Available online at www.sciencedirect.

com

ScienceDirect
Procedia CIRP 99 (2021) 478–483
www.elsevier.com/locate/procedia

14th CIRP Conference on Intelligent Computation in Manufacturing Engineering, CIRP ICME ˈ20

Programming of Welding Robots in Shipbuilding

Alexander Zych*
Fraunhofer IGP, Albert-Einstein-Str. 30, 18059 Rostock, Germany
* Corresponding author. Tel.: +49-381-496-8243; fax: +49-381-496-8212. E-mail address: [email protected]

Abstract

The use of robots in shipbuilding in particular for welding plays an important role for shipyards to reduce costs and overcome the lack of
skilled welders. When using robots in shipbuilding one major subject is to find a suitable method for robot programming because one-off
production in combination with conventional methods often leads to an unfavorable ratio of programming time to welding time. This paper
gives an overview of different robot programming methods in shipbuilding starting from CAD-based methods through hybrid solutions with the
use of sensor data to sensor-based approaches completely integrated in the production process.
© 2021 The Authors. Published by Elsevier B.V.
This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0)
Peer-review under responsibility of the scientific committee of the 14th CIRP Conference on Intelligent Computation in Manufacturing Engineering,
15-17 July 2020.
Keywords: shipbuilding; welding; robot programming

1. Introduction shipyards the proportion of robots for welding applications is

still very low. There are many reasons for this situation. On
In the last decades, welding has been the most important the one hand, shipbuilding has special demands on robotic
joining technology in shipbuilding and unaffected by welding applications compared to other industries (see section
developments in the field of alternative materials like 2). These demands often lead to comparatively high
composites this is expected to remain true in the future. For investment costs for conventional robotic production lines.
that reason, the mechanization and automation of welding Especially small and medium shipyards can thereby often not
processes plays a major role for yards with regard to their achieve a high utilization of these robot systems. This results
competitiveness on the world market. First industrial robot in long amortization periods. Furthermore, conventional
welding applications were implemented in the 1980s in Japan. welding robot systems in shipbuilding are usually inflexible
Especially bigger yards in Europe, the USA and South Korea for instance with regard to changes in component range or
followed this development in the 1990s. The main reasons for enhancement of the functionality (e.g. automatic inspection
using welding robots at that time were the lack of qualified tasks).
welders, the increase of work safety and quality as well as the On the other hand, robot programming typically generates
large proportion of labor costs [1]. high operational costs due to the small batch or one-off
Since then, labor costs have doubled whereas robot prices production in shipbuilding. Typically robot programs are only
have been halved [2]. Furthermore significant advancements used once. This leads to an unfavorable ratio of programming
of welding robot technology have been made with regard to time to production time when using conventional robot
the robotic hardware, sensor technology and robot programming methods tailored to the needs of serial
programming methods [3-5]. By contrast, the number of production (e.g. in the automotive sector). Despite numerous
welding robots in the shipbuilding industry did not increase in approaches to reduce robot programming costs in shipbuilding
the same way as in other industrial sectors even though the over the last decades [7-11], the development of cost efficient
reasons for using welding robots mentioned above are even programming methods is still highly relevant today and a key
more relevant today [6]. Especially in small and medium sized factor to increase the number of robots used in shipbuilding.

2212-8271 © 2021 The Authors. Published by Elsevier B.V.

This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0)
Peer-review under responsibility of the scientific committee of the 14th CIRP Conference on Intelligent Computation in Manufacturing Engineering,
15-17 July 2020.
10.1016/j.procir.2021.03.107

This is a resupply of March 2023 as the template used in the publication of the original article contained errors. The content of the article has remained unaffected.
 Alexander Zych et al. / Procedia CIRP 99 (2021) 478–483 479

