Harnessing a direct scan-to-modeling approach for structural mechanics

Abstract

To ensure the safety and reliability operation of infrastructures, accurate and rapid condition assessment of in-service infrastructures is extremely imperative. Image technologies and techniques, as a non-destructive evaluation (NDE) tool, have created new opportunities to achieve this. It is able to resolve sub-millimeter details including surface flaws and defects, providing a foundation for quantitatively linking observable features to operational performance. This linkage is possible when the quantified feature is directly integrated into tools suitable for describing mechanical response, such as the finite element method. However, the workflow associated with this integration is indirect and presents challenges. This work creates a pathway for a direct scan-to-model strategy suitable for translating condition data derived from a 3D laser scanning system into a computation model capable of describing the mechanical response of the component. The geometric characterization capabilities of the proposed approach and the influence of scan resolution on the geometric characterization are explored. Results of this study provide the foundation of a computational framework for establishing the fundamental link between visually observable geometric changes and the numerical models that engineers use to understand the performance of engineered systems.

Keywords

laser scanning three-dimensional (3D) reconstruction optimization scan capability resolution

Introduction

For aging structures and infrastructures, the as-is conditions might vary significantly from as-designed conditions due to accumulated damage. To ensure the safety operation of infrastructures, it is imperative to know the health condition of the infrastructures during their service life. Common critical challenges for this are the complexity associated with measuring the current condition state, and the lack of mechanisms to translate these measurements into the descriptions of performance needed for decision-making. Recent advances in measurement techniques have created opportunities for improving this. One such advancement that aligns with the visual assessment approach is 3D laser scanning (3DLS), which is also known as light detection and ranging (LiDAR). It has been increasingly used since the 1990s (Bosché et al., 2013), which leverages time of flight principles to describe the distance between the scanner and an object of interest (Lemmens, 2011). The benefits of 3DLS can be highlighted as eliminating in situ sensor instrumentation, artificial light sources and wiring expenses (Dabous and Feroz, 2020). Additionally, it is able to collect large amounts of spatial data in very short periods of time and with a high level of geometric precision.

3DLS represents a form of non-destructive evaluation (NDE) by which the complex geometric features of a physical component can be collected in a non-contact manner and recreated in a digital form (Toth et al., 2013; Toth and Zivack, 2014). Within the civil infrastructure community, by comparing the complex 3D shape before and after scan, 3DLS has primarily focused on quantifying structure damage, such as section loss (Dolhon et al., 2013; Guldur et al., 2015; Li et al., 2008), corrosion (Fernandez et al., 2016; Guldur and Hajjar, 2017), cracks (Guldur et al., 2015; Guldur and Hajjar, 2017), displacement (Park et al., 2007), spalling (Guldur et al., 2015; Guldur and Hajjar, 2017), deflection (Gordon and Lichti, 2007). What’s more, 3DLS is also applied across numerous fields such as historical building heritage protection (Armesto-González et al., 2010; Asteris and Plevris, 2015), detection of pavement distresses (potholes, large-area utility cuts or patches) (Chang et al., 2005), blast impact evaluation (Watson et al., 2011), and reconstruct as-build 3D BIM models (Bosché et al., 2013; Sepasgozar, et al., 2018). These applications highlight some of the potential for leveraging 3D data by comparing the structures’ geometry at different times, but the application of the information extracted from 3DLS is not restricted to these and its use can be extended to wider horizons. One specific application is the integration of these high resolution and detailed measurements into models capable of simulating the effects of measured damage on mechanics response (Gheitasi and Harris, 2015b). However, only limited research has explored the concept of further leveraging these datasets to describe structural performance of damaged systems. The following few examples highlight some of the most recent studies relevant to this concept, which is to translate the datasets derived from 3DLS into finite element model (FEM) for analysis.

