Deep sea mining equipment design system based on improved genetic algorithm

Abstract

The application relates to the technical field of deep-sea mining equipment, and discloses a deep-sea mining equipment design system based on an improved genetic algorithm, which can improve the design efficiency and accuracy of the deep-sea mining equipment, simplifies the coupling strength between disciplines, improves the optimization efficiency while ensuring the accuracy, and comprises a data input layer, a discipline decoupling layer, an optimization analysis layer, an intelligent solving layer and an operation interaction layer.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of deep-sea mining equipment, in particular, to a deep-sea mining equipment design system based on an improved genetic algorithm. BACKGROUND

[0002] Deep-sea mining equipment design involves multi-disciplinary coupling problems such as structure, power, energy consumption, and environmental adaptability. Decoupling between multiple disciplines is difficult, and there is currently no systematic solution theory and method. To ensure the safety of the equipment, the existing design usually follows the conservative "functionality first" principle, resulting in a lack of overall design consideration in the design of deep-sea mining equipment, especially deep-sea mining vehicles. This makes the landed mining equipment have a high degree of redundancy or defects, significantly increasing the cost of manufacturing and maintenance of mining equipment, affecting the economy of mining activities, and seriously hindering the development of deep-sea mining.

[0003] Specifically, there are the following key problems:

[0004] 1. The complex operating environment of the deep sea requires that mining equipment must consider multiple external factors during design, combined with the complexity of the structure and materials of the mining equipment, ultimately resulting in the need to input a large number of constant parameters to build the underlying environment. On the one hand, the deep-sea environment, especially the seabed, has complex sediment and seawater characteristics, making it difficult to obtain constant parameters. On the other hand, there are numerous types of constant parameters, lack of standardized templates and automated verification mechanisms, and inconsistent data semantics across disciplines, which can cause model conflicts, making it difficult to coordinate and exchange between multiple design software, and significantly reducing design efficiency.

[0005] 2. The multi-disciplinary coupling in the design of deep-sea mining equipment is complex, and existing tools rely on fixed decoupling strategies, lack quantitative analysis of the dynamic coupling strength between disciplines, and are difficult to adjust the decoupling priority. The complex coupling relationship directly leads to long model solving time and increased computational cost.

[0006] 3. In the design of deep-sea mining equipment, different systems and disciplines have different requirements for the accuracy of the established models. If they are uniformly processed according to low precision, the design quality will be affected. If they are uniformly simulated according to high precision, the workload of solving will be significantly increased. Currently, there is a lack of mixed optimization strategy of high-precision simulation and low-precision proxy model, and existing technologies cannot verify the convergence and accuracy of the model from a global perspective, affecting the credibility of the optimization results.

[0007] 4. Existing optimization algorithms are mainly NSGA-II and NCGA. In deep-sea equipment design, NSGA-II, in high-dimensional multi-objective (4 or more) optimization, may suffer from a lack of global diversity due to the local density of adjacent solutions. As the number of objectives increases, the computational complexity grows exponentially, leading to uneven distribution of the solution set and a tendency to get trapped in local optima. NCGA requires preset niche radius or density thresholds, and parameter selection has a significant impact on the results. Furthermore, it is difficult to effectively delineate niche regions in high-dimensional problems. Both of these algorithms have limited search capabilities in high-dimensional problems, requiring multiple iterations to converge, resulting in high computational costs. In addition, their dynamic constraint adjustment capabilities are insufficient, making it difficult to balance convergence speed and diversity, especially in scenarios with strong multidisciplinary coupling, which can easily lead to iterative oscillations or non-convergence. Summary of the Invention

[0008] One of the objectives of this invention is to propose a deep-sea mining equipment design system based on an improved genetic algorithm, which can improve the design efficiency and accuracy of deep-sea mining equipment, simplify the coupling between disciplines, and improve optimization efficiency while ensuring accuracy.

[0009] The technical solution of the present invention is as follows:

[0010] A deep-sea mining equipment design system based on an improved genetic algorithm includes:

[0011] Data Input Layer: Used to configure templated parameter ports, allowing users to define parameterized input templates and set input constants, input variables, constraints, and optimization objectives;

[0012] Subject decoupling layer: includes subject establishment module and coupling analysis module, responsible for the integration and coupling of various subject models;

[0013] Optimization and analysis layer: This includes the experimental design module and the data modeling module, which are responsible for executing experimental design and optimization algorithms and supporting user modeling decisions;

[0014] Intelligent solution layer: Adopts a multi-stage optimization principle, optimizes the objectives in stages, first optimizes a single objective and then expands to multiple objectives; designs optimization strategies for the surrogate model, can automatically perform experimental design sampling and update the surrogate model, and automatically call the quasi-Newton method (L-BFGS), SPEA2, and improved NSGA-III optimization algorithms.

[0015] Operational Interaction Layer: Utilizing PySide interface design technology to provide UI controls and functions, ensuring efficient operation through multidisciplinary design optimization. It efficiently manages the thread pool using QThreadPool and QRunnable technologies, introduces a cloud data storage module for parallel processing to avoid frequent thread creation and destruction, and supports collaborative editing and conflict detection by experts in mechanical, control, and environmental fields.

[0016] Further, the data input layer sets the main input constant parameter and the secondary input constant parameter as input constants;

[0017] The input variables are divided into structural design variables, power and energy design variables, and support system variables;

[0018] The structural design variables are divided into length, width, height, external features, topological structure, local structure freedom, and connector strength redundancy of mining equipment;

[0019] The power and energy design variables are divided into propulsion device power and layout, battery capacity, and energy distribution strategy;

[0020] The support system variables are divided into umbilical cable armor thickness and maximum load of deployment and recovery device;

[0021] The constraint conditions are divided into behavior constraints, strength constraints, and discipline consistency constraints;

[0022] The optimization objectives include the weight and distribution of deep-sea mining equipment, travel-recovery resistance, and operation energy consumption.

[0023] Further, the data input layer includes the following operation steps:

[0024] S110: The main input constant parameter is input by the designer;

[0025] S120: A deep-sea environment database is established, the secondary input constant parameter is input, and the automatic prediction and input of the multi-element characteristic parameters of seawater, sediments, and nodules at different depths in the collection area are completed based on the deep-sea environment function of the polymetallic nodule mining area with the main input constant parameter as the independent variable;

[0026] S130: The secondary constant range check and logic consistency verification are performed by combining machine learning embedded automatic verification rules.

[0027] Further, step S130 includes:

[0028] S131: After the user inputs the main constant parameter, a pre-trained random forest model is called to predict the legality probability p of the input constant parameter combination; if the legality probability p is greater than or equal to the threshold value of 0.9, the test is passed;

[0029] If the legality probability p is less than the threshold value of 0.9, the range check rule library comparison is triggered, the out-of-limit or parameter mismatch is prompted, and the recalculation or manual review process is triggered;

[0030] S132: After the automatic prediction of the secondary input constant by the environment function, an isolation forest model is used to calculate the anomaly score q; if the anomaly score q is less than or equal to 0.6, the parameters are considered reasonable;

[0031] If the abnormal score q>0.6, start logical consistency verification, check parameter relevance, generate a visual report for complex abnormalities, highlight conflicting parameters and recommend correction suggestions, and realize visual manual review and error correction of parameters;

[0032] S133: The correlation feature is that the hydrostatic pressure has a positive correlation with the seawater depth, and the approved pressure depth ratio is 0.01 MPa / m; the seabed temperature is 5-25℃ at the surface-2000m, which is adjusted according to the sea environment and air temperature of the mining area; 2-4℃ at 2000-3500m, 1-2℃ at 3500-4000m, and 4000-6000m slightly increases due to the adiabatic self-pressure effect of seawater, which is adjusted according to the statistical status of seabed hydrothermal of the mining area;

[0033] The seawater salinity tends to be uniform after exceeding the halocline, about 34‰-35‰;

[0034] The seawater density increases with the increase of depth, 4000-6000m is 1025-1030kg / m3, the seawater viscosity increases with the decrease of temperature and slightly increases with the increase of salinity, 4000-6000m viscosity is 1.37-1.69×10⁻³Pa·s;

[0035] 4000-6000m ocean current velocity is affected by density difference and geostrophic flow, between 3cm / s-30cm / s.