2. Hull structure production and requirements for robotic 2.1. Component dimensions
welding
In shipbuilding, due to large component dimensions, the
In general, the hull structure of ships is composed of plates use of welding robots with a reach of about 1.5 to 2 m requires
and profiles (e.g. bulb, flat, T). Further components are used additional positioning systems. Preference is given to the
to increase stiffness between plates and profiles (e.g. brackets, application of the following positioning systems (see also [13-
and collar plates) or to ensure tightness in special ship sections 16]):
(sealing plates). By introduction of sectional construction, the
production of hull structure has been divided into different  Robot gantries (3 axis respectively 2 axis for flat
levels (see Fig. 1). In later production levels the component components)
dimensions naturally increase as well as their tolerances. The  Crane driven systems (to position robots in boxes of open
increasing tolerances are attributable mainly to deformations sections)
caused by heat input of welding operations [12]. By contrast,  Mobile solutions (particularly to work in closed sections)
automation level is decreasing in later production stages,
among other reasons due to the high dimensions and In the production of micro panels and panels, robots gantry
tolerances. are used. Furthermore, there are two basic system concepts for
Apart from some mobile applications in closed sections, production. Firstly, the location of workpieces is stationary
welding robots are mainly used in the production of micro and the working area of the robot gantry is divided into
panels, panels, curved panels and open volume sections. The different tasks areas (positioning of components, robotic
production process of these components may generally be welding and finalizing tasks). Tasks areas are switched in
described as follows: accordance with the progress of work. Secondly, workpieces
are transferred between different work stations (e.g. by means
 Positioning and tack welding of individual components of roller conveyor). Fig. 2 shows a corresponding example of
(manually / mechanized by positioning devices) a micro panel line. Work stations 1 and 2 are used for
 Robotic welding of seams between individual components workpiece positioning followed by a buffer station. On
(usually with six-axis articulated robots) stations 4 to 6 two robot gantries are used for welding. The
 Finalizing tasks (manual welding of unfinished or by robot other stations are used for finalizing tasks.
unreachable seams, weld seam inspection...)
2.2. Tolerances
Exceptions to this process occur, for example, when using
laser hybrid welding. In this case the positioning of profiles is Deviations of component positions and geometry require
usually automated in addition with a clamping system to the use of sensor information to adapt robot trajectories. Kah
minimize the gaps between plates and profiles. In this process, et al. provide an overview of different sensors in robotic arc
no tack welding is needed. The welding process starts directly welding [4]. Before welding, a seam finding (workpiece
after clamping the profile. calibration) has to be performed to compensate positional
Compared to conventional robotic path welding tolerances. The most common approach is touch sensing by
applications, shipbuilding applications are characterized by wire tip or gas nozzle. Using the wire tip allows searching in
some specific requirements. To be mentioned in particular in cutoffs, but wire cutting is required to obtain good precision.
this context are the large component dimensions, the high Optical sensors are also used for seam finding. However,
tolerances with regard to workpiece location as well as using additional sensors such as light section sensors increases
geometry and the partially limited working space for robots to the interference contour of the welding tool and therefore
operate in (e.g. boxes with an area of 0.5x 0.5 m). limits the usability especially in confined areas.

Fig. 1. Hull structure production: Levels / Components

Fig. 2. Layout of micro panel line at Warnow Yard

This is a resupply of March 2023 as the template used in the publication of the original article contained errors. The content of the article has remained unaffected.
 480 Alexander Zych et al. / Procedia CIRP 99 (2021) 478–483

