Ship structure ice level strengthening method based on artificial intelligence

Abstract

The invention discloses a ship structure ice level strengthening method based on artificial intelligence, which comprises the following steps of: acquiring basic parameter information of various types of ships, and performing finite element simulation modeling and calculation to obtain stress distribution, displacement and total strain of a ship body structure under the combination of various ship types and ice conditions; performing linear strength direct calculation or non-linear strength direct calculation on the obtained data, analyzing a potential failure mode of the ship, and forming a structure ice-resistant grade database in combination with evaluation of the structure ice-resistant grade; an artificial intelligence model is established, and the structural ice-resistant grade database is used for training; basic parameters of a target ship in actual navigation are input into the trained artificial intelligence model, and an ice-level strengthening scheme matched with the target ship is generated through feature extraction, rule reasoning and decision output. According to the method, automation and intellectualization of ice level strengthening design under different ship types and ice area levels are achieved, and the safety, economical efficiency and efficiency of polar navigation ship structure design are improved.

Description

Technical Field

[0001] This invention belongs to the field of ship structure design and intelligent computing technology, specifically involving a ship structure ice-class strengthening method that combines artificial intelligence and structural strength specifications. It is used in the design and modification of ships navigating in polar or cold sea areas to intelligently generate ice-class strengthening schemes that meet the specifications based on ship type, structural characteristics and ice zone level. Background Technology

[0002] With the development of Arctic shipping routes and polar shipping, more and more ships need to navigate in ice-covered environments. To ensure the safety and durability of ships under ice loads, ice-class strengthening of the hull structure is usually required. Current ice-class strengthening methods mainly rely on manual experience or classification society standards, requiring designers to perform calculations and judgments on a case-by-case basis for different ship types and navigation areas. However, existing methods have the following shortcomings:

[0003] (1) The design process is complicated, the strength calculation methods are diverse, and the calculation formulas and empirical rules corresponding to different ship types and structures are different, lacking a unified and efficient application approach;

[0004] (2) The calculation of ice load often differs greatly from the actual ice conditions, making it difficult to reflect the environmental changes in the navigation area in real time;

[0005] (3) Traditional reinforcement design is mostly based on human judgment, which is easily affected by the experience level of the designers, resulting in large differences in the schemes and making it difficult to guarantee optimality and economy.

[0006] Therefore, how to combine standardized structural strength calculation methods with artificial intelligence technology to establish an intelligent ice-class strengthening method that can adapt to different ship types and ice zone levels has become an urgent technical problem to be solved. Summary of the Invention

[0007] The purpose of this invention is to overcome the problems of low design efficiency, reliance on human experience, and lack of intelligence and personalization in existing ship structure ice-class strengthening methods. It proposes a ship structure ice-class strengthening method that combines artificial intelligence and structural strength specifications to achieve automation, intelligence and optimization of ship ice zone structure design.

[0008] The present invention achieves the above-mentioned technical objectives through the following technical means.

[0009] An AI-based method for strengthening ship structures to withstand ice class:

[0010] Obtain basic parameter information for various types of ships, including ship type, dimensions, structural layout, material properties of major components, and ice level of the target navigation area;

[0011] Based on the basic parameter information of various types of ships, through finite element simulation modeling and calculation, the stress distribution, displacement and total strain of the hull structure under various ship types and ice conditions are obtained. The obtained data are directly calculated by linear strength or nonlinear strength to analyze the potential failure modes of the ship. Combined with the evaluation of the ice resistance level of the structure, a database of the ice resistance level of the structure is formed.

[0012] An artificial intelligence model was established and trained using the aforementioned structure's ice resistance level database.

[0013] The structural parameters and ice condition characteristics of the target ship during actual navigation are input into the trained artificial intelligence model. Through feature extraction, rule reasoning, and decision output, an ice-level enhancement scheme adapted to the target ship is generated and transmitted to the target ship.

