Online modeling method for ship ladder cable motion based on dynamic neural network model

Abstract

The present application relates to the field of neural network modeling technology, and specifically to a method for online modeling of the motion of a ship ladder's accompanying cable based on a dynamic neural network model. The method comprises: updating all weights in the initial neural network model based on the second-order mixed partial derivatives of the fusion characteristic function between any two physical tensors at all time steps before the current moment, and combining the interaction between the heat tensor and the electromagnetic tensor among all physical tensors to obtain an optimized neural network; analyzing the average distribution and discreteness of all weights between each pair of nodes in the optimized neural network at all time steps before the current moment, screening out dynamic node pairs, and updating the weights between all dynamic node pairs to obtain a ship ladder dynamic neural network. The present application aims to improve the efficiency and accuracy of online modeling of the motion of a ship ladder's accompanying cable, and to solve the key technical problem of the lack of efficient and intelligent prediction of the cable motion situation.

Description

Technical Field

[0001] The present application relates to the technical field of neural network modeling, and in particular to an online modeling method for the movement of a ship ladder accompanying cable based on a dynamic neural network model. Background Art

[0002] Marine elevators (or simply ship ladders) are specialized electromechanical equipment permanently installed on ships, providing vertical or oblique transport for passengers, crew, or cargo. Ocean-going ship ladders must reliably withstand complex sea conditions such as typhoons, high waves, and localized weather conditions. The accompanying cable of the ship ladder is particularly important, as it ensures power supply and information transmission between the elevator car and the control cabinet. Due to the multi-dimensional dynamic response of ships during navigation, such as swaying and heaving, caused by wind and wave disturbances, the accompanying cable is forced to swing. This can cause interference, damage, or even breakage, between the cable and hoistway walls, guide rails, and elevator car, leading to safety accidents.

[0003] Online modeling allows for real-time monitoring of cable motion, identifying potential safety hazards promptly and preventing accidents caused by cable failures, thereby ensuring the safety of the ladder and personnel. Data and model analysis captured through online modeling can also predict the cable's future operating status. For example, by analyzing the cable's motion trends and stress conditions, the potential points and timing of subsequent cable failures can be predicted, allowing for proactive maintenance measures to reduce the probability of failure and improve system reliability. Existing technologies, however, often model mechanical, thermal, and electromagnetic fields independently, ignoring coupling effects. Traditional neural networks require training and updating of all weights, failing to meet real-time requirements. This reduces the efficiency and accuracy of online modeling of the ladder's accompanying cable motion and makes it impossible to dynamically predict cable motion in real time. Summary of the Invention

[0004] In order to solve the above technical problems, the present application provides an online modeling method for the movement of a ship ladder accompanying cable based on a dynamic neural network model to solve the existing problems.

[0005] The online modeling method of the ship ladder cable movement based on the dynamic neural network model of this application adopts the following technical solutions:

[0006] One embodiment of the present application provides an online modeling method for the movement of a ship ladder accompanying cable based on a dynamic neural network model, the method comprising the following steps:

[0007] Real-time acquisition of various physical field data of the cable during the operation of the ship ladder accompanying cable, including mechanical field, thermal field, and electromagnetic field data. The various physical field data are respectively used as inputs of a multi-branch neural network, and various physical tensors are extracted from the neural network branches and transmitted to the fully connected layer for fusion to obtain an initial neural network model. The fusion process includes: obtaining a coupling coefficient matrix between any two physical tensors at the current time step, and fusing the any two physical tensors based on the coupling coefficient matrix to obtain a fusion characteristic function between the any two physical tensors at the current time step;

[0008] According to the weights between the nodes in the initial neural network, the initial neural network model is dynamically topologically updated to obtain an optimized neural network. The optimized neural network is then trained to obtain a ship ladder neural network model. The specific process is as follows:

[0009] Based on the second-order mixed partial derivatives of the fusion characteristic function between any two physical tensors at all time steps before the current moment, and combined with the interaction between the heat tensor and the electromagnetic tensor among all physical tensors, the coupling loss of the initial neural network model at the current time step is determined, and all weights in the initial neural network model are updated to adjust the topological structure of the initial neural network model to obtain an optimized neural network;

[0010] The average distribution and discreteness of all weights between each pair of nodes in the optimized neural network in all time steps before the current time step are analyzed, dynamic node pairs are selected from all pairs of nodes in the optimized neural network, and the weights between all dynamic node pairs are updated to obtain the dynamic neural network of the ship ladder.

