Content uploaded by Gabriella Rossi
Author content
All content in this area was uploaded by Gabriella Rossi on Sep 16, 2020
Content may be subject to copyright.
Precision Partner
Enhancing GFRC craftsmanship with industry 4.0 factory-floor feedback
Paul Nicholas1, Gabriella Rossi2, Iliana Papadopoulou3,
Martin Tamke4, Nikolaj Aalund Brandt5, Leif Jessen Hansen6
1,2,3,4CITA / KADK 5BBFiberbeton 6TickCad
1,2,3,4{paul.nicholas|garo|ipap|martin.tamke}@kadk.dk 5nikolaj@bbfiberbeton.
dk 6ljh@tickcad.dk
This paper presents a novel human-machine collaborative approach to automatic
quality-control of Glass-Fiber Reinforced Concrete (GFRC) molds directly on the
factory floor. The framework introduces Industry 4.0 technologies to enhance the
ability of skilled craftsmen to make molds through the provision of horizontal
feedback regarding dimensional tolerances. Where digital tools are seldom used
in the fabrication of GFRC molds, and expert craftsmen are not digital experts,
our implementation of automated registration and feedback processes enables
craftsmen to be integrated into and gain value from the digital production chain.
In this paper, we describe the in-progress framework, Precision Partner, which
connects 3d scanning and point cloud registration of geometrically complex and
varied one off elements to factory floor dimensional feedback. We firstly
introduce the production context of GFRC molds, as well as industry standards
for production feedback. We then detail our methods, and report the results of a
case study that tests the framework on the case of a balcony element.
Keywords: 3d Scanning, GFRC, Feedback, Automation, Human in the loop,
Digital Chain
INTRODUCTION
The construction Industry is the second least digi-
tized in the world [1]. Despite the increasing ubiq-
uity of digital tools within design, simulation and
management workflows (Claypool et al. 2019), most
fabrication, assembly and construction practices rely
on hand-work and craft expertise, whether in fac-
tories or on-site. This gap becomes relevant in the
case of projects that involve complex geometries, in
which dimensional feedback and quality control is
central. While automated quality-control processes
have been long established in production lines (Matt
et al. 2020) to inspect for faults in high-volume
production, implementing registration and feedback
processes in hand craft shop-floor environments, the
typical mode of production in the building industry,
to inspect geometrically complex one-off elements
remains a challenge.
Adapting Industry 4.0 technology to the factory
floor would allow for skilled hand-craft processes to
D2.T10.S2. ROBOTIC TECTONICS, AUTOMATION AND INTERACTION - Volume 2 - eCAADe 38 |631
be integrated with, and gain value from, the digital
information chain. Skilled craftsmen perform cen-
tral and critical roles in many manufacturing and con-
struction processes, particularly within manufactur-
ing tasks where the level of complexity and varia-
tion makes automation highly difficult or inefficient
(Autor 2015). Within this type of manufacturing ac-
tivity, human craft skills are able to contribute dex-
terity, flexibility, sensory and decision-making abil-
ity [2], and are recognised to outperform current au-
tonomous and automated systems (Autor 2015). Be-
cause of this, it is important to understand how cur-
rent Industry 4.0 research can extend beyond its cur-
rent focus of establishing connections between ma-
chines and systems, and to consider the human in the
loop. In this paper we explore how Industry 4.0 can
enable new types of interface between craft exper-
tise and digital design information through the idea
of digital assistance, and the proposition that cou-
pled human-automated activities can assist expert
craftsmen in becoming even more effective and ef-
ficient.
In this paper, we describe a framework for au-
tomating feedback processes from 3d scanning to as-
sist craftsmen in evaluating the geometric quality of
their work. We contextualise our research through
description of a specific manufacturing process -
GFRC manufacturing - and against analysis of state of
the art in production feedback. We introduce a new,
prototypical framework in the context of expert mold
makers creating molds for precast fiber-reinforced
concrete building elements. Through feedback from
3d scanning, this framework enables craftsmen to
assess their production against digitally established
critical dimensions at any point throughout the mak-
ing process, and to make spot measures that they
would not be able to perform via physical measur-
ing techniques. The prototypical framework aims to-
wards a seamless solution, with the objective that
automation should reduce any disturbance of the
making process to a minimum, so that gaining feed-
back does not require expert knowledge. Key aspects
of this ambition are the development of registra-
tion protocols and the creation of a customised inter-
face. The framework is developed and evaluated in
an interdisciplinary collaboration of researchers from
CITA, GFRC producer BBFiberbeton and the metrol-
ogy provider TickCAD.
