Self-adaptive seabed load optimization method and device

Abstract

The invention provides a self-adaptive seabed load optimization method and device, and relates to the technical field of ocean engineering. The method comprises the following steps: acquiring physical parameters of a seabed soil sample, load distribution of ocean engineering equipment and mechanical response data in an interaction process of the ocean engineering equipment and a seabed; and constructing a seabed mechanical simulation model. Based on the seabed mechanical simulation model and the simulation result data set, training a reinforcement learning model, and generating an optimal load distribution scheme; predicting seabed response results under different load conditions through the trained neural network; and dynamically adjusting the load distribution of the ocean engineering equipment according to the seabed response result and the optimal load distribution scheme. According to the method, real-time self-adaptive optimization of load distribution of the ocean engineering equipment can be realized, the operation stability and efficiency of the ocean engineering equipment are improved, seabed deformation can be remarkably reduced, and the influence on the environment is reduced.

Description

Technical Field

[0001] The present invention relates to the field of marine engineering technology, and in particular to an adaptive seabed load optimization method and device. Background Art

[0002] Marine engineering equipment such as subsea robots, tracked equipment and sleds are widely used in marine resource exploitation and maintenance tasks. Their load distribution directly affects the stress and deformation characteristics of the seabed, which is related to the stability of equipment operation and seabed environmental protection.

[0003] However, existing technologies still have shortcomings in optimizing load distribution and are difficult to cope with complex and changeable seabed environments. Existing seabed mechanical analysis mostly uses static or linear models. Although they are suitable for simple scenarios, they lack accuracy when dealing with the diversity of seabed soil and dynamic operation of equipment. For example, the stress distribution and deformation characteristics of soft soil and hard soil are significantly different, and traditional models are difficult to reflect the nonlinear interaction between equipment and the seabed. In addition, most optimization methods are based on fixed parameter settings and lack real-time adjustment capabilities. They are difficult to adapt to changes in soil parameters and environmental conditions during operation, resulting in reduced equipment operating efficiency and possible damage to the seabed environment. On the other hand, seabed load optimization involves complex mechanical calculations. In multi-variable coupling and dynamic load scenarios, traditional methods have low computational efficiency and poor real-time performance, and cannot quickly respond to actual needs.

[0004] To this end, there is an urgent need for an adaptive optimization method that combines big data, reinforcement learning and neural networks. Through dynamic modeling, real-time optimization and efficient feedback mechanism, intelligent adjustment of load distribution can be achieved, thereby improving equipment operation efficiency, reducing seabed deformation, and providing reliable technical support for complex marine environments. Summary of the invention

[0005] The purpose of the present invention is to provide an adaptive seabed load optimization method and device, which are used to solve the problems that the existing seabed optimization methods cannot reflect the nonlinear interaction between equipment and seabed; lack of real-time adjustment capability, resulting in reduced equipment operating efficiency, and may cause damage to the seabed environment; and the traditional methods have low calculation efficiency, poor real-time performance, and cannot quickly respond to actual needs. The method can realize intelligent adjustment of load distribution, reflect the nonlinear interaction between equipment and seabed, thereby improving equipment operating efficiency, reducing seabed deformation, and providing reliable technical support for complex marine environments.

[0006] In order to achieve the above objectives, in a first aspect, the present invention provides an adaptive seabed load optimization method, comprising: Step 1: Obtain the physical parameters of seabed soil samples, load distribution of marine engineering equipment, and mechanical response data during the interaction between marine engineering equipment and seabed; Step 2: Using physical parameters and load distribution of marine engineering equipment as input and mechanical response data as output, a seabed mechanical simulation model is constructed through simulation software to simulate the interaction between marine engineering equipment and the seabed under different load conditions, and obtain a simulation result data set; Step 3: Define the basic elements of the preset reinforcement learning model, build a virtual training environment based on the seabed mechanics simulation model and the simulation result data set, simulate the operation of marine engineering equipment under different load distribution strategies, train the reinforcement learning model, and optimize the load distribution strategy through multiple interactions and iterations with the virtual training environment to generate the optimal load distribution solution; Step 4: Using physical parameters and load distribution of marine engineering equipment as input parameters and mechanical response data as output parameters, a neural network is constructed, and the neural network is trained using an error back propagation algorithm. The trained neural network predicts the seabed response results under different load conditions; Step 5: Dynamically adjust the load distribution of the marine engineering equipment according to the seabed response results and the optimal load distribution scheme.