During the welding process, geometric tolerances can be as positions of external robot axis (task space problem). To
compensated by seam tracking. Arc sensing does not require cope with this partly infinite large solution space, they
additional sensors. However, arc sensing is not suitable for suggested dividing the problem into smaller sub-problems,
low seam volumes. As an alternative, laser-based light section which can be sequentially solved. Taking this into account,
sensors can be used. Some light section sensors also allow the welding task space in shipbuilding can be divided at least in
measurement of the welding gap and the corresponding two sub-problems. Firstly, finding appropriate torch poses
adaptation of welding parameters. without collision (torch angles optimized with regard to
requirements of welding process). This can be done within the
3. Welding robot programming methods process planning step. Secondly, finding appropriate robot
configurations and positions of external robot axis. This will
Besides the in section 2 mentioned requirements to robotic be part of the trajectory planning step.
welding in shipbuilding, one major challenge is, due to small Given this general procedure in robot programming for
batch production, a cost-efficient programming of welding welding applications, different methods used in shipbuilding
robots. Thus some of the first robotic applications failed due are presented in the below sub-sections.
to high programming effort and costs [17]. In general, robot
programming methods can be divided into two groups: online 3.1. CAD-based programming
programming and offline programming (OLP). The first
welding robots in shipbuilding were programmed online After implementation of first robotic welding systems in
(Teach-In) [18]. Because of high cost and effort, online shipbuilding and drawbacks due to insufficient programming
programming was quickly superseded by OLP methods. methods, producers focused on the development and
Recently, new developments in the area of collaborating advancement of offline programming systems based on CAD
robots and new teaching concepts have made robotic welding data. The main goal of these developments was to reduce the
in combination with online programming in the area of metal effort in robot programming.
constructions more attractive, even form small and medium With regard to the programming steps of generating
enterprises and small batch production. However, these workpiece geometry and creating tags, different solutions
approaches cannot be transferred directly to the shipbuilding were implemented. On the one hand, CAM solutions were
industry due to the large component dimensions. realized to transfer CAD information directly into the robot
Pan et al. introduced general key steps of offline programming system [9,17]. On the other hand, solutions were
programming [5]: developed to input the necessary information in the form of a
parametric workpiece description based on drawings [17,19].
 Generation of 3D CAD model (optionally being supplied Usually, the second approach was used in conjunction with the
with data from 2D CAD Model or 3D Scanner) use of macro technology (see below).
 Tag creation (Extracting robot position from part Further improvements were made in the area of process and
geometry) trajectory planning. A common approach reducing the
 Trajectory planning programming effort was using macros (also called templates)
 Process planning [13,18,20,21]. In robot programming, macros are predefined
 Post-processing movement patterns for frequently repeating assemblies of
 Simulation (optional) components and variations with regard to their dimensions.
 Calibration These variations are considered by input of parameters. By
using macro technology, programming time was decreased
These steps could also be transferred to other programming significantly. However, some applications had a large amount
methods. However, for welding applications in shipbuilding at of macros. This led to a significant effort in maintenance and
least some modifications should be made as described below. care of these macros as well as in the selection of an
The first programming step should be of more generic appropriate macro while programming. Furthermore macro-
nature, as existing programming methods do not necessarily based programming usually did not include collision check
use CAD data (see section 3.2 and 3.3). As an alternative, and simulation of robot movements. Along with the macro
“Generation of workpiece geometry” could be used. technology alternative OLP systems were developed,
Furthermore, the order of step 3 and 4 should be changed, including trajectory planning based on CAD data and
since process planning can also include sub-steps as the automatic collision avoidance [17,22,23].
division of seams due to technological requirements and a task Further developments have merged both approaches, macro
subdivision und distribution to several robots. These sub-tasks technology and collision-free trajectory planning [9,16,24].
can have an impact on trajectory planning. Pan et al. also The aim of these systems, often referred as Automated Offline
proposed to order welds with regard to optimized cycle time Programming (AOLP), was to further minimize user input
[5]. However, in shipbuilding, the order of weld seams should within the complete process chain of programming [25]. One
be optimized with regard to minimal welding deformations. example is the automatic creation of tags (seam identification)
Larkin et al. related to the general key steps of offline based on predefined rules. As a result, ratio of programming
programming mentioned above [11]. In this respect, they time to production time could be reduced to approx. 1:15.
addressed the problem of only partially constrained weld paths However, respective OLP systems have some
due to degrees of freedom with regard to torch angles as well disadvantages, especially for small and medium shipyards.

This is a resupply of March 2023 as the template used in the publication of the original article contained errors. The content of the article has remained unaffected.
 Alexander Zych et al. / Procedia CIRP 99 (2021) 478–483 481