[0014] Furthermore, in the direct linear strength calculation, a dual check is performed on the ship's yield strength and buckling strength; in the yield strength check, the stress level of the main supporting components is evaluated, and for the web, bulkheads, and deck plates, the nominal shear stress is checked to ensure it does not exceed the allowable value. Among them, R eh For the upper yield strength, the nominal Von Mises stress of the panels with strong ribs and horizontal trusses is checked to be no greater than the allowable value of 1.15R. eh In the buckling strength verification, based on the net member dimensions after deducting corrosion increments, the plate structure subjected to biaxial compressive and shear stresses is evaluated, and the buckling utilization factor η of the structure is calculated. act If η act If the value is greater than 1, increase the plate thickness and continue to adjust the working stress of the structure based on the thickness until η is reached. act ≤1.0.

[0015] Furthermore, in the direct calculation of nonlinear strength, a bilinear stress-strain relationship is used to describe the material properties, and the overload factor CF corresponding to the ice level is introduced into the calculation. O .

[0016] Furthermore, in the direct calculation of nonlinear strength, a bilinear material model and ice load are used to obtain the ice pressure-deformation curve. The elastic and plastic components are separated from the curve to calculate the permanent deformation, while the range of the plastic region and the peak plastic strain are statistically analyzed.

[0017] Furthermore, in the direct calculation of nonlinear strength, the lateral permanent deformation of the structure does not exceed 0.3% of the spacing between its supporting members, the out-of-plane permanent deformation of the structure is controlled within 8 mm, and the plastic strain of the structure does not exceed 0.05.

[0018] Furthermore, for plastic strain verification, after loading to the specified ice load in the nonlinear finite element analysis, the plastic strain of the main supporting components is extracted, and its allowable upper limit of not exceeding 0.05 is verified. If the verification standard is not met, the stiffness and load-bearing capacity of the components are improved by increasing the thickness of the outer plate, reducing the spacing of ribs or longitudinal girders, increasing the section modulus of the skeleton, or adjusting the arrangement of reinforcing components. After adjustment, the linear and nonlinear strength direct calculations are performed again, and the permanent deformation and plastic strain verification are repeated until the plastic strain is not greater than 0.05.

[0019] Furthermore, the structural ice resistance rating database supports closed-loop updates: during actual navigation, the collected hull stress, plastic strain and permanent deformation of key ice-covered sections, hull plate thickness wear in ice-covered areas, and ice conditions in the navigation area are continuously monitored and collected. The monitored data are then used to recalculate and verify the data, regenerate high-fidelity finite element samples, and add them to the constructed structural ice resistance rating database.

[0020] Furthermore, the artificial intelligence model adopts a deep learning-based multi-input neural network structure, including a data input layer, a feature extraction layer, a rule reasoning layer, and a decision output layer. The data input layer receives ship structural parameters and ice condition characteristics. The feature extraction layer uses a deep neural network to perform nonlinear mapping on the ship structural parameters and ice condition characteristics respectively, and then performs fusion processing to obtain high-dimensional features representing the coupling relationship between "ship parameters and ice condition". Based on the above high-dimensional coupling features, the rule reasoning layer performs constraint judgment on the candidate strengthening variable combinations generated by the artificial intelligence model. When a candidate strengthening variable combination does not meet any standard criterion or rule constraint, the rule reasoning layer eliminates or corrects the combination, and only the strengthening variable combination that meets all constraints is passed to the decision output layer as a feasible solution. The decision output layer outputs the strengthening variable combination corresponding to the optimal ice level strengthening scheme in the structural ice resistance level database.

[0021] Furthermore, the feature extraction layer includes multi-layer fully connected neural network sub-modules corresponding to structural parameters and ice condition features, respectively. After normalizing the input features, each sub-module completes feature encoding through multi-layer nonlinear activation functions to obtain structural feature vectors and ice condition feature vectors. These vectors are then concatenated in the feature fusion stage and input to the shared hidden layer to form a unified high-dimensional coupled feature representation.

[0022] Furthermore, the ice-strengthening scheme is formed by a combination of various adjustable strengthening variables, including: increasing the net thickness of the outer plate of the side and bow areas in ice-load sensitive regions; adding ice-strengthening ribs between the longitudinals and ribs, and reducing the spacing between the ribs and longitudinals or rearranging the ribs and longitudinals; changing the type and cross-sectional dimensions of the ribs and longitudinals while meeting structural layout principles and construction constraints; optimizing the arrangement, span, and number of T-sections located in the ice-load area; and installing local ice-strengthening plates or local stiffening ribs in the side, transverse bulkhead, and bow areas.