[0007] According to an adaptive seabed load optimization method provided by the present invention, the physical parameters include soil density, internal friction angle, cohesion, and compression modulus, and the mechanical response data include seabed stress distribution, contact pressure distribution, and seabed surface deformation.

[0008] According to an adaptive seabed load optimization method provided by the present invention, the marine engineering equipment includes a seabed robot, a crawler device and a sled.

[0009] According to an adaptive seabed load optimization method provided by the present invention, step 1 also includes: cleaning the physical parameter and mechanical response data, removing noise and outliers, extracting key variables, and standardizing the data, and then storing them in a data lake and a data warehouse, and using a distributed storage system to realize data storage and management.

[0010] According to an adaptive seabed load optimization method provided by the present invention, in step 2, a seabed mechanics simulation model is constructed using a discrete element method or a finite element method, and the simulation software uses EDEM or ABAQUS.

[0011] According to an adaptive seabed load optimization method provided by the present invention, in step 3, the basic elements include state space, action space and reward function, the state space is the current marine engineering equipment load distribution, seabed stress distribution, contact pressure distribution, seabed surface deformation, the action space is to adjust the load distribution strategy, and the reward function is to minimize the seabed surface deformation and the energy consumption of the marine engineering equipment.

[0012] According to an adaptive seabed load optimization method provided by the present invention, in step 3, a deep Q learning algorithm or a policy gradient algorithm is used to train a reinforcement learning model.

[0013] According to an adaptive seabed load optimization method provided by the present invention, the neural network includes a multi-layer feedforward neural network or a convolutional neural network.

[0014] According to an adaptive seabed load optimization method provided by the present invention, the prediction accuracy of the neural network is controlled within 5%.

[0015] In a second aspect, the present invention provides an adaptive seabed load optimization device, comprising: The acquisition module is used to obtain the physical parameters of seabed soil samples, the load distribution of marine engineering equipment, and the mechanical response data of the interaction between marine engineering equipment and the seabed; The mechanical simulation module is used to take physical parameters and load distribution of marine engineering equipment as input and mechanical response data as output. It constructs a seabed mechanical simulation model through simulation software, simulates the interaction between marine engineering equipment and the seabed under different load conditions, and obtains a simulation result data set; The reinforcement learning optimization module is used to define the basic elements of the preset reinforcement learning model, build a virtual training environment based on the seabed mechanics simulation model and simulation result data set, simulate the operation of marine engineering equipment under different load distribution strategies, train the reinforcement learning model, and optimize the load distribution strategy through multiple interactions and iterations with the virtual training environment to generate the optimal load distribution plan; The modeling and prediction module is used to construct a neural network with physical parameters and load distribution of marine engineering equipment as input parameters and mechanical response data as output parameters, and train the neural network using the error back propagation algorithm. The trained neural network can predict the seabed response results under different load conditions; The adaptive module is used to dynamically adjust the load distribution of marine engineering equipment according to the seabed response results and the optimal load distribution plan.

