Ship structure standardization design data analysis method and system based on AI

Abstract

The invention provides an AI-based ship structure standardized design data analysis method and system, and the method comprises the steps: S1, extracting feature data based on a historical ship type structure three-dimensional model, and building a training database; s2, selecting a machine learning algorithm based on the feature data of the training database, and training a machine learning model; s3, for the newly designed ship model, extracting feature data of the structural member based on the current design model; s4, on the basis of the machine learning model trained in the S2, inputting the feature data extracted in the S3, and outputting various design prediction data of the structural component by the machine learning model, and S5, arranging the predicted standard component at a corresponding position in three-dimensional design software on the basis of the various design prediction data, and completing standardized design; and S6, inputting data of the newly designed ship model into the training database, and repeatedly executing S1 and S2 to train the machine learning model to realize iterative optimization of the ship model.

Description

Technical Field

[0001] The present application belongs to the technical field of ship structure design, and in particular, relates to an AI-based ship structure standardized design data analysis method and system. Background Art

[0002] Traditional ship structure design mostly relies on manual analysis, which usually has the following problems: 1) Low efficiency of manual design: The number of ship structure parts is huge and the features are complex. Taking a liquefied gas ship as an example, the number of its structural parts can reach 120,000, of which patching accounts for about 18%, involving more than 40 types. Manual analysis can easily lead to errors and inefficiency. 2) Poor design accuracy: Traditional methods are difficult to effectively associate multi-dimensional features, such as plate thickness, material, opening location, surrounding structure specifications, etc., resulting in inaccurate standardized classification. 3) Poor dynamic adaptability: When a new ship type or design rule changes, the traditional method needs to be manually readjusted, historical design data is not effectively utilized, and there is a lack of automated iteration mechanism, and it is impossible to improve the quality of subsequent designs through iterative learning.

[0003] Therefore, there is an urgent need for a standardized ship structure design method that integrates AI technology, supports dynamic iteration, and is deeply integrated with design tools. Summary of the invention

[0004] In view of the above-mentioned shortcomings of the prior art, the purpose of this application is to provide an AI-based ship structure standardized design data analysis method and system to significantly improve the efficiency and standardization level of ship design.

[0005] In a first aspect, the present application provides an AI-based ship structure standardized design data analysis method, comprising the following steps:

[0006] S1: Extract feature data and establish a training database based on the structural 3D model of historical ship types;

[0007] S2: Based on the feature data of the training database, select a machine learning algorithm and train a machine learning model;

[0008] S3: For the newly designed ship model, based on the current design model, all characteristic data of the structural parts, including position information, are extracted;

[0009] S4: Based on the machine learning model trained in S2, the feature data extracted in S3 is input, and the machine learning model outputs various design prediction data of the structural parts;

[0010] S5: Based on various design prediction data, the predicted standard parts are arranged at corresponding positions in the 3D design software to complete the standardized design;

[0011] S6: Enter the data of the newly designed ship model into the training database, and repeat S1 and S2 to train the machine learning model to achieve iterative optimization of the ship model.

[0012] In an optional implementation manner, step S1 includes:

[0013] S11: In the 3D design software, based on the structural 3D model of the historical ship type, the structured feature data and label data are extracted;

[0014] S12: Perform data cleaning on the extracted data, complete missing feature data, and correct or delete erroneous feature data;

[0015] S13: Establish a training database and save the data cleaned in S12 to the database.

[0016] In an optional implementation manner, the feature data includes structured feature data and label data; wherein,

[0017] The structured feature data includes: the name, structure type, size, area, plate thickness, material, weight, number of boundaries, number and value of arcs, number and type, size and position of openings, type and specification of surrounding structures, number of profiles attached to the structure and their specification and position coordinates;

[0018] The label data includes: the standard part type, standard part name and position coordinates of the structural part.

[0019] In an optional implementation, in step S2, based on the feature data of the training database, different machine learning models are trained according to ship type, region, and standard part type.