[0036] Further, the discipline decoupling layer includes the following running steps:

[0037] S210: The deep sea mining equipment main body is divided into structural discipline, configuration discipline, propulsion discipline and energy consumption discipline by the discipline establishment module;

[0038] S220: The discipline establishment module determines the data flow of constant and input variable transmission to each discipline and system layer by analyzing the characteristics of each component;

[0039] S230: Sobol index and correlation coefficient matrix are introduced for the coupling analysis module to quantify the coupling strength between disciplines, Sobol index<0.2 is defined as weakly coupled discipline, Sobol index 0.2-0.4 is defined as moderately coupled discipline, and Sobol index>0.4 is defined as strongly coupled discipline;

[0040] The correlation coefficient between input variables |r|>0.5 is considered to be strongly correlated, and |r|<0.5 is considered to be weakly correlated;

[0041] High Sobol index or strongly correlated disciplines are preferentially processed, and discipline coupling structure proxy model is established according to sensitive parameters to guide decoupling priority;

[0042] Jacobi iteration is used for weakly coupled disciplines and is executed in parallel, while serial coordination is used for strongly coupled disciplines.

[0043] Further, the step S220 comprises:

[0044] S221: In the design, the structure discipline transmits the equipment geometry, topology constraints to the configuration discipline, the transmission factor is set to 0.65-0.85, transmits the structure weight and distribution to the propulsion discipline, the transmission factor is set to 0.85-0.95, transmits the weight and distribution to the energy consumption discipline, the transmission factor is set to 0.85-0.95;

[0045] S222: The configuration discipline transmits the layout feasibility feedback to the structure discipline, the transmission factor is set to 0.55-0.65, transmits the fluid dynamics shape parameters, propulsion benefits to the propulsion discipline, the transmission factor is set to 0.45-0.55, transmits the equipment space layout, energy distribution strategy to the energy consumption discipline, the transmission factor is set to 0.75-0.90;

[0046] S223: The propulsion discipline transmits the structure strength requirement of the propeller installation position to the structure discipline, the transmission factor is set to 0.68-0.78, transmits the space occupation constraint to the configuration discipline, the transmission factor is set to 0.85-0.95, transmits the real-time power demand curve, power consumption characteristics to the energy consumption discipline, the transmission factor is set to 0.8-0.9;

[0047] S224: The energy consumption discipline transmits the sensitivity of energy consumption to weight distribution to the structure discipline, the transmission factor is set to 0.65-0.75, transmits the cable energy supply equipment, energy consumption equipment layout to the configuration discipline, the transmission factor is set to 0.35-0.65, transmits the maximum allowable power threshold under the energy consumption constraint to the propulsion discipline, the transmission factor is set to 0.85-0.95.

[0048] Further, the optimization analysis layer supports the integration of process models constructed by Isight, ModelCenter, Tosca, Optistruct;

[0049] And set the text parser, system command executor, calculator, and finite element software special interface components;

[0050] The test design module mainly carries out deep-sea equipment structural mechanics test, deep-sea equipment corrosion resistance evaluation test, fatigue test, deep-sea equipment operation hydrodynamic test, flow field simulation and analysis test; different test design strategies are adopted according to different optimization target quantities and coupling relationships; when the optimization target quantity n is less than 3 and the optimization targets are not located in the same discipline or the mutual influence coefficient c is less than 0.3, full-factor design test is adopted; when the optimization target quantity n is less than 3 and the optimization targets are located in the same discipline or the mutual influence coefficient c is greater than 0.3, orthogonal low workload design test is adopted; when the optimization target quantity n is greater than 3, Latin hypercube or central composite design test is adopted;

[0051] The data modeling module is provided with a parameterized integrated interface of commercial software or self-programmed programs for multidisciplinary design simulation, including SolidWorks, ANSYS, LongRuan4D-GIS, OpenFOAM, digital twin interactive control platform and Simulink; parameterized integration and automatic calling of self-programmed programs and commercial CAD / CAE programs are realized;

[0052] The data modeling module provides polynomial response surface model algorithm, and has an interface of proxy model algorithm, based on which external proxy model algorithm can be embedded into the platform; multi-fidelity optimization strategy is adopted, high-precision simulation and low-precision proxy model are mixed, high-precision simulation prediction accounts for 65%-75% of the global target weight, low-precision proxy model prediction accounts for 25%-35% of the global target weight, and when the deviation d of low-precision proxy model and high-precision simulation result exceeds 12%, the weight of low-precision proxy model is reduced to 0.1;

[0053] According to the sensitivity of each system unit to disciplines, only high-sensitivity parameters are optimized with high precision, high-precision whole-system simulation is performed once every 5 local optimizations, and the overall convergence is based on the fact that the change rate of the objective function is less than 1% for three consecutive iterations and the cross-disciplinary constraint violation is less than 5%.

[0054] Further, the data modeling module includes multidisciplinary design:

[0055] High-precision simulation is preferentially established for structural disciplines in detailed stress, strain and fatigue analysis of three-dimensional geometry, plasticity, creep and anisotropy of materials, fluid load and structural response; low-precision proxy model is preferentially adopted for fast statics analysis of bar and shell elements;

[0056] High-precision simulation modeling is performed on the propeller auxiliary propulsion device by the propulsion discipline, and the preliminary relationship between propulsion power and propulsion speed adopts a low-precision proxy model;

[0057] High-precision simulation is adopted for dynamic energy consumption simulation during the load change process of start-up / impact working conditions by the energy consumption discipline, and low-precision proxy model is adopted for preliminary prediction of energy consumption per unit of mining quantity.

[0058] The configuration discipline adopts a low-precision surrogate model in the preliminary optimization stage of overall size, arrangement and center of gravity, and adopts a high-precision simulation in the configuration driving-recovery resistance analysis.

[0059] Further, the improved NSGA-III optimization algorithm in the intelligent solving layer comprises the following steps:

[0060] S310: initializing a population: generating an initial near-optimal and diverse population by using chaotic Logistic mapping based on historical data and a set of good points;

[0061] S320: evaluating fitness: evaluating each individual and calculating its fitness value under multiple optimization objectives;

[0062] S330: fast non-dominated sorting: saving the dominated and dominated number sets, sorting all individuals in the dominated number set, and regarding an individual as a non-dominated individual if the individual domination number is zero, and setting the Pareto level to the current highest level plus one;

[0063] S340: calculating crowding degree: calculating the crowding degree of each individual to maintain the diversity of the population for subsequent elite selection;

[0064] S350: generating offspring: dynamically adjusting parameters according to population diversity, generating the next generation population through genetic operations, reserving 5% to 10% of elite individuals to ensure population diversity and convergence, performing local disturbance and acceptance criterion judgment on the generated offspring individuals, selecting the stage by replacing roulette selection with simulated annealing acceptance probability, setting the initial temperature of simulated annealing to 100 to 300, and setting the temperature decay coefficient to 0.90 to 0.99;

[0065] S360: merging the population: merging the parent and offspring populations to form a new population;

[0066] S370: elite selection: dividing the merged and mixed population into 3 to 6 subgroups, selecting individuals with high fitness and low crowding degree as elite individuals of the next generation population, preferentially reserving feasible solutions, and gradually relaxing the constraint tolerance by using a dynamic constraint method, and merging after parallel distributed optimization;

[0067] S380: repeating iteration: repeating the above steps until the set stop condition is reached;

[0068] S390: obtaining Pareto optimal solution: the individuals in the final population are the Pareto optimal solution set, and each solution cannot be completely dominated by other solutions in all objectives.

[0069] Further, the data input layer adopts HDFS and MinIO as a distributed storage strategy and form, utilizes a Memcached memory database and an efficient index strategy, and establishes a Spark computing framework;

[0070] The data input layer dynamically generates a data structure, adopts a JSON general format, a text format, ontology definition data semantics and formulates cross-disciplinary data standards, classifies data by recording data sources, types and units to form a data attribution list, and establishes a metadata framework;

[0071] The data input layer sets multiple weights and gradient constraints in the constraint condition, and adopts a multi-level constraint limiting strategy of priority safety>environment>cost>energy consumption optimization;

[0072] The coupling analysis module of the discipline decoupling layer is provided with a random forest classification model, which is trained in advance by historical optimization data to identify coupling parameter combinations prone to oscillation or non-convergence; at the same time, a real-time monitoring panel is provided to display coupling residuals, iteration times, discipline model calling states, and built-in benchmark cases of typical coupling problems for verifying decoupling algorithms and performance tuning;

[0073] The model of the data modeling module of the optimization analysis layer is set according to the complexity and nonlinearity of the test data;

[0074] The intelligent solving layer sets a quasi-Newton method algorithm for solving a single-objective optimization, sets a SPEA2 algorithm for optimizing a multi-objective optimization model with 2-3 objectives, and when the number of objectives exceeds 3, an improved NSGA-III algorithm is introduced;

[0075] The working interface of the operation interaction layer is divided into three interfaces of design, monitoring and output: the design interface includes a classification tree structure component library, a flow structure tree, a global / local variable management area, a parameter mapping relationship management area, a priority setting area and a resource quota selection area; the monitoring interface includes a task timeline view area and a progress monitoring column; and the output interface includes a multi-objective optimization result chart area, a data sandbox area and a key parameter influence and optimal scheme recommendation report generation area.

