Large hyperbolic upright lockrand metal roof plate arrangement deepening method

Abstract

The invention discloses a large hyperbolic upright lockrand metal roof arrangement deepening method, and particularly relates to the technical field of building construction, and the method comprises the steps of S1, curved surface pretreatment, S2, intelligent arrangement generation, S3, AR enhancement verification, and S4, dynamic parameter updating. According to the method, curved surface features are accurately analyzed through curved surface preprocessing to determine the plate arrangement distance, intelligent plate arrangement is carried out to generate an optimized plate arrangement path, the material utilization rate is improved, automatic numbering and sorting are carried out, AR enhancement verification is carried out to display the fitting condition in real time, error prompt and operation guidance are provided, dynamic parameter updating is carried out according to construction data, and a scheme is adjusted in time. The method effectively solves the problems of limitation of a layout algorithm, data islands, low field verification efficiency and the like of a traditional method, achieves the effects of improving the design precision, improving the construction efficiency and reducing the cost, and has a remarkable application value in the building industry.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of building construction, more particularly, the present application relates to a large-scale hyperbolic standing lock edge metal roof panel deepening method. BACKGROUND

[0002] At present, in the context of the continuous development of the construction industry, in large-scale construction projects, hyperbolic metal roof has been widely used in large-scale construction projects due to its unique shape and excellent performance.

[0003] However, the panel deepening design has always been a difficulty in the engineering field, and although the existing technology has made some progress with the help of Dynamo parameterized modeling and adaptive family technology, there are still many problems that cannot be ignored, such as: 1. Limitations of panel algorithm: Currently, the panel algorithm based on Dynamo parameterized modeling is mostly based on preset rules, and when facing complex hyperbolic metal roof with complex curvature, this fixed rule paneling method cannot dynamically optimize the panel distribution according to the real-time changes of the curved surface. 2. Data island problem is prominent: There is a serious obstacle to data interaction between deepening design and construction links, and currently, the data transmission between the two links often relies on manual intervention, which cannot realize real-time synchronization, and the information update of construction progress and material state is not timely, which makes it difficult for designers to adjust the paneling scheme according to the actual construction situation. 3. Low efficiency of on-site verification: On-site verification mainly relies on manual measurement and three-dimensional model comparison, manual measurement is not only low in efficiency and large in error, but also in complex hyperbolic roof environment, it is difficult to measure every installation position comprehensively and accurately, and when comparing the measurement results with the three-dimensional model, there is also a lack of efficient analysis tools, which cannot quickly locate the installation deviation.

[0004] In view of the above situation, the present application provides a large-scale hyperbolic standing lock edge metal roof panel deepening method. SUMMARY

[0005] In order to overcome the above-mentioned defects of the prior art, the present application provides a large-scale hyperbolic standing lock edge metal roof panel deepening method to solve the problems raised in the background art.

[0006] To achieve the above purpose, the present application provides the following technical scheme: a large-scale hyperbolic standing lock edge metal roof panel deepening method, specifically comprising the following steps: S1, surface preprocessing S1.1, feature recognition and region division, a model composed of convolutional layers, pooling layers and fully connected layers is constructed using a convolutional neural network, the curvature and slope characteristics of the Rhino model surface are identified according to the formula, a heat map is generated, single-curved, double-curved and transition regions are divided, and the optimal spacing of the plates in each region is determined as a set value with a floating range of ±5%; The formula is: y = f(x), wherein y is the feature map value of the i-th layer, j-th row and k-th column, and x is the convolution kernel weight of the i-th layer, and b is the bias term; S1.2, mesh division and report generation, efficient mesh division algorithm is used for surface mesh division and curvature calculation, and three-dimensional visual surface feature report is generated; S2, intelligent plate arrangement generation S2.1, plate arrangement path optimization, taking the fitness function as the core, dynamically optimizing the plate arrangement path with the help of genetic algorithm, generating an adaptive family list containing rotation angle and length parameters, using multi-objective optimization, using cloud parameter engine for collaborative editing and version control, accessing RFID data, and using ≤3m short plate first; The fitness function is: f = w1 * f1 + w2 * f2 + w3 * f3, wherein f1 is the material utilization rate function, f2 is the construction convenience function, and w1, w2 and w3 are weight coefficients, the value range is 0-1 and; S2.2, plate number and sequence generation, using the automatic segmentation algorithm of special-shaped plates, automatically generating plate numbers and installation sequence; S3, AR augmented verification S3.1, fit condition display and error prompt, using AR equipment, real-time rendering through WebGL format mobile lightweight model, positioning based on SLAM technology, displaying virtual plate arrangement and entity fit condition, outputting error prompt, generating rectification mark and associating BIM model; S3.2, reference line and animation assistance, projecting virtual reference line and aligning with entity, superimposing operation guidance animation for construction personnel; S4, dynamic parameter updating, integrating construction progress and material data, using reinforcement learning to correct the plate arrangement scheme online according to the state transition equation, at the same time, setting construction error trigger mechanism, detecting conflicts and recommending solutions with the help of graph neural network, and feeding back the measured data to the plate arrangement algorithm to form a closed loop; The reinforcement learning state transition equation is: s t+1 = f(st, a t), wherein st is the current state, at is the action taken, and st+1 is the next state.