[0020] In an optional implementation manner, step S3 includes:

[0021] S31: In the three-dimensional design software, based on the three-dimensional structural model of the historical ship type, extract the same data as in step S12;

[0022] S32: Perform data cleaning on the extracted data, complete missing feature data, and correct or delete erroneous feature data.

[0023] In an optional implementation, the machine learning algorithm includes one or more of a decision tree, a random forest, or a gradient boosting tree.

[0024] In an optional implementation, in the analysis of hull structure data, a linear regression algorithm is used to predict the continuous properties of the weight and plate thickness of the structural parts; a ridge regression algorithm and a lasso regression algorithm are used to improve the accuracy of the prediction.

[0025] In an optional implementation, in the analysis of hull structure data, a K-means clustering algorithm is used to find similar groups of structural parts to achieve inventory management; a hierarchical clustering algorithm is used to provide a clustering strategy to find different levels of structural similarity.

[0026] In the second aspect, the present application provides an AI-based ship structure standardized design data analysis system, including:

[0027] A data extraction module, used to extract feature data from the three-dimensional model;

[0028] Model training module, used to train and optimize machine learning models;

[0029] Prediction module, used to generate prediction results of standard parts type and location;

[0030] 3D model maintenance module for automatic layout of standard parts;

[0031] A visualization module is used to display the comparative analysis of model prediction results and manual correction suggestions.

[0032] Compared with the prior art, the technical solution provided by this application has the following beneficial effects:

[0033] The technical solution provided in this application is based on AI technology to analyze the standardized design data of ship structures. By taking advantage of the advantages of artificial intelligence technology, it can automatically process and analyze complex ship data and provide in-depth insights for enterprises. At the same time, the method also adopts standardized design ideas to make the data analysis process more standardized and value-added, thereby improving the accuracy and efficiency of data analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 Flow chart of the ship structure standardized design data analysis method provided for this application;

[0035] Figure 2 The diagram of the ship structure standardization design data analysis system provided for this application;

[0036] Figure 3 Schematic diagram of the data attributes for the standardized design of ship structures provided for this application. DETAILED DESCRIPTION

[0037] The following describes the implementation methods of the present application through specific examples. Those skilled in the art can easily understand other advantages and principles of the present application from the contents disclosed in this specification. The present application can also be implemented or applied through other different specific implementation methods. The details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present application.

[0038] It should be noted that the structures, proportions, sizes, etc. illustrated in the drawings of this specification are only used to match the contents disclosed in the specification for people familiar with this technology to understand and read, and are not used to limit the restrictive conditions for the implementation of this patent, so they have no substantial technical significance. Any structural modification, change in proportional relationship or adjustment of size, without affecting the effects and purposes that can be achieved by this patent, should still fall within the scope of the technical content disclosed by this patent. At the same time, the terms such as "upper", "lower", "left", "right", "middle" and so on quoted in this specification are only for the convenience of description, and are not used to limit the scope of the implementation of this patent. The change or adjustment of their relative relationship should also be regarded as the scope of the implementation of this patent without substantial change in the technical content.

[0039] Embodiment 1:

[0040] This embodiment provides an AI-based ship structure standardized design data analysis method, comprising the following steps:

[0041] S1: Based on the structural 3D model of historical ship types, feature data is extracted and a training database is established.

[0042] S11: In the 3D design software, based on the 3D structural model of the historical ship type, extract structured feature data and label data. Further, the 3D structural model includes profiles, brackets, patch plates, components, etc., and the feature data includes structured feature data and label data; wherein the structured feature data includes: the name of the structural part, the structure type, size, area, plate thickness, material, weight, number of boundaries, number of arcs and value, number of openings and type, size and position, surrounding structure type and specification, number of profiles attached to the structural part and its specification position coordinates; label data includes: standard part type, standard part name and position coordinates of the structural part.

