Method for automatic modeling of complex structures

Abstract

The invention is a method for the automatic modeling of complex structures based on point cloud surveys. The method determines the actual geometric and static characteristics of the structure by processing the point cloud, with particular regard to the cross-sectional centers of gravity of the structural elements and the static frame derived from them. The method uses parametric processing, which allows the quantification of geometric deviations relative to the reference geometry and the automatic generation of a model that reflects the actual shape of the structure.

Description

[0001] METHOD FOR AUTOMATIC MODELING OF COMPLEX STRUCTURES

[0002] The subject of the invention is a method for the automatic modeling of complex structures based on point cloud surveys using parametric design methods. The invention preferably can be used, for example, to automate the modeling process of bridges based on 3D surveys, particularly to determine the actual static frame and surface model of the structure.

[0003] During the maintenance and operation of existing bridges, it is in many cases necessary to survey and evaluate their current condition (e.g., inspections of the support structure or interventions to improve their technical condition during service life). Traditional survey methods are extremely time-consuming, which is why digital methods (e.g., drone surveying, laser scanning, or photogrammetry) have come to the front nowadays, resulting in a 3-dimension set of points, the so-called point cloud, of the bridge. This data can be used to create 3D models, for which there are several methods, but in most cases, this requires significant manual work and time. This can be attributed to two main reasons; one of the most challenging tasks is the classification and breakdown of the point cloud into subsets corresponding to structural elements, for which although there are automated processes, the human factor cannot yet be completely eliminated. Furthermore, depending on the type of surveying equipment used and the surveying method, the point cloud has a certain degree of inaccuracy, which must be considered during modeling. It is particularly difficult to determine with sufficient accuracy the cross-sectional centers of gravity for complex, multi-element support structures and the static framework connecting them based on the point cloud.

[0004] The production of point cloud files form ing the input data for modeling is generally known. Various tools and methods are available for this purpose, for example on the website https: / / leicon.hu / termekek / 3d-pontfelho- egoldasok / .

[0005] A well-known 3D modeling method is the Scan-to-BIM method, which is offered, for example, on the website https: / / felmerespontosan.hu / , where the point cloud is inserted into a BIM software and the model is created using manual (handheld) tools. A point cloud usually contains many unnecessary points, the individual elements (e.g., a wall, a slab) have a significant dispersion, and the survey generally records other points from the environment in addition to the structural element to be surveyed. During modeling, points are manually selected from the point cloud to determine the position and geometric dimensions of each element, therefore, the accuracy and quality of a model created using this method depends primarily on the person doing the modeling and the modeling technique used, and it is an extremely time-consuming process. Scan-to-BIM methods generally result in architectural or geometric models and do not provide an automatic procedure for determining the static framework of structural elements, i.e. , the cross-sectional centers of gravity and the axis lines derived from them.

[0006] Currently, there is no known methodology available that would automatically generate a model based on the survey, which, in addition to being usable for a number of other purposes, would provide a more accurate and comprehensive picture of the geometry of the structure, unlike traditional survey methods, including the geometric imperfections that determine the actual static behavior of the structure.

[0007] A model generated directly from the survey can be created in a short time, compared with the original plans, can be used to detect changes in shape and deformations, and to generate a model of the existing structure for renovation or modernization purposes during preparation. Nowadays, the modeling of the renovation of existing structures is typically based on design data and in many cases does not consider the actual geometry, shape, and geometric imperfections of the structure. Ignoring this has led to significant technical problems in the past as well. Therefore, it is of paramount importance to provide an automatic procedure that is able to process point cloud-based survey data and generate a model in a reproducible manner, even in the case of complex structures. The method according to the invention determines the actual static frame and surface model of the structural elements from the point cloud and then compares the resulting model with a reference geometry, allowing the quantification of geometric imperfections. We developed the method according to the invention for steel and composite beam bridges, considering the existing domestic bridges, with the help of case studies. We used parametric design (or visual programming) for processing, which allows dynamic changes to the input data of the model, therefore extending the developed method to multiple types of structures. We recognized that the actual 3D and BIM models and static frames of the structure can be automatically created from the point cloud if we divide the sub-point clouds of the structural elements into parametrically defined sections (sampling ranges) perpendicular to the longitudinal axis, by projecting the resulting sets of intersection points onto a plane, we determine the actual contour of the crosssection and, from this, the cross-sectional center of gravity, and then, from the series of centers of gravity - or, in the case of steel structures, from the elementary lines of gravity broken down into subsets - we automatically generate the static axis or, in the case of surface elements, the middle surface.