[0007] Preferably, in the intelligent plate arrangement generation process of S2 step, the distributed parameter engine supports cloud Dynamo node library based on microservice architecture, multiple design teams can perform parallel operation through the node library, and adjust the plate arrangement parameters at the same time; In the dynamic parameter updating link of S4 step, a construction data real-time visual board is provided, which integrates multi-dimensional indexes of progress, quality and cost, and helps the management personnel quickly master the overall state of the project, and provides data support for dynamically adjusting the plate arrangement scheme.

[0008] Preferably, the adaptive family parameters generated in S2 step include a rotation angle of 1-5° with an accuracy of 0.1°, a plate length allowable deviation of ±2mm, and a cutting width allowable deviation of ±5mm, for improving the accuracy of plate arrangement and material utilization.

[0009] Preferably, the AI algorithm involved in S1, S2 and S4 steps contains a CNN surface feature extraction model with an identification accuracy of more than 95%, which provides an accurate data basis for surface feature identification, plate arrangement scheme optimization and dynamic parameter updating. In S4 step, the graph neural network conflict detection model processes thousands of spatial relationship data per second, and gives at least 3 solutions within 10 seconds, for improving the efficiency of conflict detection and solution.

[0010] Preferably, the BIM collaborative cloud platform relied on in S4 step is provided with an IOT data access interface, supports RFID and sensor device access, can obtain various data of the construction site in real time, and at the same time, adopts a message queue mechanism to complete data synchronization within 1 second, ensuring the timeliness and consistency of the data of each participant. The BIM collaborative cloud platform has a distributed version control and permission management system, and different personnel can operate the BIM model according to the permissions, ensuring the security and traceability of the data.

[0011] Preferably, in the AR augmented verification system of S3 step, a mobile terminal application supporting Android / iOS is contained, for improving the operation convenience of the construction personnel, and the system has a multi-modal interactive interface with an interaction accuracy of more than 90%, and has the functions of automatic generation and export of error analysis report, and the report generation time is not more than 3 minutes, for providing convenience for quality control and problem rectification in the construction process.

[0012] Preferably, in S4 step, the construction error triggering mechanism is that the construction progress lags behind more than 10%, the material inventory is less than 50 pieces, and the deviation between the measured data and the model is more than ±5mm, for triggering parameter updating to ensure the smooth progress of the construction.

[0013] Preferably, in S1 step, the surface grid division accuracy is 0.1mm, and the generated three-dimensional visual surface feature report is used to help the design and construction personnel understand the surface characteristics within 10 minutes.