[0043] Taking a certain type of ship as an example, the structured feature data of the ship includes:

[0044] Geometric attributes: plate name (such as "cargo hold area plate"), type (flat / arc), size (length × width × thickness: 1200mm × 800mm × 12mm), area (0.96m 2 ), plate thickness (12mm), material (AH36 steel), weight (92kg), number of boundaries (4), number of openings (2), opening diameter (50mm and 80mm), opening position coordinates (X=300mm, Y=200mm; X=900mm, Y=600mm);

[0045] Topological attributes: adjacent structure type (longitudinal, transverse bulkhead), profile connection method (welding), surrounding structure specifications (plate thickness 10mm, material DH36);

[0046] Label data: standard part type (such as "standard patch plate type A"), name (such as "A-12-50-80"), position coordinates (X=2050mm, Y=4800mm, Z=1200mm).

[0047] Through multi-dimensional feature extraction, the properties of structural parts are comprehensively summarized to provide high-information input for the model. The label data can clarify the type and location of standard parts, directly link them to design goals, and improve model training efficiency.

[0048] S12: Perform data cleaning on the extracted data, complete the missing feature data, and correct or delete the erroneous feature data.

[0049] Let's continue to take a certain type of ship as an example:

[0050] Missing value processing: For the 5% missing "opening position" data, interpolation is performed using the opening distribution law of adjacent patch panels;

[0051] Outlier correction: Delete abnormal records with plate thickness exceeding 30mm and correct material labeling errors, accounting for about 0.3%;

[0052] Noise filtering: outlier data with weight deviation exceeding ±20% are eliminated, accounting for about 1.2%.

[0053] Through the above data processing operations, the data quality can be significantly improved, the interference of noise on model training can be reduced, and data integrity can be ensured through interpolation and correction to avoid model deviation due to missing values.

[0054] S13: Establish a training database and save the data cleaned in S12 to the database.

[0055] As an example, the cleaned data is stored in a MySQL database and classified by fields (ship type, region, feature attributes, and tags), forming a total of 12,000 valid training data. Structured storage facilitates fast retrieval and model calling, and classification management supports on-demand training of sub-models, such as sub-models that are further divided by ship type or region.

[0056] S2: Based on the feature data of the training database, select a machine learning algorithm and train the machine learning model.

[0057] Specifically, based on the characteristic data of the training database, different machine learning models are trained according to ship type, region, and standard part type.

[0058] Optionally, a decision tree algorithm is used. In the standardized design of ship hull structures, the algorithm classifies structural parts into standard parts types according to characteristics such as structure type, size, and material.

[0059] Optionally, a random forest algorithm is used. In the analysis of hull structure data, the algorithm can handle a large number of features and effectively reduce overfitting and improve classification accuracy.

[0060] Optionally, a gradient boosting tree algorithm (such as XGBoost, LightGBM) is used. In the standardized design of hull structures, XGBoost or LightGBM can predict the type of standard parts based on complex feature combinations and interactions, which is particularly suitable for processing large-scale and high-dimensional data sets.

[0061] Optionally, a linear regression algorithm is used. In the analysis of hull structure data, this algorithm is used to predict continuous properties such as weight and plate thickness of structural parts, but it is necessary to pay attention to handling possible nonlinear relationships and outliers.

[0062] Optionally, Ridge Regression and Lasso regression algorithms are used. In the analysis of hull structure data, these two regression methods can help select important features and improve the accuracy of prediction. It can be understood that Lasso regression is particularly suitable for scenarios where the number of features needs to be reduced.

[0063] Optionally, a convolutional neural network algorithm (CNN) is used. Although direct application to hull structure data may require preprocessing, such as converting CAD drawings into image operations, CNN can be used to identify and analyze structural features in drawings, such as profile layout, opening locations, etc. In addition, the hull structure data can also be converted into a two-dimensional or three-dimensional grid form and CNN can be used for feature extraction and classification.