[0076] The beneficial effects of the present application are:

[0077] This invention establishes an environmental database at the data input layer and performs automatic prediction of secondary constant parameters and data classification to ultimately improve the design efficiency and accuracy of deep-sea mining equipment. At the discipline decoupling layer, a coupling proxy model based on discipline sensitivity is set up to simplify the coupling strength between disciplines and avoid the design process from becoming bogged down in complex decoupling processes. At the optimization analysis layer, multi-simulation optimization is implemented, using a combination of high-precision simulation and low-precision proxy models to improve optimization efficiency while ensuring accuracy. At the intelligent solution layer, different algorithms are used to solve problems of varying complexity. Furthermore, the NSGA-III genetic algorithm is improved for the multidisciplinary model of deep-sea mining equipment to avoid getting trapped in local optima when solving complex models. Attached Figure Description

[0078] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0079] Figure 1 This is a general framework diagram of the present invention;

[0080] Figure 2 This is a diagram of the data input layer structure of the present invention;

[0081] Figure 3 This is a structural diagram of the subject decoupling layer of the present invention;

[0082] Figure 4 This is a structural diagram of the optimization analysis layer of the present invention;

[0083] Figure 5 This is a structural diagram of the intelligent solution layer of the present invention. Detailed Implementation

[0084] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. 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 of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0085] Example 1

[0086] like Figure 1 As shown, this embodiment provides a deep-sea mining equipment design system based on an improved genetic algorithm. It is mainly used for designing deep-sea mining vehicles and their supporting operational systems. The system is written in Python, with front-end application development using frameworks such as Vue, and the interface and business logic built using Javascript. It includes five major frameworks: a data input layer, a subject decoupling layer, an optimization analysis layer, an intelligent solution layer, and an operation interaction layer.

[0087] Specifically, the data input layer is configured to configure a parameterized port, allow a user to define a parameterized input template, set input constants, input variables, constraint conditions, and optimization objectives.

[0088] The data input layer adopts HDFS and MinIO as a distributed storage strategy and form, uses a Memcached in-memory database and an efficient indexing strategy to accelerate high-frequency data access, establishes a Spark computing framework, and improves large-scale data processing capability.

[0089] The data input layer dynamically generates a data structure to adapt to different optimization scenarios, adopts a JSON general format, a text format, and an ontology to define data semantics and formulate interdisciplinary data standards to ensure interoperability. The data attribution list is formed by recording data sources, types, and units to complete data classification, establish a metadata framework, and enhance traceability and consistency.

[0090] The data input layer sets multiple weights and gradient constraints in the constraint conditions, adopts a multi-level constraint limiting strategy of priority safety > environment > cost > energy consumption optimization to ensure that the multi-disciplinary joint simulation obtains a trade-off optimization result.

[0091] The above data input layer sets the input constants as primary input constant parameters and secondary input constant parameters, divides the input variables into structural design variables, power and energy design variables, and support system variables, divides the structural design variables into length, width, height, external features, topology, local structure freedom, and connector strength redundancy of mining equipment (deep-sea mining vehicle, umbilical cable, ROV, AUV), divides the power and energy design variables into propulsion device power and layout, battery capacity, and energy distribution strategy (operation mode and standby energy consumption ratio), divides the support system variables into umbilical cable armor thickness and maximum load of deployment and recovery device, divides the constraint conditions into behavior constraints (travel-recovery resistance, thrust, operation energy consumption (power consumption), weight and weight distribution, reserve buoyancy, and envelope volume), strength constraints (strength and stability, stress), and discipline consistency constraints, and the optimization objectives include deep-sea mining equipment weight and its distribution, travel-recovery resistance, operation energy consumption (power consumption), and corresponding economic optimization (high-strength material cost and manufacturing complexity).

[0092] The data input layer includes the following operation steps:

[0093] S110: The designer inputs the primary input constant parameters, including seawater depth, sediment cross shear strength, structural material type, corrosion resistance requirement, and redundancy design standard.

[0094] S120: Establish a deep-sea environment database, input secondary input constant parameters, and complete automatic prediction and input of multi-element characteristic parameters of seawater, sediments, and nodules at different depths in the collection area based on the deep-sea environment function of the polymetallic nodule mining area with the primary input constant parameters as the independent variable. The automatically predicted and input secondary constants include environmental constants (hydrostatic pressure, seabed temperature, seawater salinity, seawater density and viscosity, and ocean current speed), and sediment characteristic constants (shear strength, cohesion, and internal friction angle).

[0095] S130: Combine machine learning embedded automated verification rules to perform secondary constant range checking and logical consistency verification.

[0096] Specifically, step S130 includes:

[0097] S131: After the user inputs the primary constant parameters, a pre-trained random forest model is called to predict the legality probability p of the input constant parameter combination. If the legality probability p is greater than or equal to the threshold value of 0.9, the verification is passed.

[0098] If the legality probability p is less than the threshold value of 0.9, the range checking rule library comparison is triggered, and the out-of-limit or parameter mismatch is prompted, and the recalculation or manual review process is triggered.

[0099] S132: After the automatic prediction of the secondary input constants by the environmental function, an isolation forest model is used to calculate the anomaly score q, which is based on the average path length of the sample in all isolation trees, .

[0100] wherein, : the path length (number of splits) of the sample in a certain isolated tree;

[0101] : the average path length of the sample in all trees;

[0102] : the average path length expectation when the sample size is n (which can be considered as a normalization constant); The calculation formula is:

[0103] . is the harmonic mean of n, (Euler-Mascheroni constant); Since 95% of historical anomalies need to be covered, if the anomaly score q is less than or equal to 0.6, the parameters are considered reasonable.

[0104]

[0105] ​​​​If the abnormal score q > 0.6, start the logic consistency verification, check the parameter relevance, generate a visual report for complex abnormalities, highlight the contradictory parameters and recommend correction suggestions, and realize the visual manual checking and error correction of parameters;

[0106] S133: The correlation feature is that the hydrostatic pressure has a positive correlation with the seawater depth, and the approved pressure depth ratio is 0.01 MPa / m; the seabed temperature is 5-25℃ at the surface-2000m, which is adjusted according to the sea environment and air temperature of the mining area; 2-4℃ at 2000-3500m, 1-2℃ at 3500-4000m, and 4000-6000m slightly increases due to the adiabatic self-pressure effect of seawater, which is adjusted according to the statistical status of seabed hydrothermal in the mining area;

[0107] The seawater salinity tends to be uniform after exceeding the halocline (usually at 100-300m), about 34‰-35‰;

[0108] The seawater density increases with the increase of depth, about 1025-1030kg / m3 at 4000-6000m, and the seawater viscosity increases with the decrease of temperature and slightly increases with the increase of salinity, about 1.37-1.69×10⁻³Pa·s at 4000-6000m;

[0109] The 4000-6000m ocean current speed is affected by the density difference and geostrophic flow, generally between a few centimeters per second and a few tens of centimeters per second, i.e. 3-30cm / s.

[0110] Discipline decoupling layer: including discipline establishment module and coupling analysis module, responsible for the integration and coupling of each discipline model.