Yan et al. (2017) classified point clouds for each component, and then identified proper geometry characteristics for each component. After generating finite element meshes for each component, assembled them together and the FEM was generated. The proposed method can create finite element meshes for the primary structural members of bridges, including piers, girders and decks, but cannot be extended to structural components such as bracing, bearing and connections. Conde-Carnero et al. (2016) used auxiliary planes fitting to the point cloud and then combined with an extrusion technique for volumetric object creation. Compared with the common approach, which makes use of triangular meshes to model the object’s surfaces or volume, it can avoid errors caused by poor smoothing or sharp feature edges, but it may lead to errors when the components are not symmetric or damaged. Castellazzi et al. (2015) extracted sections from large point cloud databases to form point cloud slicing. Then the stacking of these slices generated the 3D geometry. However, when the surface was irregular (curved) or not planar to an axis direction, the resulting FE mesh was a jagged representation of the original geometry. Ghahremani et al. (2018) proposed a methodology to locally update a finite element model of a damaged structural component, which can address the problem of the excessive computational costs caused by the undue number of finite elements in the final model. Nevertheless, due to the inherent sensitivity of the algorithm to voxel size, it suffered from a lack of accuracy in modeling small and detailed regions.

These works mentioned above do provide a comprehensive illustration of collecting and translating completed 3D point cloud datasets obtained by 3DLS into a FEM for simulation, but a limitation that remains is three-fold: 1) the ability to resolve highly localized features, 2) the translation of these features into a useable model representation accurately and effectively and 3) the capability to generate a suitable mesh from the highly irregular geometries derived from the scan datasets. As a step towards achieving these goals, this manuscript proposes a method to transform structured light laser scanning (SLLS) data into finite element models capable of describing the complete mechanical response of structural components, which providing an answer to the question, “Can the geometry of complex damage be directly integrated into numerical simulation?” Then to explore the geometry characterization capabilities of the proposed method and the effects of resolution on model approximation, two experiments are designed and the results are discussed. These two components aim to answer two fundamental questions, “What level of damage detail representation can be achieved via scanning?” and “What is the effect of resolution on geometric characterization?” Finally, conclusions are made in Section 5.

SLLS to finite element model framework

With an overall objective of leveraging SLLS data to develop an accurate numerical model of a structural component, a series of intermediate steps to translate the surface measurement data into complete solid models were explored in this study. An illustration of the entire workflow is shown in Figure 1. Some of the terminologies describe in the workflow are described as follows:

Figure 1.

Workflow of translation the scanned data into finite element analysis.

Data collection: A number of commercial laser scanners are available on the market and provide the ability to resolve high-resolution geometries. These laser scanners are divided into three categories, handheld laser scanners (HLS), terrestrial laser scanners and airborne laser scanners. Handheld laser scanners measure distances of several meters with the accuracy of submillimeter. They are portable and convenient which are prone to be used in the lab or for localized scanning. Terrestrial laser scanners measure distances between several meters and thousands of meters with the accuracy of submillimeter. Compared with terrestrial laser scanners, they can acquire and process a large number of data efficiently, so they are widely used for the scanning of actual structures. Airborne laser scanners are widely used in aerial photogrammetry, the flight height of the airborne laser scanner is generally several kilometers.

The NDE is primary a localized condition evaluation (Dong and Catbas, 2020), such as cracks, corrosion, spalling, delamination and so on. For localized defects and small specimens, the short-range portable laser scanner is a more practical and suitable solution. Another distinct advantage of the short-range device is the real-time visualization capabilities. However, little literature can be accessible on the application of portable laser scanner. This work leverages a handheld laser scanner, Creaform Handyscan 3D 700, as the platform for collecting scan data with a goal of translating the raw scanned data sets into structural finite element model suitable for describing behavior accurately and effectively. It is a portable short-range SLLS device capable of collecting and providing real-time visualizations of 3D objects as shown in Figure 2. The scanner relies on 3D stereo-photography with triangulation provided by targets and a mesh (7 × 7 lines) of lasers and is capable of acquiring more than half a million data points per second. Table 1 provides a summary of the manufacturer specifications, but it should be noted that this system was studied extensively, using geometrical accuracy tests, by the Laboratory for Photogrammetry and Laser Scanning at HafenCity University Hamburg and was determined to yield comparable results to classic static structured light systems in a portable form factor (Creaform, 2018).