[0008] The essence of the invention is that the static frame and surface model of the structural elements are determined not from pre-assumed geometric axes, but by algorithmic processing of section-based geometric information derived from the point cloud.

[0009] The objective is achieved by the method described in claim 1.

[0010] The preferred embodiments are described in the subclaims.

[0011] The invention is described in more detail below with the aid of the attached figures, using an embodiment example.

[0012] Figure 1 illustrates the steps required to generate the point cloud necessary for implementing the invention.

[0013] Figure 2 shows the steps necessary to produce a static frame.

[0014] Figure 3 describes the generation of the surface model.

[0015] The method according to the invention is primarily used for the automatic determination of the static frame of complex structures and, secondarily, for the automatic generation of the 3D surface model of the structure based on the static frame.

[0016] Parametric modeling was used to develop the method, which allows the process to be automatically adapted and extended to different structure types by modifying the input parameters.

[0017] Figure 1 shows the steps involved in generating the point cloud required to implement the invention. In step SP2, we create the raw point cloud after the SP1 3D survey - preparation, which we then clean up in step SP3, removal of unnecessary elements, by removing unnecessary elements (e.g., the environment), thus significantly reducing the size of the file. Further cleaning of the point cloud is performed in the SP4 noise filtering step, and if necessary, the file size can be further reduced by lowering the resolution. Finally, we obtain the SP5 point cloud to be examined.

[0018] Figure 2 shows the steps involved in creating the static frame. To do this, we break down the SS1 raw point cloud (if it is a complex structure and not a single element) into subsets corresponding to the individual structural elements in the SS2 subset breakdown step. Then, in the SS4-A and SS4-B determination steps, we determine whether each element should be considered as a bar or a surface element, based on the geometric characteristics of the point cloud.

[0019] In the case of bar elements, in the SS5-A step, we divide the partial point cloud of the structural element into parametrically distributed sections perpendicular to its longitudinal axis, determine the cross-section contour and its center of gravity from the set of points of each section, and then form the SS6-A static axis from the series of centers of gravity.

[0020] In the case of surface elements, in step SS5-B, we examine the point cloud with sections perpendicular to the surface normal and determine the SS6-B middle surface based on the resulting section points.

[0021] We export the resulting geometry, consisting of lines and polygons, which is the SS7 static frame geometry, to a structural analysis software, where we check the connections between the elements and, if necessary, supplement or modify the model.

[0022] The generation of the surface model can be followed in Figure 3. To do this, we break down the raw point cloud of the complex structure SM1 (if it is a complex structure and not a single element) into subsets corresponding to the individual structural elements in the SM2 subset break down step. In the SM4 fitting step, we fit lines that help determine the geometric axis to the SM3 sub-point clouds corresponding to the structural elements, which specifies the longitudinal axis of the given element. In the SM5 cuboid generation step, we create cuboids in a plane perpendicular to the longitudinal axis, which we use in the SM6 cutting step to cut out individual slices of the point cloud. The thickness, number, and distance between the sections (i.e. , the cuboids) to be cut out of the point cloud can be set parametrically for each element. The points found in the resulting point cloud sections are aligned into a plane in the SM7 move to a plane step. This gives us a more accurate result than if we were to examine only a single section of the point cloud, as in many cases, due to measurement inaccuracies, the points necessary to determine the actual cross-section may be missing from the cross-section. The more sections we create, the more detailed and accurate the model will be. By combining the point sets from the corresponding cross-sections, the cross-section is determined not based on a single geometric cross-section, but based on a parametrically controlled, combined evaluation of multiple spatial samplings, which allows for the determination of statistically stable geometric characteristics even in the case of noisy or incomplete point clouds.

[0023] Then, the process differs depending on the building material of the structural element. In the case of steel structures, the SM8-A point sets forming the crosssections are then broken down into subsets corresponding to the individual elements, typically plates (e.g., in the case of an I-beam, the bottom and top flanges and the web), in the SM9-A breakdown step. For the resulting point cloud subsets, in the SM10-A fitting straight lines step we determine the lines of gravity associated with the plates, from which the SM11 -A surface creation step forms a surface to obtain the SM12 3D surface model of the structural element. In the case of reinforced concrete or wooden structures, in the SM9-B contour fitting step we fit a contour to the SM8-B point sets forming the cross-sections and then connect these line chains with surfaces along the axis of the structural element in the SM10-B surface creation step to obtain the SM12 3D surface model.