[0023] Compared with the prior art, the present invention has the following beneficial effects:

[0024] (1) This invention combines the structural ice-resistance level database formed by finite element simulation with artificial intelligence model to realize the automation and intelligence of ice-resistance strengthening design of ship structure, avoiding the limitations of traditional methods that rely on human experience and single standards.

[0025] (2) Based on deep learning and rule reasoning, the present invention can generate targeted reinforcement schemes under normative constraints, thereby improving the design results in terms of security, economy and transparency.

[0026] (3) By establishing a finite element simulation database, this invention enhances the consistency between ice load calculation and actual navigation environment. Attached Figure Description

[0027] Figure 1 This is a flowchart of the ship structure ice-class strengthening method of the present invention;

[0028] Figure 2 Flowchart for building a structural ice resistance rating database;

[0029] Figure 3 This is a structural block diagram of the AI ​​ice-level enhancement system of the present invention;

[0030] Figure 4 This is a diagram showing the relationship between the AI ​​model calling specifications and optimization decisions in this invention.

[0031] Figure 5 Schematic diagram of ice-class reinforcement scheme for ship hull structure;

[0032] In the diagram, 1-outer plating, 2-longitudinal rib, 3-rib, 4-middle girder, 5-transverse support beam, 6-ship side deck frame, 7-deck longitudinal rib, 8-deck frame, 9-ice zone reinforcing rib, 10-ice zone reinforcing plate, 11-ice zone area. Detailed Implementation

[0033] This invention proposes a method for strengthening the ice rating of ship structures by combining artificial intelligence and structural strength specifications. It aims to address the problems of existing methods for strengthening the ice rating of ship structures, which rely on manual experience and single specifications, and lack intelligent and personalized design. The method is described in detail below with specific embodiments.

[0034] like Figure 1 As shown, the ship structure ice-level strengthening method of the present invention includes the following steps in sequence: obtaining ship parameters, finite element simulation calculation and constructing a structural ice-resistance level database, training and reasoning with an artificial intelligence model, and outputting an ice-level strengthening scheme under the constraints of specifications.

[0035] First, basic parameter information for various types of ships is collected. These parameters include hull type, dimensions (overall length, length between perpendiculars, beam, depth, draft, displacement, deadweight tonnage, waterline length), structural layout, material properties of major components, and ice class of the target navigation area. This data can typically be obtained from design drawings, classification society data, and technical documents provided by the shipyard. Based on this, this invention establishes a structural ice resistance rating database through finite element simulation calculations. In finite element simulation software, various ship types (such as oil tankers, bulk carriers, icebreakers, and research vessels), diverse scales and material properties, as well as different ice conditions (including parameters such as ice thickness, ice impact angle, and ice strength) are modeled and set. Through simulation calculations, the stress distribution, displacement, and total strain of the hull structure under various ship types and ice conditions can be obtained. Then, according to the linear strength direct calculation and nonlinear strength direct calculation standard formulas of the "Guidelines for Direct Calculation of Structural Strength under Ice Load (2025)," the data obtained from the finite element software are calculated to analyze potential failure modes (i.e., failure to meet the requirements of yield strength, buckling strength, permanent deformation, and plastic strain in the "Guidelines"), and the ice resistance level of the structure is evaluated in conjunction with the "Guidelines." Finally, a large-scale database is formed as the training basis for artificial intelligence models.

