Method for determining initial range of ship type optimization based on knowledge reasoning

Abstract

The application relates to the field of hull line intelligent design, and particularly discloses a ship type optimization initial range determination method based on knowledge reasoning, which comprises the following steps: S1, constructing a Yangtze River cargo ship line design knowledge base, wherein the knowledge base comprises a parent type instance library, a design rule library and a line knowledge graph library; S2, based on the design ship demand input by a user, determining applicable parameterized geometric models and initial line optimization ranges through similar ship type retrieval; S3, calculating the best optimization range of ship body local line parameters, wherein the best optimization range is determined in combination with parameter sensitivity analysis and knowledge reasoning results; S4, constructing a total resistance approximation model, and performing fine optimization on the local line parameters in the best optimization range by using an optimization algorithm; and S5, generating a geometric model of the optimized ship body line, and verifying the resistance performance through CFD simulation. The application solves the problems of blindness and low design efficiency of the traditional ship type optimization range, and effectively improves the local line optimization accuracy and resistance performance of the Yangtze River cargo ship.

Description

A method for determining the initial range of ship morphology optimization based on knowledge reasoning Technical Field

[0001] This invention relates to the field of intelligent hull design technology, specifically a method for determining the initial range of hull optimization based on knowledge reasoning. Background Technology

[0002] In recent years, with the advancement of my country's high-quality development strategy for Yangtze River shipping and the implementation of green ship standards such as the International Maritime Organization's Energy Efficiency Design Index (EEDI), the demand for energy conservation and emission reduction for Yangtze River cargo ships has become increasingly urgent. As a core element determining a ship's sailing performance, especially its resistance performance, the optimized design of the hull line has become a key link in the development of green ship types.

[0003] Currently, simulation-based design optimization (SBDO) has gradually replaced traditional manual design and become the mainstream technology for hull design. This method has significantly improved the automation level of hull design by combining parametric geometric modeling, computational fluid dynamics (CFD) simulation and intelligent optimization algorithms. However, there are still technical bottlenecks that need to be overcome in the core link of fine optimization of local hull lines of Yangtze River cargo ships—"determining the optimal optimization range".

[0004] From the perspective of current technological development, hull line optimization has formed a mature process of "parametric modeling - approximate model construction - optimization solution." In terms of parametric modeling, tools such as CAESES have achieved fully parametric expressions of hull bow and stern configurations (such as upright bows and twin tail fins), allowing hull surface deformation to be driven by geometric parameters. In the field of approximate models and optimization algorithms, models such as BP neural networks and Kriging have been used to replace time-consuming CFD calculations to improve optimization efficiency, and multi-objective optimization algorithms such as NSGA-II are widely used in solving objectives such as resistance minimization. Meanwhile, the application of knowledge engineering technology in ship design is gradually expanding. The construction of knowledge bases such as parent model instance libraries and design rule libraries provides support from real-world experience and regulatory knowledge for hull line design. For example, the initial hull form can be determined through parent model instance retrieval, and the approximate value range of hull line parameters can be obtained using the design rule library.

[0005] However, existing technologies still have significant shortcomings in the "optimal optimization range calculation" stage of fine-grained local profile optimization for Yangtze River cargo ships: First, the determination of the optimization range lacks a dual-driven mechanism of "knowledge-sensitivity". Current technologies mostly directly adopt the full variable range of the parametric model, or rely solely on the empirical intervals given by the design rule base, without combining the sensitivity of local profile parameters to resistance performance for screening. The parameter range provided by the design rule base often covers all feasible regions, but some intervals have a weak impact on resistance changes. Directly using them for optimization will lead to redundancy in the design space, increasing the difficulty of constructing approximate models and the amount of optimization calculations. Second, there is insufficient coordination between local profile optimization and knowledge reasoning. Existing knowledge engineering applications often remain at the initial ship type selection (e.g., parent type instance retrieval) or global parameter determination (e.g., longitudinal position of the center of buoyancy), failing to combine knowledge reasoning results (e.g., recommended range from the design rule base) with parameter sensitivity analysis to accurately locate the "optimal optimization sub-interval" that plays a crucial role in local hull performance (e.g., bow wave-making resistance, stern viscous pressure resistance). This leads to local optimization easily falling into inefficient searches or parameter coupling interference. Thirdly, the adaptation to the unique hull characteristics of Yangtze River cargo ships is insufficient. Yangtze River cargo ships are mostly medium-to-low speed, full-bodied hulls (block coefficient 0.8-0.9). The impact of local hull parameter changes (e.g., concealed bulbous bow, twin-fin tunnel) on resistance differs from that of seagoing vessels. Existing optimization range determination methods do not address the characteristics of this type of ship and lack dedicated calculation logic that combines the Yangtze River cargo ship hull knowledge graph with local parameter sensitivity, making it difficult to guarantee the accuracy of the optimization range.

[0006] Furthermore, in the practice of fine-tuning local profiles, existing technologies also face the problem of an "efficiency-performance" imbalance: if the optimization range is too wide, it will lead to a surge in the sample size of the approximate model and an increase in the number of iterations of the optimization algorithm, significantly extending the design cycle; if the range is narrowed down solely based on experience, it may miss key intervals that are sensitive to resistance, causing the optimization results to fall into local optima. For example, for the optimization of the inclination angle of the stern tunnel of Yangtze River cargo ships, existing methods may directly adopt the full range of 10°-25° allowed by the parametric model, without using knowledge reasoning to select the 15°-20° interval recommended by the design rule base, and without combining CFD sensitivity analysis to further focus on the 16°-18° sub-interval that has the most significant impact on stern viscous drag, ultimately resulting in redundant optimization processes or limited performance improvement.

