-temperature anomaly region, morphological operations are performed on the high-temperature anomaly region to determine the hot spot area and temperature rise gradient of each electrical device, and then based on the hot spot area and temperature rise gradient, the thermal characteristics of each electrical device are determined; The thermal characteristics and several electrical parameters of each electrical device are fused to determine the electrothermal fusion data of each electrical device, and the state risk mode of the ship's electric propulsion system is predicted based on the electrothermal fusion data; Based on several electrical parameters and thermal characteristics of each of the electrical devices, and the state risk mode of the ship's electric propulsion system, the state monitoring results of the ship's electric propulsion system are determined.

2. The method for monitoring the condition of a ship's electric propulsion system based on infrared thermal imaging as described in claim 1, characterized in that, The step of acquiring several electrical parameters for each of the electrical devices, projecting the electrical parameters onto the corresponding thermal distribution image, and determining the electrical offset center of each thermal distribution image includes: Obtain several electrical parameters for each of the electrical devices and the rated electrical parameters corresponding to each of the electrical parameters; The electrical deviation of each electrical parameter is determined based on the ratio of each electrical parameter to the corresponding rated electrical parameter. Acquire the device image of each electrical device, and register the device image and thermal distribution image of each electrical device based on the SIFT algorithm to determine the electrothermal transformation matrix of each electrical device; Based on the transformation matrix and the measurement position of each electrical parameter in the electrical equipment, several electrical parameters of each electrical equipment are projected onto the corresponding thermal distribution image to determine the electrical projection point of each electrical parameter on the corresponding thermal distribution image. Based on the coordinates of the electrical projection points on the thermal distribution image and the corresponding electrical deviation, the electrical projection points on each thermal distribution image are clustered to determine the electrical deviation center of each thermal distribution image.

3. The method for monitoring the condition of a ship's electric propulsion system based on infrared thermal imaging as described in claim 1, characterized in that, The step of performing morphological operations on the high-temperature anomaly regions based on the structural elements of each high-temperature anomaly region to determine the hot spot area and temperature rise gradient of each electrical device, and then determining the thermal characteristics of each electrical device based on the hot spot area and temperature rise gradient, includes: Based on the structural elements of each high-temperature anomaly region, morphological opening and closing operations are performed on the high-temperature anomaly regions to obtain the first high-temperature anomaly region corresponding to each high-temperature anomaly region. Connectivity analysis is performed on several first high-temperature anomaly regions in each of the thermal distribution images to determine the hot spot area of ​​each thermal distribution image; then, the hot spot area of ​​each electrical device is determined based on the hot spot area of ​​each thermal distribution image. The temperature rise gradient of each of the first high-temperature anomaly regions is calculated based on a preset gradient operator; Based on the area of ​​each first high temperature anomaly region in each of the thermal distribution images, the temperature rise gradient weight of each first high temperature anomaly region is determined. Based on the temperature rise gradient weight, the temperature rise gradient of the first high temperature anomaly region in each of the thermal distribution images is weighted and summed to determine the temperature rise gradient of each of the thermal distribution images, and then the temperature rise gradient of each of the electrical devices is determined. The thermal characteristics of each electrical device are determined based on the hot spot area and temperature rise gradient of each device.

4. The method for monitoring the condition of a ship's electric propulsion system based on infrared thermal imaging as described in claim 1 or 3, characterized in that, The process of fusing the thermal characteristics and several electrical parameters of each electrical device to determine the electrothermal fusion data of each electrical device, and predicting the state risk mode of the ship's electric propulsion system based on the electrothermal fusion data, includes: Obtain the electrothermal knowledge graph of the ship's electric propulsion system; Based on the electrothermal knowledge graph, the thermal characteristics and several electrical parameters of each electrical device are fused to determine the electrothermal fusion data of each electrical device; The electrothermal fusion data of each electrical device is input into the pre-trained overheating prediction model to determine the overheating prediction result of each electrical device. Keyword extraction is performed on the overheating prediction results of each electrical device to determine the state risk similarity between any two electrical devices, and the state risk pattern of the ship's electric propulsion system is inferred based on the state risk similarity.