[0024] To summarize the operation of the method, the invention allows us to automatically generate the actual static frame of the structural elements from the point cloud resulting from the 3D survey using a parametric design method based on predefined geometric and mathematical rules, and to create the 3D surface model of the structure based on this. Section-based processing, the determination of cross-sectional contours and centers of gravity, and, in the case of steel structures, the lines of gravity associated with the plates provide novel geometric information that enables the accurate determination of the actual axis and surface geometry required for static calculations.

[0025] Each step of the process is based on the geometric characteristics derived from the previous step, in particular the contour, center of gravity, and axis information derived from the parametrically generated section series, so that the entire model is automatically adapted by modifying the input parameters. The resulting static frame and surface model can be used directly in structural analysis or can be used to generate beam models or solid models via further processing. The method according to the invention requires minimal manual intervention and ensures the determination of the actual geometry in a reproducible manner, even for complex structures.

[0026] The method was developed using the example of a sinusoidal web steel beam, so its operating principle will be explained below using this example.

[0027] The point cloud was created using laser scanner survey, which was, in the first step, processed using Autodesk Recap software. Due to the operation of the laser scanner, it was necessary to remove unnecessary parts (environment), therefore significantly reducing the file size. Further cleaning (noise filtering) of the point cloud was performed using Cloudcompare software, but Autodesk Recap software is also suitable for this task. The point cloud was exported in ,e57 format.

[0028] In the next step, we imported the cleaned point cloud containing only the beam into Rhinoceros 7 software.

[0029] In the next step, we fit a straight line approximating the longitudinal direction of the beam to the points of the beam point cloud, which gives the geometric longitudinal axis of the beam, and then create cuboids perpendicular to this longitudinal axis, with which we cut out each slice of the point cloud.

[0030] Since the Grasshopper algorithm has steps that do not need to be modified later (such as the creation of cutting cuboids), we can hide them using the Cluster command, making the algorithm more transparent, and free of unnecessary elements. Thus, only the parametrically adjustable elements appear in the algorithm. When creating the intersections, the thickness of the cuboids, their number, the distance between them, and their starting point are flexibly adjustable values. We defined these values with a mathematical sequence.

[0031] In the present example, we cut 1 cm thick slices from the point cloud, and then in the next step, we bring the set of points belonging to each intersection into a plane. This gives us a more accurate result than if we were to examine only a single slice of the point cloud, as in many cases, due to measurement inaccuracies, the points necessary to determine the actual cross-section may be missing from the cross-section. There are missing points in both the bottom and top flanges, even when considering one cm slices.

[0032] The point sets forming the cross-sections must then be broken down into subsets corresponding to the individual elements, in this case the bottom and top flanges and the web. In this example, we solved this by determining a Z coordinate based on the known main dimensions of the beam, and the algorithm detects the points located below this coordinate as the points of the bottom flange. We created this command using a simple Python code. In the same way, we defined the points of the top flange based on a Z coordinate value, and then the points between the two coordinate values form the points of the web. Finally, by fitting straight lines to the resulting point cloud subsets, we obtain the line of gravity (axis of gravity) of the flanges and the web of the beam interpreted in the given cross-section.

[0033] Then, we fit surfaces to the center of gravity axes obtained by fitting using the Loft command. Since the lines fitted to the web often do not intersect with the axes of the bottom and top flanges, we can obtain the complete surface model of the beam by extending the surfaces obtained from the fitted lines of the web to the surfaces of the bottom and top flanges.

[0034] Grasshopper allows exporting in numerous formats. The purpose of the embodiment example presented here was to assess the actual geometry of the beam, which, when compared to the ideal geometry, allows us to determine the degree of geometric imperfection. This procedure allows us to perform structural analysis on the actual shape, which gives more accurate results than examining the ideal shape. The numerical analysis is performed using the ABAQUS software, which uses the point cloud forming the actual geometry as input data, thus we resolved the created surface model into points and exported it to an Excel file.

[0035] At the same time, we also have the option of direct export to AxisVM finite element software via a live connection, which is also suitable for running structural analysis.

[0036] In addition to the export options described above, Grasshopper can work with numerous other software programs, either through a live connection or using various file formats. In connection with the Scan-to-BIM process, it is also worth mentioning the live connection between Archicad and Grasshopper, which allows us to create architectural models based on point clouds. The center of gravity axes of the structural elements created using the method described above can be used directly as reference axes for Archicad elements (e.g., walls, beams, columns, etc.), allowing us to automatically generate architectural models from point clouds. Grasshopper can also work with Autodesk Revit software and several other programs (e.g., Tekla Structures, ConSteel, FEM-Design, etc.), which further expands the scope of application of the method.