[0111] Among them, the coupling analysis module is provided with a random forest classification model, which is trained in advance to identify the coupling parameter combination that is easy to cause oscillation or non-convergence through historical optimization data; at the same time, a real-time monitoring panel is provided to display the coupling residual, the number of iterations, the state of calling the discipline model, and the built-in benchmark case of typical coupling problems (fluid dynamics-structure coupling) for verifying the decoupling algorithm and performance tuning;

[0112] The above-mentioned discipline decoupling layer includes the following running steps:

[0113] S210: The deep sea mining equipment main body is divided into structure discipline, configuration discipline, propulsion discipline and energy consumption discipline through the discipline establishment module;

[0114] S220: The discipline establishment module determines the data flow of constant and input variable from each component to each discipline and system layer by analyzing the characteristics of each component;

[0115] Step S220 specifically includes:

[0116] S221: In the design, the structure discipline transmits the equipment geometry, topological constraints (maximum envelope volume) to the configuration discipline, the transmission factor is set to 0.65-0.85 (this embodiment is set to 0.65), transmits the structure weight and distribution to the propulsion discipline, the transmission factor is set to 0.85-0.95 (this embodiment is set to 0.9), transmits the weight and distribution to the energy consumption discipline, the transmission factor is set to 0.85-0.95 (this embodiment is set to 0.85);

[0117] S222: The configuration discipline transmits the layout feasibility feedback to the structure discipline, the transmission factor is set to 0.55-0.65 (this embodiment is set to 0.45), transmits the fluid dynamics shape parameters and propulsion benefits to the propulsion discipline, the transmission factor is set to 0.45-0.55 (this embodiment is set to 0.5), transmits the equipment space layout and energy distribution strategy to the energy consumption discipline, the transmission factor is set to 0.75-0.90 (this embodiment is set to 0.9);

[0118] S223: The propulsion discipline transmits the structure strength requirement of the propeller installation position to the structure discipline, the transmission factor is set to 0.68-0.78 (this embodiment is set to 0.7), transmits the space occupation constraint to the configuration discipline, the transmission factor is set to 0.85-0.95 (this embodiment is set to 0.85), transmits the real-time power demand curve and power consumption characteristics to the energy consumption discipline, the transmission factor is set to 0.8-0.9 (this embodiment is set to 0.9);

[0119] S224: The energy consumption discipline transmits the sensitivity of energy consumption to weight distribution to the structure discipline, the transmission factor is set to 0.65-0.75 (this embodiment is set to 0.7), transmits the cable power supply equipment and energy consumption equipment layout to the configuration discipline, the transmission factor is set to 0.35-0.65 (this embodiment is set to 0.6), transmits the maximum allowed power threshold under the energy consumption constraint to the propulsion discipline, the transmission factor is set to 0.85-0.95 (this embodiment is set to 0.9).

[0120] S230: Introduce Sobol index and correlation coefficient matrix for coupling analysis module, quantify the coupling strength between disciplines, Sobol index <0.2 is defined as weakly coupled discipline, Sobol index 0.2-0.4 is defined as moderately coupled discipline, Sobol index >0.4 is defined as strongly coupled discipline;

[0121] The correlation coefficient between input variables |r|>0.5 is considered to be strongly correlated, |r|<0.5 is considered to be weakly correlated;

[0122] Prioritize disciplines with high Sobol index (>0.3) or strong correlation (|r|>0.5), establish a discipline coupling structure proxy model based on sensitivity parameters to guide decoupling priority;

[0123] Jacobi iteration is used for weakly coupled disciplines and serial coordination is used for strongly coupled disciplines.

[0124] Optimization analysis layer: including experimental design module and data modeling module, responsible for performing experimental design and optimization algorithm and supporting user modeling decisions.

[0125] The above optimization analysis layer supports the integration of process models built by Isight, ModelCenter, Tosca, Optistruct, to help customers migrate models from legacy multidisciplinary optimization platforms;

[0126] And set up text parser, system command executor, calculator, and finite element software dedicated interface components;

[0127] The above experimental design module mainly carries out deep-sea equipment structural mechanics test, deep-sea equipment corrosion resistance evaluation test, fatigue test, deep-sea equipment operation hydrodynamic test, flow field simulation and analysis test; according to different optimization target quantity, coupling relationship, different test design strategies are adopted, when the optimization target quantity n < 3, and the optimization target is not located in the same discipline or the mutual influence coefficient c < 0.3, full factorial design test is adopted; when the optimization target quantity n < 3, and the optimization target is located in the same discipline or the mutual influence coefficient c > 0.3, orthogonal low workload design test is adopted; when the optimization target quantity n > 3, Latin hypercube or central composite design test is adopted.

[0128] The above data modeling module is provided with parameterized integration interface for commercial software or self-programming of multidisciplinary design simulation, including SolidWorks, ANSYS (strength and stability simulation of mining equipment shell and frame, umbilical cable connection parts), LongRuan4D-GIS (dynamic fusion of seawater parameters and seabed sediment characteristics, establishment of environment-equipment interaction model), OpenFOAM (simulation of fluid resistance and propulsion efficiency of deep-sea mining car, ROV\AUV in complex ocean current, optimization of shape design and propeller layout), digital twin interactive control platform (real-time simulation of deep-sea mining equipment operation state, support for multi-device collaborative operation and emergency obstacle avoidance), Simulink (optimization of battery capacity and energy distribution strategy); realizing parameterized integration and automatic calling of self-programming and commercial CAD / CAE program;

[0129] And the data modeling module provides a polynomial response surface model algorithm, and is additionally provided with an agent model algorithm interface, based on which an external agent model algorithm can be embedded into the platform; a multi-fidelity optimization strategy is adopted, high-precision simulation (error <5%) and low-precision agent model (error <20%) are mixed, high-precision simulation prediction accounts for 65%-75% of the global target weight (75% in the embodiment), and low-precision agent model prediction accounts for 25%-35% of the global target weight (20% in the embodiment), when the deviation d of the low-precision agent model and the high-precision simulation result exceeds 12%, the weight of the low-precision agent model is reduced to 0.1.

[0130] And according to the sensitivity of each system unit to the discipline, only the high-sensitivity parameters (>0.5) are subjected to high-precision optimization, and high-precision full-system simulation is performed once every 5 local optimizations, and the overall convergence is based on the fact that the change rate of the objective function is <1% and the cross-disciplinary constraint violation is <5% for three consecutive iterations.

[0131] Among them, the model of the data modeling module is set according to the complexity and nonlinearity of the test data; when the variable dimension is ≤5 or the Sobol interaction index is <0.05 or the linear misfit test is ≥0.1, a linear polynomial response surface model is suitable; when 5<variable dimension≤15 or Sobol interaction index≥0.1 or 0.05≤linear misfit test<0.3, a quadratic polynomial response surface model is suitable; when variable dimension>15 or Sobol interaction index≥0.3 or linear misfit test<0.05, a high-order polynomial response surface model is suitable.

[0132] In addition, the above-mentioned data modeling module includes multi-disciplinary design:

[0133] The structure discipline preferentially establishes high-precision simulation for detailed stress, strain and fatigue analysis in three-dimensional geometry, material plasticity, creep, anisotropy, fluid load and structure response; low-precision agent models are preferentially used for fast static analysis of bar and shell elements;

[0134] The propulsion discipline performs high-precision simulation modeling on the propeller auxiliary propulsion device, and a low-precision agent model is used for the preliminary relationship between propulsion power and propulsion speed;

[0135] The energy consumption discipline uses high-precision simulation for dynamic energy consumption simulation during the load change process of start-up / impact working conditions, and a low-precision agent model is used for preliminary prediction of unit mining energy consumption;

[0136] The configuration discipline uses a low-precision agent model during the preliminary optimization of overall size, arrangement and center of gravity, and uses high-precision simulation when performing configuration driving-recovery resistance analysis.

[0137] Intelligent solving layer: adopt multi-stage optimization principle, optimize target in stages, optimize single target first and then extend to multi-target; design optimization strategy for proxy model, automatically perform test design sampling and update proxy model, and automatically call quasi-Newton method (L-BFGS), SPEA2 (strength Pareto evolutionary algorithm 2) and improved NSGA-III optimization algorithm.

[0138] Specifically, the quasi-Newton method (L-BFGS) algorithm is used to solve single-target optimization, the SPEA2 algorithm is set for optimizing a multi-target optimization model with 2-3 targets, and the improved NSGA-III algorithm is introduced when the number of targets exceeds 3.