[0011] Preferably, the mechanical field data, thermal field data, and electromagnetic field data are respectively used as inputs of the convolutional layer in the multi-branch neural network, and the neural network branches of the mechanical field, the thermal field, and the electromagnetic field are output.

[0012] Preferably, the expression of the coupling coefficient matrix between any two physical tensors at the current time step is: Where, Represents the coupling coefficient matrix between the p-th physical tensor and the q-th physical tensor at the current time step; Represents the preset weight matrix at the current time step; Indicates the result of concatenating the p-th physical tensor and the q-th physical tensor at the current time step; Represents the normalization function.

[0013] Preferably, the expression of the fusion characteristic function between any two physical tensors at the current time step is: ; Represents the fusion characteristic function between the p-th physical tensor and the q-th physical tensor at the current time step; Represents the result of the fusion operation on the p-th physical tensor and the q-th physical tensor at the current time step.

[0014] Preferably, the expression of the coupling loss of the initial neural network model at the current time step is: Where, Represents the coupling loss of the initial neural network model at the current time step; Represents the cumulative sum of the second-order mixed partial derivatives of the fusion characteristic functions between all physical tensors at the time step t before the current time step; Represents the gradient of the heat field data at time step t before the current time step; represents the gradient of the electromagnetic field intensity at time step t; k represents a preset value; N represents the number of all time steps before the current time step during the initial neural network model training process; express norm.

[0015] Preferably, updating all weights in the initial neural network model includes:

[0016] The formula for updating the weight between node i and node j in the initial neural network model is: Where, 、 Represent the weights between nodes i and j in the neural network at time step t+1 and time step t respectively; represents the learning rate at time step t, where the initial learning rate is a preset first value; represents the first-order partial derivative of the coupling loss with respect to the weight between node i and node j; represents the error between the output of the initial neural network model at time step t and the actual measured cable state, where the cable state includes: cable position, velocity, and acceleration; express norm; Represents an adaptive learning rate function based on the error norm.

[0017] Preferably, adjusting the topological structure of the initial neural network model includes:

[0018] After the initial neural network training is completed, if the weight between node i and node j in the initial neural network is less than the preset threshold, the connection between node i and node j is deleted. Otherwise, the connection between node i and node j is retained, and the weights between all nodes are traversed to obtain the optimized neural network topology structure.

[0019] Preferably, the step of selecting dynamic node pairs from all pairs of nodes in the optimized neural network includes:

[0020] Calculate the mean and standard deviation of the weights of each pair of nodes in the optimized neural network before the current time step in all iterations, and record the corresponding nodes whose mean is greater than the preset first value and the standard deviation is greater than the preset second value as the dynamic node pair in the optimized neural network at the current time step.

[0021] Preferably, updating the weights between all dynamic node pairs includes:

[0022] The weight adjustment value of the dynamic node pair d at the next moment of the current time step The expression is: Where, Represents the weight between the dynamic node pair d at the current time step; represents the learning rate at the current time step, wherein the initial learning rate in the weight update process between dynamic node pairs is a preset second value; represents the gradient of the weight between the dynamic node pair d at the current moment on the coupling loss L; Represents a constrained projection operator.

[0023] Preferably, obtaining the ship ladder dynamic neural network includes:

[0024] In the optimization neural network, all dynamic node pairs are obtained at different time steps, and the weights of all dynamic node pairs are updated. The neural network model after the optimization neural network training is completed is used as the ship ladder dynamic neural network.