Figure 2.

Creaform Handyscan 700 and illustration of basic operation. (a) Overview of scanning (b) Realtime data acquisition.

Table 1.

Manufacturer specifications for creaform HandySCAN 3D 700 (Creaform, 2018).

Parameter	Value
Accuracy	Up to 0.030 mm
Resolution	0.050 mm
Weight	0.94 kg
Measurement rate	480k measurements/s
Light source	7 laser crosses (+1 extra line)
Scanning area	275 × 250 mm
Stand-off distance	300 mm
Depth of field	250 mm

The scanner is a line of sight tool, which means that the object or portion of the object being measured must be visible by the imaging sensors and illuminated by the laser beam(s). The physical and visible geometry of the measured component captured by a scanner can be represented in two forms: triangular mesh or point cloud using its compatible acquisition software VXelements (VXelements Software, 2017) as shown in Figure 3. Resolution is described as the level of detail the scanner is able to acquire. For triangle mesh surface, the resolution refers to the dimension of the triangle sides used to construct the scanned surface, which ranges from 0.2 mm∼10 mm. For point cloud surface, it refers to the distance between each data point and ranges from 0.1 mm∼5 mm. In this study, all scanned data is presented as triangle mesh surface. The dimensions of the damage can be detected depend on the resolution used when processing the scanning data. The proposed method include cleaning the scanning data, combining multiple data sets into a single set, establishing a 3D geometric model and forming a finite element model. These operations may alter the surface area and the volume of the 3D model, but will not refer to the damage parts.

Figure 3.

Representation of data collection by different scan methods. (a) Triangle mesh surface (b) Point cloud surface.

Optimization: The collected data usually contains environmental noise and unwanted features, which need to be cleaned and effectively removed from the scene. The optimization eliminates these features and enables the refinement of the scan. The processes associated with optimization are described below:

• Data cleaning – removal of isolated patches by a statistical approach proposed by Rusu et al. (2008). First of all, the mean μ and SD σ of the Euclidian distance of every point from its k-nearest neighbors are calculated. Then points fall outside μ ± ασ are regarded as noise and removed. The removed points are isolated patches, thus the actual defects will not be excluded.

• Fill holes – fill missing holes with poly-faces based on the local mesh shape by an algorithm presented in (Wang and Oliveira, 2003). The strategy can fill every hole that is smaller than the given radius.

• Registration – registration is a preprocessing step required if multiple scans are taken to capture a scene, it combines multiple data sets into a single set by the iterative closet point (ICP) algorithm (Besl and McKay, 1992). For example, scans for both the back and front of the specimen were collected separately and registered to create a single combined scan of the entire specimen.

In this study, these operations mentioned above can also be done in a semiautomatic manner through the built-in functions in either VXelements (VXelements Software, 2017) or Geomagic Design X (Geomagic Design X Software, 2016) software packages.

3D Geometric Model: After optimization, a 3D geometric model is established.

Surface Meshing: Creation of curved networks on the mesh and fitting of surface patches to the network which maintain accuracy to the underlying mesh. After surface meshing, a 3D freeform surface body is created.

Finite Element Model: Export the surface body into a finite element software, such as ANSYS Workbench (ANSYS Workbench Software, 2018), the finite element model is obtained. A visual description of the proposed scan to model workflow is provided in Figure 4.

Figure 4.

Visual description of the proposed scan to model workflow. (a) Data collection of the front side (b) Data collection of the back side (c) Optimization two sides (d) Registration (e) Model creation (f) Surface meshing for finite element analysis.

The proposed method is universal for the scanning data obtained by any device, but the scanner used in this study HandySCAN 3D 700 is not. This scanner can scan an object regardless of the material or the dimension of the object and do not require specific illumination conditions, but the distance between the scanner and the object should be less than 3 m.