[0014] Technical effects and advantages of the present application: 1、The present application utilizes convolutional neural network to identify curved surface characteristics, accurately divides single curve, double curve and transition area, and determines the optimal panel spacing, combines high-precision mesh division and curvature calculation, provides accurate data basis for panel design, and the generated three-dimensional visual curved surface characteristic report helps designers and construction personnel quickly understand the curved surface characteristics, reduces design errors, at the same time, the adaptive family parameter is accurately set, such as the rotation angle accuracy is 0.1°, the allowable deviation of the plate length is ±2mm and the allowable deviation of the cutting width is ±5mm, which further improves the accuracy of the panel, makes the plate better fit the double curved roof surface, and improves the overall quality of the metal roof; 2、Multiple design teams can perform parallel jobs through cloud Dynamo node library, and adjust panel parameters, which greatly shortens the design cycle, in the intelligent panel generation process, the special-shaped plate automatic segmentation algorithm and the function of automatically generating plate number and installation sequence are adopted, which reduces the time and errors of manual processing, the AR augmented verification system can show the virtual panel and the entity fitting situation in real time, provide operation guidance animation, and quickly detect and prompt errors, which helps construction personnel to more efficiently perform installation work, reduces the construction difficulty, and speeds up the construction progress. In addition, the dynamic parameter updating mechanism can adjust the panel scheme in time according to the construction progress, material inventory and measured data, so as to avoid the delay of the design and construction; 3、Through the fitness function of multi-objective optimization, the panel path is optimized by combining the genetic algorithm, the short plate ≤3m is preferentially used, the material utilization rate is improved, the material waste is reduced, the accurate panel design and accurate control in the construction process reduce the rework cost caused by plate mismatching, installation errors and the like, the IOT data access and message queue mechanism of the BIM collaborative cloud platform realize the real-time synchronization and efficient management of the construction site data, reduce the management cost, at the same time, the panel deepening design efficiency is improved, the project cycle is shortened, and the equipment rental cost and labor cost are indirectly reduced. BRIEF DESCRIPTION OF DRAWINGS

[0015] Figure 1 The whole flowchart of the present application. DETAILED DESCRIPTION

[0016] The present application provides a large double curved standing seam metal roof panel deepening method, which specifically comprises the following steps: S1、Curved surface preprocessing S1.1、Feature recognition and region division, specifically: Model construction: select a suitable deep learning framework (such as TensorFlow or PyTorch), construct a convolutional neural network model, which includes convolutional layers, pooling layers and fully connected layers, to process the curved surface data of the Rhino model; Data preparation: Extract the geometric data of the surface from the Rhino model and preprocess it to convert it into a format suitable for CNN model input, such as discretizing the surface into a grid data of certain resolution and labeling the corresponding curvature and slope information; Model training: Train the constructed CNN model using a large amount of double-curved roof surface data with annotations. During the training process, adjust the convolution kernel weights and bias terms to make the model learn the relationship between the surface curvature and slope features and the feature map values until the model reaches a stable and high recognition accuracy; Feature recognition and region division: Input the Rhino model surface data to be processed into the trained CNN model. The model identifies the surface curvature and slope features according to the formula and generates a surface complexity heat map. According to the feature distribution of the heat map, single-curved, double-curved and transition regions are divided, for example, regions with a curvature change rate within a certain threshold range are single-curved regions, regions with a curvature change rate exceeding the threshold and having a specific change trend are double-curved regions, and the transition between the two is a transition region. For different regions, combined with engineering experience and material characteristics, the optimal panel spacing is determined as a floating value of ±5% based on the set value, for example, for single-curved regions with small curvature changes, the panel spacing can be appropriately increased to improve construction efficiency; while for double-curved regions with complex curvature changes, the panel spacing is reduced to ensure the fit of the panel to the surface.