[0036] In actual implementation, such as Figure 2As shown, the construction of the structural ice resistance rating database includes finite element modeling, numerical simulation condition (load, boundary conditions, ice conditions) setting, direct linear strength calculation to generate low-fidelity samples, direct nonlinear strength calculation to generate high-fidelity samples, quality control and criterion verification, and sample normalization labeling and database management. The database samples are generated and managed according to the strategy of "linear wide coverage + nonlinear high precision". First, the hull is modeled in finite element software according to the ship's technical documents. The corresponding working conditions are set as shown in Table 1. The range settings of local component models and overall compartment models are used to avoid boundary effects. Strong frames, longitudinal girder and deck plates are modeled using shell elements. Loads are applied according to the load plate size, position and direction given in the "Guide". When the load plate is inconsistent with the simulation software mesh, the ice pressure is equivalent to keep the total force consistent. Linear analysis is used to evaluate yield and buckling responses in batches and screen weak combinations. Nonlinear analysis uses a bilinear material model and ice load to obtain ice pressure-deformation curves. Elastic and plastic components are separated from the curves to calculate permanent deformation. At the same time, the range of plastic zone and peak plastic strain are statistically analyzed. Finite element method (FEM) software discretizes a continuous structure into a computational model composed of elements and nodes. After applying loads and boundary constraints, it solves the overall equilibrium equations of the structure to obtain the displacement response of each node in each degree of freedom direction. After obtaining the nodal displacements, it calculates the displacement gradient within each element based on the element's displacement interpolation function, thereby obtaining the total strain distribution of the structure. This total strain is then converted into corresponding stresses, yielding the stress distribution of the structure under various loading conditions. When considering material nonlinearity, the FEM software updates the stress-strain relationship based on the material model during loading and obtains the nonlinear response of the structure in incremental solutions.

[0037] Table 1 Verification Status

[0038]

[0039] In the direct linear strength calculation, to ensure that the modeling and calculation meet the technical requirements of the Guidelines, the modeling scope, mesh generation, load application, and boundary conditions of all samples are strictly performed in accordance with the specifications. The meshing scope of the local component model and the overall compartment model refers to the modeling requirements in Section 2.2.1 of the Guidelines, and the boundary constraints and loading methods are set according to the principles in Section 2.2.3 of the Guidelines to ensure that the calculation results can truly reflect the stress state of the structure. The application of ice loads follows the calculation formula given in Section 2.3.1 of the Guidelines. (where: P) adj For the adjusted ice pressure, P avg Let w be the average ice pressure, b be the width of the load plate, b be the height of the load plate, and AF be the hull area factor. mesh b is the mesh boundary width that is closest to the width of the load plate. meshThe mesh boundary height is defined as the closest to the width of the load plate, thus achieving consistency between the load distribution and physical conditions in the finite element calculation. After obtaining the stress distribution and mid-surface stress of the shell elements using finite element software, the yield strength and buckling strength of the ship are double-checked, and the check results are recorded together with the hull structural parameters for subsequent learning and inference. Specifically, in the yield strength check, the stress level of the main supporting components is evaluated by the designer or the calculation program. For the web plate, bulkhead plates, and deck plates, the nominal shear stress is checked to be no greater than the allowable value R. eh / (R) eh (For the upper yield strength); for panels with strong ribs and horizontal trusses, the nominal Von Mises stress is checked to be no greater than the allowable value of 1.15R. eh The corresponding verification results and stress indices are written into the database as sample data for the artificial intelligence model. Based on this, buckling strength verification is further performed during the database construction phase: plate structures subjected to biaxial compressive and shear stresses are evaluated based on the net member dimensions after deducting corrosion increments, using standard formulas. (where: σ) A The corrected working stress is σ, where σ is the calculated working stress and t is the plate thickness in the model. c Thickness correction is applied to the working stress (for corrosion increment), and the buckling utilization factor of the structure is calculated. (In the formula: For the applied equivalent stress, (for equivalent buckling stress), if η act A value greater than 1 indicates that the buckling load capacity of the plate is insufficient at the current plate thickness. The plate thickness should be appropriately increased, and the working stress should be further corrected based on the thickness until η is reached. act ≤1.0, to ensure that stability requirements are met; the above buckling check results are also recorded as sample data.

[0040] In the direct calculation of nonlinear strength, the bilinear stress-strain relationship shown in Section 3.3.1 of the Guidelines is used to describe the material properties, and the overload factor CF corresponding to the ice class is introduced into the calculation. O To reflect the nonlinear response under ice loads, the assessment of lateral permanent deformation, out-of-plane permanent deformation, and plastic strain of the structure was performed in accordance with the provisions of Sections 3.6.2 and 3.6.3 of the Guidelines, respectively, ensuring that the calculation results met the corresponding safety criteria. Specifically, lateral permanent deformation did not exceed 0.3% of the spacing between its supporting members, out-of-plane permanent deformation was controlled within 8 mm, and plastic strain did not exceed 0.05. After all calculations were completed, only samples that met the above-mentioned code limits and convergence requirements were entered into the database to ensure the accuracy, verifiability, and consistency with the code.