MANUFACTURING ARCHITECTURAL GFRC
ELEMENTS
Within architecture, GFRC is used to manufacture
lightweight, prefabricated cladding elements, facade
panels, and balconies with varying forms and shapes
[3]. GFRC elements can be categorised along a scale
of increasing geometric and manufacturing com-
plexity as either rectilinear, faceted and free-form,
and their manufacture requires a series of stages to
be connected together. The typical process of man-
ufacture starts with an analysis of the client’s input
data (2d CAD or 3d BIM), detailing and optimisation
of the design according to budget, manufacturing
and other constraints and the definition of quality cri-
teria for the later element, including surface quality
and tolerances in dimensions.
In a second stage - process planning - the mold
is designed and drawings of its parts are generated.
Most molds are made from flat sheet materials, com-
bining wood and plastic, cut to shape (Fig. 1). Be-
cause of the high variability in GFRC element design,
and resultant high geometric variability design, mak-
ing and assembling of molds is not automatised and
the resulting quality relies on the skills of craftsmen.
In order to effect overall cost and efficiency the de-
sign of GFRC elements and molds is tweaked, so that
a mold can be reused many times or easily adapted
to changing geometries in the production of an ele-
ment.
The 25% of molds at our industry partner, which
are faceted or free-form are more challenging, as 2D
drawings can fail to describe the complexity of the
form regardless of the number of details included.
Here, CNC routing can be used for the production of
curvilinear sub elements of molds or the 3d milling of
entire mold surfaces (Kreysler, 2017).
632 |eCAADe 38 - D2.T10.S2. ROBOTIC TECTONICS, AUTOMATION AND INTERACTION - Volume 2
Finally GFRC is sprayed in layers into the mold
and rolled a handheld roller (Fig.1), which requires
high level of craftsmanship again, since the strength
of the concrete is determined from the correct mix-
ture and compacting the concrete slurry with the fi-
bres (Brandt 2020).
Quality Control is imperative at all stages of pro-
duction. At two stages geometric discrepancies be-
tween the built mold and its specification are mea-
sured: 1) during the manufacturing of the mold and
2) after finishing the element. The use of contact-
type measurement devices (ruler & tape measures)
determine the achievable precision and type of mea-
surements. Complex mold geometries limit their use.
The permissible geometric discrepancies between
designed and built vary with the type of building el-
ements. Façade elements require relatively high pre-
cision, especially in the position of mounting points
(aim: functioning attachment), and for edges of el-
ements (aim: aesthetically pleasing gap dimensions
and alignment with adjacent elements). The toler-
ances for length are: +-3mm for elements < 3m, +-
5mm for elements < 9m, and 8mm for linear straight-
ness (Kim et al. 2020). Others quality aspects, like
straightness and squareness, have higher permissi-
ble tolerances and are in practice hardly measured.
The above workflow shows a disconnection be-
tween the physical production and the digital ele-
ment specification since the whole manufacturing
process relies on the moldmaker’s execution of 2d
technical drawings.
Figure 1
Mold building and
mold casting, both
specialized hand
craft processes
STATE OF THE ART IN
PRODUCTION FEEDBACK
The quality of the finished GFRC element is depen-
dent upon the quality achieved in each of the steps
from design to fabrication. According to the NRC (Na-
tional Research Council 1995) the ability to produce
quality products hinges on four key competencies:
modeling of process form and precision levels, de-
sign tolerancing of parts and products, selecting pro-
duction processes that match part specifications, and
applying quantitative measurement methods for in-
spection and process control. The first two-process
modeling and design tolerancing-are of primary im-
portance and drive the second two. The first three
take place pre-production, while inspection can only
take place during and post production.