[0025] This application has at least the following beneficial effects:

[0026] This application dynamically adjusts the network layer connection weights according to real-time physical field data, realizes the coordinated evolution of model structure and physical state, and significantly improves the model's adaptability to dynamic environments; further, it introduces the cross-coupling term of mechanical-thermal-electromagnetic fields, quantifies the interaction of multiple physical fields through tensor fusion, reduces prediction deviation, and improves the prediction accuracy of the model; further, it adopts a local weight update strategy. When the environment changes drastically, only the sub-network affected by the environmental change is locally trained without the need to retrain the entire model, which reduces the computational complexity, improves the efficiency and accuracy of online modeling of the ship ladder cable movement, and solves the key technical problem of the lack of efficient and intelligent prediction of the cable movement situation. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present application or the prior art, the following is a brief introduction to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0028] Figure 1 A flowchart of the steps of an online modeling method for the movement of a ship ladder accompanying cable based on a dynamic neural network model provided in one embodiment of the present application;

[0029] Figure 2 A schematic diagram of the process of acquiring a dynamic neural network model of a ship ladder provided in one embodiment of the present application. DETAILED DESCRIPTION

[0030] To further illustrate the technical means and effectiveness of this application to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effectiveness of the online modeling method for the motion of a ship ladder's accompanying cable based on a dynamic neural network model proposed in this application. In the following description, references to different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.

[0031] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

[0032] The specific scheme of the on-line modeling method of the ship ladder accompanying cable movement based on the dynamic neural network model provided by the present application is described in detail below with reference to the accompanying drawings.

[0033] An embodiment of the present application provides an online modeling method for the movement of a ship ladder cable based on a dynamic neural network model. Specifically, the following online modeling method for the movement of a ship ladder cable based on a dynamic neural network model is provided. Figure 1 , the method comprises the following steps:

[0034] Step S1: Real-time acquisition of various physical field data of the cable during the operation of the ship ladder accompanying cable. The various physical field data include: mechanical field, thermal field and electromagnetic field data. The various physical field data are respectively used as inputs of the multi-branch neural network, and various physical tensors are extracted from the neural network branches and transmitted to the fully connected layer for fusion to obtain an initial neural network model. The fusion process includes: obtaining the coupling coefficient matrix between any two physical tensors at the current time step, and fusing the any two physical tensors based on the coupling coefficient matrix to obtain the fusion characteristic function between any two physical tensors at the current time step.

[0035] The neural network used in this embodiment is a convolutional neural network. In actual application, as other implementation methods, implementers may also use other neural networks, and this embodiment does not impose any special restrictions. Among them, convolutional neural networks are well-known technologies, and their specific principles will not be described in detail.

[0036] The input layer of the neural network is responsible for receiving and processing data from multiple sensors, providing a foundation for subsequent model analysis and processing. During actual operation, the ship's accompanying cable is affected by a variety of factors. To fully obtain cable status information, multi-source sensors are required for data collection. In this embodiment, these sensors include:

[0037] Mechanical sensor: used to measure the mechanical stress on the cable and obtain relevant data of the cable under mechanical action in real time.

[0038] Temperature sensors are located at various locations along the cable to monitor temperature changes in real time. Since cables generate heat during operation due to current flow, friction, and other factors, temperature changes significantly impact the performance and lifespan of the cable. Therefore, real-time cable temperature data is collected to provide a basis for analyzing the cable's thermal status.

[0039] Electromagnetic field sensor: used to detect the distribution and changes of the electromagnetic field around the cable, obtain electromagnetic field strength data, and provide data support for analyzing the electromagnetic characteristics of the cable.

[0040] Among them, the acquisition frequency of physical field data in this embodiment is 1 Hz, so in the following neural network training process, each time step is also set to 1s. In actual application, the implementer can also set it by himself according to the specific situation. This embodiment does not impose any special restrictions.