[0041] When it is necessary to evaluate the plastic response and ultimate bearing capacity of a structure, incremental loading and unloading analysis is performed using a nonlinear strength direct calculation method to obtain the ice pressure-deformation curve of the structure, and then applied according to the code formula. (where: CF) O Where AF is the overload factor, P is the hull area factor, and P is the overload factor. avg To calculate the ice load P, which includes the overload factor, for the design ice load, the following calculations are performed. e The plastic response was evaluated under this action, and the results were entered into the database as high-fidelity samples. Based on the above nonlinear analysis results, the lateral permanent deformation and out-of-plane permanent deformation of the components were quantitatively evaluated during the database construction phase to ensure that the former does not exceed 0.3% of the spacing between its supporting components and the latter does not exceed the millimeter level given in the specification. The permanent deformation was checked, and the relevant deformation indicators and check conclusions were stored in the database. Meanwhile, for plastic strain verification, the plastic strain of the main supporting components is extracted after loading to the specified ice load in the nonlinear finite element analysis, and its allowable upper limit of no more than 0.05 is verified. If the verification standard is not met, the stiffness and load-bearing capacity of the components need to be improved by increasing the thickness of the outer plate, reducing the spacing of ribs or longitudinal girders, increasing the section modulus of the skeleton, or adjusting the arrangement of reinforcing components. After adjustment, the linear and nonlinear strength direct calculations are re-performed, and the permanent deformation and plastic strain verification are repeated until the plastic strain is no more than 0.05, thereby preventing the structure from failing due to excessive plastic flow. The plastic strain index and verification conclusion are recorded as database samples for the learning of artificial intelligence models and scheme evaluation.

[0042] Building upon the linear strength calculations, the nonlinear strength direct calculation method establishes a nonlinear finite element model for the controlling load conditions and critical structural locations. A bilinear material model is employed, and a design ice load including an overload factor is applied. The nonlinear response of the structure is obtained through incremental loading. Based on this, ice pressure-deformation curves, permanent deformation, and plastic strain are extracted and evaluated according to relevant standards to assess the actual load-bearing capacity of the structure under ice loads. Through this process, the system can adaptively select between linear and nonlinear methods for different ship types, rather than being limited to a single calculation approach.

[0043] All samples must pass through a quality control process before being input into the AI ​​model. Permanent deformation and plastic strain must meet the limits specified in the guidelines. Qualified samples are accompanied by metadata tags such as ship type, component geometry, material parameters, load and mesh information, version, and convergence criteria to ensure traceability and verification. The database supports closed-loop updates: During actual navigation, the system continuously monitors and collects data such as hull stress, plastic strain and permanent deformation of key ice-covered sections, hull plate thickness and wear in ice-covered areas, and ice conditions in the navigation area. The obtained monitoring data is then used for recalculation and verification, thereby achieving iterative optimization of ice-class reinforcement schemes. In accordance with the provisions of the Guidelines on the criterion for permanent deformation and plastic strain, when monitoring results indicate that the lateral permanent deformation of a major supporting component (e.g., a T-shaped member) under ice load exceeds 0.3% of the spacing between its supporting components, or the out-of-plane permanent deformation exceeds 8 mm, or the plastic strain is greater than 0.05, the component is deemed to have exceeded the allowable range of the Guidelines. Within the framework of linear and nonlinear direct calculation modeling and load case arrangement specified in the Guidelines, and combined with the monitored actual ice load level (i.e., hull stress) and plate thickness wear, high-fidelity finite element samples are regenerated. The new calculation results, along with updated material parameters, load cases, and criterion conclusions, are written into the structural ice-resistant database. Through this closed-loop update mechanism, the database samples can be dynamically expanded and corrected as the operating environment and material state evolve. Based on this, the artificial intelligence model continuously corrects and optimizes subsequent ice-level strengthening schemes, thereby ensuring that the recommended schemes always meet the code safety criterion and have engineering verifiability.