[0016] The technical solution of the present invention has at least the following technical effects: The present invention provides an adaptive seabed load optimization method and device, the method comprising: obtaining the physical parameters of seabed soil samples, the load distribution of marine engineering equipment, and the mechanical response data during the interaction between marine engineering equipment and the seabed; and constructing a seabed mechanical simulation model. Based on the seabed mechanical simulation model and the simulation result data set, a reinforcement learning model is trained to generate an optimal load distribution scheme; the seabed response results under different load conditions are predicted by the trained neural network; and the load distribution of the marine engineering equipment is dynamically adjusted according to the seabed response results and the optimal load distribution scheme. The present invention realizes high-precision seabed load optimization and real-time dynamic adjustment through the organic combination of data acquisition and preprocessing, mechanical simulation modeling, reinforcement learning optimization, and neural network prediction. Compared with traditional technologies, it has the characteristics of high modeling accuracy, strong real-time performance, and strong adaptability, and can effectively meet the optimization needs in complex seabed environments, providing strong technical support for the design and operation of marine engineering equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0018] In the attached picture: Figure 1 It is a flow chart of the adaptive seabed load optimization method of the present invention; Figure 2 It is a structural block diagram of the adaptive seabed load optimization device of the present invention. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0020] Some embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.

[0021] See also Figure 1The embodiment of the present invention provides an adaptive seabed load optimization method based on big data, reinforcement learning and neural network, which realizes real-time adaptive optimization of load distribution of marine engineering equipment through the combination of high-precision mechanical modeling and intelligent optimization strategy. It can not only improve the stability and efficiency of equipment operation, but also significantly reduce seabed deformation and reduce the impact on the environment. The present invention is suitable for complex and changeable seabed environments, especially in scenarios where the operating conditions of marine engineering equipment are complex and the load optimization requirements change dynamically. The method includes: Step 1: Obtain the physical parameters of seabed soil samples, load distribution of marine engineering equipment, and mechanical response data during the interaction between marine engineering equipment and seabed; Specifically, the physical parameters of seabed soil samples include soil density, internal friction angle, cohesion, compression modulus, etc., which are used to describe the mechanical properties of soil under different seabed conditions. The mechanical response data during the interaction between marine engineering equipment and the seabed include seabed stress distribution, contact pressure distribution, seabed surface deformation, etc., which are used to characterize the mechanical relationship between the interaction between marine engineering equipment and the seabed. Marine engineering equipment includes submarine robots, tracked equipment, and sleds.

[0022] In some embodiments, step 1 can collect soil samples in different seabed areas through sampling equipment, record the geographical location and basic characteristics of the soil samples, and thus obtain the physical parameters of the seabed soil samples. Subsequently, the sampling equipment uses pressure sensors, displacement sensors, strain sensors, etc. to collect real-time mechanical response data during the interaction between marine engineering equipment and the seabed. To ensure data quality, the collected raw data needs to be preprocessed, that is, step 1 also includes: cleaning the physical parameters and mechanical response data, removing noise and outliers, and ensuring the accuracy and reliability of the data. Next, key variables such as stress distribution, load changes, and seabed surface deformation are extracted from the cleaned data through feature extraction technology for subsequent analysis and modeling. Then, data standardization is performed to uniformly convert the extracted feature data into a standard format and range to eliminate the impact of different dimensions on model training. The standardized data is transmitted to the storage management module and stored in the data lake and data warehouse, and distributed storage systems (such as Hadoop HDFS and Amazon S3) are used to achieve efficient storage and management of data. Finally, all pre-processed data are integrated to generate high-quality data sets in a unified format, providing data support for subsequent mechanical simulation, reinforcement learning optimization, and neural network prediction. The entire process makes full use of big data processing tools (such as Apache Spark) and distributed storage technology to ensure the efficiency and scalability of data processing.

[0023] It should be noted that data lakes can store raw and semi-structured data and support a variety of data types and formats; data warehouses can store structured and processed data to optimize query performance and data analysis efficiency; distributed storage systems can provide high availability and fault tolerance of data and support linear expansion of data. Through the ETL (Extract, Transform, Load) process, using big data processing tools such as Apache Spark, data is extracted, transformed and loaded from the data lake into the data warehouse for data cleaning, conversion and aggregation operations, and then data access interfaces, including APIs and query tools, are provided for subsequent application modules (such as seabed mechanics simulation models, reinforcement learning model optimization, neural network construction and prediction) to achieve rapid data acquisition and sharing. The entire big data storage and management architecture ensures centralized management, rapid access and efficient processing of data through efficient data flow and distributed computing technology, giving full play to the key role of big data technology in the storage management module.