[0139] Specifically, the improved NSGA-III optimization algorithm in the intelligent solving layer includes the following steps:

[0140] S310: Initialize population: based on historical data and good point set, generate initial near-optimal diversity population by using chaotic Logistic mapping (the chaotic factor of Logistic mapping is designed to be 3.57-4.0, and in this embodiment, it is designed to be 4.0), including 80-120 individuals (in this embodiment, 120 individuals), each individual is composed of input variables; in addition, it can also be the cross-sectional area, thickness, length, width, height, support structure form and frame size of deep sea equipment components in structural discipline, the installation position, connection mode and relative position relationship of functional components in configuration discipline on deep sea equipment, the installation position and direction of propulsion device in propulsion discipline on deep sea equipment, transmission ratio, efficiency and energy distribution mode in energy consumption discipline; use neural network to predict individual mass, and remove good point set for individuals with abnormal mass;

[0141] S320: Evaluate fitness: evaluate each individual and calculate its fitness value under multiple optimization targets. The fitness value can be calculated by the objective function, and the repeatedly called objective function is cached to avoid repeated calculation;

[0142] S330: Fast non-dominated sorting: save the dominated number and dominated number set, sort all individuals in the dominated number set, if the individual dominates the number is zero, it is considered that the individual is a non-dominated individual, and the Pareto level is set to the current highest level plus one; based on the above fast non-dominated sorting logic, the individuals in the population are non-dominated sorted, and the individuals are divided into different Pareto front levels to ensure that each individual has the opportunity to become a Pareto optimal solution;

[0143] S340: Calculate the crowding degree: in order to maintain the diversity of the population, the crowding degree of each individual is calculated, which is used for subsequent elite selection;

[0144] S350: Generate offspring: dynamically adjust parameters according to population diversity, generate next generation population through genetic operation (crossing probability is set to 0.85~0.95, this embodiment is set to 0.9; mutation probability is set to 1 / dimension~3 / dimension, this embodiment is set to 2 / dimension, high dimension can be uniformly set to 0.02~0.05, this embodiment is set to 0.02), while retaining 5%~10% proportion (this embodiment retains 10%) of elite individuals to ensure population diversity and convergence, generated offspring individuals are subjected to local disturbance and acceptance criterion judgment of simulated annealing, selection stage, simulated annealing acceptance probability is used instead of roulette selection, simulated annealing initial temperature is set to 100~300, this embodiment is set to 100, temperature decay coefficient is set to 0.90~0.99, preferably set to 0.90;

[0145] Or divide the population into genetic subgroups and particle swarm (PSO), genetic subgroups perform crossing and mutation, PSO subgroups update velocity, particle swarm inertia weight is set to 0.4~0.9, this embodiment is set to 0.5-0.8, and the real action is dynamically adjusted:

[0146] When the variable dimension is less than 80, it is set to 0.4~0.6, and this embodiment is set to 0.5;

[0147] When the variable dimension is 80~100, it is set to 0.6~0.8, and this embodiment is set to 0.7;

[0148] When the variable dimension is 100~120, it is set to 0.8~0.9, and this embodiment is set to 0.8, and the two regularly exchange excellent individuals;

[0149] Or use "DE / rand / 1" or "DE / best / 1" strategy of differential evolution to generate mutant individuals, differential mutation coefficient is set to 0.4~0.8 (preferably set to 0.5), and local development ability is enhanced;

[0150] S360: Merge population: merge parent and offspring populations to form a new population, at this time only update the non-dominated relationship of new individuals, but not re-sort globally, reduce calculation redundancy;

[0151] S370: Elite selection: divide the merged and mixed population into 3~6 subgroups, select individuals with high fitness and low crowding degree as elite individuals of the next generation population, preferentially retain feasible solutions, and gradually relax the constraint tolerance by using dynamic constraint method, the initial value of dynamic constraint tolerance is initially 0.01, and the iteration increases by 0.005~0.01, this embodiment increases by 0.005, and the parallel distributed optimization is completed and merged;

[0152] S380: Repeat iteration: Repeat the above steps until the set stop condition is reached (such as reaching the maximum number of iterations, the maximum number of iterations is set to 100-300, this embodiment is set to 300 (when the variable dimension is less than 80, it is set to 100-150, when the variable dimension is 80-100, it is set to 150-200, when the variable dimension is 100-120, it is set to 200-300), the number of convergence is greater than or equal to 10);

[0153] S390: Obtain Pareto optimal solution (Pareto optimal, also known as Pareto efficiency): the individuals in the final obtained population are the Pareto optimal solution set, and each solution cannot be completely dominated by other solutions in all objectives.

[0154] Through the above steps, the improved NSGA-III can effectively search for the Pareto optimal solution set of the multi-objective optimization problem, provide a diversity and balanced solution for the designer, and help them make better decisions in the design process.

[0155] Operation interaction layer: PySide interface design technology is used to provide UI controls and functions to ensure the efficiency of multidisciplinary design optimization operation, QThreadPool and QRunnable technology are used to efficiently manage thread pool, cloud data storage module is introduced for parallel processing to avoid frequent creation and destruction of threads, and experts in mechanical, control, environmental and other fields are supported for collaborative editing and conflict detection.

[0156] At the same time, the working interface of the operation interaction layer is divided into design, monitoring and output interfaces: the design interface includes a classification tree structure component library, a flow structure tree, a global / local variable management area, a parameter mapping relationship management area, a priority setting area and a resource quota selection area; the monitoring interface includes a task timeline view area and a progress monitoring bar; the output interface includes a multi-objective optimization result chart area, a data sandbox area, a key parameter influence and an optimal scheme recommendation report generation area.

[0157] Table 1 shows the technical index results of the final deep-sea mining vehicle scheme designed in application example 1 of the present application:

[0158]

[0159] Example 2:

[0160] This embodiment is specially designed for deep-sea mining vehicles, and differs from example 1 in that:

[0161] 1. In the steps of the discipline coupling layer, S221: the transfer factor is set to 0.85, the structure weight and distribution are transferred to the propulsion discipline, the transfer factor is set to 0.95, the weight and distribution are transferred to the energy consumption discipline, and the transfer factor is set to 0.95;

[0162] In S222, the transfer factor is set to 0.55 to transfer the fluid dynamics shape parameters, propulsion benefits to the propulsion discipline, the transfer factor is set to 0.5 to transfer the equipment space layout, energy distribution strategy to the energy consumption discipline, and the transfer factor is set to 0.8;

[0163] In S223, the transfer factor is set to 0.68 to transfer the space occupation constraint to the configuration discipline, the transfer factor is set to 0.95 to transfer the real-time power demand curve, power consumption characteristics to the energy consumption discipline, and the transfer factor is set to 0.8;

[0164] In S224, the transfer factor is set to 0.65 to transfer the cable power supply equipment, energy consumption equipment layout to the configuration discipline, the transfer factor is set to 0.4 to transfer the maximum allowable power threshold under the energy consumption constraint to the propulsion discipline, and the transfer factor is set to 0.85.

[0165] 2. In the data modeling module in the optimization analysis layer, the polynomial response surface model algorithm is provided, the high-precision simulation prediction accounts for 65% of the global target weight, and the low-precision proxy model prediction accounts for 35% of the global target weight. When the deviation d of the low-precision proxy model and the high-precision simulation result exceeds 12%, the weight of the low-precision proxy model is reduced to 0.1.

[0166] 3. In step S310 of the intelligent solving layer, the Logistic mapping chaos factor is designed to be 3.58, including 80 individuals.

[0167] In S350, the crossover probability of genetic operation is set to 0.8; the mutation probability is set to 1 / dimension, and the high dimension can be uniformly set to 0.02;

[0168] At the same time, 5% of the elite individuals are reserved to ensure population diversity and convergence;

[0169] The initial temperature of the simulated annealing is set to 300, and the temperature attenuation coefficient is set to 0.95;

[0170] The particle swarm inertia weight is set to 0.4, and is also dynamically adjusted, the variable dimension is less than 80, the variable dimension is 80-100, the variable dimension is 100-120, and the two regularly exchange excellent individuals; or use the "DE / rand / 1" or "DE / best / 1" strategy of differential evolution to generate a mutant individual, and the differential mutation coefficient is set to 0.48 to enhance the local development capability.