[0041] In order to eliminate the impact of data dimension on subsequent content analysis, all collected data are normalized to obtain normalized values. In this embodiment, the maximum and minimum value normalization method is used to normalize the data. In actual application, as other implementation methods, implementers may also use other normalization methods such as the z-score normalization method. Regarding the selection of normalization method, this embodiment does not impose any special restrictions. Among them, the maximum and minimum value normalization method is a well-known technology, and its specific principles are not repeated here.

[0042] Furthermore, all kinds of physical field data are used as inputs of a multi-branch neural network to obtain neural network branches of various physical fields, specifically: mechanical field data, thermal field data, and electromagnetic field data are respectively used as inputs of the convolutional layer in the multi-branch neural network, and the neural network branches of the mechanical field, the thermal field, and the electromagnetic field are output; further, tensors of various physical field data are extracted from the neural network branches and transmitted to the fully connected layer for fusion to obtain an initial neural network model, wherein, for the sake of ease of expression, the tensors of the physical field data are recorded as physical tensors, that is, the tensors of the mechanical field data, the thermal field data, and the electromagnetic field data layer are recorded as mechanical tensors, thermal tensors, and electromagnetic tensors, respectively.

[0043] Among them, the process of extracting physical tensors from all neural networks of the neural network is a well-known technology, and the specific principle and process will not be repeated here.

[0044] In addition, the specific contents of various neural network branches are as follows:

[0045] (1) Mechanical field neural network branch: Using a 1D convolution layer, the convolution operation is performed on the 1D data by sliding the convolution kernel, which can effectively extract local strain features. The convolution operation can capture local information. In the mechanical stress analysis of the ladder accompanying cable, this means that it can comprehensively consider the strain of the cable within a certain range. For example, when detecting cable bending or stretching, it can integrate the stress change information of adjacent positions, thereby more comprehensively capturing the characteristic pattern of stress distribution.

[0046] After the mechanical tensor from the input layer enters the 1D convolutional layer, the convolution kernel slides point by point over the data, performing convolution calculations on each local region. In this way, the raw mechanical field data is converted into more representative local strain features. These features highlight the deformation characteristics and patterns of the cable under mechanical stress, providing key information for subsequent fusion with other physical field feature points and model analysis. In this example, the convolution kernel size is set to 5 and the stride is set to 1.

[0047] (2) Neural network branch of thermal field: A fully connected layer is used to map the temperature gradient distribution. In this embodiment, the number of neurons is set to 32, which effectively transforms the input temperature data and learns the gradient change characteristics of the temperature at different positions or at different times of the cable, thereby accurately reflecting the distribution of the thermal field.

[0048] The input heat tensor enters the fully connected layer, where each neuron is connected to all elements of the input data and weighted summed. It is then transformed nonlinearly through an activation function. This allows the fully connected layer to map the raw temperature data into a 32-dimensional feature space, where each dimension represents a characteristic of the temperature length, such as the trend of temperature change and temperature differences at different locations. These characteristics are crucial for understanding the thermal state of the cable and its interaction with other physical fields.

[0049] (3) Electromagnetic field neural network branch: Fast Fourier transform is used as the frequency domain change layer to extract the spectral characteristics of noise. In the complex electromagnetic environment where the ship ladder accompanying cable is located, the electromagnetic field data contains rich frequency information. Fast Fourier transform can efficiently convert the electromagnetic signal in the time domain to the frequency domain, revealing the distribution of different frequency components in the signal. By analyzing the characteristic values of the noise spectrum, the frequency characteristics and intensity distribution of electromagnetic interference can be understood, which is very critical for evaluating the operating status and performance of the cable in the electromagnetic environment.

[0050] After the input electromagnetic tensor enters the frequency domain transformation layer, it is processed by the Fast Fourier Transform algorithm, converting the time domain signal into a frequency domain representation. The resulting frequency domain data contains amplitude and phase information at different frequencies. This information constitutes the noise spectrum characteristics of the electromagnetic field, which can intuitively demonstrate the characteristics and changes of the electromagnetic environment around the cable, providing important electromagnetic information for subsequent multi-physics field coupling analysis.