[0037] Unlike the known examples, in the method we propose, the section-based geometric information derived from the point cloud determines the axes, cross- sectional lines of gravity, and contours of the structural elements, which is performed by the algorithm we developed. With the help of Grasshopper, we use mathematical and logical rules to determine the axes or contours of the elements, which not only speeds up modeling but also increases accuracy, and the geometry of the model created in this way matches the geometry of the constructed structure.

[0038] The method described in the invention significantly reduces the time required for surveying and modeling structures, thereby saving costs. Furthermore, the BIM methodology can be applied by assigning the appropriate information content to the model elements. Within this framework, automatic quantity statements and updated plans can be obtained from the model and can be used to perform and monitor operational and maintenance tasks throughout the further service life of the structure.

[0039] The model created is suitable for analyzing different design variants, therefore enabling more economical and sustainable solutions. In addition, the static frame that can be automatically generated from the model allows for structural analysis considering the actual shape of the structure.

[0040] The novelty of the method lies in the fact that the sectional information obtained from the point cloud is used in the parametric model in such a way that the cross- sectional contours, centers of gravity, and lines of gravity can be automatically determined, and from these, the system can essentially automatically create the actual static frame and surface model of the structural elements, requiring only the specification of the necessary input parameters. The parametric generation of section sequences, as well as the division of contours into subsets and the determination of their center of gravity axes, results in a technical assembly that ensures the reproducible determination of the actual geometry even in the case of complex structures.

[0041] The use of a parametric environment allows the same algorithm to be adapted to different structures with similar designs, as each step of the process is automatically modified to match the input point cloud. Therefore, a 3D survey of the structure is sufficient, and the created point cloud serves directly as input for the model generation.

[0042] The model produced by this method specifies the actual axes and surfaces of the structural elements, which provides a basis for performing static calculations and examining geometric imperfections. In the case of surveys carried out at several points in time, the structural changes and deformations can be determined numerically by comparing the models, which provides an additional technical advantage for assessing the condition of existing structures and monitoring their structural behavior.

Claims

What is claimed is1. Method for the automatic, material-independent modeling of complex structures based on point cloud surveys using a parametric design method to determine the actual geometric and static characteristics of the structure, in which- a point cloud is created from the structure by means of a survey, and it is pre- processed by noise filtering and removing points that do not belong to the structure under investigation,- the point cloud into sub-point clouds corresponding to the structural elements is divided,- based on the sub-point clouds, a straight line or plane defining a geometric direction as an orientation auxiliary geometry to the structural elements is fitted, characterized in that- for bar elements, groups of points located at a given distance from each other from the sub-point cloud are selected using spatially extended cuboid sampling regions with a parametrically defined position and size in a direction perpendicular to the auxiliary geometry,- for surface elements, point groups originating from sampling regions defined with respect to the auxiliary geometry or the normal of the surface are projected onto a common plane and automatically generate the middle surface of the surface element based on them,- the point groups from the corresponding sampling regions are projected onto a common plane and treat them as a combined sampling point set,- the cross-section contour from the combined sampling point set are automatically determined, then the center of gravity of the cross-section based on the contour is automatically calculated, and- the static axis or middle surface of the structural element from the series of cross- sectional centers of gravity determined along the structural element are automatically formed, then the determined static axis or middle surface is comparedwith a predefined reference geometry and quantify the deviations as geometric imperfections.

2. The method according to claim 1 , characterized in that, when creating the cuboid sampling domains, the thickness, number of cuboids, and the distance between them are defined parametrically using a mathematical series, and by changing these parameters, the statistical stability of the cross-sectional point set and the accuracy of the specified static axis are controlled.

3. The method according to any of claim 1 or 2, characterized in that the point groups from related sampling domains are combined to compensate for local inaccuracies arising from incomplete or noisy point clouds, and the determination of the cross-sectional contour is based on the combined geometric information of point groups from multiple sampling domains.

4. The method according to any of claims 1-3, characterized in that, in the case of steel bar elements, the cross-sectional point set is divided into subsets corresponding to the plates of the structural element, and the lines of gravity belonging to the plates are determined from the straight lines fitted to the individual subsets, from which the composite static axis of the structural element is formed.

5. The method according to any of claims1-4, characterized in that, in the case of a structure consisting of several bar elements, the nodes of the static frame are generated by extending the defined static axes and automatically determining their intersections.

6. The method according to any of claims1-5, characterized in that a geometrically nonlinear static analysis is performed using the geometric imperfections determined according to claim 1 as input parameters for the static model.

Images