Experimental design

The condition state of structural components has significant variation ranging from fine localized cracking from fatigue to volumetric reductions associated with corrosion (Figure 5). However, the measurement detail required to describe the influence of condition state on performance has not been well studied. A series of studies conducted by Gheitasi et al. (Gheitasi et al., 2015a, 2015b) demonstrated the influence of idealized damage states on bridge structural performance, but these representations were highly idealized and not inclusive of the stochastic nature of damage progression in real structures. For any structural system, the ideal measurement resolution is linked with the smallest geometric detail that influences either local or global response of the structural components. For example, a hairline crack could likely produce significant stress concentrations at the local scale, but if the location of this crack occurred in a redundant or low stress region, the impact may not be significant. On the contrary, section loss located in the areas with high bearing stresses could result in instability and local buckling failures. While the highest measurement resolution is ideal, there are a number of limiting factors associated with line of sight measurement approaches such as the inability to characterize internal condition state and parts in the geometry which are too small, complex or hidden by other parts. An example of this limitation is shown in Figure 6, which demonstrates that while the scanner is able to capture high resolution detail, the technique is not appropriate for all scenarios (e.g. tight crevices, etc.). During initial trial scanning efforts, it was found that using the higher resolution settings for the scanner yields improvements in the surface detail that can be described, but a trade-off becomes the need significant computation time and the potential for noisy results in the surface model. On the contrary, a lower resolution scanner setting may result in inaccurate geometries. This trade-off balance was not explored in detail in this work, but is worth highlighting as a point for consideration.

Figure 5.

Typical concrete structural deterioration. (a) Cracks (b) Spalling and reinforcement corrosion.

Figure 6.

Illustration of limitations in line of sight for SLLS. (a) Stair-stepped specimen (b) Dimension of the stair-stepped specimen (mm) (c) Full scan of stair-stepped specimen (d) Scan of detail of interest.

The experimental study explored in this work evaluates the geometric characterization capabilities of the proposed approach for translating the derived 3D scans into models suitable for describing the mechanical response of structural components. Besides, the influence of scan resolution on the geometric characterization was also evaluated. To achieve these, two experiments are designed as follows.

Geometric characterization capability

To understand what kind of defects can be detected, four specimens with different defects were selected for evaluation including a solid (intact) prismatic shape (A), a prismatic shape with a rectangular notch defect (B), a prismatic shape with a hemispherical notch defect (C) and a prismatic shape with three distinct V shape notch defects (D) as shown in Figure 7. The actual structural damage is as shown in Figure 5, which is irregular and non-uniform thickness. The dimensions of the actual damage are extremely difficult to measure. It is also difficult to know the reliability of the proposed method by comparing the scanned geometric model with the real geometric model. Therefore, we made four specimens with different defects in the laboratory instead of scanning the actual damaged structure to explore the reliability of the proposed method. These specimens were fabricated in the University of Virginia Physics machine shop by machining bar stock sections of 6061 aluminum, the accuracy of the instrument is ± 0.1 mm. First of all, scanned models of the four different specimens were obtained by the proposed method with the same resolution of 0.6 mm. Then the surface area and volume of the scanned model are compared with the designed one for manufacturing developed in Solidworks (Solidworks Software, 2018), which can reflect the deviation between the two models globally. The scanned model is compared with the designed specimen other than the manufacture specimen. This is due to measuring the actual dimensions by hands may bring bigger errors than the manufacture process (the accuracy of the cutting instrument is ±0.1 mm). Besides, for each of the specimen, the poly-vertices’ deviation between the scanned model and the designed model is presented, which can reflect the deviation locally. The maximum, minimum, absolute values of average and standard deviations are also listed.

Figure 7.

Reference specimens with different defects. (a) Machined specimens (b) Dimension of the reference specimens (mm).

Scan resolution for geometric characterization

To evaluate the influence of scan resolution on the geometric characterization, finite element models of specimen C are generated by four different scan resolutions: 0.2 mm, 0.6 mm, 1.0 mm and 2.0 mm. Similarly, the surface area and volume of the scanned model are compared with the designed one for manufacturing developed in Solidworks (Solidworks Software, 2018). For each of the specimen, the poly-vertices’ deviation between designed model and scanned model is presented. The maximum, minimum, absolute values of average and SD are also listed. All the results are shown and discussed in the next section.