5. The method for monitoring the condition of a ship's electric propulsion system based on infrared thermal imaging as described in claim 4, characterized in that, The process of fusing the thermal characteristics and several electrical parameters of each electrical device based on the electrothermal knowledge graph to determine the electrothermal fusion data of each electrical device includes: Based on the hot spot area and temperature rise gradient of each electrical device, the electrothermal knowledge graph is identified to determine a number of hot spot area nodes corresponding to the hot spot area and a number of temperature rise gradient nodes corresponding to the temperature rise gradient. Based on the hot spot area node and the temperature rise gradient node, the starting point and ending point of the electric heating path are determined, and the electric heating knowledge graph is searched for connected paths to determine several electric heating paths for each electrical device. The electrical parameter nodes on each of the electrothermal paths are statistically analyzed to generate a list of electrical parameter nodes for each of the electrical devices; Based on the list of electrical parameter nodes, several electrical parameters of each electrical device are filtered to determine the associated electrical parameters of each electrical device; The associated electrical parameters, hot spot area, and temperature rise gradient of each electrical device are spliced ​​together to generate electrothermal fusion data for each electrical device.

6. The method for monitoring the condition of a ship's electric propulsion system based on infrared thermal imaging as described in claim 4, characterized in that, The process of extracting keywords from the overheat prediction results of each electrical device, determining the state risk similarity between any two electrical devices, and inferring the state risk pattern of the ship's electric propulsion system based on the state risk similarity includes: Based on a preset text extraction algorithm, keywords are extracted from the overheating prediction results of each electrical device to determine the keyword set for each electrical device. The correlation score between each two sets of keywords is calculated based on the Jaccard similarity coefficient algorithm to determine the similarity of the state risk of each two electrical devices. Based on the similarity of the state risk of every two electrical devices, a similarity matrix of the ship's electric propulsion system is constructed. The key electrical equipment of the ship's electric propulsion system is identified by extracting features from the similarity matrix using the PCA algorithm. The keyword set of the ship's electric propulsion system is determined based on the union of the keyword sets of all the key electrical equipment. The keyword set of the ship electric propulsion system is arranged based on a preset semantic model to determine the state risk tendency of the ship electric propulsion system. Based on the state risk tendency query of the ship electric propulsion system, the pre-acquired risk tendency operation and maintenance database is used to determine the load adjustment strategy and early warning strategy corresponding to the ship electric propulsion system. Based on the state risk tendency, load regulation strategy, and early warning strategy of the ship's electric propulsion system, the state risk mode of the ship's electric propulsion system is determined.