The field of manufacturing engineering has de-
veloped multiple approaches to ensure and improve
quality and efficiency in production. A crucial ele-
ment is early and repeated testing and feedback at
each stage of development and production of ele-
ments (Matisoff 1986) . This feedback can be organ-
ised within a single organisational level of production
(Horizontal Control Loop) or between multiple or-
ganisational levels (Vertical Control Loops) (Schmitt
et al. 2011). Horizontal Control describes how one
process is being controlled and how this control loop
interacts with other control loops (e.g. end-of-line
inspection on shop floor level) Vertical Control de-
scribes a multi-scalar approach of how control loops
of lower levels are monitored, controlled and de-
signed (e.g. management assessment of business
processes). The need for higher quality production
methods, which can adapt quickly to changes in pro-
duction and are efficient at minimal batch sizes - as
in the call for an Industry 4.0 (Kagermann 2015) - has
led to an extension of thinking about quality control
in industry. This expands consideration to include 1)
the organisational level, where feedback is extended
to span customer, production and design in Closed-
Loop Quality Systems [4], 2) temporal aspects, with
an urge for continuous and real time feedback on hu-
man and machine processes (Even-Chaim 2019) and
D2.T10.S2. ROBOTIC TECTONICS, AUTOMATION AND INTERACTION - Volume 2 - eCAADe 38 |633
3) the further development of flexible manufacturing
systems (De Toni & Tonchia 1998). These last items
resonate with research in architectural construction
exploring online feedback systems for robotic man-
ufacturing with (Sutjipto et al. 2018) or without hu-
mans in the loop (Nicholas et al. 2017).
The triggering moment for feedback in any qual-
ity loop is the act of data generation, the registration
of the measured data to a reference model and the
analysis and feedback of deviations between these
(Martin Tamke et al. 2016).A vast array of large-scale
metrology technology exists for the capturing of di-
mensional data(Peggs et al. 2009). Ìn the context
of GFRC manufacturing, this list is narrowed down
to fast, contactless and mobile techniques 3d Laser
Scanner and Digital Photogrammetry. This is due to
the specifics of GFRC production:
• Large dimensions, weight, complex geometry
with few geometric features, undercuts, narrow
spaces and uniform color of molds and GFRC el-
ements
• Small batch size of production runs, which re-
quires frequent 3d capture
• Dusty production environments - necessity to
move gear gear off site, when not in use
• Financial constraints, require measurements to
be carried out by production floor personnel
(need for ease of use and high automatisation)
Both Laser Scanners and Digital Photogrammetry
provide sufficient precision to match the minimal
production tolerances of casting (1mm) (National Re-
search Council 1995) and the required tolerances in
the construction industry (M. Tamke et al. 2014).
Laser Scanning is used in research for control of sur-
face qualities in concrete production (Puri & Turkan,
2016; Wang et al. 2016) and onsite control of rebar
and concrete formwork (Kim et al. 2020).
Once scan data is generated, it needs to be pro-
cessed to enable dimensional feedback. However re-
lated industrial products [6,7], are generic tools made
for use in mechanical industries. In order to achieve
the required precision, these products must be op-
erated by specialists who adapt the measurement to
the specific object, postprocess data and execute de-
viation control in separate specialist software pack-
ages. And while technological developments allow
for a high degree of automatisation in the registra-
tion of large sets of point cloud data (Chen & Medioni
1992; Ochmann et al. 2014; Schnabel et al. 2007),
the automated analysis of dimensional deviations re-
mains a field of research (Martin Tamke et al. 2016).
METHODS
In this paper we present a new, prototypical frame-
work, termed Precision Partner - that enables the
craft practice of mold-makers by establishing an au-
tomated connection to precise, digital information.
In the previous section we have established the re-
quirements for this novel approach to horizontal pro-
duction feedback.
The workflow is established through software
and hardware prototyping in direct dialogue and
discussion with industry partners. We also de-
velop a larger research methodology for industrial re-
search collaboration, concerning real-case data gath-
ering, quick Design/Development/Testing and feed-
back and engagement with the craftsmen and engi-
neers to guide development and testing. Each soft-
ware and hardware implementation is evaluated for
its independent functionality and also for fit into the
Precision Partner framework as a whole.