[0017] S1.2, mesh division and report generation, specifically: Mesh division algorithm selection: Use efficient mesh division algorithms such as the front propagation method and Delaunay triangulation to divide the surface of the divided region. These algorithms can generate high-quality meshes according to the geometry and features of the surface, ensuring that the mesh division accuracy reaches 0.1mm. During the division process, the curvature change of the surface is fully considered, and the grid is appropriately densified in areas with large curvature to improve the accuracy of subsequent curvature calculation; Curvature calculation: Based on the divided grid, use differential geometry principles to calculate the curvature of each grid element, for example, by calculating the normal and tangent vectors of the surface at the grid nodes, and then obtaining the Gaussian curvature and mean curvature and other curvature indicators; Report generation: Integrate the mesh division results and curvature calculation data to generate a three-dimensional visual surface feature report. The report displays the surface features of different regions in an intuitive three-dimensional model, such as using different colors to identify different curvature regions, and also includes detailed textual explanations, including the curvature range of each region, panel spacing recommendations, and other information, to help design and construction personnel quickly understand the surface characteristics within 10 minutes and provide clear guidance for subsequent paneling work; S2, intelligent paneling generation S2.1, paneling path optimization, specifically: Fitness function construction: According to the material utilization function and the construction convenience function, the fitness function is constructed, wherein and are weight coefficients, the value range is 0-1 and, the material utilization function can be determined by calculating the ratio of the covering area of the plate on the curved surface to the theoretical maximum covering area, and the construction convenience function can be quantified by considering factors such as the complexity of plate splicing and installation difficulty, for example, reducing the number of special-shaped plates can improve the construction convenience, according to the actual needs of the project, reasonably adjust the value of and, to balance the material utilization and construction convenience; Genetic algorithm implementation: genetic algorithm is used to dynamically optimize the plate arrangement path, first, a group of initial plate arrangement schemes are randomly generated as the population, each plate arrangement scheme contains parameters such as the rotation angle and length of the plate, the fitness value of each individual (plate arrangement scheme) in the population is calculated, and selection, crossover and mutation operations are performed according to the fitness value, to generate a new population, after multiple generations of evolution, the population gradually converges to the optimal or near-optimal plate arrangement scheme; Cloud collaboration and data access: The cloud Dynamo node library based on micro-service architecture is used to realize distributed collaborative editing and version control, multiple design teams can simultaneously call different function Dynamo nodes through the node library to adjust the plate arrangement parameters, realize parallel operation, at the same time, access the RFID data of the construction site material yard, real-time obtain the size, quantity and location information of the materials, preferentially select short plates with length ≤3m to improve material utilization and reduce cost; S2.2, plate numbering and sequence generation, specifically: Automatic segmentation algorithm for special-shaped plates: for special-shaped plates generated in areas with complex curved shapes, an automatic segmentation algorithm based on geometric shape analysis is used, which divides the special-shaped plates into reasonable shapes according to the geometric characteristics of the curved surface and the plate arrangement rules, ensuring that the segmented plates can meet the design requirements while being convenient for processing and installation; Numbering and sequence generation: according to the optimized plate arrangement scheme, a unique number is automatically generated for each plate, and its installation sequence is determined, the numbering rule should be convenient for construction personnel to identify and find, and the installation sequence is determined according to the actual situation of the construction site and the construction process requirements, for example, from one end of the roof to the other end, or according to a certain specific construction process, to improve construction efficiency and reduce rework caused by installation errors.

[0018] S3, AR augmented verification S3.1, fit condition display and error prompt, specifically: Mobile application development: Develop AR mobile applications supporting Android / iOS systems, convert BIM models to WebGL format, realize real-time rendering of mobile lightweight models, use SLAM technology to quickly locate the position of AR devices in the construction site, and accurately match virtual paneling models with physical structures; Error detection and prompt: During construction, construction personnel check the fit of virtual paneling and physical structures through AR devices, the system detects the deviation between the two in real time, and when the deviation exceeds the set threshold (such as ±2mm), error prompt information is output, and the error position is marked in a prominent color on the AR interface, at the same time, generate rectification markers and associate them to the BIM model, to facilitate construction personnel to rectify in time; S3.2, reference line and animation assistance, specifically: Virtual reference line projection: Project virtual reference lines on the AR interface to align them with the physical structure, which provides clear installation benchmarks for construction personnel, helping them quickly and accurately determine the installation position of the panel, improving installation accuracy; Operation guidance animation superposition: For the installation of complex nodes, superimpose construction personnel operation guidance animation, which details the installation steps and key points, construction personnel can repeatedly watch the animation through the AR device, quickly master the installation skills, reduce the installation time, and improve the construction quality; S4, dynamic parameter update, specifically: Paneling scheme correction Data integration: Through the BIM collaborative cloud platform, integrate construction progress and material data, use IOT data access interface to access various sensors and RFID devices in the construction site, and real-time obtain construction progress information (such as the number and position of installed panels), material inventory data (such as the remaining number of different specifications of panels), and measured data (such as the actual position and angle deviation of installed panels); Reinforcement learning correction: Use reinforcement learning to correct the paneling scheme according to the state transition equation online, where is the current state, including construction progress, material inventory and measured data, is the action taken, such as adjusting the paneling spacing or replacing the panel specification, is the next state, according to the current state, select the optimal action through the reinforcement learning algorithm to adjust the paneling scheme to adapt to the actual situation of the construction site; Conflict detection and triggering mechanism Conflict detection: Use graph neural networks to detect spatial conflicts between paneling models and components such as rafters and gutters, graph neural networks can quickly process large amounts of spatial relationship data, process thousands of spatial relationship data per second, and give at least 3 solution recommendations within 10 seconds, for example, when detecting a conflict between the paneling model and the rafter, recommend adjusting the position, angle or replacing the panel specification of the panel. Trigger mechanism: set construction error trigger mechanism, when the construction progress lags behind more than 10%, the material inventory is less than 50, the measured data deviates more than ± 5mm from the model, the parameter update is automatically triggered, the system adjusts the plate arrangement scheme according to the trigger condition and the conflict detection result, ensures the smooth progress of construction. At the same time, using the message queue mechanism, the parameter update information is synchronized to all participants in real time, ensuring the timeliness and consistency of the data of each participant.