[0044] like Figure 3 As shown, the AI-powered ice-level enhancement system of this invention is based on a structural ice-resistance level database. It inputs target ship parameters and ice condition characteristics into an AI model, and through feature extraction, rule reasoning, and decision output, automatically generates an ice-level enhancement scheme adapted to the target ship. The AI ​​model used in this invention is a deep learning-based Multiple-input Neural Network (MINN) structure, which consists of a data input layer, a feature extraction layer, a rule reasoning layer, and a decision output layer. The data input layer receives ship structural parameters and ice condition characteristics. The feature extraction layer uses a deep neural network to perform nonlinear mapping and fusion processing on the ship structural parameters and ice condition characteristics respectively, to obtain high-dimensional features representing the coupling relationship between "ship parameters and ice conditions." The feature extraction layer includes multi-layer fully connected neural network sub-modules corresponding to structural parameters and ice condition characteristics respectively. Each sub-module normalizes the input features and completes feature encoding through multi-layer nonlinear activation functions. The resulting structural feature vector and ice condition feature vector are concatenated in the feature fusion stage and input to a shared hidden layer to form a unified high-dimensional coupled feature representation, used to represent the nonlinear coupling relationship between ship structural parameters and ice conditions. Figure 4As shown, the rule-based reasoning layer embeds standard formulas and an empirical rule base to constrain and screen candidate strengthening variable combinations and determine their feasibility. Based on the aforementioned high-dimensional coupling characteristics, the rule-based reasoning layer constrains and determines the candidate strengthening variable combinations generated by the artificial intelligence model. These constraints include criterion conditions such as yield strength, buckling strength, permanent deformation, and plastic strain set according to the "Guidelines," as well as pre-established engineering empirical rules (engineering empirical rules are empirical constraint rules formed in ship ice-class design and structural strengthening practice, used to guide the selection of strengthening measures, determine the rationality of their arrangement, and assess the feasibility of engineering implementation; they are used to further screen and optimize candidate strengthening schemes while meeting the standard formulas). This embodiment... In this model, the application of engineering experience rules includes plate thickness, rib spacing, stiffening, and profiles. Their structural layout, reinforcement priority, component proportions, and layout rationality must meet design experience requirements. When a candidate combination of reinforcement variables does not meet any specification criterion or rule constraint, the rule reasoning layer eliminates or modifies the combination, only passing the combination of reinforcement variables that satisfies all constraints as a feasible solution to the decision output layer. The decision output layer outputs the combination of reinforcement variables corresponding to the optimal ice-class reinforcement scheme in the structural ice-resistance level database. These reinforcement variables include at least outer plate thickness, rib / girder spacing, stiffener type and size, and T-section arrangement, thereby automatically generating ice-class reinforcement schemes for the target vessel. The model can balance the influence of ship type parameters and external working conditions on the results during the training phase, ensuring the integrity and accuracy of information transmission. Model training adopts a phased strategy, including a pre-training phase and a fine-tuning phase. The pre-training phase utilizes a large-scale low-fidelity sample database generated based on the linear strength direct calculation method for initial training to establish the mapping relationship between ship structural parameters and ice-resistant response. The fine-tuning phase employs high-fidelity sample data obtained based on the nonlinear strength direct calculation method for retraining to improve prediction accuracy and prevent overfitting. During training, the backpropagation algorithm is used for parameter optimization, with the mean absolute error (MAE) loss function selected to measure the deviation between the model output and simulation results. Hyperparameter optimization is performed using the Bayesian optimization algorithm, obtaining optimal training parameters through a combination of learning rate and decay rate. The training dataset is divided into three parts: a training set, a validation set, and a test set, used for model weight updates, hyperparameter adjustments, and generalization performance evaluation, respectively. Validation shows that the model achieves a goodness of fit of 0.998 on the high-fidelity database, stably predicting structural responses under different ship types and ice condition combinations, providing high-precision data support for ice-class reinforcement design.

