Intelligent design method for hydraulic retaining wall based on BIM and AI

Abstract

The invention provides an intelligent design method for a hydraulic retaining wall based on BIM and AI, and the method converts the design, calculation, drawing and calculation work of the current hydraulic retaining wall from a working mode of data sealing and mutual splitting into a working mode of data sharing and cooperative coupling, updates parameters through one key, and improves the working efficiency of the hydraulic retaining wall. And automation and intelligentization of the design process are realized through parameter driving, AI optimization and model linkage. The method has the advantages that the quality is controllable, the resources are reused, the efficiency is improved, the time for designing, checking and reviewing is saved, the work of several weeks is shortened to several minutes, and the quality and efficiency are improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent design technology for water conservancy and hydropower projects, specifically to an intelligent design method for hydraulic retaining walls based on Building Information Modeling (BIM) and Artificial Intelligence (AI). Background Technology

[0002] Hydraulic retaining walls are earth-retaining structures used in water conservancy and hydropower projects to withstand earth pressure and prevent soil collapse. In addition to the general earth-retaining function of retaining walls, hydraulic retaining walls also have multiple functions such as bank connection, water blocking, water diversion, and lateral seepage prevention. They are widely used in reservoir projects, water diversion projects, hydropower stations, and various canal system structures.

[0003] Retaining walls are involved in the design of almost all hydraulic structures. Their core task is to ensure the wall's long-term stability and reliability, meeting the overall functional requirements of the hydraulic project while minimizing material usage and reducing construction costs. The design of hydraulic retaining walls generally includes steps such as selecting the structural type, determining cross-sectional dimensions, overall stability calculation, structural strength calculation, and detailed structural design. These steps are interconnected and require multiple iterations for optimization. For example, determining cross-sectional dimensions and overall stability calculation often involves determining dimensions, performing stability calculations, adjusting dimensions, and recalculating, with each adjustment involving modifications to multiple dimensions. Furthermore, drawing annotation is crucial; it's a key factor in verifying the rationality of the structural layout. Without accurate annotations, quantity calculations and construction cannot be performed based on the design drawings. In traditional design methods, manual annotation often requires significant effort, and the large amount of manual work is prone to errors, requiring considerable time and effort for review and verification.

[0004] Overall, due to the strong correlation between the various steps and the computational complexity, current hydraulic retaining wall design methods rely heavily on manual experience, with a large proportion involving repetitive mechanical labor, specifically manifesting in the following shortcomings: (1) Low efficiency: Manually determining dimensions and drawing drawings is time-consuming and labor-intensive; (2) Poor collaboration: Data is isolated in the design, calculation and drawing stages, and modifications require multiple repetitive steps; (3) Low tolerance for error: Mechanical operation is prone to oversights and the review process is delayed.

[0005] While existing digital technologies can assist in calculations, a fully intelligent solution for the entire design process of hydraulic retaining walls has not yet been developed, leaving significant room for improvement in work efficiency. Therefore, there is an urgent need for a method that integrates parametric design, AI optimization, and BIM collaboration to enhance design efficiency and quality. Summary of the Invention

[0006] The purpose of this invention is to provide a BIM and AI-based intelligent design method for hydraulic retaining walls that automates and intelligentizes the design process through parameter-driven, AI-optimized, and model-linked approaches.

[0007] To achieve the above objectives, the present invention is implemented through the following technical solution: A smart design method for hydraulic retaining walls based on BIM and AI includes the following steps: S1) Customize the cross-sectional design parameters according to the characteristics of various retaining wall structures.

[0008] Customized cross-sectional design parameters are developed based on the characteristics of each type of cross-section. The cross-sectional types of hydraulic retaining walls include equal gravity type, semi-gravity type, counterweight type, buttress type, empty box type, and cantilever type; the types of cross-sectional design parameters include width, height, and slope, which are divided into input parameters and calculated parameters in terms of assignment method.

[0009] In the process of customizing cross-sectional design parameters, engineering experience is incorporated into the setting of parameter values. For example, the base width of gravity retaining walls is usually 1 / 3 to 1 / 2 of the wall height; the thickness of cantilever walls is generally not less than 1 / 10 to 1 / 12 of the wall height.

[0010] S2) Develop a design script, input the cross-sectional design parameters to automatically generate the retaining wall cross-section, and output the corner point coordinates and numbers to generate a standardized geometric data file.

[0011] Use Tongyi Lingma to assist in programming and develop AI design scripts.