[0007] While the integration of knowledge engineering and SBDO technology has provided new directions for hull line design, existing research has not yet broken through the core bottleneck of "optimal optimization range calculation": the lack of a systematic method that can integrate knowledge reasoning (parent model instance library, design rule library) and parameter sensitivity analysis makes it impossible to provide accurate and efficient optimization space definition for fine-grained local hull line optimization of Yangtze River cargo ships. This gap makes it difficult for the hull line design of Yangtze River cargo ships to balance "optimization efficiency" and "performance improvement," thus hindering the rapid development of green ship types.

[0008] In summary, although the current Yangtze River cargo ship hull optimization technology has achieved initial breakthroughs in automation and intelligence, there are still problems such as insufficient coordination of knowledge and sensitivity and blind scope definition in the "determination of the optimal optimization range" stage of fine optimization of local hull lines.

[0009] Therefore, it is urgent to construct an optimal optimization range calculation method that combines knowledge reasoning and parameter sensitivity analysis to provide a precise design space for fine optimization of local hull lines, thereby solving the problem of balancing efficiency and performance in the design of Yangtze River cargo ship hull lines and promoting the development of green ship types. Summary of the Invention

[0010] This invention addresses the aforementioned problems in the prior art by providing a method for determining the initial range of ship hull optimization based on knowledge reasoning. This method effectively solves the problems of blind optimization range and disconnect between knowledge and optimization in traditional methods, and effectively improves the efficiency of local hull hull optimization and ship resistance performance of Yangtze River cargo ships.

[0011] To achieve the above objectives, this invention proposes a method for determining the initial range of ship morphology optimization based on knowledge reasoning, comprising:

[0012] S1. Construct a knowledge base for the design of Yangtze River cargo ship hull lines, which includes a parent model instance library, a design rule library, and a hull line knowledge graph library.

[0013] S2. Based on the user-inputted design requirements for the ship, the applicable parametric geometric model and the initial hull optimization range are determined through similar ship type retrieval;

[0014] S3. Calculate the optimal optimization range of the local hull line parameters, wherein the optimal optimization range is determined by combining the results of parameter sensitivity analysis and knowledge reasoning.

[0015] S4. Construct an approximate model of total resistance and use an optimization algorithm to perform fine optimization on the local profile parameters within the optimal optimization range;

[0016] S5. Generate the optimized geometric model of the hull shape and verify the resistance performance through CFD simulation.

[0017] Preferably, in S1, the specific steps for constructing the Yangtze River cargo ship hull form design knowledge base are as follows:

[0018] S11. Construct a parent model instance library: Collect actual ship design data of Yangtze River cargo ships, extract ship characteristic attributes and structural attributes, and store the two types of attributes in a relational database; the ship characteristic attributes include ship number, ship type, bow shape, stern shape, main dimensional parameters, hull form coefficients, and design speed; the structural attributes include the parametric geometric model of the corresponding ship type, the applicable hull form knowledge graph type, and the applicable design rule base type;

[0019] S12. Construct a design rule base: Organize ship design manuals, literature and optimization simulation data, divide the hull line knowledge into curve type, parameter range type and empirical formula type, and use production rules to represent and store it;

[0020] S13. Construct a knowledge graph library of hull lines: Using hull form factor and scale ratio as input dimensions and key hull line geometric parameters as output dimensions, generate a gridded dataset through experimental design and CFD simulation, construct and store a knowledge graph that reflects the mapping relationship between hull line parameters and resistance performance.

[0021] Preferably, in S12, the method for validating the design rule base is as follows: select the actual ship parameters of multiple Yangtze River cargo ships, compare the fit between the actual ship form parameters and the recommended range of the design rule base, and modify the rule content for parameters that do not fit by combining the actual ship design constraints.

[0022] Preferably, in S2, the specific steps for retrieving similar ship types based on user-inputted design ship requirements are as follows:

[0023] S21. Obtain user-inputted design vessel requirements: including vessel type, bow and stern shape, main dimensions, hull form coefficients, and design speed;

[0024] S22. Calculate the similarity between the designed ship and instances in the parent model instance library: distinguish between qualitative and quantitative attributes to calculate the distance. After standardizing the quantitative attributes, use the nearest neighbor search algorithm to calculate the comprehensive distance, and then obtain the similarity based on the comprehensive distance.

[0025] S23. Determine the applicable parametric model and initial optimization range: Based on the similarity results, determine the applicable parametric geometric model for the designed ship, and use the range of hull line parameters obtained by reasoning from the design rule base as the initial optimization range.

[0026] Preferably, in S22, the specific method for calculating the similarity is as follows:

[0027] S221. Qualitative attribute distance calculation: For qualitative attributes such as ship type, bow shape, and stern shape, the distance for the same attribute is set to the preset minimum value, the distance for similar attributes is set to the preset median value, and the distance for different attributes is set to the preset maximum value.

[0028] S222, Quantitative Attribute Standardization: For quantitative attributes such as main dimensions and ship type coefficients, extreme value standardization is used to process them and map the attribute values ​​to a preset range; S223, Comprehensive Similarity Calculation: The comprehensive distance between the design ship and the instance is calculated using the distance formula, and then the comprehensive distance is converted into similarity using the similarity formula.