The development of the framework (Fig. 2) is
guided by and tested against three main criteria.
First, the Precision Partner framework needs to
be open-ended, where components can be contin-
uously added. This reflects the high level of variabil-
ity in GFRC manufacture. A digital baseline approach,
to handle incoming project information and gener-
ate construction information, as well as set the stage
for registration and deviation feedback, is at the core
of the framework. Second, the framework neces-
sitates an easy integration with existing produc-
tion pipelines without too much disruption for
the business. For instance, differently from indus-
trial product manufacturing, GFRC molds are made in
634 |eCAADe 38 - D2.T10.S2. ROBOTIC TECTONICS, AUTOMATION AND INTERACTION - Volume 2
Figure 2
Relationship
between the digital
components and
the human
activities in the
Precision Partner.
Variability of the
types of human
interaction - station
based, physical (3d
scanning) and on
the fly.
an adaptive production context, where the factory fa-
cilities provide a flexible production sequence, rather
than a standard production chain, depending on the
size and amount of elements to be produced. This is
why we develop multiple flexible scanning scenarios.
Finally and most importantly, the framework needs
to be accessible to a workforce without special-
ized computer skills, since it will be used by shop-
floor craftsmen. We design a simple interactive UI
for on-the-fly dimension check, where the data pro-
cessing is fully automated, allowing the craftsman to
easily assess the quality of a component and react to
any deficiencies. We expand upon the technicalities
of each of these three development criterions in the
sections below.
Digital modelling
The core of our production feedback lies in being
able to compare the real element, to it’s digital rep-
resentation. For that, we require a standard set of
outputs: The geometry of the mold or casted ele-
ment in ply, with the geometry oriented orthogo-
nally in XYZ space, as well as a list of the critical di-
mensions that need to be checked, such as height,
length, width and thickness. We accommodate for
element variability while ensuring a baseline qual-
ity. In fact, since element design and specification
comes in different data formats from one project to
the other, we develop input pipelines specific to dif-
ferent 3d model formats. However, when the GRC el-
ement is rectilinear and geometrically simple the in-
put data from the client is provided as 2d drawing.
As these are not fitting to a 3d approach, they re-
quire heavy manual editing in order to create con-
sistent quality and labeling. We therefore develop
a parametric configurator. This produces from sim-
ple numerical inputs,both a 2d technical drawing set
and a cutting list to build the mold, as well as the
necessary 3d model of the element and the mold.
This drastically increases design-to-production pro-
ductivity and guarantees consistent data quality for
the following operations. It also allows very easily to
create partial digital models, to compare against par-
tial digital scans during the mold-building process.
D2.T10.S2. ROBOTIC TECTONICS, AUTOMATION AND INTERACTION - Volume 2 - eCAADe 38 |635
Figure 3
Back-end cloud
processing
workflows:
processing of the
acquired phyisical
scan, alignement
with digital model,
comparison and
feedback
Scanning Protocols
Horizontal feedback on the factory floor requires flex-
ible scanning protocols in order to be minimally dis-
ruptive to the ongoing production chain. At differ-
ent instances of the production, different parts of the
mold can be scanned to check the quality of the built
piece. Later, the whole mold can be checked prior
to casting, and the element can be checked after un-
molding. For that we develop two different scanning
scenarios: partial or small scans and complete scans.
Partial Scans - Handheld. Handheld scanners op-
erate through structured light. By shooting up to
16 frames per second, they are able to reconstruct
the surface they are scanning on the fly using pho-
togrammetry. Their use by craftsmen requires little
training. Scans can be made of specific edges, or spe-
cific mounting points, which then feed into the regis-
tration workflows described below. A craftsman can
also use printed trackers to check for a specific di-
mension of interest, that could be hard to measure
636 |eCAADe 38 - D2.T10.S2. ROBOTIC TECTONICS, AUTOMATION AND INTERACTION - Volume 2
with a ruler for example. We envision that these scans
would be used for fast intermediate checks during
the mould building process.