[0024] Step 2: Using physical parameters and load distribution of marine engineering equipment as input and mechanical response data as output, a seabed mechanical simulation model is constructed through simulation software to simulate the interaction between marine engineering equipment and the seabed under different load conditions, and obtain a simulation result data set; Specifically, simulation software (such as EDEM, ABAQUS) can be used to build a seabed mechanics simulation model based on the discrete element method (DEM) or the finite element method (FEM), simulate the interaction between marine engineering equipment and the seabed under different load conditions, obtain a simulation result data set, and store the simulation result data set in the storage management module.

[0025] In some embodiments, first, according to the seabed soil and equipment characteristics, a suitable simulation method, such as discrete element method or finite element method, is selected. Then, the input of the seabed mechanics simulation model is set, including soil density, internal friction angle, cohesion, compression modulus, and load distribution of marine engineering equipment (such as load distribution of tracks and sleds). Next, a seabed mechanics simulation model is constructed using simulation software (such as EDEM, ABAQUS) to simulate the interaction process between marine engineering equipment and the seabed. After the construction of the seabed mechanics simulation model is completed, the seabed mechanics simulation model is run to simulate key mechanical responses such as seabed stress distribution, contact pressure distribution, and seabed surface deformation under different load conditions, and the simulation results are obtained. Subsequently, the simulation results are compared with the actual data, and the seabed mechanics simulation model is calibrated and verified to ensure the accuracy and reliability of the seabed mechanics simulation model. If the seabed mechanics simulation model fails to pass the calibration, the input needs to be adjusted until it passes the calibration and verification, and a simulation result data set covering multiple working conditions is generated, and it is stored in the storage management module for subsequent reinforcement learning model optimization and neural network modeling and prediction. The entire process utilizes distributed computing frameworks (such as Apache Spark) and big data storage technologies to achieve efficient processing and management of large amounts of simulation data, ensuring the high accuracy and reliability of the seabed mechanics simulation model.

[0026] By adjusting the input, the seabed mechanics simulation model can simulate the mechanical response under different seabed conditions (such as soft soil, hard soil or multi-layer composite soil environment) and generate simulation data sets covering various working conditions. Furthermore, the seabed mechanics simulation model is verified by calibration and comparison with actual data to ensure that the results of the model output are accurate and reliable. The seabed mechanics simulation model provides rich sample support for the training of subsequent reinforcement learning models and is the theoretical basis for achieving load distribution optimization.

[0027] Step 3: Define the basic elements of the preset reinforcement learning model, build a virtual training environment based on the seabed mechanics simulation model and the simulation result data set, simulate the operation of marine engineering equipment under different load distribution strategies, train the reinforcement learning model, and optimize the load distribution strategy through multiple interactions and iterations with the virtual training environment to generate the optimal load distribution solution; Specifically, the basic elements of the reinforcement learning model include state space, action space and reward function. The state space is the current load distribution of marine engineering equipment, seabed stress distribution, contact pressure distribution, and seabed surface deformation. The action space is the strategy for adjusting the load distribution. The reward function is to minimize the seabed surface deformation and the energy consumption of marine engineering equipment. The reinforcement learning model can be trained using the Deep Q-learning algorithm or the policy gradient algorithm.