Usually, system design is not open and implementing new OLP systems e.g. with regard to the arrangement of
block shapes or structures requires considerable effort and components in the shop floor. For that reason, these systems
high-level programming skills, that are often unavailable are primarily suited for the production of micro panels.
within small and medium shipyards [4,10,26].
Finally, OLP systems do not consider deviations of 3.3. Sensor-based programming
workpiece positions and geometry. Large shape deviation due
to welding distortions might lead to collisions between robot In contrast to the previously introduced programming
and workpiece. Furthermore, using offline programs in methods, the approach of sensor-based programming is
production usually requires work-intensive alignment of completely independent of CAD data and routine user inputs.
workpieces according to the predefined locations. Fig. 3 shows the essential hardware and software components
of the sensor based robot programming method. In the
3.2. Hybrid programming methods following, the respective programming process is described.
In the first step, three-dimensional digitalization of the
About 15 years ago and inspired by the development of tack-welded components is carried out. For this purpose the
workshop-oriented programming in the machine tool industry, 3D sensor system usually is being moved to different scanning
programming of welding robots started to partly shift back positions within the working area to digitalize the complete
from office (OLP) to production [27]. The aim of this components with minimal shadings. For positioning of the
development was to use knowledge of workers in the shop sensor either the welding robot gantry or a separate kinematic
floor area about the production processes as well as sensor device may be used. Depending on the type of components
information of component’s locations in the programming (e.g. micro panels / double bottoms) and the used 3D sensor
process. However, these programming methods still use key hardware, digitalization process may be divided into two
components of OLP systems in the programming process. steps. A preliminary, low-resolution acquisition to determine
Holamo and Ruottu introduced a programming system overall component structure and a detailed acquisition stage
using information of a 2D vision system to determine panel with optimized sensor positions according to the current
positions in the shop floor [8]. The robot programming system component positions.
uses either CAD-data or user input to determine seam Thereafter, automatic processing of acquired 3D data is
parameters (tag creation). Such programming method is performed. As a result, shape and location of individual parts
characterized by a ratio of programming time to production of are identified followed by the determination of seams to be
approx. 1:16 and thus comparable to advanced OLP systems welded (tag creation). After identification of component
[28]. geometry and seams, an initial sequence of weld seams is
Similar approaches are introduced by Ferreira et al. and determined based on predefined rules.
Kang et al. [26,29]. The system presented by Ferreira uses 2D In the next step, results of data processing are visualized
machine vision to identify workpiece locations and CAD data within a graphical user interface. At this point, the operator
to obtain workpiece geometry. Only minimal user input is has the opportunity to make modifications (e.g. sequence,
needed in program creation (selection of CAD model and weld directions) required for specific components, which
seams / allocation of welding parameters). The system differ from standard. Finally, the operator is releasing the
introduced by Kang uses a stereo vision system with cameras current job for production and respective production data is
installed on a positioning device to cover the entire working converted into a robot program by means of a postprocessor.
area. CAD data is used to provide necessary workpiece First industrial application using the sensor-based robot
information. Information of the vision system is used to programming approach was realized in 2008 at Warnow Werft
identify workpiece positions as well as shape data (e.g. in Rostock [30]. Therefore, the existing micro panel line was
positions of profiles on the plate). During robot program upgraded by installing required components (3D laser scanner,
creation some user input is needed (e.g. allocation of CAD positioning device, control system) on a buffer station ahead
data to vision data or adaptation of welding positions). The of the welding robots (see Fig. 2 station no. 3). Fig. 4 shows
system also includes functionality to allocate jobs to different the hardware components of the programming system. A 3D
robots with interference check between the individual robots laser scanner (No. 1) is installed in a protective housing due to
to avoid collision. heavy contamination of ambient air. The present production
A different programming approach was introduced by line concept allows a complete decoupling of programming
Galindo et al. [10]. This method is based on the macro and robotic welding tasks so that there are no non-productive
technology and uses sensor information to obtain the specific times due to the programming process.
workpiece geometry. Different macros (here called templates)
represent different shapes of workpieces and are defined
independently of their geometry. A portable measuring arm is
used by an operator in the shop floor to determine actual
workpiece positions and dimensions after selecting the
appropriate template. Due to comparatively short time needed
for measurement, one operator can program several robots.
The introduced hybrid programming methods are
characterized by a higher degree of flexibility compared to
Fig. 3. System components for the sensor-based robot programming

This is a resupply of March 2023 as the template used in the publication of the original article contained errors. The content of the article has remained unaffected.
 482 Alexander Zych et al. / Procedia CIRP 99 (2021) 478–483

Furthermore, postprocessor’s functionality had to be

extended with regard to process and trajectory planning
methods to cope with the high risk of collision in confined
spaces and the infinite solution space due to redundancy of the
kinematic structure. To this end, the task space problem was
divided into sub tasks and further restrictions were considered
e.g. minimization of hose package twisting. On the considered
test line, a ratio of programming time to production of approx.
1:12 was achieved, despite making use of the robot gantry for
3D data acquisition and investigating rather complicated
components.
Compared to other programming methods, sensor-based
robot programming offers several advantages:

 Consideration of actual component geometry (e.g. welding

distortions) and component positions
 Independence from preceding steps in the designing and
technological work
 Integration of programming into production process
 High flexibility of production (e.g. production sequence,
work piece locations)