7. A ship electric propulsion system condition monitoring system based on infrared thermal imaging, characterized in that, include: The module includes a high-temperature anomaly area identification module, an electrical parameter projection module, an electrical temperature rise feature point identification module, a structural element construction module, an electrothermal fusion prediction module, and a state monitoring result generation module. The high-temperature anomaly area identification module is used to acquire thermal distribution images of various electrical devices in the ship's electric propulsion system based on infrared thermal imaging technology, and to identify several high-temperature anomaly areas in each thermal distribution image. The electrical parameter projection module is used to acquire several electrical parameters of each electrical device, project the electrical parameters onto the corresponding thermal distribution image, and determine the electrical offset center of each thermal distribution image; The electrical temperature rise feature point recognition module is used to construct the connection edge from the electrical offset center to each of the high temperature anomaly regions in each of the thermal distribution images; The electrothermal variation coefficient of each connecting edge is calculated based on the thermal distribution image, and then several electrothermal temperature rise feature points are identified based on the electrothermal variation coefficient. The electrical temperature rise feature point identification module includes: an electrical temperature rise feature point identification unit; the electrical temperature rise feature point identification unit is used to obtain the centroid of each high-temperature anomaly region in each thermal distribution image; connect the electrical offset center with the centroid of each high-temperature anomaly region to construct a connection edge from the electrical offset center to each high-temperature anomaly region; determine the temperature value of each pixel on each connection edge based on the thermal distribution image; calculate the standard deviation and average temperature value of all pixels on each connection edge based on the temperature value of each pixel on each connection edge; determine the electrothermal variation coefficient of each connection edge based on the ratio of the standard deviation and average temperature value of the pixels on each connection edge; and filter the centroids of the high-temperature anomaly regions based on the electrothermal variation coefficient of each connection edge to determine several electrical temperature rise feature points; the electrical temperature rise feature point identification unit includes: an electrical temperature rise feature point identification subunit; the electrical temperature rise feature point identification subunit... The unit is used to accumulate and average the electrothermal variation coefficients of each of the connecting edges to determine the average electrothermal variation coefficient of the heat distribution image; select the first connecting edge with the largest electrothermal variation coefficient from the connecting edges, and take the centroid of the high-temperature anomaly region corresponding to the first connecting edge as the maximum electrothermal temperature rise point of the heat distribution image; select the second connecting edge with the smallest electrothermal variation coefficient from the connecting edges, and take the centroid of the high-temperature anomaly region corresponding to the second connecting edge as the minimum electrothermal temperature rise point of the heat distribution image; determine the absolute deviation of the electrothermal variation coefficient of each connecting edge based on the electrothermal variation coefficient of each connecting edge and the average electrothermal variation coefficient; select the third connecting edge with the smallest absolute deviation of the electrothermal variation coefficient from the connecting edges, and take the centroid of the high-temperature anomaly region corresponding to the third connecting edge as the average electrothermal temperature rise point of the heat distribution image; and determine several electrothermal temperature rise feature points based on the maximum electrothermal temperature rise point, the minimum electrothermal temperature rise point, and the average electrothermal temperature rise point. The structural element construction module is used to determine the shape of the structural element of each heat distribution image based on the plurality of electrical temperature rise feature points; and to scale the shape of the structural element based on the electrical temperature variation coefficient to determine the structural element of each high temperature anomaly region; and to perform morphological processing on the high temperature anomaly region based on the structural element to determine the thermal characteristics of each electrical device; the structural element construction module includes: a structural element construction unit; the structural element construction unit is used to construct a geometric triangle of each heat distribution image with the electrical temperature rise maximum point, electrical temperature rise minimum point and electrical temperature rise average point as vertices; to use the geometric triangle of each heat distribution image as the structural element shape of each heat distribution image, and to use the area of ​​the geometric triangle of each heat distribution image as the structural element reference area of ​​each heat distribution image; to obtain the area of ​​each high temperature anomaly region in each heat distribution image, and to determine the thermal characteristics of each electrical device based on each The ratio of the area of ​​the high-temperature anomaly region to the reference area of ​​the structural element of the corresponding thermal distribution image is used to determine the area change rate of each high-temperature anomaly region. Based on the electrothermal variation coefficient of the connecting edge corresponding to each high-temperature anomaly region and the area change rate, a scaling factor for each high-temperature anomaly region is determined. The scaling factor of each high-temperature anomaly region is multiplied by the reference area of ​​the structural element of the corresponding thermal distribution image to scale the shape of the structural element of the corresponding thermal distribution image, thereby determining the area of ​​the structural element of each high-temperature anomaly region. Based on the area of ​​the structural element of each high-temperature anomaly region and the shape of the structural element of the corresponding thermal distribution image, a structural element for each high-temperature anomaly region is constructed. Morphological operations are performed on the high-temperature anomaly regions based on their structural elements to determine the hotspot area and temperature rise gradient of each electrical device, and then the thermal characteristics of each electrical device are determined based on the hotspot area and temperature rise gradient. The electrothermal fusion prediction module is used to fuse the thermal characteristics and several electrical parameters of each electrical device to determine the electrothermal fusion data of each electrical device, and predict the state risk mode of the ship's electric propulsion system based on the electrothermal fusion data. The condition monitoring result generation module is used to determine the condition monitoring result of the ship's electric propulsion system based on several electrical parameters and thermal characteristics of each electrical device, as well as the condition risk mode of the ship's electric propulsion system.

Images