[0028] In some embodiments, first, the basic elements of a preset reinforcement learning model are defined. Then, based on the seabed mechanics simulation model and the simulation result data set, a virtual training environment is constructed to simulate the operation of marine engineering equipment under different load distribution strategies. Then, a deep Q learning algorithm or a policy gradient algorithm is used to train the reinforcement learning model, and through multiple interactions and iterations with the virtual training environment, the load distribution strategy is optimized, and the performance of the load distribution strategy is gradually improved, and finally the optimal load distribution scheme is obtained. The optimized load distribution strategy is applied to the load adjustment of actual marine engineering equipment, and the load distribution strategy is continuously optimized through a real-time actual data feedback mechanism to form a closed-loop iterative optimization process. The entire reinforcement learning optimization process utilizes massive simulation data in big data storage to ensure the comprehensiveness of model training and the efficiency of strategy optimization, and give full play to the key role of big data technology in data-driven decision-making. Through the adaptive ability of reinforcement learning, the present invention can dynamically adjust the load distribution parameters to cope with complex and changeable seabed environments.

[0029] Step 4: Using physical parameters and load distribution of marine engineering equipment as input parameters and mechanical response data as output parameters, a neural network is constructed, and the neural network is trained using an error back propagation algorithm. The trained neural network predicts the seabed response results under different load conditions; Specifically, the neural network includes a multi-layer feedforward neural network or a convolutional neural network.

[0030] In some embodiments, a neural network is first constructed, including an input layer, multiple hidden layers, and an output layer. The input layer is used to receive the physical parameters of the seabed soil sample and the load distribution of the marine engineering equipment; the output layer is used to generate the seabed response results under different load conditions, including the seabed stress distribution, contact pressure distribution, and the prediction results of the seabed surface deformation. Then, the neural network is trained using the error back propagation algorithm to optimize the parameters of the neural network, and a regularization term is introduced to avoid overfitting, so as to ensure the generalization ability and prediction accuracy of the neural network. After the training is completed, the neural network is verified using an independent verification data set (actual data) to ensure that the prediction accuracy error of the neural network is controlled within 5%. If the neural network performs poorly, it is necessary to return to the training for parameter adjustment. After verification, a neural network that has been trained and verified is obtained, and rapid prediction of the seabed response results under different load conditions is achieved. The entire neural network modeling and prediction process uses massive simulation data in big data storage, combined with distributed computing frameworks (such as Apache Spark and TensorFlow distributed training), to achieve efficient learning and accurate prediction of complex nonlinear relationships, which fully reflects the key role of big data technology in deep learning model training and optimization.

[0031] Preferably, through training, the prediction error of the neural network is controlled within 5%, and the computational efficiency is improved by dozens of times compared with traditional mechanical simulation. The neural network provides an efficient and reliable computing tool for real-time optimization, significantly shortening the response time of load optimization.

[0032] Step 5: Dynamically adjust the load distribution of the marine engineering equipment according to the seabed response results and the optimal load distribution scheme.

[0033] It should be noted that the load distribution strategy of marine engineering equipment is dynamically adjusted according to the predicted seabed response results and the optimal load distribution scheme to ensure that the marine engineering equipment is always in the optimal operating state. In addition, through the closed-loop feedback mechanism, the operation strategy of marine engineering equipment can be automatically adjusted according to real-time actual data to adapt to changes in the seabed soil, path and other environments, reduce damage to the seabed and improve the operating stability of marine engineering equipment. Multi-objective optimization can be achieved, which can not only minimize seabed deformation and reduce environmental impact, but also maximize energy utilization efficiency, thereby improving overall operating efficiency. Combining the advantages of neural networks and reinforcement learning models, it can respond in real time in complex dynamic environments to ensure efficient and stable operation of marine engineering equipment under different working conditions.