[0051] Furthermore, different types of tensors are fused. The specific process is as follows:

[0052] In the actual operating environment of ship ladder cables, mechanical stress, temperature fluctuations, and electromagnetic interference often interact with each other. Bending or stretching the cable can cause changes in its internal resistance, which in turn affects current distribution and electromagnetic field characteristics. Friction and other factors can also generate heat, changing the cable's temperature. The coupled tensor fusion layer, through this calculation method, fully accounts for the complex coupling relationships between these physical fields, organically integrating the characteristics of each. Compared to traditional methods that model each physical field independently, this method more realistically reflects the actual operating state of the cable, reduces prediction errors caused by ignoring the effects of multi-field coupling, and provides more accurate and effective feature information for subsequent dynamic topology adjustments and cable status prediction.

[0053] In order to integrate different physical field characteristics, the coupling coefficient matrix is defined , where Represents the coupling coefficient matrix between the p-th physical tensor and the q-th physical tensor at the current time step; Represents the preset weight matrix at the current time step; Indicates the result of concatenating the p-th physical tensor and the q-th physical tensor at the current time step; Represents the normalization function.

[0054] It should be noted that the setting process of the preset weight matrix includes: when the neural network training iteration has not started at the initial moment, the value of the weight matrix is randomly set, and subsequently during the neural network training process, the weight matrix will be updated according to back propagation; among them, back propagation is a well-known technology in the neural network training process, and its specific principles will not be repeated.

[0055] In addition, the process of splicing different types of physical tensors is a well-known technology, and the specific process will not be repeated here.

[0056] Furthermore, the arbitrary two physical tensors are fused based on the coupling coefficient matrix to obtain the fusion characteristic function between the arbitrary two physical tensors. The specific calculation formula is:

[0057] The fusion characteristic function between the p-th physical tensor and the q-th physical tensor at the current time step The expression is: ; Represents the result of the fusion operation on the p-th physical tensor and the q-th physical tensor at the current time step.

[0058] Among them, the process of fusing different types of physical tensors in this embodiment needs to be fused according to the characteristics of different physical tensors. The fusion operation is a well-known technology and its specific principles will not be repeated here.

[0059] At this point, the physical tensors are fused to obtain the fused characteristic function.

[0060] Step S2: According to the connection between different physical tensors, the initial neural network model is dynamically topologically updated to obtain an optimized neural network, and the optimized neural network is trained to obtain a ship ladder neural network model.

[0061] S201: Based on the second-order mixed partial derivatives of the fusion characteristic function between any two physical tensors at all time steps before the current moment, and combined with the interaction between the heat tensor and the electromagnetic tensor among all physical tensors, determine the coupling loss of the initial neural network model at the current time step, update all weights in the initial neural network model to adjust the topological structure of the initial neural network model, and obtain an optimized neural network.

[0062] The cable environment is highly variable, and the fusion characteristics of the multi-physics data collected in real time are also constantly changing. By dynamically adjusting the topology based on these fusion characteristics, the model can adapt to these changes in real time, adjusting its structure to better fit the new data. This ensures high prediction accuracy under different operating conditions, such as varying tidal conditions and ship loads.

[0063] Therefore, according to the second-order mixed partial derivatives of the fusion characteristic function between any two physical tensors, and combined with the interaction between the heat tensor and the electromagnetic tensor in all physical tensors, the coupling loss of the initial neural network model is determined. According to the coupling loss, combined with back propagation, the weights in the initial neural network model are dynamically updated, specifically:

[0064] Coupling loss of the initial neural network model at the current time step The expression is: Where, Represents the cumulative sum of the second-order mixed partial derivatives of the fusion characteristic functions between all physical tensors at the time step t before the current time step; Represents the gradient of the heat field data at time step t before the current time step; represents the gradient of the electromagnetic field intensity at time step t; k represents a preset value; N represents the number of all time steps before the current time step during the initial neural network model training process; express norm.