By using 3D laser scanner technology, data acquisition is

independent of ambient light conditions and well suited for the
digitalization of shop-primed steel components. However, for
the digitalization of highly reflective surfaces such as
Fig. 4. 3D data acquisition on buffer station of micro panel line
aluminum or stainless steel, 3D laser scanners have essentially
proved unsuitable.
Following this approach, further micro panel production
lines were upgraded or rather newly installed. Based on the 4. Conclusion and future work
respective production line concept, the 3D sensor was
mounted on the welding gantry. However, using alternative Different methods for the programming of welding robots
sensor technology, downtime of welding robot due to data in shipbuilding have been presented. At present, the majority
acquisition could be reduced to less than 3 minutes [31]. of welding applications is programmed using offline
Since first sensor-based programming systems were only programming systems. The advancement of respective
suited for micro panel production, the programming method systems with regard to automation of programming steps has
was extended also for the production of volume sections [32]. led to lower programming costs even in small batch
The major challenge to this approach was to deal with the production. However, OLP systems still require highly
significantly more complex component structures with regard qualified operators and high investment and maintenance
to 3D data acquisition, data processing and trajectory planning costs. Thus, they are in particular appropriate for bigger
since the robot will have to work in confined spaces with a shipyards with several robotic production lines. Furthermore,
high risk of collision. Therefore, adaptive scanning strategies tolerances with regard to workpiece location and geometry
have been developed as well as iterative segmentation cannot be considered, which might lead to problems in
procedures. Fig. 5 shows the result of such point cloud production.
segmentation procedure with identified individual parts of a Other programming approaches, such as hybrid or sensor-
box. based methods, present a good alternative especially for small
and medium sized shipyards or where high demands are made
with regard to flexibility. By integrating robot programming
into the production process and using at least actual workpiece
positions, individual production processes can be optimized.
In spite of many major improvements in programming of
welding robots, further research is required especially with
regard to increase flexibility of overall robotic applications.
Some of these future issues are listed below.

 Use of alternative 3D sensor technology for sensor-based

programming to enable acquisition of high reflective
materials (e.g. by use of dynamic machine vision systems)

Fig. 5. Segmented point cloud of a box

This is a resupply of March 2023 as the template used in the publication of the original article contained errors. The content of the article has remained unaffected.
 Alexander Zych et al. / Procedia CIRP 99 (2021) 478–483 483