[0029] Preferably, in S3, the method for calculating the optimal range of the local hull line parameters is as follows:

[0030] S31. Determine the local profile parameters to be optimized: Select parameters that are sensitive to drag performance and whose value range meets the preset conditions from the local profile parameters obtained by reasoning from the design rule base;

[0031] S32. Calculate the parameter sensitivity coefficient: Using the controlled variable method, take multiple sets of gradient values ​​for each local profile parameter within the initial optimization range, calculate the total resistance change rate under each set of gradients through CFD, and determine the parameter sensitivity coefficient based on the change rate.

[0032] S33. Determine the optimal optimization range: Take the intersection of the parameter range obtained by reasoning from the design rule base and the significant influence interval reflected by the sensitivity coefficient as the optimal optimization range of the local profile parameter.

[0033] Preferably, in S31, when screening the local profile parameters to be optimized, parameters with parameter coupling relationships need to be excluded; if key profile parameters are coupled, the parameter with a more significant impact on drag performance should be fixed first, and then the optimization range of another parameter should be determined.

[0034] Preferably, in S4, the specific steps for constructing the approximate model of total resistance and performing fine optimization are as follows:

[0035] S41. Sample Sampling: Using experimental design methods, samples are taken within the optimal optimization range. The number of samples is matched with the number of local profile parameters to be optimized, so that the samples are evenly distributed in the design space.

[0036] S42. Construct an approximate model: Using local profile parameters as input and total resistance as output, construct an approximate model of total resistance using a neural network algorithm, and verify the model fitting effect through preset accuracy indicators;

[0037] S43. Perform fine optimization: Select a multi-objective optimization algorithm with the goal of minimizing total resistance, and combine it with layout constraints to iteratively search for the optimal combination of local profile parameters within the best optimization range.

[0038] S44. Output the optimal parameter combination: From the optimal solution set obtained by optimization, select the local profile parameter combination that satisfies all constraints as the final optimization result.

[0039] Preferably, in S42, the process of constructing the approximate model includes:

[0040] S421. Data preprocessing: Standardize the sampled local profile parameters and corresponding total resistance values ​​to eliminate the impact of magnitude differences on model training.

[0041] S422, Model Training: Set the number of neurons in the input layer, hidden layer, and output layer of the neural network, select the activation function and optimization algorithm, and iteratively train the model until the validation set loss meets the preset convergence condition;

[0042] S423. Accuracy Verification: Select a portion of the sampled data as a test set and verify the model's prediction accuracy for total resistance using preset error evaluation indicators.

[0043] Preferably, in step S5, the specific steps for generating the geometric model and verifying the drag performance are as follows:

[0044] S51. Generate an optimized ship geometry model: Substitute the optimized local profile parameters into the applicable parametric geometry model, drive the generation of the hull surface through parametric modeling software, and output a standard format geometry model file.

[0045] S52, CFD simulation verification: The computational domain and mesh are constructed using CFD software, the turbulence model, boundary conditions and solution parameters are set, and the total resistance of the ship before and after optimization is calculated.

[0046] S53. Performance Comparison Analysis: Compare the resistance performance data of the ship type before and after optimization, and verify the optimization effect by combining the simulation results of hull surface pressure distribution and free surface wave generation.

[0047] Therefore, this invention proposes a method for determining the initial range of ship morphology optimization based on knowledge reasoning, which has the following beneficial effects:

[0048] (1) Accurately define the optimal optimization range of the local profile. By conducting resistance sensitivity analysis on the local profile parameters and combining the knowledge reasoning results of the design rule base, the intersection of the two intervals is taken to determine the optimal optimization range. This avoids the problem of blindly broad or too narrow range definition in traditional optimization, making the fine optimization of the local profile more targeted, reducing invalid optimization searches, and improving optimization efficiency.

[0049] (2) Achieve full-process collaboration of design-optimization-verification. Relying on the developed intelligent hull line design system, it can automatically complete the search of similar ship types and the reasoning of initial hull line parameters. The reasoning results can be directly connected to the construction of the total resistance approximation model without manual parameter conversion. After optimization, the resistance performance is quickly verified through CFD simulation, forming a complete closed loop of "knowledge reasoning-fine optimization-performance verification", which greatly simplifies the design process and shortens the design cycle of Yangtze River cargo ship hull lines.

[0050] (3) Adapt to the unique configuration requirements of Yangtze River cargo ships. The knowledge base is specifically built for Yangtze River cargo ships. The parent model instance library contains parameterized models of common Yangtze River cargo ship configurations such as upright bow and double tail fin. The hull line knowledge graph is also adapted to the ship type characteristics of Yangtze River cargo ships. During optimization, unique local parameters such as stern tunnel and tail fin can be adjusted in a targeted manner to solve the problem of insufficient adaptability of general optimization methods to Yangtze River cargo ship configurations and ensure that the optimized ship type meets the actual needs of Yangtze River shipping.

[0051] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0052] Figure 1 is a framework diagram of intelligent hull line design for the knowledge-based reasoning method for determining the initial range of hull optimization according to the present invention.

[0053] Figure 2 is a diagram of the input interface of the similar ship type retrieval module of the ship type optimization initial range determination method based on knowledge reasoning of the present invention;

[0054] Figure 3 is a flowchart of the application of intelligent reasoning in the hull form optimization initial range determination method based on knowledge reasoning of the present invention.

[0055] Figure 4 is a comparison of the inferred ship type and the optimized ship type surface of the ship type optimization initial range determination method based on knowledge reasoning of the present invention.