[0019] In summary, the present application aims at the large hyperbolic standing lock edge metal roof plate arrangement deepening problem, realizes accurate and efficient plate arrangement through multi-technology integration, and its working principle is as follows: Firstly, the convolutional neural network is used to construct a specific model to identify the curvature and slope characteristics of the Rhino model surface, and the regions are divided and the optimal plate arrangement spacing of each region is determined according to the identification results, and the grid division and curvature calculation are carried out, and the visual report is generated, and the surface pretreatment is completed, which provides basic data for subsequent plate arrangement; Then, taking the fitness function including material utilization rate and construction convenience as the core, the genetic algorithm is used to optimize the plate arrangement path, the adaptive family list is generated, the cloud parameter engine is used for collaborative design and version management, the RFID data is connected to optimize material use, and the function of automatic segmentation and numbering sequence of special-shaped plates is generated, realizing intelligent plate arrangement generation; In the construction verification link, with the help of AR equipment and WebGL format mobile lightweight model, based on SLAM technology positioning, the virtual plate arrangement and entity fitting are displayed, the error prompt is output and the BIM model is associated, and the virtual reference line is projected, the operation guidance animation is superimposed to assist construction; Finally, the construction progress and material data are integrated, the reinforcement learning is used to correct the plate arrangement scheme according to the state transition equation, the graph neural network is used to detect conflicts and recommend schemes, the parameter update is triggered when the construction progress, material inventory and measured data do not meet the set conditions, forming a dynamic parameter update closed loop, ensuring that the plate arrangement scheme meets the construction actual situation, so as to realize efficient and accurate plate arrangement deepening design and construction of large hyperbolic standing lock edge metal roof.

[0020] The above only describes the preferred embodiments of the present application and is not used to limit the present application, any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for deepening large-scale hyperbolic standing seam metal roof panels, characterized by: The specific steps include: S1. Surface preprocessing S1.

1. Feature recognition and region segmentation: A convolutional neural network is used to construct a model consisting of convolutional layers, pooling layers, and fully connected layers. The curvature and slope characteristics of the Rhino model surface are identified according to the formula. A heat map is generated to divide the surface into single-curved, double-curved, and transitional regions. The optimal panel spacing in each region is determined to be ±5% of the set value. The formula is: , where is the feature map value of the th row and th column of the th layer, is the convolution kernel weight of the th layer, and is the bias term; S1.

2. Meshing and report generation: Use efficient meshing algorithms to perform surface meshing and curvature calculations, and generate 3D visual surface feature reports. S2. Intelligent layout generation S2.

1. Optimize the layout path. Using the fitness function as the core, we dynamically optimize the layout path with the help of a genetic algorithm, generate an adaptive family list containing rotation angle and length parameters, adopt multi-objective optimization, utilize a cloud-based parameter engine for collaborative editing and version control, access RFID data, and prioritize the use of short-length panels ≤3m. The fitness function is: , where is the material utilization function, is the construction convenience function, and is the weight coefficient, ranging from 0 to 1; S2.2, Plate numbering and sequence generation, using the special-shaped plate automatic segmentation algorithm to automatically generate plate numbers and installation sequence; S3, AR enhanced verification S3.