[0045] In practical applications, the artificial intelligence model first determines the similarity between the target ship's hull type and samples in the structural ice resistance rating database, and selects the corresponding sample set or weights accordingly. This is used to predict the target ship's structural response under given ice conditions based on the structural strength calculations already completed in the database, and to recommend ice-level reinforcement schemes. After the strength calculations are completed, the AI ​​model, based on the structural strength calculations performed according to the guidelines and under the constraints of the rule-based reasoning layer, proposes specific ice-level reinforcement schemes for the hull structure. These ice-level reinforcement schemes correspond one-to-one with the reinforcement variables used during the establishment of the structural ice resistance rating database. The ice-level reinforcement schemes are formed by combining multiple adjustable reinforcement variables, mainly including: increasing the net thickness of the outer plating in ice-sensitive areas such as the sides and bow to improve the local load-bearing capacity of the outer plating under ice loads; and adding ice-strengthening ribs (such as...) between the longitudinal ribs and the ribs. Figure 5 As shown), by reducing the spacing between ribs and longitudinals or rearranging them, the overall bending and shear resistance of the side structure is enhanced; under the premise of meeting structural layout principles and construction constraints, the type and cross-sectional dimensions of the ribs and longitudinals are changed to improve the section modulus and bearing stiffness of the main supporting components; for T-sections located in the ice load area, their arrangement, span, and quantity are optimized to improve the coordinated load-bearing capacity of the longitudinal and transverse support systems on the side; in ice load concentrated areas such as the side, transverse bulkheads, and foreposts, local ice-strengthening plates (such as...) are installed. Figure 5 (As shown) or local stiffening ribs, forming targeted reinforcement measures for ice-loaded areas. In the above reinforcement scheme, the adjustment of various reinforcement variables is based on the requirements of the "Guidelines" to ensure that permanent deformation and plastic strain meet the fixed requirements of the "Guidelines", namely, lateral permanent deformation exceeding 0.3% of the spacing between its supporting members, out-of-plane permanent deformation exceeding 8 mm, and plastic strain greater than 0.05. The artificial intelligence model learns the applicable scope and combination rules of the above standardized reinforcement measures by analyzing samples in the database of "ship parameters - ice condition - strength calculation results - reinforcement variable combination". In the prediction stage, it automatically selects and combines one or more sets of reinforcement variables for the target ship to form an ice-class reinforcement scheme that meets the standard requirements and is suitable for the ship type and target ice class.

[0046] In summary, this invention, by combining artificial intelligence with structural strength standards, constructs a complete process from ship parameter acquisition, ice condition identification, strength calculation to reinforcement scheme output. Compared with traditional methods, this invention can achieve personalized and intelligent ice-class reinforcement design under different ship types, navigation areas, and material conditions, significantly improving the safety and economy of polar vessels.

[0047] The embodiments described above are preferred embodiments of the present invention, but the present invention is not limited to the above embodiments. Any obvious improvements, substitutions or modifications that can be made by those skilled in the art without departing from the essence of the present invention shall fall within the protection scope of the present invention.

Claims

1. A method for strengthening the ice class of ship structures based on artificial intelligence, characterized in that: Obtain basic parameter information for various types of ships, including ship type, dimensions, structural layout, material properties of major components, and ice level of the target navigation area; Based on the basic parameter information of various types of ships, through finite element simulation modeling and calculation, the stress distribution, displacement and total strain of the hull structure under various ship types and ice conditions are obtained. The obtained data are directly calculated by linear strength or nonlinear strength to analyze the potential failure modes of the ship. Combined with the evaluation of the ice resistance level of the structure, a database of the ice resistance level of the structure is formed. An artificial intelligence model was established and trained using the aforementioned structure's ice resistance level database. The structural parameters and ice condition characteristics of the target ship during actual navigation are input into the trained artificial intelligence model. Through feature extraction, rule reasoning, and decision output, an ice-level enhancement scheme adapted to the target ship is generated and transmitted to the target ship.

2. The method for strengthening the ice class of ship structures based on artificial intelligence according to claim 1, characterized in that, In the direct linear strength calculation, the yield strength and buckling strength of the ship are double-checked. In the yield strength check, the stress level of the main supporting components is evaluated, and for the web, bulkheads, and deck plates, the nominal shear stress is checked to ensure that it does not exceed the allowable value. , where R eh For the upper yield strength, the nominal Von Mises stress of the panels with strong ribs and horizontal trusses is checked to be no greater than the allowable value of 1.15R. eh In the buckling strength verification, based on the net member dimensions after deducting corrosion increments, the plate structure subjected to biaxial compressive and shear stresses is evaluated, and the buckling utilization factor η of the structure is calculated. act If η act If the value is greater than 1, increase the plate thickness and continue to adjust the working stress of the structure based on the thickness until η is reached. act ≤1.

0.