[0056] Figure 5 is a CFD calculation grid distribution diagram of the initial range determination method for ship type optimization based on knowledge reasoning of the present invention. Detailed Implementation

[0057] To make the technical solutions, advantages, and objectives of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below. The described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention without creative effort are within the protection scope of this application.

[0058] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains.

[0059] As shown in Figures 1-5, this invention proposes a method for determining the initial range of ship morphology optimization based on knowledge reasoning, including:

[0060] S1. Construct a knowledge base for the design of Yangtze River cargo ship hull lines, which includes a parent model instance library, a design rule library, and a hull line knowledge graph library.

[0061] The specific steps for constructing a knowledge base for the design of cargo ship hull lines on the Yangtze River are as follows:

[0062] S11. Construct a parent model instance library: Collect actual ship design data of Yangtze River cargo ships, extract ship characteristic attributes and structural attributes, and store the two types of attributes in a relational database; the ship characteristic attributes include ship number, ship type, bow type, stern type, main dimensional parameters, hull form coefficients and design speed; the structural attributes include the parametric geometric model of the corresponding ship type, the applicable line knowledge graph type and the applicable design rule base type;

[0063] S12. Construct a design rule base: Organize ship design manuals, literature and optimization simulation data, divide the hull line knowledge into curve type, parameter range type and empirical formula type, and use production rules to represent and store it;

[0064] S13. Construct a knowledge graph library of hull lines: Using hull form factor and scale ratio as input dimensions and key hull line geometric parameters as output dimensions, generate a gridded dataset through experimental design and CFD simulation, construct and store a knowledge graph that reflects the mapping relationship between hull line parameters and resistance performance.

[0065] The method for validating the design rule base is as follows: Select parameters of multiple Yangtze River cargo ships, compare the fit between the actual ship form parameters and the recommended range of the design rule base, and revise the rule content for parameters that do not fit by combining the actual ship design constraints.

[0066] S2. Based on the user-inputted design requirements for the ship, the applicable parametric geometric model and the initial hull optimization range are determined through similar ship type retrieval;

[0067] The specific steps for retrieving similar ship types based on user-inputted design requirements are as follows:

[0068] S21. Obtain user-inputted design vessel requirements: including vessel type, bow and stern shape, main dimensions, hull form coefficients, and design speed;

[0069] S22. Calculate the similarity between the designed ship and instances in the parent model instance library: distinguish between qualitative and quantitative attributes to calculate the distance. After standardizing the quantitative attributes, use the nearest neighbor search algorithm to calculate the comprehensive distance, and then obtain the similarity based on the comprehensive distance.

[0070] The specific method for calculating similarity is as follows:

[0071] S221. Qualitative attribute distance calculation: For qualitative attributes such as ship type, bow shape, and stern shape, the distance for the same attribute is set to the preset minimum value, the distance for similar attributes is set to the preset median value, and the distance for different attributes is set to the preset maximum value.

[0072] S222, Quantitative Attribute Standardization: For quantitative attributes such as main dimensions and ship form coefficients, extreme value standardization method is used to process the attribute values ​​and map them to a preset range;

[0073] S223. Comprehensive Similarity Calculation: The comprehensive distance between the designed ship and the instance is calculated using the distance formula, and then the comprehensive distance is converted into similarity using the similarity formula.

[0074] S23. Determine the applicable parametric model and initial optimization range: Based on the similarity results, determine the applicable parametric geometric model for the designed ship, and use the range of hull line parameters obtained by reasoning from the design rule base as the initial optimization range.

[0075] S3. Calculate the optimal optimization range of the local hull line parameters. The optimal optimization range is determined by combining the results of parameter sensitivity analysis and knowledge reasoning.

[0076] The specific method for calculating the optimal range of local hull line parameters is as follows:

[0077] S31. Determine the local profile parameters to be optimized: Select parameters that are sensitive to drag performance and whose value range meets the preset conditions from the local profile parameters obtained by reasoning from the design rule base;

[0078] When screening local profile parameters to be optimized, parameters with parameter coupling relationships should be excluded; if key profile parameters are coupled, the parameter with a more significant impact on drag performance should be fixed first, and then the optimization range of the other parameter should be determined.

[0079] S32. Calculate the parameter sensitivity coefficient: Using the controlled variable method, take multiple sets of gradient values ​​for each local profile parameter within the initial optimization range, calculate the total resistance change rate under each set of gradients through CFD, and determine the parameter sensitivity coefficient based on the change rate.

[0080] S33. Determine the optimal optimization range: Take the intersection of the parameter range obtained by reasoning from the design rule base and the significant influence interval reflected by the sensitivity coefficient as the optimal optimization range of the local profile parameter.

[0081] S4. Construct an approximate model of total resistance and use an optimization algorithm to perform fine optimization on the local profile parameters within the optimal optimization range;

[0082] The specific steps for constructing the approximate model of total resistance and performing fine-grained optimization are as follows:

[0083] S41. Sample Sampling: Using experimental design methods, samples are taken within the optimal optimization range. The number of samples is matched with the number of local profile parameters to be optimized, so that the samples are evenly distributed in the design space.

[0084] S42. Construct an approximate model: Using local profile parameters as input and total resistance as output, construct an approximate model of total resistance using a neural network algorithm, and verify the model fitting effect through preset accuracy indicators;

[0085] The process of constructing an approximate model includes:

[0086] S421. Data preprocessing: Standardize the sampled local profile parameters and corresponding total resistance values ​​to eliminate the impact of magnitude differences on model training.