[0171] Table 2 is the final deep-sea mining vehicle scheme technical index result designed in this embodiment:

[0172]

[0173] Example 3:

[0174] This embodiment is used to design mining operation auxiliary AUV / ROV, which is different from example 1 in that:

[0175] 1. In step S221 of the discipline decoupling layer: the transfer factor is set to 0.65, the structure weight and distribution are transferred to the propulsion discipline, the transfer factor is set to 0.85, the weight and distribution are transferred to the energy consumption discipline, and the transfer factor is set to 0.85;

[0176] In S222, the transfer factor is set to 0.55, the fluid dynamics shape parameters and propulsion benefits are transferred to the propulsion discipline, the transfer factor is set to 0.45, the equipment space layout and energy distribution strategy are transferred to the energy consumption discipline, and the transfer factor is set to 0.90;

[0177] In S223, the transfer factor is set to 0.78, the space occupation constraint is transferred to the configuration discipline, the transfer factor is set to 0.95, the real-time power demand curve and power consumption characteristics are transferred to the energy consumption discipline, and the transfer factor is set to 0.8;

[0178] In S224, the transfer factor is set to 0.7, the cable power supply equipment and energy consumption equipment layout are transferred to the configuration discipline, the transfer factor is set to 0.6, the maximum allowable power threshold under energy consumption constraints is transferred to the propulsion discipline, and the transfer factor is set to 0.95.

[0179] 2. The polynomial response surface model algorithm provided by the data modeling module in the optimization analysis layer predicts 70% of the global target weight with high precision simulation, and 30% of the global target weight with low precision proxy model prediction. When the deviation d between the low precision proxy model and the high precision simulation result exceeds 12%, the weight of the low precision proxy model is reduced to 0.1. The model of the data modeling module is set according to the complexity and nonlinearity of the test data.

[0180] 3. In step S310 of the intelligent solving layer, the Logistic mapping chaos factor is designed to be 3.8, including 100 individuals;

[0181] In S350, the crossover probability of genetic operation is set to 0.95; the mutation probability is set to 3 / dimension, and the high dimension can be uniformly set to 0.05;

[0182] At the same time, 10% of the elite individuals are reserved to ensure population diversity and convergence;

[0183] The initial temperature of simulated annealing is set to 200, and the temperature attenuation coefficient is set to 0.99;

[0184] The inertia weight of the particle swarm is set to 0.9, and the inertia is dynamically adjusted, the variable dimension is less than 80, the variable dimension is set to 0.6, the variable dimension is 80-100, the variable dimension is set to 0.8, the variable dimension is 100-120, the variable dimension is set to 0.9, and the two regularly exchange excellent individuals; or use the "DE / rand / 1" or "DE / best / 1" strategy of differential evolution to generate mutant individuals, and the differential mutation coefficient is set to 0.65, and the local development ability is enhanced.

[0185] Table 3 is the final deep-sea mining vehicle scheme technical index result designed in the embodiment:

[0186]

[0187] Comparative Example 1:

[0188] A deep-sea mining vehicle designed by using a general-purpose multidisciplinary design platform, an X ship performance comprehensive optimization platform. Comparative Example 1 is customized for ship sea surface environment, and each model does not consider seabed sediment, seabed seawater submerged environment, CAE model needs to be manually reconstructed, and openness is weak. Its core capability is concentrated in the synergy of fluid mechanics and static structure of two disciplines, but it lacks the support of key disciplines such as dynamic response, control and manufacturing process. The platform is uneven in the fidelity of deep-sea operation equipment model optimization, the proxy model has high precision, but the fidelity effect of complex strong coupling problems (such as fluid-solid resonance) is significantly insufficient. The calculation efficiency is in the order of optimization cycle, which can be accelerated by distributed parallel calculation, but the algorithm convergence needs a large number of iterations, and there is a certain efficiency bottleneck. The degree of automation is limited, and only 40% of manual intervention is reduced through mature CAD / CAE tool chain. It does not support robust optimization (no failure probability control ability), has no sensitivity analysis function, and multi-objective optimization is limited to two objectives, which cannot meet the requirements of large-scale system and component collaborative optimization of deep-sea equipment. The openness is weak, and only typical CAE model import is supported. The platform is suitable for single-discipline-dominated, weakly coupled deterministic design scenarios such as ship fluid shape optimization, but it cannot meet the requirements of high reliability or deep-sea complex system level.

[0189] Table 4 is the final deep-sea mining vehicle scheme technical index result designed in Comparative Example 1:

[0190]

[0191] Comparative Example 2:

[0192] A deep-sea mining vehicle designed by using a general-purpose multidisciplinary design platform, a multidisciplinary optimization platform for complex industrial products, supports the synergy of three disciplines of structural mechanics, aerodynamics, and electromagnetism, but lacks disciplines in key fields such as control and propulsion for deep-sea mining operations. The use advantage is that the established proxy model has high accuracy; but the strong coupling discipline lacks sufficient fidelity, the simplified and calculated model has poor capability, and the optimization period is as long as a month. The platform uses distributed parallel processing logic and acceleration strategies, and the models can be well linked to reduce 80% of manual intervention, achieve good full-process automation, and integrate CAD / CAE tool chains to realize task automatic scheduling. The platform has single-system multi-objective optimization capability (supports target ≥ 3), but lacks an intelligent solution screening mechanism; provides primary robustness optimization (failure probability ≤ 1e-3), which does not meet the high reliability scenario requirements of the deep-sea operation environment; only supports local unit sensitivity analysis, and cannot handle system-level coupled sensitivity. The automation advantage is offset by inefficient calculation and weak coupling processing capability, and is only suitable for non-strong coupling and low dynamic response vehicle component optimization (such as traditional chassis structure design).

[0193] Table 5 is the final technical index result of the deep-sea mining vehicle designed in Comparative Example 2:

[0194]

[0195] From the results, it can be seen that the deep-sea mining vehicle designed in the embodiment achieves high-level lightweight, with a lightweight coefficient of 4.26-4.33. The ground pressure of the deep-sea mining vehicles designed in Examples 1-3 is reduced by 20%, and the ground pressure unevenness is significantly reduced to 0.05-0.12, fully ensuring the operation stability of the walking mechanism. In addition, the thrust / resistance ratio is 1.21-1.35, avoiding the shortcomings of insufficient thrust reserve and redundant thrust setting in Comparative Examples 1 and 2. In terms of energy consumption, the embodiment achieves an energy consumption of 4.4-5.6kW*h / t through multidisciplinary optimization of the energy consumption discipline, achieving low-energy operation of less than 6kW*h / t.

[0196] The following table is the comparison result of the advancement of the design platform in the embodiment and the comparative example:

[0197]

[0198] Through comparison, it can be seen that Examples 1-3 achieve all-round breakthroughs in the advancement of the multidisciplinary optimization design platform, which is suitable for deep-sea equipment design environment, while the comparative examples lack the ability to adapt to deep-sea mining equipment design in many aspects, specifically in:

[0199] In terms of breadth and depth, it covers five disciplines such as static-dynamic structure, fluid, control, propulsion and manufacturing process, which significantly exceeds the limitations of Comparative Example 1 (two disciplines) and Comparative Example 2 (three disciplines); through automatic switching of agent model and high-precision CAE simulation, the defects of insufficient fidelity of complex model in Comparative Example 1 and poor fidelity effect of strongly coupled disciplines in Comparative Example 2 are solved, and multi-gradient technology is used to efficiently process nonlinear strongly coupled problems. In terms of efficiency and performance, the embodiment is on par with Comparative Example 1 and far superior to Comparative Example 2 in terms of monthly time consumption, while the CPU / GPU hybrid parallel acceleration ratio reaches 80 times (Comparative Example 1 is 60 times, and Comparative Example 2 is 50 times), and the algorithm convergence speed (50-200 iterations) is significantly improved compared to Comparative Example 1 (100-500 times) and Comparative Example 2 (300-400 times). In terms of method and architecture, the embodiment compresses the failure probability after robust optimization to (2 orders of magnitude lower than deterministic optimization), which is more reliable than Comparative Example 1 ( ) and Comparative Example 2 ( ), and supports optimization of more than three objectives and system-level-subsystem-level collaboration, and its full-system sensitivity analysis capability also fills the gap of Comparative Example 1. In addition, the embodiment is customized for CAE tool chain in deep sea complex environment, supports integration of multiple legacy models, which is more technically challenging than the adaptation of regular scenarios in Comparative Example 1 / 2; the multi-process interactive interface further optimizes the user experience, realizing efficient collaborative design of complex engineering problems.