[0012] Furthermore, the design script was developed using the Civil 3D Subassembly Composer (part editor) visual programming platform, and the specific process is as follows: S21) Parametric Logic Construction: In the visualization environment, by defining the topological relationships and geometric constraints (such as relative distance, slope, vector direction, etc.) between feature points, connecting lines, and closed shapes, a parametric component of the retaining wall with dynamic adjustment capabilities (.pkt file) is constructed. S22) Numbering and Identification: During the construction process, standardized logical codes (Codes) are assigned to the key corner points, edge lines and different material areas of the cross section, which serve as unique "numbers" for subsequent scripts to identify geometric elements; S23) Script Conversion and Generation: The system reads and parses the above component files, extracts the internal geometric construction logic and coding mapping relationship, and converts them into an automated drawing script described in Python language (based on graphics libraries such as ezdxf); the script can receive parameters passed in from the front end, and automatically calculate coordinates and draw accurate retaining wall cross-section diagrams based on the parsed topological logic.

[0013] S3) Research the data interface of professional computing software, convert geometric data files into multi-line text file calculation templates with annotations, and then store the text file calculation templates in the local AI knowledge base.

[0014] Many professional calculation software programs for hydraulic retaining walls, both domestically and internationally, can solve most engineering calculation problems, greatly improving the efficiency of hydraulic engineers and the accuracy of calculation results. This study investigates the data formats of mainstream professional calculation software, converting geometric data files into multi-line text files containing information such as geometric shape, loads, and material parameters. Comments are added to each line, providing explanations and interpretations of the data file; for example, the comment after the coordinate line of corner point P1 is "P1.X, P1.Y". This text file calculation template is stored in a local AI knowledge base for subsequent automatic and targeted modifications by the AI.

[0015] Furthermore, the conversion process is completed within a self-developed backend data processing script, which acts as an intermediary bridge between the "geometric modeling end" and the "structural calculation end." The conversion process mainly includes two core steps: format reorganization and semantic enhancement. ① Format Reorganization: The system first parses the geometric data file generated in step S2) (such as extracting the key point coordinates of the cross section, wall height, bottom width, slope ratio and other geometric feature values); then, according to the standard data stream format required by the industry-standard structural calculation kernel (such as mature calculation engines such as Lizheng Geotechnical), it maps these geometric feature values ​​and fills them into the specified positions of the text stream to construct the basic calculation data file.

[0016] ② Semantic Enhancement: This is the key innovation of this invention. Traditional computational data files (such as Lizheng) usually consist only of pure numerical sequences (such as "6.500000, 0.660000..."), which are difficult for large AI models to directly understand in terms of their physical meaning. Therefore, the conversion script automatically adds natural language annotations before key data paragraphs while generating data lines (for example, adding "# wall unit weight" before "23.000000", and adding "# concrete strength modulus" before "2100.000000").

[0017] S4) Based on the parameters and coordinates of the current scheme, AI rewrites the text file calculation template and dynamically generates a new calculation model to ensure that the new calculation model is consistent with the design parameters.

[0018] Deepseek R1 automatically replaces key information such as corner coordinates to generate a new calculation model adapted to the current section, thereby avoiding unnecessary parameter input (due to the large number of parameters and the large amount of preprocessing work, designers will engage in mechanical repetitive work).

[0019] Specifically, the AI ​​command flow drives the overall stability calculation of anti-slip, anti-tilting and base stress, and uses the gradient descent method to derive multiple near parameter combinations, iteratively calculates stability and strength indices, and obtains the optimal solution.

[0020] Furthermore, the technical terms designed in step S4) are explained in detail as follows: "Current scheme" refers to the specific design input conditions determined by the designers at the current specific work site (such as chainage K0+100), including specific values ​​such as retaining wall height, ground slope, and geological parameters (friction angle, bearing capacity).

[0021] The significance of AI rewriting: Since the calculation template (generated in step S3) is merely a "skeleton" containing logical placeholders, while the "current scheme" is the flesh and blood, the role of AI is to perform "semantic-level data filling." It can understand the meaning of annotations in the template (such as "#wall base width") and accurately fill in the specific geometric and geological parameters of the current scheme into the corresponding positions in the template. In the traditional mode, designers need to manually input hundreds of parameters in the calculation software interface, which is prone to errors and time-consuming; AI rewriting achieves millisecond-level automatic mapping "from geometric model to calculation model," eliminating a large amount of mechanical manual preprocessing work.

[0022] A detailed description of "command flow driven": Since commonly used mature computing kernels in the industry (such as Lizheng Geotechnical) are typically closed systems without open underlying API data interfaces, this invention employs a "non-intrusive external drive" technology. Specifically, the system, in the background, simulates manual operation through operating system-level instructions, automatically launches the executable program of the computing engine, passes in the AI-generated computing data file as a parameter, silently executes the computing task, and finally automatically captures the generated computing result file. This method achieves automated invocation without cracking the original software.