 Integration of further automatic production steps (e.g. Symposia on Intelligent Industrial Automation and Soft Computing.
automatic weld seam inspection with robots) Genova, Italy. Vol.1; 1999.
[15] Lee D et al. Development of a mobile robotic system for working in the
 Increased mobility of robotic systems to enable the use in double-hulled structure of a ship. In: Robotics and Computer-Integrated
different locations within a yard and thereby increase Manufacturing 26; 2010. p 13–23
utilization of robots (especially important for smaller [16] Lin W, Luo H. Robotic Welding. In: Nee A (eds). Handbook of
yards) Manufacturing Engineering and Technology. London: Springer; 2015
[17] Sagatun S I, Kjelstad K E. Robot technology in the shipyard production
environment. In: Journal of Ship Production, 12(1); 1996. p 39–48
Furthermore, alternative programming and operational [18] Di Filippo G, Manzon L, Maschio P. An Integrated Steel Workshop for
concepts, like augmented reality (see e.g. Ni [33]), offer Shipbuilding: A Real Application of Automation. In: Journal of Ship
promising prospects in case of adaptation to yard-specific Production, 14(4); 1998. p 225–237
requirements. [19] Rooks B. Robot welding in shipbuilding. In: The Industrial Robot 24(6);
1997. p 413–417
[20] Skjoelstrup C E, Ostergaard S. Shipbuilding Using Automated Welding
References Processes. In: Welding Review International 13(1); 1994
[21] Reeve R, Rongo R, Blomquist P. Flexible Robotics for Shipbuilding. In:
[1] Boeckholt R. Welding mechanization and automation in shipbuilding Journal of Ship Production, 13(1); 1997. p 36–47
worldwide. Cambridge: Abington Publishing; 1996 [22] Overgaard L, Petersen H G, Perram J W. A general algorithm for
[2] Tilley J. Automation, Robotics, And The Factory Of The Future. dynamics control of multi-link robots. In: International Journal of
McKinsey&Company; 2017; https://www.mckinsey.com/business- Robotics Research 14; 1995. p 281-294
functions/operations/our-insights/automation-robotics-and-the-factory-of- [23] Blasko G J, Howser B C, Moniak, D J. Evaluation of Hitachi Zosen
the-future# Welding Robots for Shipbuilding. In: Journal of Ship Production, 9(1);
[3] Pires JN, Loureiro A, Bölmsjo G. Welding robots: technology, system 1993. p 58–67
issues and applications. London: Springer; 2006. [24] Larkin N. et al. Offline programming for short batch robotic welding. In:
[4] Kah P et al. Robotic arc welding sensors and programming in industrial 16th Joining of Materials (JOM) conference 2012; 2011. p 1-6
applications. In: International Journal of Mechanical and Materials [25] Bedaka, A K, Vidal J, Lin C Y. Automatic Robot Path Integration Using
Engineering 10 (13); 2015. p 1-6 Three-Dimensional Vision and Offline Programming. In:The International
[5] Pan Z et al. Recent progress on programming methods for industrial Journal of Advanced Manufacturing Technology 102(8); 2019. p 1-16
robots. In: Robotics and Computer-Integrated Manufacturing 28; 2012. [26] Ferreira L A et al. Offline CAD-based robot programming and welding
p 87-94 parametrization of a flexible and adaptive robotic cell using enriched
[6] Lee D. Robots in the shipbuilding industry. In : Robotics and Computer- CAD/CAM system for shipbuilding. In: Procedia Manuf. 11; 2017.
Integrated Manufacturing 30(5); 2014. p 442–450 p 215–223
[7] Ang Jr M H, Lin W, Lim S –Y. A walk-through programmed robot for [27] Morocutti M. Actual Trend in Automated Shipbuilding Plants. In: Drews
welding in shipyards. In: Industrial Robot: An International Journal 26(5); P (eds.) Mechatronics and Robotics 2004, Aachen: Eysoldt; 2004
1999. p 377-388 [28] NN. Pema Vision Robot System – More Power to Production. In:
[8] Holamo O P, Ruottu K. Machine Vision System-Aided Robot Welding of PEMANEWS - Pemamek Customer Magazine; 2008
Micropanels. In: DVS-Berichte 237, Düsseldorf: DVS-Verlag; 2005. [29] Kang S W et al. Development of multi welding robot system for sub
p 516-519 assembly in shipbuilding. IFAC Proc. 41(2); 2008. p 5273–5278
[9] Jacobsen N J, Jacobsen C H. Generating a Job Description for Motion [30] Wanner M - C. Zych A, Pfletscher U. Automatic Robot Programming
Planning. In: IFAC Proceedings Volumes 40 (3); 2007. p 24-29 System for Welding Robots in Micro-panel Production. In:
[10] Galindo P L et al. Development of a Customized Interface for a Robotic Schiffbauforschung 47(1); 2008. p 39-48
Welding Application at Navantia Shipbuilding Company. In: Ollero A et [31] Seyffarth P, Gaede R. Image Processing for Automated Robotic
al., editors. ROBOT 2017: Third Iberian Robotics Conference. Advances Welding. In: Robotic Welding, Intelligence and Automation; 2011. p 15-
in Intelligent Systems and Computing 694. Springer; 2018 21
[11] Larkin N et al. Automated Programming for Robotic Welding. In: [32] Harmel A, Zych, A. Sensor-Based Robot Programming for Automated
Transactions on Intelligent Welding Manufacturing. Singapore: Springer; Manufacturing of High Orthogonal Volume Structures. In: Tagungsband
2018. p 48-59 des 3. Kongresses Montage Handhabung Industrieroboter; 2018. p 93-101
[12] Mandal N R. Ship Construction and Welding. Singapore: Springer; 2017 [33] Ni D et al. Haptic and visual augmented reality interface for
[13] Okumoto Y. Advanced Welding Robot System to Ship Hull Assembly. programming welding robots. In: Advances in Manufacturing 5(3); 2017.
In: Journal of ship Production 13 (2); 1997. p 101-110 p 191–198
[14] Bostelman R, Jacoff A, Bunch R. Delivery of an advanced double-
hullship welding system using Robocrane. Third International ICSC

This is a resupply of March 2023 as the template used in the publication of the original article contained errors. The content of the article has remained unaffected.