[0200] In summary, the above embodiment has the following technical effects:

[0201] 1. The data input layer allows users to customize parameterized input templates through templated parameter ports, dividing constants into two categories: primary and secondary. Primary constant parameters (such as seawater depth and sediment cross shear strength) are directly input by designers, while secondary constant parameters (such as hydrostatic pressure and sediment cohesion) are automatically predicted based on polymetallic nodule mining area environmental functions, reducing the burden of manual input. The deep sea environment database integrates historical data and machine learning verification rules to perform range checks and logical consistency verification on parameters, and triggers recalculation or manual review for abnormal data. A common format is used to define data semantics, and a unified data standard is established to solve the problem of inconsistent semantics; metadata framework is used to record data sources, types and units to enhance traceability; distributed storage and Spark computing framework support large-scale data collaborative processing to ensure cross-platform interoperability. Ultimately, the problem of complex and lack of standardized input of deep sea environmental parameters is solved.

[0202] 2. The subject decoupling layer introduces the Sobol index and correlation coefficient matrix to classify the coupling strength between subjects into three categories: weak, medium, and strong. Weakly coupled subjects are optimized using Jacobi iterative parallel execution; strongly coupled subjects are optimized using serial coordinated optimization. The random forest classification model is trained using historical data to identify coupling parameter combinations prone to oscillations in advance, reducing the risk of non-convergence. Built-in benchmark cases for typical problems such as fluid dynamics-structure coupling are used to verify the performance of the decoupling algorithm; a real-time monitoring panel displays key indicators such as coupling residuals and iteration counts, supporting dynamic tuning. Dynamic coupling strength quantification and decoupling priority adjustment are implemented, solving the problems of complex multi-disciplinary coupling and rigid decoupling strategies.

[0203] 3. A multi-fidelity optimization strategy was developed, which combines high-precision simulation and low-precision surrogate models in the optimization analysis layer. Weights are dynamically allocated based on sensitivity, with high-precision simulation being prioritized for highly sensitive parameters and surrogate models used for low-sensitivity parameters. A bias correction method was designed. The experimental design module selects strategies (full factorial, Latin hypercube, etc.) based on the number of targets to reduce redundant experiments. A scheme for local high-precision optimization and global simulation verification was designed, achieving periodic full-system verification and convergence control, thus solving the problem of balancing model accuracy and computational efficiency.

[0204] 4. A diverse initial population is generated using chaotic logistic mapping to avoid local optima; simulated annealing, PSO, and differential evolution (DE) are embedded to enhance local exploitation capabilities; the constraint tolerance is gradually relaxed, prioritizing the retention of feasible solutions. A multi-stage optimization and parallel computation approach is designed, starting with a single objective and then expanding to multiple objectives, using staged optimization to reduce complexity; the population is divided into 3-6 subgroups for parallel optimization, and efficiency is improved through elite selection and subgroup partitioning. Ultimately, the improved NSGA-III algorithm effectively overcomes the limitations of the NSGA-II and NCGA algorithms in deep-sea mining equipment design.

[0205] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A deep-sea mining equipment design system based on an improved genetic algorithm, characterized by, The application relates to a multi-disciplinary design optimization system and method. The data input layer is used for configuring a parameterized template port, allowing a user to define a parameterized input template, set input constants, input variables, constraint conditions and optimization targets; The subject decoupling layer includes a subject establishing module and a coupling analysis module, and is responsible for integration and coupling of various subject models; The subject decoupling layer includes the following operation steps: S210: the subject establishing module is used to divide a deep-sea mining equipment main body into a structure subject, a configuration subject, a propulsion subject and an energy consumption subject; S220: the subject establishing module is used to analyze characteristics of each component to determine data flow of constant and input variable transmission to each subject and a system layer; S230: the coupling analysis module is introduced with Sobol indexes and a correlation coefficient matrix to quantify coupling strength between subjects, and when the Sobol index is less than 0.2, the subject is defined as a weakly coupled subject, when the Sobol index is 0.2-0.4, the subject is defined as a moderately coupled subject, and when the Sobol index is greater than 0.4, the subject is defined as a strongly coupled subject; When a correlation coefficient |r| between input variables is greater than 0.5, the input variables are considered to be strongly correlated, and when the correlation coefficient |r| is less than 0.5, the input variables are considered to be weakly correlated; High Sobol indexes or strongly correlated subjects are preferentially processed, a subject coupling structure proxy model is established according to sensitivity parameters, and decoupling priority is guided; Jacobi iteration is used for parallel execution of weakly coupled subjects, and serial coordination is used for strongly coupled subjects; The optimization analysis layer includes a test design module and a data modeling module, and is responsible for executing test design and optimization algorithms and supporting user modeling decisions; The intelligent solving layer adopts a multi-stage optimization principle, divides optimization targets into stages, optimizes single targets first and then extends to multi-targets, designs an optimization strategy for a proxy model, can automatically perform test design sampling and update the proxy model, and automatically calls a quasi-Newton method, an SPEA2, and an improved NSGA-III optimization algorithm; The operation interactive layer adopts a PySide interface design technology to provide UI controls and functions, guarantees operation efficiency of multi-disciplinary design optimization, efficiently manages a thread pool through QThreadPool and QRunnable technologies, introduces a cloud data storage module to perform parallel processing and avoid frequent creation and destruction of threads, and supports experts in mechanical, control and environmental fields to perform collaborative editing and conflict detection.

2. The deep-sea mining equipment design system based on improved genetic algorithm according to claim 1, characterized in that, The data input layer sets input constants as main input constant parameters and secondary input constant parameters; Input variables are divided into structure design variables, power and energy design variables and support system variables; The structure design variables are divided into length, width, height, external features, topological structure, local structure freedom and connecting piece strength redundancy of the mining equipment; The power and energy design variables are divided into propulsion device power and layout, battery capacity and energy distribution strategy; The support system variables are divided into umbilical cable armor thickness and maximum load of a laying and recovering device; Constraint conditions are divided into behavior constraints, strength constraints and subject consistency constraints; Optimization targets include weight and distribution of the deep-sea mining equipment, running-recovering resistance and operation energy consumption.

3. The deep-sea mining equipment design system based on improved genetic algorithm according to claim 2, characterized in that, The data input layer includes the following operation steps: S110: main input constant parameters are input by a designer. S120: Establish a deep-sea environment database, input secondary input constant parameters, and complete automatic prediction and input of multi-element characteristic parameters of seawater, sediments, and nodules at different depths in the collection area based on the deep-sea environment function of the polymetallic nodule mining area with the primary input constant parameters as the independent variable; S130: Combine machine learning embedded automatic verification rules to perform secondary constant range checking and logical consistency verification.

4. The deep-sea mining equipment design system based on improved genetic algorithm of claim 3, wherein, Step S130 includes: S131: After the user inputs the primary constant parameters, a pre-trained random forest model is called to predict the legality probability p of the input constant parameter combination; if the legality probability p is greater than or equal to 0.9, the test is passed; If the legality probability p is less than 0.9, the range checking rule library is compared, and out-of-range or parameter mismatch is prompted, and a recalculation or manual review process is triggered; S132: After the environmental function automatically predicts the secondary input constant, an isolation forest model is used to calculate the anomaly score q; if the anomaly score q is less than or equal to 0.6, the parameters are considered reasonable; If the anomaly score q is greater than 0.6, logical consistency verification is started, parameter correlation is checked, a visual report is generated for complex anomalies, conflicting parameters are highlighted, and correction suggestions are recommended, achieving parameter visualization, manual review, and error correction; S133: The correlation characteristics are: the hydrostatic pressure has a positive correlation with the seawater depth, and the nodule pressure depth ratio is 0.01 MPa / m; the seabed temperature is 5-25°C at the surface-2000 m, adjusted according to the environmental conditions and air temperature of the mining area; 2-4°C at 2000-3500 m, 1-2°C at 3500-4000 m, and slightly higher at 4000-6000 m due to the adiabatic self-pressure effect of seawater, adjusted according to the statistics of seabed hydrothermal conditions in the mining area; The seawater salinity tends to be uniform at 34-35‰ after exceeding the halocline; The density of seawater increases with depth, and is 1025-1030 kg / m3 at 4000-6000 m 3 The viscosity of seawater increases with temperature decrease and slightly increases with salinity increase, and is 1.37-1.69 x 10 ​ The ocean current speed is between 3 cm / s and 30 cm / s due to the influence of density difference and geostrophic flow at 4000-6000 m.