[0065] It should be noted that the value of k is artificially set and is a material constant. In this embodiment, the value of k is 500 MPa. In actual application, the implementer can also set it by himself based on the specific situation. This embodiment does not impose any special restrictions.

[0066] The calculation process of the gradient of the thermal field data and the gradient of the electromagnetic field intensity at time step t are both well-known technologies, and the specific calculation steps are not repeated here; and Norm is a well-known technology, and the specific principle will not be described in detail.

[0067] Furthermore, all weights in the initial neural network model are updated, including:

[0068] The formula for updating the weight between node i and node j in the initial neural network model is: Where, 、 Represent the weights between nodes i and j in the neural network at time step t+1 and time step t respectively; represents the learning rate at time step t, where the initial learning rate is a preset first value; represents the first-order partial derivative of the coupling loss with respect to the weight between node i and node j; represents the error between the output of the initial neural network model at time step t and the actual measured cable state, where the cable state includes: cable position, velocity, and acceleration; express norm; Represents an adaptive learning rate function based on the error norm.

[0069] It should be noted that in this embodiment, the initial learning rate is set The value of is 0.001, that is, the preset first value is 0.001. In the subsequent iteration process, the learning rate will adaptively change during the iteration process. The process of adaptive updating of the learning rate is a well-known technology, and the specific principle will not be repeated here.

[0070] It should be noted that the actual measured cable status includes the position, velocity, and acceleration of the cable, which can be measured by various sensors.

[0071] Among them, the adaptive learning rate function based on the error norm is a well-known technology, and the specific principle is not repeated here.

[0072] Furthermore, after the initial neural network training is completed, if the weight between node i and node j in the initial neural network is less than a preset threshold, it is considered that the connection between nodes i and j is too weak and contributes little to the information transmission and feature learning of the model. The connection between node i and node j is deleted. Otherwise, the connection between node i and node j is retained, and the weights between all nodes are traversed to obtain the updated topological structure of the initial neural network as the optimized neural network.

[0073] It should be noted that the value of the preset threshold is set manually. In this embodiment, the value of the preset threshold is 0.6. In actual application, the implementer can also set it by himself according to the specific situation. This embodiment does not impose any special restrictions.

[0074] S202: Analyze the average distribution and discreteness of all weights between each pair of nodes in the optimized neural network in all time steps before the current time step, select dynamic node pairs from all pairs of nodes in the optimized neural network, and update the weights between all dynamic node pairs to obtain the dynamic neural network of the ship ladder.

[0075] The network weights are divided into static and dynamic blocks. The static block contains weight parameters that are relatively stable during model training and have a significant impact on the overall structure and basic characteristics. These parameters do not require frequent updates during most of the model's runtime because they represent the inherent characteristics and basic laws of the cable under general conditions. The dynamic block contains weight parameters that are easily affected by environmental changes and require timely adjustment based on new data.

[0076] Therefore, we analyze the average distribution and discreteness of all weights between each pair of nodes in the optimized neural network at all time steps before the current time step, and select dynamic node pairs from all pairs of nodes in the optimized neural network, specifically:

[0077] Calculate the mean and standard deviation of the weights of each pair of nodes in the optimized neural network before the current time step in all iterations, and record the corresponding nodes whose mean is greater than the preset first value and the standard deviation is greater than the preset second value as the dynamic node pair in the optimized neural network at the current time step.

[0078] It should be noted that the values of the preset first numerical value and the preset second numerical value are both set manually. In this embodiment, the value of the preset first numerical value is 0.1, and the value of the preset second numerical value is 0.05. The implementer can also set them according to the specific situation. This embodiment does not impose any special restrictions.