[0064] Optionally, the recurrent neural network algorithm (RNN) and the long short-term memory network algorithm (LSTM) are used. In the analysis of hull structure data, the application of RNN and LSTM may be relatively limited because they are more suitable for processing data with obvious sequence properties, such as time series analysis, natural language processing, etc. However, if there is a sequence dependency in the hull structure data, such as the assembly order of parts, these algorithms can be considered for analysis.

[0065] Optionally, a principal component analysis algorithm (PCA) is used. In the analysis of hull structure data, this algorithm can help identify the most important feature combinations and reduce the dimension of the data set, thereby simplifying the subsequent analysis and modeling process.

[0066] Optionally, a mutual information algorithm (MI) is used. In the standardized design of hull structures, MI can be used to select the features most relevant to the target variable, such as selecting the type of standard parts, for further analysis, which helps to reduce unnecessary computational burden and improve the accuracy of the model.

[0067] Optionally, a K-means clustering algorithm is used. In the analysis of hull structure data, the algorithm can be used to find similar groups of structural parts to support inventory management and optimized design.

[0068] Optionally, a hierarchical clustering algorithm is used. In the analysis of hull structure data, hierarchical clustering can provide a more flexible clustering strategy and help discover structural similarities at different levels.

[0069] Optionally, a genetic algorithm (GA) is used. In the design of ship hull structure, the algorithm can be used to optimize parameters such as size, shape and material selection of structural parts to achieve goals such as lightweight design, cost reduction and performance improvement.

[0070] Optionally, a particle swarm optimization algorithm (PSO) is used. Similar to GA, PSO can also be used for the optimization design of ship hull structure. By adjusting the speed and position parameters of the particles, PSO can explore the solution space and find the optimal solution that meets the design requirements.

[0071] Optionally, a rule engine can be used to implement complex business logic and decision-making processes in the standardized design of ship hull structures. For example, it can automatically determine the type and classification of standard parts based on the size, material, and purpose of the structural parts.

[0072] Optionally, an expert system can be used. In the design of ship hull structures, the expert system can integrate the knowledge and experience of domain experts to provide intelligent assistance and decision support for designers. For example, it can recommend appropriate structural component types and layout solutions based on the complexity and design requirements of the hull structure.

[0073] As an example, a hierarchical model training method with a three-level model is used. Hierarchical training can reduce model complexity and improve classification accuracy, for example:

[0074] Level 1 model (ship type classification): uses the random forest algorithm, with input features including plate thickness, opening diameter, and material, and outputs ship type classification (liquefied gas carrier / bulk carrier), with an accuracy rate of 95%;

[0075] Secondary model (regional classification): Using the XGBoost algorithm, the characteristic differences between the cargo area and the cabin area (such as opening density and surrounding structure specifications) were input, and the regional classification accuracy was 92%;

[0076] Level 3 model (standard parts classification): uses the LightGBM algorithm, inputs all feature data, and outputs the standard type of patching, with an accuracy rate of 93%.

[0077] Furthermore, for the layered training method, we will continue to implement auxiliary algorithm application

[0078] Lasso regression: filter key features (“plate thickness”, “opening diameter”, “material” contribution 70%), simplify the model, and reduce redundant features.

[0079] K-means clustering (k=5): The plate replacements are divided into five groups, which supports inventory optimization (group 3 corresponds to "type A plate replacement", and the inventory requirement is reduced by 15%), reducing the spare parts cost.

[0080] S3: For the newly designed ship model, all feature data of the structural parts, including position information, are extracted based on the current design model.

[0081] S31: In the 3D design software, based on the structural 3D model of the historical ship type, extract the same data as in step S12; for example, extract feature data at a certain position in the cargo hold area of ​​a newly designed liquefied gas ship: plate thickness 12mm, opening diameter 50mm and 80mm, material AH36, adjacent structure is longitudinal bone, plate thickness 10mm.