[0087] S422, Model Training: Set the number of neurons in the input layer, hidden layer, and output layer of the neural network, select the activation function and optimization algorithm, and iteratively train the model until the validation set loss meets the preset convergence condition;

[0088] S423. Accuracy Verification: Select a portion of the sampled data as a test set and verify the model's prediction accuracy for total resistance using preset error evaluation indicators.

[0089] S43. Perform fine optimization: Select a multi-objective optimization algorithm with the goal of minimizing total resistance, and combine it with layout constraints to iteratively search for the optimal combination of local profile parameters within the best optimization range.

[0090] S44. Output the optimal parameter combination: From the optimal solution set obtained by optimization, select the local profile parameter combination that satisfies all constraints as the final optimization result.

[0091] S5. Generate the optimized geometric model of the hull shape and verify the resistance performance through CFD simulation.

[0092] The specific steps for generating the geometric model and verifying the drag performance are as follows:

[0093] S51. Generate an optimized ship geometry model: Substitute the optimized local profile parameters into the applicable parametric geometry model, drive the generation of the hull surface through parametric modeling software, and output a standard format geometry model file.

[0094] S52, CFD simulation verification: The computational domain and mesh are constructed using CFD software, the turbulence model, boundary conditions and solution parameters are set, and the total resistance of the ship before and after optimization is calculated.

[0095] S53. Performance Comparison Analysis: Compare the resistance performance data of the ship type before and after optimization, and verify the optimization effect by combining the simulation results of hull surface pressure distribution and free surface wave generation.

[0096] Example 1

[0097] As shown in Figures 1-5, this invention provides a method for determining the initial range of ship type optimization based on knowledge reasoning. Taking the design of a bulk carrier's form line in the Yangtze River Basin as the application object, this embodiment aims at the "design of a 13,000-ton bulk carrier's form line in the Yangtze River" and executes the entire process of "knowledge base construction - similar ship type retrieval - optimal optimization range calculation - fine optimization - CFD verification". The specific implementation process is as follows:

[0098] S1. Construct a knowledge base for the design of Yangtze River cargo ship hull lines;

[0099] S11. Construct the parent type instance library;

[0100] Design data for actual ships, including bulk carriers, container ships, and oil tankers, were collected from the Yangtze River basin. This included 21 typical ship types such as 7500t bulk carriers and 600TEU container ships. Two types of attributes were extracted and stored in a MySQL database.

[0101] Ship type characteristics and attributes: including ship number, ship type, bow type, stern type, length between perpendiculars, length-to-beam ratio, width-to-draft ratio, block coefficient, and design speed. Bow types include upright bow, bulbous bow, and sloping bow. Stern types include twin stern and twin tail fin. Example data is shown in Table 1.

[0102] Structural attributes: Associate the CAESES parametric model file path, applicable hull line knowledge graph type (upright bow-double tail fin graph), and applicable design rule base type (cargo ship rule base) for each ship, enabling one-click association retrieval of "feature-structure".

[0103] Table 1. Parent Model Instance Library - Examples of Ship Type Feature Attributes

[0104]

[0105] S12. Construct a design rule base;

[0106] By reviewing ship design manuals, literature, and optimization simulation data, hull form knowledge is categorized into curve types, parameter range types, and empirical formula types, and represented and stored using production rule rules.

[0107] The curve-type knowledge includes the curve of the vertical position of the center of buoyancy changing with the rhombus coefficient. Extract the type value points and store them as a rule table. When needed, linear interpolation can be used to obtain the recommended value under any rhombus coefficient.

[0108] The parameter range knowledge includes the parameters of the upright head and the hidden sphere head, the minimum width-to-height ratio, and the recommended value range of the cross section area coefficient at the raised point, which are stored in the form of "IF (square coefficient ≥ 0.8) THEN (parameter ∈ [X1, X2])".

[0109] The empirical formulas that relate the inlet section length ratio to the Froude number can be stored directly, but the Froude number calculation parameters of the design vessel need to be substituted into them.

[0110] S13. Construct a linear knowledge graph base;

[0111] Using the squareness coefficient CB, aspect ratio L / B, and width-to-draft ratio B / T as input dimensions, and the longitudinal position of the center of buoyancy Lcb, the inner fatness of the caudal fin inskegy_b, the caudal shaft spacing ratio b_B, and the head width contraction ratio ydwl_ratio as output dimensions, the following steps are performed to construct the system:

[0112] First, a Latin hypercube experimental design was used to generate 310 sets of samples within the typical ranges of CB, L / B, and B / T.

[0113] Next, for each sample group, the total resistance is calculated using STAR-CCM+, and the sample parameters with the lowest resistance are selected as the optimal parameters.

[0114] The correspondence between input dimension and optimal parameter is stored as a graph dataset, forming a knowledge graph that reflects the mapping relationship between scale ratio, profile parameters and drag performance. Example data is shown in Table 2.

[0115] Table 2. Knowledge Graph Dataset for Corrugated Lines

[0116]

[0117] S2. Based on the user-inputted design requirements for the ship, the applicable parametric geometric model and the initial hull optimization range are determined through similar ship type retrieval;

[0118] S21. Obtain the user's input requirements for the design ship;

[0119] Users input their ship design requirements through the intelligent design system interface developed in Qt, including:

[0120] Vessel type: bulk carrier; Bow type: straight bow; Stern type: twin stern;

[0121] Main dimensions: 127m length between perpendiculars, 21.8m width, and 5.5m design draft;

[0122] Ship type coefficient: block coefficient 0.8654; design speed 18.5km / h.