3. The method for strengthening the ice class of ship structures based on artificial intelligence according to claim 2, characterized in that, In the direct calculation of nonlinear strength, a bilinear stress-strain relationship is used to describe the material properties, and an overload factor CF corresponding to the ice class is introduced into the calculation. O .

4. The method for strengthening the ice class of ship structures based on artificial intelligence according to claim 3, characterized in that, In the direct calculation of nonlinear strength, a bilinear material model and ice load are used to obtain the ice pressure-deformation curve. The elastic and plastic components are separated from the curve to calculate the permanent deformation, while the range of the plastic region and the peak plastic strain are statistically analyzed.

5. The method for strengthening the ice class of ship structures based on artificial intelligence according to claim 4, characterized in that, In the direct calculation of nonlinear strength, the lateral permanent deformation of the structure shall not exceed 0.3% of the spacing between its supporting members, the out-of-plane permanent deformation of the structure shall be controlled within 8 mm, and the plastic strain of the structure shall not exceed 0.

05.

6. The method for strengthening the ice class of ship structures based on artificial intelligence according to claim 5, characterized in that, For plastic strain verification, after loading to the specified ice load in nonlinear finite element analysis, the plastic strain of the main supporting components is extracted and verified to be no greater than 0.

05. If the verification standard is not met, the stiffness and load-bearing capacity of the components are improved by increasing the thickness of the outer plate, reducing the spacing of ribs or longitudinal girder, increasing the section modulus of the skeleton, or adjusting the arrangement of reinforcing components. After adjustment, the linear and nonlinear strength direct calculations are performed again, and the permanent deformation and plastic strain verification are repeated until the plastic strain is no greater than 0.

05.

7. The method for strengthening the ice class of ship structures based on artificial intelligence according to claim 6, characterized in that, The structural ice resistance database supports closed-loop updates: During actual navigation, the collected data on hull stress, plastic strain and permanent deformation of key ice-covered sections, hull plate thickness wear in ice-covered areas, and ice conditions in the navigation area are continuously monitored and verified. High-fidelity finite element samples are then regenerated and input into the structural ice resistance database.

8. The method for strengthening the ice class of ship structures based on artificial intelligence according to claim 1, characterized in that, The artificial intelligence model employs a deep learning-based multi-input neural network structure, including a data input layer, a feature extraction layer, a rule inference layer, and a decision output layer. The data input layer receives ship structural parameters and ice condition characteristics. The feature extraction layer uses a deep neural network to perform nonlinear mapping on the ship structural parameters and ice condition characteristics, followed by fusion processing to obtain high-dimensional features representing the coupling relationship between "ship parameters and ice condition." Based on these high-dimensional coupling features, the rule inference layer constrains and determines the candidate strengthening variable combinations generated by the artificial intelligence model. When a candidate strengthening variable combination does not meet any standard criterion or rule constraint, the rule inference layer eliminates or corrects the combination, only passing the strengthening variable combination that satisfies all constraints as a feasible solution to the decision output layer. The decision output layer outputs the strengthening variable combination corresponding to the optimal ice-level strengthening scheme in the structural ice resistance level database.

9. The method for strengthening the ice class of ship structures based on artificial intelligence according to claim 8, characterized in that, The feature extraction layer includes multi-layer fully connected neural network sub-modules corresponding to structural parameters and ice condition features, respectively. After normalizing the input features, each sub-module completes feature encoding through multi-layer nonlinear activation functions to obtain structural feature vectors and ice condition feature vectors. These vectors are then concatenated in the feature fusion stage and input into the shared hidden layer to form a unified high-dimensional coupled feature representation.

10. The method for strengthening the ice class of ship structures based on artificial intelligence according to claim 1, characterized in that, The ice-strengthening scheme is formed by a combination of various adjustable strengthening variables, including: increasing the net thickness of the outer plating in the ice-load sensitive areas of the hull side and bow; adding ice-strengthening ribs between the longitudinals and ribs, and reducing the spacing between the ribs and longitudinals or rearranging the ribs and longitudinals; changing the type and cross-sectional dimensions of the ribs and longitudinals while meeting structural layout principles and construction constraints; optimizing the arrangement, span, and number of T-sections located in the ice-load area; and installing local ice-strengthening plates or local stiffening ribs in the hull side, transverse bulkhead, and bow areas.

Images