Results

Geometric characterization capability

To numerically evaluate the geometric characterization capabilities, the four specimens shown in Figure 7 were scanned with the resolution of 0.6 mm. Comparisons of the surface area (S) and the volume (V) calculated in Solidworks (Solidworks Software, 2018) are made between the scanned specimen and the designed specimen. As mentioned earlier, the comparisons are made between the scanned and the designed specimen due to that measuring the actual dimensions by hands may bring bigger errors than the manufacture process (the accuracy of the cutting instrument is ±0.1 mm). This difference is important as it provides a clear illustration of the potential deviation formed during the construction of the 3D geometric model by the proposed method. The comparison results are shown in Table 2. The scanning deviation is defined by equation (1). For each of the specimen, the poly-vertices’ deviation between the scanned model and designed model is presented in Figure 8, which can reflect the deviation of the scanned specimen locally. The maximum deviation D_max, minimum deviation D_min, absolute values of average deviations D_avg and the SD σ are listed in Table 3.

S c a n n i n g d e v i a t i o n = \frac{S c a n n e d V a l u e - D e s i g n e d V a l u e}{D e s i g n e d V a l u e} \times 100 %

(1)

Table 2.

Comparison of the designed value and scanned value of Specimens A∼D.

Specimens	A		B		C		D
Specimens	S (mm²)	V (mm³)	S (mm²)	V (mm³)	S (mm²)	V (mm³)	S (mm²)	V (mm³)
Designed value	67097	786579	32258	188451	32120	190210	30847	172064
Scanned value	66546	788753	31174	187645	31978	197520	29862	171242
Scanning Deviation (%)	−0.82	0.28	−3.36	−0.43	−0.44	3.84	−3.19	−0.48

Figure 8.

Deviations between the designed model and scanned model (mm). (a) Specimen A (b) Specimen B (c) Specimen C (d) Specimen D.

Table 3.

Accuracy analyzer between the designed model and scanned model of specimens A∼D.

Specimens	A	B	C	D
Max deviation D_max (mm)	0.1416	2.8284	1.2239	3.6172
Min deviation D_min (mm)	−1.5282	−1.3558	−1.6321	−1.5764
Absolute value of verage deviation D_avg (mm)	0.0035	0.0004	0.0086	0.0054
SD σ (mm)	0.1110	0.1116	0.1548	0.1111

When comparing the performance of the geometric representation derived from the scanning, the results indicate that the highest accuracy attained is scanning straight line segments, which is specimen A (solid prismatic shape), with the maximum deviation observed being less than 1%. For specimens with curved or broken line defects, the deviations are all less than 4%, which indicates that the proposed workflow is reliable.

Figure 8 illustrates the poly-vertices’ deviation between the designed model and the scanned model. The red part indicates that the designed model is covered with the scanned model but the blue part indicates the opposite. The green part represents the segments whose deviation is less than 0.1 mm. The range of the colorbar is D_avg ± 2σ, which includes more than 94% of the poly-vertices. According to the contour in Figure 8, the overlapping models are mostly green. Moreover, the absolute values of average deviations D_avg in Table 3 are all less than 0.01 mm. These indicate that the designed model and the scanned model have a high coincidence. For the four specimens, the edges of the overlapping models are all blue. This is due to the limitation of the scanner, which has a bad representation of sharp edges. What’s more, for specimen D, the middle V-notched defect with the angle of 90° has the maximum deviation, the right V-notched of 127° comes second and the left V-notch of 143° has the minimum deviation. Therefore, it can be concluded that the larger the included angle is, the easier the scanner can be accessible to and more accuracy scanned model can be obtained. Similarly, for simple shapes or more accessible defects such as specimen A and C, the precision of the scanned model is higher than those that have more complicated or less accessible defects like specimen B and D. For the four specimens of A∼D, the standard deviations in Table 3 are all between 0.1 and 0.2, which indicates that the scanned model is consistent well with the designed model.