[0123] S22. Calculate the similarity between the designed ship and instances in the parent model instance library;

[0124] S221. Qualitative attribute distance calculation: For a bulk carrier with a straight bow that matches the parent model, the distance is 0; for a twin-tailed stern, the distance to the twin-fin stern in the parent model is set to 0.5, and the distance to the single-tailed stern is set to 1.

[0125] S222. Standardization of Quantitative Attributes: For quantitative attributes such as length between perpendiculars and aspect ratio, Max-Min standardization is used to eliminate differences in magnitude. The calculation formula is as follows:

[0126] ;

[0127] In the formula, For the actual value of the attribute, This represents the theoretical maximum value of this attribute in the parent type library. This is the theoretical minimum value of this attribute in the parent type library;

[0128] S223. Comprehensive Similarity Calculation: The comprehensive distance between the designed ship and the instance in the parent model library is calculated using Euclidean distance, and then converted into a similarity score using the similarity formula. The similarity calculation formula is as follows:

[0129] ;

[0130] In the formula, For the design of the ship A standardized attribute, For the first instance of the parent type A standardized attribute, Total number of attributes;

[0131] The calculation results show that the design vessel has the highest similarity (0.93) with the 12,000t bulk carrier (ship number 1) in the parent model library, and a similarity of 0.735 with the 7,500t bulk carrier (ship number 2).

[0132] S23. Determine the applicable parametric model and the initial optimization range;

[0133] Since the similarity is ≥0.9, output the structural attributes of the 12000t bulk carrier:

[0134] Applicable parametric model: CAESES upright bow-twin stern bulk carrier parametric model;

[0135] Applicable hull form knowledge graph: Hull form knowledge graph for cargo ships with upright bows and double fins;

[0136] Initial optimization range: The range of values ​​for local profile parameters (such as the minimum width-to-height ratio of the concealed sphere 0.77-0.83, and the tunnel inclination angle 15°-20°) is derived from the design rule base.

[0137] S3. Calculate the optimal optimization range for local hull profile parameters;

[0138] S31. Determine the local profile parameters to be optimized;

[0139] From the local parameters within the initial range, exclude parameters that are strongly coupled with other parameters, and then filter for parameters that are sensitive to drag performance:

[0140] Head parameters: minimum width-to-height ratio of the stealth spherical head, cross-sectional area coefficient at the raised head, and curvature of the stealth spherical head;

[0141] Tail parameters: tunnel tilt angle, tail plate height ratio, tail fin tilt angle, and outer tail fin flank enlargement.

[0142] S32. Calculate the parameter sensitivity coefficient;

[0143] Using the controlled variable method, five sets of gradient values ​​were taken for each parameter within the initial range. Other parameters were fixed, and the total drag rate of change under each set of gradients was calculated using STAR-CCM+. The sensitivity coefficient calculation formula is as follows:

[0144] ;

[0145] In the formula, The relative rate of change of total resistance. This represents the relative rate of change of the parameter.

[0146] The calculation results show that the sensitivity coefficients of the minimum width-to-height ratio of the stealth bulb (S=0.032), the tunnel tilt angle (S=0.028), and the tail fin tilt angle (S=0.025) are ≥0.02, which have a significant impact on drag; the sensitivity coefficients of the other parameters are <0.02, which have a weak impact on drag.

[0147] S33. Determine the optimal optimization range;

[0148] The optimal optimization range is the intersection of the "initial range of the design rule base" and the "range where sensitivity has a significant impact".

[0149] Minimum width-to-height ratio of the stealth sphere: initial range 0.77-0.83, significant sensitivity range 0.78-0.82, optimal range 0.78-0.82;

[0150] Tunnel inclination angle: initial range 15°-20°, significant sensitivity range 16°-18°, optimal range 16°-18°;

[0151] Tail fin tilt angle: initial range 14°-20°, significant sensitivity range 16°-19°, optimal range 16°-19°;

[0152] The remaining parameters are fixed at the median of the initial range (e.g., the curvature of the invisible sphere is 0.91) and do not participate in subsequent optimization.

[0153] S4. Construct an approximate model of total resistance and use an optimization algorithm to perform fine optimization on the local profile parameters within the optimal optimization range;

[0154] S41, Sample collection;

[0155] Within the optimal optimization range, which includes three parameters to be optimized, 30 sets of samples are generated according to the principle of parameter number × 10, ensuring that the samples are evenly distributed in the design space and avoiding local clustering. Example samples are shown in Table 3:

[0156] Table 3 Sample of parameters to be optimized

[0157]

[0158] S42. Construct a BP neural network approximation model of total resistance;

[0159] S421. Data preprocessing: The sampled local profile parameters and the corresponding total drag values ​​calculated by CFD are standardized. Max-Min standardization is used to eliminate the influence of magnitude differences on model training.

[0160] S422, Model Training: Set the number of neurons in the input, hidden, and output layers of the neural network, select the activation function and optimization algorithm, and iteratively train the model until the validation set loss meets the preset convergence condition. Specifically:

[0161] Input layer: 3 neurons, corresponding to 3 parameters to be optimized;

[0162] Hidden layers: according to empirical formula It was determined to be 9 neurons;

[0163] In the formula, The number of neurons in the input layer =3, Represents the number of neurons in the output layer =1, This represents an empirical adjustment factor used to fine-tune the number of neurons in the hidden layer based on the actual task scenario. =5;

[0164] Output layer: 1 neuron, corresponding to total resistance;

[0165] Activation functions: The hidden layer uses the Sigmoid function, and the output layer uses a linear function;

[0166] Model training: The loss function is the mean squared error (MSE), the learning rate is 0.01, and the number of iterations is 5000. Training is stopped when the validation set loss does not decrease for 50 consecutive iterations.