1. Display of fitting status and error prompts: Using AR devices, a lightweight model in WebGL format is rendered in real time on a mobile terminal. Based on SLAM technology positioning, the virtual panel layout and the physical fit are displayed, error prompts are output, and correction marks are generated and associated with the BIM model. S3.2, reference line and animation assistance, projecting virtual reference lines to align with the entity, and superimposing construction personnel operation guidance animation; S4: Dynamic parameter updates integrate construction progress and material data, using reinforcement learning to online correct the layout plan based on the state transition equation. Simultaneously, a construction error trigger mechanism is set up, and a graph neural network is used to detect conflicts and recommend solutions. Measured data is fed back to the layout algorithm to form a closed loop. The state transfer equation for reinforcement learning is: , where is the current state, is the action taken, and is the next state.

2. The method for deepening large-scale hyperbolic standing seam metal roof panels according to claim 1, characterized in that: During the intelligent layout generation process in step S2, the distributed parameter engine supports a cloud-based Dynamo node library based on a microservices architecture. Multiple design teams can use this node library to work in parallel and adjust layout parameters simultaneously. In the dynamic parameter update link of step S4, a real-time visualization dashboard of construction data is provided. The real-time visualization dashboard of construction data integrates multi-dimensional indicators of progress, quality and cost, which helps managers to quickly grasp the overall status of the project and provide data support for dynamic adjustment of the layout plan.

3. The method for deepening large-scale hyperbolic standing seam metal roof panels according to claim 1, characterized in that: The adaptive family parameters generated in step S2 include a rotation angle of 1-5° with an accuracy of 0.1°, a plate length tolerance of ±2 mm, and a cutting width tolerance of ±5 mm, which are used to improve the accuracy of plate arrangement and material utilization.

4. The method for deepening large-scale hyperbolic standing seam metal roof panels according to claim 1, characterized in that: The AI ​​algorithms involved in steps S1, S2, and S4 include a CNN surface feature extraction model with an accuracy rate exceeding 95%, providing an accurate data foundation for surface feature recognition, layout optimization, and dynamic parameter updates. In step S4, the graph neural network conflict detection model processes thousands of spatial relationship data per second and provides at least 3 solutions within 10 seconds, which improves the efficiency of conflict detection and resolution.

5. The method for deepening large-scale hyperbolic standing seam metal roof panels according to claim 1, characterized in that: The BIM collaborative cloud platform used in step S4 is equipped with an IOT data access interface that supports RFID and sensor device access, enabling real-time acquisition of various construction site data. Furthermore, the message queue mechanism enables data synchronization within 1 second, ensuring the timeliness and consistency of data from all participants. The BIM collaborative cloud platform has a distributed version control and permission management system. Different personnel can operate the BIM model based on their permissions to ensure data security and traceability.

6. The method for deepening large-scale hyperbolic standing seam metal roof panels according to claim 1, characterized in that: The AR enhanced verification system in step S3 includes a mobile application that supports Android / iOS to improve the convenience of operation for construction workers. The system has a multimodal interactive interface with an interaction accuracy rate of over 90%. It also has the function of automatically generating and exporting error analysis reports. The report generation time does not exceed 3 minutes, which provides convenience for quality control and problem rectification during the construction process.

7. The method for deepening large-scale hyperbolic standing seam metal roof panels according to claim 1, characterized in that: In step S4, the construction error trigger mechanism is to trigger parameter updates when the construction progress lags behind by more than 10%, the material inventory is less than 50 pieces, or the deviation between the measured data and the model exceeds ±5mm. This is used to timely adjust the layout plan to ensure smooth construction.

8. The method for deepening large-scale hyperbolic standing seam metal roof panels according to claim 1, characterized in that: In step S1, the surface meshing accuracy reaches 0.1mm, and the generated 3D visual surface feature report is used to help designers and construction personnel understand the surface characteristics within 10 minutes.

Images