Scan resolution for geometric characterization

In characterizing the geometry of a structural component, the trade-off between accurate representations of the features of interest, necessary resolution and computation demand are critical. The results described in this section only highlight the first two of these trade-offs, with the understanding that an increase in resolution results in an increase in computational demand and volume of data. Figure 9 provide a visual illustration of the resulting data derived for one of the test specimens C - bar stock with a hemispherical notch presented by triangle mesh with four different scan resolutions, 0.2 mm, 0.6 mm, 1.0 mm and 2.0 mm. From these illustrations it is visually apparent that significant differences in model detail manifest across the various scan resolutions. With the reduction of resolution (increase in resolution parameter), the specimen’s edges become less defined and the resulting shape becomes a gross representation of the actual geometry. For example, at a resolution of 2.0 mm, the hemisphere is still apparent, but the specimen is out of shape and highly irregular along the boundaries; however, for a resolution of 0.2 mm the surface geometry is smooth and consistent in the notched region. Similar characteristics were exhibited for the other three specimens evaluated, but only a visual illustration of Specimen C is presented here.

Figure 9.

Scanned specimen C using different resolutions. (a) Resolution of 0.2 mm (b) Resolution of 0.6 mm (c) Resolution of 1.0 mm (d) Resolution of 2.0 mm.

The visual illustrations provide a foundational understanding of the scan resolution capabilities, but to numerically evaluate the performance of the various scanning resolutions, comparisons were made between the designed model and scanned model. For different resolutions, the poly-vertices’ deviation between designed model and scanned model is presented in Figure 10, which can reflect the deviation of the scanned specimen locally. The maximum deviation D’_max, minimum deviation D’_min, absolute values of average deviations D’_avg and the SD $σ'$ are listed in Table 4.

Figure 10.

Deviations of specimen C between designed model and scanned model at different resolutions. (a) Resolution of 0.2 mm (b) Resolution of 0.6 mm (c) Resolution of 1.0 mm (d) Resolution of 2.0 mm.

Table 4.

Accuracy analyzer between designed model and scanned model of Specimens C

Resolution (mm)	0.2	0.6	1.0	2.0
Max deviation D’_max (mm)	−0.6112	1.2239	−1.0359	−2.5249
Min deviation D’_min (mm)	3.1484	−1.6321	4.7939	3.9058
Absolute values of average deviation D’_avg (mm)	0.0353	0.0086	0.0094	0.0172
SD σ' (mm)	0.0809	0.1548	0.1501	0.5062

In Figure 10, the range of the right-hand colorbar is also D’_avg ± 2σ, which includes more than 94% of the poly-vertices. According to the contour in Figure 10, with the reduction of resolution (increase in resolution parameter), the green part shrinks, which indicates that the differences between the scanned model and standard model become more obvious. In Table 4, the scanned model with the highest resolution of 0.2 mm has the maximum average deviation, this is probably due to the impurities in the air are scanned as a part of the specimen, which may lead to errors at some details. The scanned model with the resolution of 0.2 mm has the minimum SD, but the scanned model with the resolution of 2.0 mm does the opposite. Additionally, the absolute values of average deviations between different resolutions do not have much difference. This indicates that the geometric center of the scanned model coincides well with that of the standard model, but when the resolution is lower than or equal to 2.0 mm, the scanned model distorts and the global deviation becomes greater. In general, the resolution should be set according to users' requirements. The highest resolution may not obtain the most accurate model but leads to an increase in computational demand and volume of data. When the resolution is lower than or equal to 2.0 mm, the scanned model distorts. Choosing the resolution between 0.2 mm and 2 mm can obtain accurate scanned model using proper processing time and data volume.