[0167] S423. Accuracy Verification: Select 10% of the sampled data as the test set and calculate the model evaluation index: MSE = 1.98 × 10⁻⁶. -4 MAPE=0.16%, R 2 =0.992, which meets the accuracy requirement (R²≥0.98).

[0168] S43, Perform fine-grained optimization;

[0169] Integrate the BP neural network approximation model into the CAESES optimization module and set the optimization parameters:

[0170] Optimization objective: Minimize total resistance;

[0171] Constraints: Displacement volume ≥ Design ship displacement volume 13175m³ 3 The stern shaft spacing ratio is 0.5-0.55, and the clearance between the stern hull and the propeller blade is ≥0.2 times the propeller diameter.

[0172] Algorithm parameters: population size 30, number of iterations 35, crossover rate 0.9, mutation rate 0.01.

[0173] After the iteration is complete, the optimal parameter combination that satisfies all constraints is selected from the Pareto optimal solution set:

[0174] The minimum width-to-height ratio of the stealth bulb is 0.805, the tunnel tilt angle is 17.2°, and the tail fin tilt angle is 17.8°.

[0175] S44. Output the optimal parameter combination. Integrate the optimal parameters with the fixed parameters to form the final local profile parameter scheme.

[0176] S5. Generate the optimized geometric model of the hull shape and verify the resistance performance through CFD simulation.

[0177] S51. Generate an optimized ship geometry model;

[0178] The final local profile parameters are substituted into the CAESES upright bow-double stern parametric model. The hull surface is generated by driving the F-Spline curve, and the optimized hull geometry model in IGES format is output. The model includes the fine surfaces of the stealthy bow, the stern tunnel, and the tail fin.

[0179] S52 and CFD simulation verification;

[0180] The calculation domain is set as follows: bow 1.5 times ship length, stern 2.5 times ship length, height 2.5 times ship length, and width 2 times ship length. The calculation is simplified using a symmetrical plane, and only the starboard half of the ship is calculated.

[0181] Mesh generation: Polyhedral mesh + prism layer is adopted. The basic mesh size is 1 / 50 of the ship length. The mesh is refined on the hull surface, free liquid surface and bow / stern area, with a final mesh size of 1.8 million.

[0182] Physical model settings: The turbulence model adopts the k-ε model, the free surface adopts the VOF method, and the boundary conditions are set as velocity inlet-pressure outlet-no-slip wall-symmetry plane;

[0183] Solution settings: time step 0.02s, stop calculation when the drag coefficient fluctuation is ≤1%.

[0184] S53, Performance Comparison Analysis;

[0185] The resistance performance of the initial ship type (mother ship), the inferred ship type generated by S2, and the optimized ship type are compared and the results are shown in Table 4.

[0186] Table 4 Comparison of Resistance Performance of Different Ship Types

[0187]

[0188] Based on the analysis of the pressure distribution on the hull surface: the optimized hull shape has a smaller high-pressure area at the bow and a gentler low-pressure area at the stern, resulting in a smaller pressure difference between the bow and stern and a reduction in viscous pressure resistance; the wave-making height on the free surface is slightly lower than that of the inference hull shape, further optimizing wave-making resistance.

[0189] Therefore, this invention provides a method for determining the initial range of ship hull optimization based on knowledge reasoning. Through a knowledge-driven, fine-tuned optimization, and CFD verification process, it realizes intelligent design of the hull lines of Yangtze River bulk carriers. Compared with the traditional SBDO method, it effectively shortens the time from demand input to optimized hull generation, effectively improves performance, and reduces the total resistance of the optimized hull compared with the initial hull, thus meeting the energy-saving and emission-reduction requirements of Yangtze River cargo ships. It effectively solves the problems of blind range, low efficiency, and limited performance improvement in traditional hull optimization, and provides efficient technical support for the green hull line design of Yangtze River cargo ships.

[0190] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for determining the initial range of ship morphology optimization based on knowledge reasoning, characterized in that, include: S1. Construct a knowledge base for the design of Yangtze River cargo ship hull lines, including a parent model instance library, a design rule library, and a hull line knowledge graph library; S2. Based on user-inputted design requirements, determine the applicable parametric geometric model and initial hull line optimization range through similar ship type retrieval; S3. Calculate the optimal optimization range for local hull line parameters, which is determined by combining parameter sensitivity analysis and knowledge reasoning results; S4. Construct an approximate total resistance model, and use optimization algorithms to perform fine-tuning of local hull line parameters within the optimal optimization range; S5. Generate the optimized hull line geometric model. The drag performance was verified through CFD simulation. In section S3, the calculation method for the optimal optimization range of the hull local profile parameters is as follows: S31, Determine the local profile parameters to be optimized: From the local profile parameters obtained through reasoning from the design rule base, select parameters that are sensitive to drag performance and whose value range meets preset conditions; S32, Calculate the parameter sensitivity coefficient: Using the controlled variable method, take multiple sets of gradient values ​​for each local profile parameter within the initial optimization range, calculate the total drag change rate under each set of gradients using CFD, and determine the parameter sensitivity coefficient based on the change rate; S33, Determine the optimal optimization range: Take... The intersection of the parameter range obtained from the rule base reasoning and the significant influence interval reflected by the sensitivity coefficient is taken as the optimal optimization range for the local profile parameter. In S31, when determining the local profile parameter to be optimized, parameters with parameter coupling relationships need to be excluded. If key profile parameters are coupled, the parameter with a more significant impact on drag performance is fixed first, and then the optimization range of another parameter is determined. In S4, the specific steps for constructing the approximate model of total drag and performing fine optimization are as follows: S41, Sample sampling: Sample within the optimal optimization range using experimental design methods, with the number of samples matching the number of local profile parameters to be optimized, so that the samples are evenly distributed in the design space; S42, Construct an approximate model: Using the local profile parameter as input and total drag as output, construct an approximate model of total drag using a neural network algorithm, and verify the model fitting effect through preset accuracy indicators; S43, Perform fine optimization: Select a multi-objective optimization algorithm, with the minimum total drag as the optimization objective, and combine the arrangement constraints to iteratively search for the optimal combination of local profile parameters within the optimal optimization range; S44, Output the optimal parameter combination: From the optimal solution set obtained by optimization, select the local profile parameter combination that satisfies all constraints as the final optimization result.