[0034] Based on the same inventive concept, another embodiment of the present invention provides an adaptive seabed load optimization device, which corresponds to the method of the above embodiment, such as Figure 2 As shown, the device comprises: The acquisition module is used to obtain the physical parameters of seabed soil samples, the load distribution of marine engineering equipment, and the mechanical response data of the interaction between marine engineering equipment and the seabed; The mechanical simulation module is used to take physical parameters and load distribution of marine engineering equipment as input and mechanical response data as output. It constructs a seabed mechanical simulation model through simulation software, simulates the interaction between marine engineering equipment and the seabed under different load conditions, and obtains a simulation result data set; The reinforcement learning optimization module is used to define the basic elements of the preset reinforcement learning model, build a virtual training environment based on the seabed mechanics simulation model and simulation result data set, simulate the operation of marine engineering equipment under different load distribution strategies, train the reinforcement learning model, and optimize the load distribution strategy through multiple interactions and iterations with the virtual training environment to generate the optimal load distribution plan; The modeling and prediction module is used to construct a neural network with physical parameters and load distribution of marine engineering equipment as input parameters and mechanical response data as output parameters, and train the neural network using the error back propagation algorithm. The trained neural network can predict the seabed response results under different load conditions; The adaptive module is used to dynamically adjust the load distribution of marine engineering equipment according to the seabed response results and the optimal load distribution plan.

[0035] In summary, the adaptive seabed load optimization method and device provided by the present invention integrates big data, reinforcement learning and neural network technology, aiming to solve the load distribution optimization problem when marine engineering equipment interacts with the seabed, reduce seabed deformation, and improve the operational stability and adaptability of marine engineering equipment. The method comprises: first, obtaining seabed soil samples, classifying different types of seabed soil, detecting their physical parameters and mechanical response data, and preprocessing the collected data in combination with big data technology, including data cleaning, feature extraction and standardization, to form a high-quality training and analysis data set. Based on the collected data, a seabed mechanical simulation model is constructed using simulation software to simulate the stress distribution, deformation law and interaction process between the equipment and the seabed of marine engineering equipment such as crawlers and sleds under the action of the seabed, obtain a large amount of mechanical response data through simulation analysis, and provide basic data support for model training and optimization. On this basis, a reinforcement learning algorithm is introduced, and the state, action and reward function are designed. Based on the parameters such as the load distribution of marine engineering equipment and the deformation of the seabed as optimization indicators, the reinforcement learning model and the simulation feedback mechanism are used to adaptively adjust the load parameters, and the operating state of the marine engineering equipment is continuously optimized. In order to further improve the computational efficiency and prediction accuracy of the model, neural network technology is combined to build a neural network for prediction by training simulation data sets, learning the nonlinear mapping relationship between load distribution and seabed response, and realizing rapid prediction of seabed mechanical behavior under different load conditions, reducing the time consumption of high-cost simulation calculations. The neural network is combined with the reinforcement learning algorithm to form an efficient adaptive optimization system, which can dynamically adjust the load parameters of marine engineering equipment according to real-time feedback data, ensure the optimal load distribution in a complex and dynamic seabed environment, and minimize seabed deformation and energy loss.

[0036] The present invention breaks through the limitations of traditional mechanical models and manual parameter optimization, and has high precision, high adaptability and real-time response capabilities. It is particularly suitable for engineering scenarios with diverse seabed soils and complex and changeable marine environments. It provides an advanced and reliable technical means for the design and application of marine engineering equipment such as submarine robots, tracked equipment, and sleds, and has broad engineering application prospects.

[0037] Those skilled in the art will readily appreciate other embodiments of the present invention after considering the specification and practicing the embodiments disclosed herein. The present invention is intended to cover any variations, uses or adaptations of the present invention that follow the general principles of the present invention and include common knowledge or customary techniques in the art that are not disclosed by the present invention. It should be understood that the present invention is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from the scope thereof. The scope of the present invention is limited only by the appended claims.