Conclusions

This study validates the possibility to leverage high-resolution SLLS as a non-destructive evaluation tool for structural health monitoring. First, methods to transform scanned raw data into finite element model are presented. Then SLLS is introduced and the geometric characterization capability and trades-off between scanning resolution and model approximation are explored. It is clear that the SLLS capability has excellent potential for quantifying and linking measurements of condition to a model for simulating performance, but the scanning workflow has some inherent limitations that may lead to some deviations between the finite element model and the original model. This framework lays the foundation for accurate integration of observable damage into a simulation framework, a key feature necessary to describe the true behavior of existing structural systems. Based on this investigation, the following findings and conclusions are drawn:

1) As the SLLS continues to evolve, there are some immediate advantages provided by the technology that can be integrated within a structural health monitoring framework: non-contact with the object, direct measurement of condition, does not require specific illumination conditions, and portable to take it to everywhere.

2) The proposed framework of translating surface measurement data into solid models is reliable. It can obtain accurate model for finite element analysis by choosing proper resolution. However, with the scanner being a line of sight tool, there are fundamental limitations with small complicated or hidden parts of the specimen since these parts are difficult to get within the laser scanner’s reach or line of sight.

3) With the tested scanner system, an increase in resolution results in improvements in the detail that can be resolved, but at a cost of increased computation time. The resolution should be set according to users' requirements and the scanning instrument. The highest resolution may not obtain the most accurate model but leads to an increase in computational demand and volume of data.

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by Central University Basic Research Fund of China (FRF-TP-20-040A1).

Data availability

The data used to support the findings of this study are available from the corresponding author upon request.

ORCID iDs

Shuzhen Yang

Devin K Harris

References

ANSYS Workbench Software (2018) ANSYS Workbench User’s Manual 18.0.

Armesto-González

Riveiro-Rodríguez

González-Aguilera

, et al. (2010) Terrestrial laser scanning intensity data applied to damage detection for historical buildings. Journal of Archaeological Science 37(12): 3037–3047. DOI: 10.1016/j.jas.2010.06.031.

Asteris

Plevris

(2015) Handbook of Research on Seismic Assessment and Rehabilitation of Historic Structures. DOI: 10.4018/978-1-4666-8286-3.

Besl

McKay

(1992) A method for registration of 3-D shapes. IEEE Transactions on Pattern Analysis and Machine Intelligence 14(2): 239–256. DOI: 10.1109/34.121791.

Bosché

Guillemet

Turkan

, et al. (2013) Tracking the built status of MEP works: assessing the value of a scan-vs-BIM system. Journal of Computing in Civil Engineering 28(4): 05014004. DOI: 10.1061/(asce)cp.1943-5487.0000343.

Castellazzi

D’Altri

Bitelli

, et al. (2015) From laser scanning to finite element analysis of complex buildings by using a semi-automatic procedure. Sensors 15(8): 18360–18380. DOI: 10.3390/s150818360.

Chang

Liu

(2005) Detection of pavement distresses using 3D laser scanning technology.In: Proceedings of the 2005 ASCE international conference on computing in civil engineering, pp. 12–15. DOI: 10.1061/40794(179)103.

Conde-Carnero

Riveiro

Arias

, et al. (2016) Exploitation of geometric data provided by laser scanning to create FEM structural models of bridges. Journal of Performance of Constructed Facilities 30(3): 04015053. DOI: 10.1061/(ASCE)CF.1943-5509.0000807.

Creaform (2018) HandySCAN 3D 700. Retrieved from https://www.creaform3d.com/en/metrology-solutions/portable-3d-scanner-handyscan-3d

10.

Dabous

Feroz

(2020) Condition monitoring of bridges with non-contact testing technologies. Automation in Construction 116(October 2019): 103224. DOI: 10.1016/j.autcon.2020.103224.

, et al. (2013) Forensic engineering 2012: gateway to a safer tomorrow. In: Proceedings of the Sixth Congress on Forensic Engineering. San Francisco, CA.

12.

Dong

Catbas

(2020) A review of computer vision–based structural health monitoring at local and global levels. Structural Health Monitoring 20(August): 692–743. DOI: 10.1177/1475921720935585.

13.

Fernandez

Bairán