[0079] Furthermore, the weights between all dynamic node pairs are updated to obtain the ship ladder dynamic neural network. Specifically:

[0080] The weight adjustment value of the dynamic node pair d at the next moment of the current time step The expression is: Where, Represents the weight between the dynamic node pair d at the current time step; represents the learning rate at the current time step, wherein the initial learning rate in the weight update process between dynamic node pairs is a preset second value; represents the gradient of the weight between the dynamic node pair d at the current moment on the coupling loss L; represents the constraint projection operator; in this embodiment, the feasible domain The range is .

[0081] It should be noted that the value of the initial learning rate in the weight update process between dynamic node pairs is set artificially. In this embodiment, the value of the initial learning rate in the weight update process between dynamic node pairs is set to 0.001, that is, the preset second value is 0.001. The implementer can also set it by himself according to the specific situation. There is no special restriction in this embodiment. In the subsequent iteration process, the learning rate will adaptively change during the iteration process. The process of adaptive update of the learning rate is a well-known technology, and the specific principle will not be repeated.

[0082] Among them, the process of calculating the gradient of the weight on the coupling loss, that is, calculating the gradient of the weight on the loss function and the constrained projection operator are both well-known technologies, and their specific principles are not repeated here.

[0083] In the optimized neural network, all dynamic node pairs are obtained at different time steps, and the weights of all dynamic node pairs are updated. The neural network model after the optimized neural network training is used as the ladder dynamic neural network.

[0084] Preferably, the schematic diagram of the process of obtaining the dynamic neural network model of the ship ladder provided in this embodiment is as follows: Figure 2 shown.

[0085] It should be noted that the order in which the embodiments of the present application are presented is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. Furthermore, the foregoing descriptions of specific embodiments of this specification are provided. Furthermore, the processes depicted in the accompanying drawings do not necessarily require the specific order or sequential sequence shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0086] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments.

[0087] The above-described embodiments are only used to illustrate the technical solutions of the present application, and not to limit them. Modifications to the technical solutions described in the aforementioned embodiments, or equivalent replacements of some of the technical features therein, do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application, and should all be included in the scope of protection of the present application.

Claims

1. The online modeling method of ship ladder cable motion based on dynamic neural network model is characterized by: The method comprises the following steps: Real-time acquisition of various physical field data of the cable during the operation of the ship ladder accompanying cable, including mechanical field, thermal field, and electromagnetic field data. The various physical field data are respectively used as inputs of a multi-branch neural network, and various physical tensors are extracted from the neural network branches and transmitted to the fully connected layer for fusion to obtain an initial neural network model. The fusion process includes: obtaining a coupling coefficient matrix between any two physical tensors at the current time step, and fusing the any two physical tensors based on the coupling coefficient matrix to obtain a fusion characteristic function between the any two physical tensors at the current time step; According to the weights between the nodes in the initial neural network, the initial neural network model is dynamically topologically updated to obtain an optimized neural network. The optimized neural network is then trained to obtain a ship ladder neural network model. The specific process is as follows: Based on the second-order mixed partial derivatives of the fusion characteristic function between any two physical tensors at all time steps before the current moment, and combined with the interaction between the heat tensor and the electromagnetic tensor among all physical tensors, the coupling loss of the initial neural network model at the current time step is determined, and all weights in the initial neural network model are updated to adjust the topological structure of the initial neural network model to obtain an optimized neural network; The average distribution and discreteness of all weights between each pair of nodes in the optimized neural network in all time steps before the current time step are analyzed, dynamic node pairs are selected from all pairs of nodes in the optimized neural network, and the weights between all dynamic node pairs are updated to obtain the dynamic neural network of the ship ladder.

2. The method for online modeling of the movement of the ship ladder accompanying cable based on the dynamic neural network model according to claim 1 is characterized in that: The mechanical field data, thermal field data, and electromagnetic field data are respectively used as the input of the convolutional layer in the multi-branch neural network, and the neural network branch of the mechanical field, the neural network branch of the thermal field, and the neural network branch of the electromagnetic field are output.