Complete Scans - Rig Based. Lidar scanner.capture
large areas quickly. This allows for a complete scan
of the mold during manufacturing or for a final qual-
ity control. We build a rig with a suspended 3d scan-
ner, which can be operated by a craftsman and po-
sitioned over the mould. The data is then passed to
the registration workflows we describe below.
Cloud Processing Workflows
In order to give feedback on dimensions, the entire
or partial scan, must be registered to its 3d model.
We develop a workflow (Fig.3) for automatic registra-
tion using the Open3d library for Python (Zhou et al.
2018). The workflow uses a combination of RANSAC
(Schnabel et al. 2007) algorithm for global regis-
tration and ICP (Chen and Medioni 1992) We pre-
pare the point clouds of the scans through noise re-
moval using nearest neighbour algorithms. We sub-
sequently analyse the density of the scans and gener-
ate a point cloud from the 3d model with similar den-
sity, as this improves the success rate of the registra-
tion drastically. Since Ransac relies on an initial ran-
dom seed, the results are previewed in the UI for con-
firmation. In the case of miss-alignment (currently
at 5% of cases, but protocols are being developed
to improve this) we developed a fallback manually-
aided alignment based on matching the bounding
boxes through simple slider-based operator inputs.
The transformation matrix is applied to the scanned
cloud to match it to the 3d model, allowing criti-
cal measures to be calculated on the scanned cloud
using nearest neighbour search based on the base-
line model. These workflows are all computed in the
back-end and need under 1 minute computing time.
UI for precision feedback
To guarantee accessibility of this tool, the whole
workflow is wrapped into a simple User Interface (UI)
built using Dash library for python [8]. The UI runs
on a web browser, and is built for interactive queries.
At startup, the operator selects the project folder,
in which cloud scan and 3d model are saved. The
tool displays the pre-defined dimensions for both the
scan and the model, and flags dimensions that are
beyond tolerance. The operator can then use the
spot-check feature to add additional dimensions, and
better diagnose the problem.
WORKFLOW EVALUATION
BALCONY CASE STUDY
In order to evaluate the performance of our work-
flow, we test it against a recurring type of element at
our industry partner: The balcony panel type. These
present a rectilinear base with edges on 3 sides. The
case study in question measures 2554x1297 mm and
is 170mm deep with a 12mm thickness. The edges
are at 90 degree with respect to the base, but present
a chamfer feature. The panel requires a 3x6 grid of
metallic mounts. The parametric configurator gen-
erates a 2d technical drawing set, as well as mold
cutting list for the mold builders. It also generates
a 3d model of the element, and a 3d model of the
mold (Fig. 4). We export these models and the crit-
ical dimensions to the Open3d workflow. The el-
ement mold is scanned both entirely and partially
(Fig 5). Scans are processed as described in the sec-
tion above. We generate a model point cloud con-
sisting of approx. 3.5million points to match an ap-
prox. 4 million point scan. We test two registration
methods on both partial and full scans, a fully auto-
mated one using Ransac and ICP and a manually aug-
mented one (Fig 6). The latter utilizes trackers to cre-
ate two sets of correspondence points. This is accom-
plished through the calculation of distances between
the three points for each set and the forming of a se-
quence (e.g. p1 corresponds to p1‘, p2 corresponds to
p2’, etc.). After that, a transformation matrix is created
in order to match the two point sets and it’s applied
to the scan. Through this the scan and the model are
roughly registered. Finally, ICP is used for local refine-
ment. The critical dimensions are computed on the
UI (Fig 7) and the operator can visualize and interact
with them.
D2.T10.S2. ROBOTIC TECTONICS, AUTOMATION AND INTERACTION - Volume 2 - eCAADe 38 |637
Figure 4
Outputs of the
mold configurator
parametric
algorithm
Figure 5
Scanning of a
balcony type mold
using Artec3d Eva
handheld scanner
DISCUSSION
The conducted experiments have revealed that the
relationship between 3d scan and 3d model is the
cornerstone of the workflow. In order to be able to
register the scans to its 3d model, they need to be
of similar nature. We have found that the geometric
features of the element have a big impact on the suc-
cess of the registration algorithm. Full automation is
successful when there is a 70-100% correspondence
between the point clouds. Partial scans with less than
50% correspondence will require partial models to be
registered to. This is possible when the model is gen-
erated from the parametric configurator. While being
a theoretical use case this would allow to produce as-
sembly sequence models, to be able to check at steps
in the mold building process.