[0023] A detailed description of the "gradient descent method": This refers to the parameter adjustment strategy when searching for the "most economical cross-section" that meets the specifications. When the calculation results of the initial cross-section do not meet the stability requirements (or the safety factor is too large, resulting in waste), the system will not blindly try calculations, but will calculate the "sensitivity gradient." For example, the system analysis finds that "increasing the wall heel width by 0.1m" has a significantly higher efficiency (gradient) in improving the "anti-slip stability coefficient" than "increasing the wall toe width by 0.1m." Based on this gradient direction, the AI ​​will prioritize generating a set of derived parameter combinations (i.e., adjacent parameter combinations) that are fine-tuned on the "highly sensitive parameters," rather than generating them randomly.

[0024] Description of the "optimal solution" and the entire calculation process: The entire optimization process is a closed-loop iterative system of "generation-evaluation-feedback". a. Initial calculation: The AI ​​generates the first calculation model based on the preliminary design parameters and drives the calculation through command flow; b. Result analysis and judgment: The system automatically analyzes the calculation result file and extracts key indicators such as anti-slip, anti-tilting, and stress.

[0025] c. Iterative optimization: If the index does not meet the specifications (e.g., anti-slip coefficient < 1.3), the system increases the key section size according to the gradient strategy and regenerates the model for calculation; if the index far exceeds the specifications (e.g., anti-slip coefficient > 2.0), the system appropriately reduces the size to save costs; d. Convergence Output: When the calculation results meet all specification requirements and the engineering quantity (cross-sectional area) is minimized, the set of parameters is the "optimal solution". The system will solidify it and output the final drawing.

[0026] S5) Input the new calculation model into the BIM model, automatically extract the engineering quantity and annotation information (referring to the quantity of steel reinforcement and concrete), realize the intelligent annotation of structural points / lines based on coding rules, and generate standardized drawings and reports.

[0027] Compared with the prior art, the present invention has the following advantages: This invention presents an intelligent design method for hydraulic retaining walls based on BIM and AI, which ensures quality control, resource reuse, and improved efficiency. It transforms the current work of designing, calculating, drawing, and quantity taking of hydraulic retaining walls from a data-closed and fragmented approach to a data-sharing and collaborative approach. One-click parameter updates save time in design, verification, and review, reducing a work that takes weeks to complete in minutes, thus improving quality and efficiency.

[0028] Specifically, even designers without prior experience in hydraulic retaining wall design can easily perform cross-sectional design, quantity calculation, drafting, and calculation report preparation for hydraulic retaining walls using this intelligent design method. By continuously accumulating parametric components and various drawing styles for hydraulic retaining walls, it supports reuse across multiple projects, reducing marginal costs. In traditional design methods, design schemes exist in the designer's mind, while drafting is the process of expressing these schemes. This design method presents design schemes in an intuitive way and can automatically extract drawings and quantities, significantly reducing the time spent. Furthermore, the time-consuming early-stage design work often involves frequent modifications; this method's multi-step interconnected design function significantly improves the level of automation, thereby achieving a general increase in efficiency. Attached Figure Description

[0029] Figure 1 This is a schematic diagram of the method flow of the present invention.

[0030] Figure 2This is a schematic diagram of the AI ​​design script flow of the present invention.

[0031] Figure 3 This is a thumbnail of the text file calculation template of the present invention. Detailed Implementation

[0032] The embodiments of the present invention will now be described in further detail with reference to the accompanying drawings.

[0033] Taking the design of a hydraulic retaining wall as an example, this invention provides an intelligent design method for hydraulic retaining walls based on BIM and AI, comprising the following steps: S1) Customize the cross-sectional design parameters according to the characteristics of various retaining wall structures.

[0034] The types of common hydraulic retaining walls are summarized in Tables 1 to 6.

[0035] Table 1. Cross-sectional parameters of gravity retaining wall (unit: cm, except for slope)

[0036] Table 2. Cross-sectional parameters of semi-gravity retaining wall (unit: cm, excluding slope).

[0037] Table 3 Cross-sectional parameters of counterweight retaining wall (unit: cm, except for slope)

[0038] Table 4 Cross-sectional parameters of empty box retaining wall (unit: cm, except for slope)

[0039] Table 5. Cross-sectional parameters of buttress retaining walls (unit: cm, excluding slope).