3. The method for online modeling of the movement of the ship ladder accompanying cable based on the dynamic neural network model according to claim 1 is characterized in that: The expression of the coupling coefficient matrix between any two physical tensors at the current time step is: Where, Represents the coupling coefficient matrix between the p-th physical tensor and the q-th physical tensor at the current time step; Represents the preset weight matrix at the current time step; Indicates the result of concatenating the p-th physical tensor and the q-th physical tensor at the current time step; Represents the normalization function.

4. The method for online modeling of the movement of the ship ladder accompanying cable based on the dynamic neural network model according to claim 3 is characterized in that: The expression of the fusion characteristic function between any two physical tensors at the current time step is: ; Represents the fusion characteristic function between the p-th physical tensor and the q-th physical tensor at the current time step; Represents the result of the fusion operation on the p-th physical tensor and the q-th physical tensor at the current time step.

5. The method for online modeling of ship ladder cable motion based on a dynamic neural network model according to claim 1, characterized in that: The expression of the coupling loss of the initial neural network model at the current time step is: Where, Represents the coupling loss of the initial neural network model at the current time step; Represents the cumulative sum of the second-order mixed partial derivatives of the fusion characteristic functions between all physical tensors at the time step t before the current time step; Represents the gradient of the heat field data at time step t before the current time step; represents the gradient of the electromagnetic field intensity at time step t; k represents a preset value; N represents the number of all time steps before the current time step during the initial neural network model training process; express norm.

6. The method for online modeling of ship ladder cable motion based on a dynamic neural network model according to claim 1, characterized in that: The updating of all weights in the initial neural network model includes: The formula for updating the weight between node i and node j in the initial neural network model is: Where, 、 Represent the weights between nodes i and j in the neural network at time step t+1 and time step t respectively; represents the learning rate at time step t, where the initial learning rate is a preset first value; represents the first-order partial derivative of the coupling loss with respect to the weight between node i and node j; represents the error between the output of the initial neural network model at time step t and the actual measured cable state, where the cable state includes: cable position, velocity, and acceleration; express norm; Represents an adaptive learning rate function based on the error norm.

7. The method for online modeling of the movement of the ship ladder accompanying cable based on the dynamic neural network model according to claim 1 is characterized in that: The adjusting the topological structure of the initial neural network model includes: After the initial neural network training is completed, if the weight between node i and node j in the initial neural network is less than the preset threshold, the connection between node i and node j is deleted. Otherwise, the connection between node i and node j is retained, and the weights between all nodes are traversed to obtain the optimized neural network topology structure.

8. The method for online modeling of the movement of the ship ladder cable based on the dynamic neural network model according to claim 1 is characterized in that: The step of selecting dynamic node pairs from all pairs of nodes in the optimized neural network includes: Calculate the mean and standard deviation of the weights of each pair of nodes in the optimized neural network before the current time step in all iterations, and record the corresponding nodes whose mean is greater than the preset first value and the standard deviation is greater than the preset second value as the dynamic node pair in the optimized neural network at the current time step.

9. The method for online modeling of ship ladder cable motion based on a dynamic neural network model according to claim 1, characterized in that: The updating of the weights between all dynamic node pairs includes: The weight adjustment value of the dynamic node pair d at the next moment of the current time step The expression is: Where, Represents the weight between the dynamic node pair d at the current time step; represents the learning rate at the current time step, wherein the initial learning rate in the weight update process between dynamic node pairs is a preset second value; represents the gradient of the weight between the dynamic node pair d at the current moment on the coupling loss L; Represents a constrained projection operator.

10. The method for online modeling of ship ladder cable motion based on a dynamic neural network model according to claim 1, characterized in that: The method of obtaining the dynamic neural network of the ship ladder includes: In the optimization neural network, all dynamic node pairs are obtained at different time steps, and the weights of all dynamic node pairs are updated. The neural network model after the optimization neural network training is completed is used as the ship ladder dynamic neural network.

Images