CONCLUSIONS AND FUTURE WORK
In this paper, we have described research towards a
new framework for factory floor quality control that
considers the human in the loop. The framework in-
troduces Industry 4.0 technologies to enhance skilled
craftsmen in the making of molds through the pro-
vision of horizontal feedback. It allows a crafts-
man to assess the geometric accuracy of their work
against a digital model on the factory floor, through
3d scanning-based feedback of either pre or user-
defined dimensions and deviations. This framework
is tested through the case of a balcony element,
where correspondence is found to be a key factor in
successful automation.
As the paper details, the seemingly straightfor-
ward activity of gaining dimensional feedback via 3d
scanning is found to require a wider, coordinated
set of digital workflows. To achieve these for the
high level of geometric complexity and variety that
characterizes GFRC production, and with non-expert
users rather than a skilled on-site technician, requires
that the framework be both automated and open
ended.
Automation is used to ensure the relationship
between 3d scan and 3d model. An automated mold
configurator tool is developed to enable a baseline
for comparison between 3d model and point-cloud.
A fully automated headless workflow correlates 3d
scans made by either handheld or rig based scanning
tools against a 3d model, using RANSAC and ICP al-
gorithms. If needed, alternative workflows offer user
input.
Open Endedness is achieved by constructing the
framework as components. This enables different in-
put pipelines, different modes of 3d scanning best
suited to different mold geometries, extents of scan-
ning, and registration either during or at the end of
the making process. Here, further work is currently
being undertaken to better enable the use of par-
tial scans through specification of partial models, as
a means to address the issue of correspondence in
the registration of in-progress work. The question of
when feedback can occur can be linked to the ques-
tion of how feedback can be provided. Here, future
work is planned to extend the framework to incorpo-
rate visual spatial feedback through overlay of point-
cloud, model, deviation information and real world
data through augmented reality.
638 |eCAADe 38 - D2.T10.S2. ROBOTIC TECTONICS, AUTOMATION AND INTERACTION - Volume 2
Figure 6
Results of two
registration
processes of 3d
model and scan:
Upper two images -
Fully automated
RANSAC and ICP.
Down two images -
Manually
augmented
Correspondence
points and ICP.
Figure 7
Our UI displayed on
a web browser
D2.T10.S2. ROBOTIC TECTONICS, AUTOMATION AND INTERACTION - Volume 2 - eCAADe 38 |639
REFERENCES
Autor, D 2015, ’Why are there still so many jobs? The His-
tory and Future of Workplace Automation’, Journal
of Economic perspectives, 29(3), pp. 3-30
Brandt, N 2020, ’Developing a new generation of quality
control for GRC’, Concrete Engineering International,
January 2020, pp. 22-24
Chen, Y and Medioni, G 99, ’Object modelling by regis-
tration of multiple range images’, Image Vis. Com-
put., 10(3), pp. 145-155
Claypool, M, Jimenez Garcia, M, Retsin, G and Soler, V
2019, Robotic Building Architecture in the Age of Au-
tomation, De Gruyter
National Research Council, T 1995, Unit Manufacturing
Processes: Issues and Opportunities in Research, The
National Academies Press
Kagermann, H 2015, ’Change Through Digitization—
Value Creation in the Age of Industry 4.0.’, in Albach,
H, Meffert, A, Pinkwart, R and Reichwald, R (eds)
2015, Management of Permanent Change, Springer