[0040] Table 6 Cross-sectional parameters of cantilever retaining wall (unit: cm, except for slope)

[0041] S2) Develop a design script that automatically generates the retaining wall cross-section by inputting cross-sectional design parameters, and outputs the corner point coordinates and numbers, generating a standardized geometric data file. Details are as follows: ======= Corner coordinates output (unit: m) ====== The coordinates of point P1 are (0.0000, 0.0000). Point P2 coordinates: (0.0000, 2.0000) Point P3 coordinates: (2.0000, 2.0000) Point P4 coordinates: (2.0000, 13.0000) Point P5 coordinates: (4.0000, 13.0000) Point P6 coordinates: (4.0000, 12.0000) Point P7 coordinates: (8.5000, 2.0000) Point P8 coordinates: (10.5000, 2.0000) Point P9 coordinates: (10.5000, 0.0000) ======== Variable Output (Unit: m) ======= Clear wall height: 10.0 Wall top width: 10 Vertical section height: 1.0 Back slope ratio: 0.45 Front toe width: 0.6 Wall heel width: 1.5 Base plate thickness: 1.0 ========================== S3) Research the data interface of professional calculation software, convert geometric data files into multi-line text file calculation templates with annotations, and then store the text file calculation templates in the local AI knowledge base. The wall top load and water pressure are as follows: / / ============Cad Item 1================ *cad_item 25. Vertical distributed load (kPa) #geom_coor 7.6 1.5 *P8.X P8.Y 6.1 1.5 *P7.X P7.Y #Layer Name Load #Line width 1 #Color 160 160 164 255 #Character height 0.5 #Soil behind the name wall #Drawing scale 0.01 #group calculation #Partial Factor 1 #Action Width (m) 1 #Loading step 0 #group Distributed load (kPa) #q_0= 212.5 #q_1= 212.5 * #F 318.75 * #Fx 0 * #Fy -318.75 * #Point of application (6.85, 0) / / ============Cad Item 2================ *cad_item 26. Liquid pressure (kPa) 7 #geom_coor 7.6 0 *P9.X P9.Y 7.6 7 *P9.X Water level behind the wall #Layer Name Load #Line width 0 #Color 160 160 164 255 #Character height 0.5 #Water level behind the name wall #Drawing scale 0.02 #Water level mark true #group calculation #Partial Factor 1 #Loading step 0 #Water level H(m) 7 *Water level behind the wall #Liquid specific gravity (kN / m3) 10 #Action Width (m) 1 like Figure 3 The image shows a partial fragment of the "text calculation book template with semantic annotations" generated in step S3), which specifically describes how the present invention transforms complex BIM geometric information into a "key-value pair" text format (such as #internal friction angle of backfill 25°) that is readable, understandable, and editable by the AI ​​large model. This is the data foundation for realizing AI intelligent rewriting in the subsequent step S4).

[0042] S4) Based on the parameters and coordinates of the current scheme, AI rewrites the text file calculation template and dynamically generates a new calculation model to ensure that the new calculation model is consistent with the design parameters.

[0043] S5) Input the new calculation model into the BIM model, automatically extract the engineering quantity and annotation information, realize the intelligent annotation of structural points / lines based on coding rules, and generate standardized drawings and reports.

[0044] The intelligent labeling rules for the cross-section of gravity retaining walls are shown in Table 7.

[0045] Table 7 Intelligent Labeling and Coding Rules for Gravity Retaining Wall Cross-Section Diagrams

[0046] The above description is only a preferred embodiment of the present invention. It should be noted that those skilled in the art can make several improvements and modifications without departing from the concept of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A smart design method for hydraulic retaining walls based on BIM and AI, characterized in that... Includes the following steps: S1) Customize the cross-sectional design parameters according to the characteristics of various retaining wall structures; S2) Develop a design script, input the cross-sectional design parameters to automatically generate the retaining wall cross-section, and output the corner point coordinates and numbers to generate a standardized geometric data file; S3) Research the data interface of professional computing software, convert geometric data files into multi-line text file calculation templates with annotations, and then store the text file calculation templates in the local AI knowledge base; S4) Based on the parameters and coordinates of the current scheme, AI rewrites the text file calculation template and dynamically generates a new calculation model to ensure that the new calculation model is consistent with the design parameters; S5) Input the new calculation model into the BIM model, automatically extract the engineering quantity and annotation information, realize the intelligent annotation of structural points / lines based on coding rules, and generate standardized drawings and reports.

2. The intelligent design method for hydraulic retaining walls based on BIM and AI according to claim 1, characterized in that: In step S4), the AI ​​command flow drives the overall stability calculation of anti-slip, anti-tilting and base stress, and uses the gradient descent method to derive multiple adjacent parameter combinations, and iteratively calculates to obtain the optimal solution.

3. The intelligent design method for hydraulic retaining walls based on BIM and AI according to claim 1, characterized in that: In step S1), the cross-sectional types of the hydraulic retaining wall include equal gravity type, semi-gravity type, counterweight type, buttress type, empty box type, and cantilever type; the cross-sectional design parameter types include width, height, and slope, which are divided into input parameters and calculated parameters in terms of the assignment method.

Images