[0082] S32: Clean the extracted data, complete missing feature data, and correct or delete erroneous feature data. After cleaning, the data is consistent with the format of the training set and has no missing or abnormal data.

[0083] S4: Based on the machine learning model trained in S2, the feature data extracted by S3 is input, and the machine learning model outputs various design prediction data of structural parts.

[0084] As an example, input feature data into the three-level model, and the output prediction result is "standard patch type A", with position coordinates: X = 2050mm, Y = 4800mm, Z = 1200mm. The model can quickly output results, shorten the design cycle by 60%, and achieve high-precision prediction, with a verified accuracy rate of ≥ 90%, greatly reducing the need for manual review.

[0085] S5: Based on a variety of design prediction data, the predicted standard parts are arranged at corresponding positions in the 3D design software to complete the standardized design.

[0086] S51: Call the "A-12-50-80" patch plate 3D model in the standard parts library through the CATIA API interface.

[0087] S52: The model is automatically assembled to the predicted coordinates and verified by the rule engine. Automated assembly reduces manual operations by at least 30%. Conflict detection is performed and the adjacent longitudinal bone spacing is detected to be 600mm, which complies with the DNV GL specification (≥500mm) and the verification is passed. The rule engine ensures design compliance and reduces rework rate.

[0088] S53: Generate a visual report, marking the predicted position and the actual assembly result (error ≤ 1mm)

[0089] S6: Enter the data of the newly designed ship model into the training database, and repeat S1 and S2 to train the machine learning model to achieve iterative optimization of the ship model.

[0090] As an example, 200 sets of newly designed patch data were entered into the training database, and the model was updated using incremental learning (online random forest). After retraining, the classification accuracy was improved to 94.5%. This step is a closed-loop optimization mechanism that can continuously improve model performance, and incremental learning can also reduce the computational cost of retraining.

[0091] This embodiment also provides an AI-based ship structure standardized design data analysis system, including:

[0092] A data extraction module, used to extract feature data from the three-dimensional model;

[0093] Model training module, used to train and optimize machine learning models;

[0094] Prediction module, used to generate prediction results of standard parts type and location;

[0095] 3D model maintenance module for automatic layout of standard parts;

[0096] A visualization module is used to display the comparative analysis of model prediction results and manual correction suggestions.

[0097] It should be noted that it should be understood that the division of the various modules of the above device is only a division of logical functions. In actual implementation, they can be fully or partially integrated into one physical entity, or they can be physically separated. And these modules can all be implemented in the form of software called by processing elements; they can also be all implemented in the form of hardware; some modules can also be implemented in the form of software called by processing elements, and some modules can be implemented in the form of hardware. For example, the model training module can be a separately established processing element, or it can be integrated in a chip of the above device. In addition, it can also be stored in the memory of the above device in the form of program code, and called and executed by a processing element of the above device. The function of the above model training module. The implementation of other modules is similar. In addition, these modules can be fully or partially integrated together, or they can be implemented independently. The processing element described here can be an integrated circuit with signal processing capabilities. In the implementation process, each step of the above method or each module above can be completed by an integrated logic circuit of hardware in the processor element or instructions in software form.

[0098] In summary, the technical solution provided in this embodiment utilizes the advantages of artificial intelligence technology. First, relevant feature data and label data are extracted from historically designed ship types through a three-dimensional model data extraction module; then these data are stored in a database as training data, and the training data is used to select and train a machine learning model in a machine learning model training module, so as to extract effective feature data from massive historical ship three-dimensional models and construct a standardized classification model; when designing a new ship, the trained machine learning model is used to intelligently predict which positions can adopt which type of standardized design, and the predicted data is provided to the three-dimensional model maintenance module. The designer selects the corresponding standard parts and arranges them in the correct position to realize intelligent design of ship standardization, and uses the AI ​​model to predict the type and position of standard parts of the newly designed ship, reduce manual intervention, and improve the accuracy and adaptability of standardized design by continuously iterating and updating the model.