2. The method for determining the initial range of ship morphology optimization based on knowledge reasoning according to claim 1, characterized in that, In S1, the specific steps for constructing the Yangtze River cargo ship hull form design knowledge base are as follows: S11, Constructing a parent model instance library: Collect actual ship design data of Yangtze River cargo ships, extract ship characteristic attributes and structural attributes, and store the two types of attributes in a relational database; the ship characteristic attributes include ship number, ship type, bow shape, stern shape, principal dimensional parameters, hull form coefficients, and design speed; the structural attributes include the parametric geometric model of the corresponding hull type, the applicable hull form knowledge graph type, and the applicable design rule base type; S12, Constructing a design rule base: Organize ship design manuals, literature, and optimization simulation data, divide hull form knowledge into curve type, parameter range type, and empirical formula type, and represent and store it using production rules; S13, Constructing a hull form knowledge graph library: Using hull form coefficients and scale ratios as input dimensions and key hull form geometric parameters as output dimensions, generate a gridded dataset through experimental design and CFD simulation, construct and store a knowledge graph reflecting the mapping relationship between hull form parameters and resistance performance.

3. The method for determining the initial range of ship morphology optimization based on knowledge reasoning according to claim 2, characterized in that, In S12, the method for validating the design rule base is as follows: select the actual ship parameters of multiple Yangtze River cargo ships, compare the fit between the actual ship form parameters and the recommended range of the design rule base, and modify the rule content for parameters that do not fit by combining the actual ship design constraints.

4. The method for determining the initial range of ship morphology optimization based on knowledge reasoning according to claim 1, characterized in that, In S2, the specific steps for retrieving similar ship types based on user-inputted design ship requirements are as follows: S21, Obtain user-inputted design ship requirements: including ship type, bow and stern shape, main dimensions, hull form coefficients, and design speed; S22, Calculate the similarity between the design ship and instances in the parent model instance library: Distinguish between qualitative and quantitative attributes to calculate the distance, standardize the quantitative attributes, use the nearest neighbor search algorithm to calculate the comprehensive distance, and then obtain the similarity based on the comprehensive distance; S23, Determine the applicable parametric model and initial optimization range: Based on the similarity results, determine the applicable parametric geometric model for the design ship, and use the hull line parameter range obtained by reasoning from the design rule base as the initial optimization range.

5. The method for determining the initial range of ship morphology optimization based on knowledge reasoning according to claim 4, characterized in that, In S22, the specific method for calculating similarity is as follows: S221, Qualitative attribute distance calculation: For qualitative attributes such as ship type, bow type, and stern type, the distance of the same attribute is set to the preset minimum value, the distance of similar attributes is set to the preset median value, and the distance of different attributes is set to the preset maximum value; S222, Quantitative attribute standardization: For quantitative attributes such as main dimensions and ship type coefficients, the extreme value standardization method is used to process them and map the attribute values ​​to the preset interval; S223, Comprehensive similarity calculation: The comprehensive distance between the design ship and the instance is calculated using the distance formula, and then the comprehensive distance is converted into similarity using the similarity formula.

6. The method for determining the initial range of ship morphology optimization based on knowledge reasoning according to claim 1, characterized in that, In S42, the process of constructing the approximate model includes: S421, data preprocessing: standardizing the sampled local profile parameters and corresponding total resistance values ​​to eliminate the influence of magnitude differences on model training; S422, model training: setting the number of neurons in the input layer, hidden layer, and output layer of the neural network, selecting activation functions and optimization algorithms, and iteratively training the model until the validation set loss meets the preset convergence condition; S423, accuracy verification: selecting a portion of the sampled data as a test set, and verifying the model's prediction accuracy for total resistance through preset error evaluation indicators.

7. The method for determining the initial range of ship morphology optimization based on knowledge reasoning according to claim 1, characterized in that, In S5, the specific steps for generating the optimized hull geometry model and verifying the resistance performance through CFD simulation are as follows: S51, Generating the optimized hull geometry model: Substitute the optimized local profile parameters into the applicable parametric geometry model, drive the hull surface generation through parametric modeling software, and output a standard format geometry model file; S52, CFD simulation verification: Use CFD software to construct the computational domain and mesh, set the turbulence model, boundary conditions, and solution parameters, and calculate the total resistance of the hull before and after optimization; S53, Performance comparison analysis: Compare the resistance performance data of the hull before and after optimization, and verify the optimization effect by combining the simulation results of hull surface pressure distribution and free surface wave generation.

Images