5. The deep-sea mining equipment design system based on improved genetic algorithm according to claim 4, characterized in that, Step S220 includes: S221: In the design, the structure discipline transmits the equipment geometric size and topological structure constraints to the configuration discipline, the transmission factor is set to 0.65-0.85, transmits the structure weight and distribution to the propulsion discipline, the transmission factor is set to 0.85-0.95, and transmits the weight and distribution to the energy consumption discipline, the transmission factor is set to 0.85-0.95; S222: The configuration discipline transmits the layout feasibility feedback to the structure discipline, the transmission factor is set to 0.55-0.65, transmits the fluid dynamics shape parameters and propulsion benefits to the propulsion discipline, the transmission factor is set to 0.45-0.55, and transmits the equipment space layout and energy distribution strategy to the energy consumption discipline, the transmission factor is set to 0.75-0.90; S223: The propulsion discipline transmits the structure strength requirement of the propeller installation position to the structure discipline, the transmission factor is set to 0.68-0.78, transmits the space occupation constraint to the configuration discipline, the transmission factor is set to 0.85-0.95, and transmits the real-time power demand curve and power consumption characteristics to the energy consumption discipline, the transmission factor is set to 0.8-0.9; S224: The energy consumption discipline transmits the sensitivity of energy consumption to weight distribution to the structure discipline, the transfer factor is set to 0.65-0.75, transmits the cable energy supply device and the energy consumption device layout to the configuration discipline, the transfer factor is set to 0.35-0.65, transmits the maximum allowable power threshold under the energy consumption constraint to the propulsion discipline, and the transfer factor is set to 0.85-0.

95.

6. The deep-sea mining equipment design system based on improved genetic algorithm according to claim 5, characterized in that, The optimization analysis layer supports the integration of process models constructed by Isight, ModelCenter, Tosca and Optistruct; And set the text parser, system command executor, calculator, and finite element software special interface components; The test design module mainly carries out deep-sea equipment structural mechanics test, deep-sea equipment corrosion resistance evaluation test, fatigue test, deep-sea equipment operation hydrodynamic test, flow field simulation and analysis test; different test design strategies are adopted according to different optimization target quantities and coupling relationships; when the optimization target quantity n is less than 3 and the optimization targets are not located in the same discipline or the mutual influence coefficient c is less than 0.3, full-factorial design test is adopted; when the optimization target quantity n is less than 3 and the optimization targets are located in the same discipline or the mutual influence coefficient c is greater than 0.3, orthogonal low workload design test is adopted; when the optimization target quantity n is greater than 3, Latin hypercube or central composite design test is adopted; The data modeling module is provided with a parameterized integration interface of commercial software or self-programmed programs for multidisciplinary design simulation, including SolidWorks, ANSYS, LongRuan4D-GIS, OpenFOAM, digital twin interactive control platform and Simulink; parameterized integration and automatic calling of self-programmed programs and commercial CAD / CAE programs are realized; The data modeling module provides polynomial response surface model algorithm, and has proxy model algorithm interface, based on which external proxy model algorithm can be embedded into the platform; multi-fidelity optimization strategy is adopted, high-precision simulation and low-precision proxy model are mixed, high-precision simulation prediction accounts for 65%-75% of the global target weight, and low-precision proxy model prediction accounts for 25%-35% of the global target weight; when the deviation d of low-precision proxy model and high-precision simulation result exceeds 12%, the weight of low-precision proxy model is reduced to 0.1; According to the sensitivity of each system unit to the discipline, only high-precision optimization is performed on high-sensitivity parameters, high-precision whole system simulation is performed once every 5 local optimizations, and the overall convergence is based on the change rate of the objective function being less than 1% for 3 consecutive iterations and the cross-discipline constraint violation being less than 5%.

7. The deep-sea mining equipment design system based on improved genetic algorithm according to claim 6, characterized in that, The multidisciplinary design in the data modeling module includes: The structure discipline preferentially establishes high-precision simulation of detailed stress, strain and fatigue analysis of three-dimensional geometry, material plasticity, creep, anisotropy, fluid load and structure response; low-precision proxy model is preferentially adopted for fast statics analysis of bar and shell elements; The propulsion discipline performs high-precision simulation modeling on the propeller auxiliary propulsion device, and adopts low-precision proxy model for the preliminary relationship between propulsion power and propulsion speed; The dynamic energy consumption simulation in the process of load change in the energy consumption discipline start / impact adopts high-precision simulation, and the preliminary prediction of energy consumption per unit of mining quantity adopts a low-precision proxy model; The low-precision proxy model is adopted in the preliminary optimization stage of the overall size, arrangement and gravity center of the configuration discipline, and the high-precision simulation is adopted in the configuration driving-recovery resistance analysis.

8. The deep-sea mining equipment design system based on improved genetic algorithm of claim 7, wherein, The improved NSGA-III optimization algorithm in the intelligent solving layer comprises the following steps: S310: initializing a population: generating an initial near-optimal diversity population based on historical data and a good point set by using chaotic Logistic mapping; S320: evaluating fitness: evaluating each individual and calculating its fitness value under multiple optimization objectives; S330: fast non-dominated sorting: saving the dominated number and dominated number set, sorting all individuals in the dominated number set, and if the individual dominated number is zero, considering that the individual is a non-dominated individual, and the Pareto level is set to the current highest level plus one; S340: calculating the crowding degree: in order to maintain the diversity of the population, the crowding degree of each individual is calculated for subsequent elite selection; S350: generating offspring: generating the next generation population through genetic operation according to the dynamic adjustment parameter of population diversity, while reserving 5% to 10% of elite individuals to ensure population diversity and convergence, generating offspring individuals for local disturbance and acceptance criterion judgment of simulated annealing, selection stage, using the acceptance probability of simulated annealing to replace roulette selection, setting the initial temperature of simulated annealing to 100 to 300, and setting the temperature decay coefficient to 0.90 to 0.99; S360: merging the population: merging the parent and child populations to form a new population; S370: elite selection: dividing the merged and mixed population into 3 to 6 subgroups, selecting individuals with high fitness and low crowding degree as elite individuals of the next generation population, preferentially reserving feasible solutions, and gradually relaxing the constraint tolerance by using a dynamic constraint method, and merging after parallel distributed optimization; S380: repeating iteration: repeating steps S310 to S370 until the set stop condition is reached; S390: obtaining the Pareto optimal solution: the individuals in the final population are the Pareto optimal solution set, and each solution cannot be completely dominated by other solutions in all objectives.

9. The deep-sea mining equipment design system based on improved genetic algorithm of claim 1, wherein, The data input layer adopts HDFS and MinIO as the distributed storage strategy and form, uses the Memcached in-memory database and high-efficiency index strategy, and establishes a Spark computing framework; The data input layer dynamically generates a data structure, adopts a JSON general format, a text format, ontology definition data semantics and formulates cross-disciplinary data standards; data classification is completed by recording data sources, types and units to form a data attribution list, and a metadata framework is established; The data input layer sets multiple weights and gradient constraints in the constraint condition, and adopts a multi-level constraint restriction strategy of priority safety>environment>cost>energy consumption optimization. The coupling analysis module of the discipline decoupling layer is provided with a random forest classification model, which is trained by historical optimization data to identify in advance the coupling parameter combinations that are prone to cause oscillation or non-convergence; meanwhile, a real-time monitoring panel is provided to display the coupling residual error, the number of iterations, the state of calling the discipline model, and the benchmark cases of typical coupling problems for verifying the decoupling algorithm and performance tuning; The model of the data modeling module of the optimization analysis layer is set according to the complexity and nonlinearity of the test data; The intelligent solving layer is provided with a quasi-Newton method algorithm for solving single-objective optimization, and a SPEA2 algorithm for optimizing a multi-objective optimization model with 2-3 objectives; when the number of objectives exceeds 3, an improved NSGA-III algorithm is introduced; The working interface of the operation interaction layer is divided into three interfaces of design, monitoring and output: the design interface includes a classification tree structure component library, a flow structure tree, a global / local variable management area, a parameter mapping relationship management area, a priority setting area and a resource quota selection area; The monitoring interface includes a task timeline view area and a progress monitoring column; and the output interface includes a multi-objective optimization result chart area, a data sandbox area and a key parameter influence and optimal scheme recommendation report generation area.

Images