[0099] The technical solution provided by this application has high industrial utilization value because it effectively overcomes various shortcomings in the existing technology.

[0100] The above embodiments are merely illustrative of the principles and effects of the present application and are not intended to limit the present application. Anyone familiar with the technology may modify or change the above embodiments without violating the spirit and scope of the present application. Therefore, all equivalent modifications or changes made by a person of ordinary skill in the art without departing from the spirit and technical ideas disclosed in the present application shall still be covered by the claims of the present application.

Claims

1. An AI-based ship structure standardized design data analysis method, characterized in that: The following steps are involved: S1: Extract feature data and establish a training database based on the structural 3D model of historical ship types; S2: Based on the feature data of the training database, select a machine learning algorithm and train a machine learning model; S3: For the newly designed ship model, based on the current design model, all characteristic data of the structural parts, including position information, are extracted; S4: Based on the machine learning model trained in S2, the feature data extracted in S3 is input, and the machine learning model outputs various design prediction data of the structural parts; S5: Based on various design prediction data, the predicted standard parts are arranged at corresponding positions in the 3D design software to complete the standardized design; S6: Enter the data of the newly designed ship model into the training database, and repeat S1 and S2 to train the machine learning model to achieve iterative optimization of the ship model.

2. The AI-based ship structure standardized design data analysis method according to claim 1 is characterized in that: Step S1 includes: S11: In the 3D design software, based on the structural 3D model of the historical ship type, the structured feature data and label data are extracted; S12: Perform data cleaning on the extracted data, complete missing feature data, and correct or delete erroneous feature data; S13: Establish a training database and save the data cleaned in S12 to the database.

3. The AI-based ship structure standardized design data analysis method according to claim 2 is characterized in that: The feature data includes structured feature data and label data; wherein, The structured feature data includes: the name of the structural part, the structural type, size, area, plate thickness, material, weight, number of boundaries, number and value of arcs, number and type, size and position of openings, type and specification of surrounding structures, number of profiles attached to the structural part and their specification and position coordinates; The label data includes: the standard part type, standard part name and position coordinates of the structural part.

4. The AI-based ship structure standardized design data analysis method according to claim 1 is characterized in that: In step S2, based on the characteristic data of the training database, different machine learning models are trained according to ship type, region, and standard part type.

5. The AI-based ship structure standardized design data analysis method according to claim 2 is characterized in that: Step S3 includes: S31: In the three-dimensional design software, based on the three-dimensional structural model of the historical ship type, extract the same data as in step S12; S32: Perform data cleaning on the extracted data, complete missing feature data, and correct or delete erroneous feature data.

6. The AI-based ship structure standardized design data analysis method according to claim 1 is characterized in that: The machine learning algorithm includes one or more of a decision tree, a random forest, or a gradient boosting tree.

7. The AI-based ship structure standardized design data analysis method according to claim 1 is characterized in that: In the analysis of hull structure data, linear regression algorithm is used to predict the continuous properties of weight and plate thickness of structural parts; ridge regression algorithm and Lasso regression algorithm are used to improve the accuracy of prediction.

8. The AI-based ship structure standardized design data analysis method according to claim 1 is characterized in that: In the analysis of hull structure data, the K-means clustering algorithm is used to find similar groups of structural parts to achieve inventory management; the hierarchical clustering algorithm is used to provide a clustering strategy to find different levels of structural similarity.

9. An AI-based ship structure standardized design data analysis system, characterized in that: include: A data extraction module, used to extract feature data from the three-dimensional model; Model training module, used to train and optimize machine learning models; Prediction module, used to generate prediction results of standard parts type and location; 3D model maintenance module for automatic layout of standard parts; A visualization module is used to display the comparative analysis of model prediction results and manual correction suggestions.

Images