Claims

1. An adaptive seabed load optimization method, characterized in that: include: Step 1: Obtain the physical parameters of seabed soil samples, load distribution of marine engineering equipment, and mechanical response data during the interaction between marine engineering equipment and seabed; Step 2: using the physical parameters and the load distribution of the marine engineering equipment as input and the mechanical response data as output, constructing a seabed mechanical simulation model through simulation software, simulating the interaction between the marine engineering equipment and the seabed under different load conditions, and obtaining a simulation result data set; Step 3: define the basic elements of the preset reinforcement learning model, build a virtual training environment based on the seabed mechanics simulation model and the simulation result data set, simulate the operation of the marine engineering equipment under different load distribution strategies, train the reinforcement learning model, optimize the load distribution strategy through multiple interactions and iterations with the virtual training environment, and generate an optimal load distribution solution; Step 4: Using the physical parameters and the load distribution of the marine engineering equipment as input parameters and the mechanical response data as output parameters, a neural network is constructed, and the neural network is trained using an error back propagation algorithm, and the seabed response results under different load conditions are predicted by the trained neural network; Step 5: dynamically adjust the load distribution of the marine engineering equipment according to the seabed response result and the optimal load distribution scheme.

2. The adaptive seabed load optimization method according to claim 1, characterized in that: The physical parameters include soil density, internal friction angle, cohesion, and compression modulus; the mechanical response data include seabed stress distribution, contact pressure distribution, and seabed surface deformation.

3. The adaptive seabed load optimization method according to claim 1, characterized in that: The marine engineering equipment includes a seabed robot, a crawler device and a sled.

4. The adaptive seabed load optimization method according to claim 2, characterized in that: The step 1 also includes: cleaning the physical parameter and mechanical response data, removing noise and outliers, extracting key variables, and standardizing the data, and then storing them in a data lake and a data warehouse, and using a distributed storage system to achieve data storage and management.

5. The adaptive seabed load optimization method according to claim 1, characterized in that: In the step 2, a discrete element method or a finite element method is used to construct a seabed mechanics simulation model, and the simulation software is EDEM or ABAQUS.

6. The adaptive seabed load optimization method according to claim 2, characterized in that: In step 3, the basic elements include state space, action space and reward function, the state space is the current load distribution of marine engineering equipment, seabed stress distribution, contact pressure distribution, and seabed surface deformation, the action space is the strategy for adjusting the load distribution, and the reward function is to minimize the seabed surface deformation and the energy consumption of marine engineering equipment.

7. The adaptive seabed load optimization method according to claim 6, characterized in that: In step 3, a deep Q learning algorithm or a policy gradient algorithm is used to train the reinforcement learning model.

8. The adaptive seabed load optimization method according to claim 1, characterized in that: The neural network includes a multi-layer feedforward neural network or a convolutional neural network.

9. The adaptive seabed load optimization method according to claim 1, characterized in that: The prediction accuracy of the neural network is controlled within 5%.

10. An adaptive seabed load optimization device, characterized in that: include: The acquisition module is used to obtain the physical parameters of seabed soil samples, the load distribution of marine engineering equipment, and the mechanical response data of the interaction between marine engineering equipment and the seabed; A mechanical simulation module is used to take the physical parameters and the load distribution of the marine engineering equipment as input and the mechanical response data as output, to construct a seabed mechanical simulation model through simulation software, to simulate the interaction between the marine engineering equipment and the seabed under different load conditions, and to obtain a simulation result data set; A reinforcement learning optimization module is used to define basic elements of a preset reinforcement learning model, construct a virtual training environment based on the seabed mechanics simulation model and the simulation result data set, simulate the operation of the marine engineering equipment under different load distribution strategies, train the reinforcement learning model, optimize the load distribution strategy through multiple interactions and iterations with the virtual training environment, and generate an optimal load distribution solution; A modeling and prediction module is used to construct a neural network using the physical parameters and the load distribution of the marine engineering equipment as input parameters and the mechanical response data as output parameters, train the neural network using an error back propagation algorithm, and predict the seabed response results under different load conditions through the trained neural network; The adaptive module is used to dynamically adjust the load distribution of the marine engineering equipment according to the seabed response result and the optimal load distribution scheme.

Images