Fachmedien Wiesbaden
Kim, MK, PutraThedja, JP and Wang, Q 2020, ’Automated
dimensional quality assessment for formwork and
rebar of reinforced concrete components using 3D
point cloud data’, Automation in Construction, 112,
pp. 103-113
Krysler, W 2017 ’Qualifying FRP composites for high-rise
building facades’, Fabricate 2017, pp. 130-137
Matisoff, BS 1986, ’Manufacturing Engineering: Def-
inition and Purpose’, in Matisoff, BS (eds) 1986,
Handbook of Electronics Manufacturing Engineering,
Springer Netherlands
Matt, DT, Modrák, V and Zsifkovits, H 2020, Industry 4.0
for SMEs: Challenges, Oppor tunities and Requirement,
Palgrave macmillan
Nicholas, P, Zwierzycki, M, Nørgaard Clausen, E, Hutchin-
son, C and Thomsen, M 2017 ’Adaptive Robotic Fab-
rication For Conditions Of Material Inconsistency In-
creasing The Geometric Accuracy Of Incrementally
Formed Metal Panels’, Fabricate 2017, pp. 114-121
Ochman, S, Vock, R, Wessel, R, Tamke, M and Klein, R
2014 ’Automatic generation of structural building
descriptions from 3D point cloud scans’, Proceed-
ings of the 2014 International Conference on Com-
puter Graphics Theory and Applications (GRAPP)
Peggs, GN, Maropoulos, PG, Hughes, EB, Forbes, AB,
Robson, S, Ziebart, M and Muralikrishnan, B 2009,
’Recent developments in large-scale dimensional
metrology’, Journal of Engineering Manufacture,
223(6), pp. 571-595
Puri, N and Turkan, Y 2016 ’Dimensional quality control
of precast concrete slabs using3D laser scan point
clouds’, Proceedings of the 16th International Confer-
ence onConstruction Applications of Virtual Reality
Schmitt, R, Monostori, L, Glöckner, H, Stiller, H and
Viharos, Z 2011 ’Designing closed quality control
loops for stableproduction systems’, Proceedings of
the 55th EOQ Congress
Schnabel, R, Wahl, R and Klein, R 2007, ’Efficient RANSAC
for point-cloud shape detection’, Computer Graphics
Forum, 26, pp. 214-226
Sutjipto, S, Tish, D, Paul, G, Vidal-Calleja, T and Schork,
T 2018 ’Towards Visual Feedback Loops for Robot-
ControlledAdditive Manufacturing’, Robotic Fabrica-
tion in Architecture, Art and Design 2018
Tamke, M, Myrup, M and Evers, HL 2014, Current state
of 3D object processing in architecturalresearch and
practice - DURAARK Deliverable No.D7.7.1, KADK, CITA
Tamke, M, Zwierzycki, M and Henrik, EL 2016 ’Track-
ing Changes in Buildings over Time - Fully Automat-
edReconstruction and Difference Detection of 3d
Scan andBIM files’, Proceedings of eCAADe 2016, pp.
643-651
De Toni, A and Tonchia, S 1998, ’Manufacturing Flexibil-
ity: a literature review’, International Journal of Pro-
duction Research, 36(6), pp. 1587-1617
Wang, Q, Kim, Mk, Cheng, JCP and Sohn, H 2016, ’Au-
tomated quality assessment of precast concrete ele-
ments with geometry irregularities using terrestrial
laser scanning’, Automation in Construction, 68, pp.
170-182
Zhou, QI, Park, J and Koltun, V 2018 ’Open3D: A Modern
Library for 3D Data Processing’, arXiv:1801.09847
[1] https://hbr.org/2016/04/a-chart-that-shows-which-i
ndustries-are-the-most-digital-and-why
[2] https://www.lexus.ae/about-us/takumi-the-60000-h
our-story-of-human-craft/
[3] https://grca.org.uk/pdf/Practical-Design-Guide-for-
GRC.pdf
[4] https://www.designnews.com/automation-motion-
control/design-can-escape-silo-through-closed-loop-q
uality-systems/151466620045975
[5] https://www.ofsystems.com/post/the-benefits-of-re
al-time-feedback-in-manufacturing
[6] https://3dmanufacturing.faro.com/us/
[7] https://www.gom.com/3d-software/gom-inspect.ht
ml
[8] https://plotly.com/dash/
640 |eCAADe 38 - D2.T10.S2. ROBOTIC TECTONICS, AUTOMATION AND INTERACTION - Volume 2
