A cost-function driven adaptive welding framework for multi-pass robotic welding
Introduction
The need to automate processes in manufacturing lies in the desire for high productivity, reduction of production costs and increase of profits through reduced labour cost [1]. Fusion welding is the process of joining two materials together, typically employing heat through solidification with or without the use of filler consumable material. Over the past decades, welding has made a considerable impact in different sectors of our world including, aviation, automotive, defence and marine through the fabrication of tanks, frigates and submarines [2].
Among the different types of welding, arc welding has evolved for over 100 years [3]. The different types of arc welding can be categorized based on the electrode method used, which is consumable or non-consumable [4]. Among the most common techniques, Gas Metal Arc Welding (GMAW) widely known as Metal Inert Gas (MIG) and Metal Active Gas (MAG) utilize a consumable filler wire. The distinction originates from the shielding gas that is employed, and this technique is suitable for both thin and thick sections resulting in high deposition rates and increased productivity [[3], [4], [5]]. The power source of these welders can be synergically controlled for a given wire material and diameter and in absolute wire feed speed or current mode, depending on which parameter the welder will control. A different variety of metals can be welded, which range from carbon steels, low alloy steels, stainless steels, aluminium alloys, copper, and nickel alloys.
In welded structures, loads are distributed between the welds of the joints. The type of joint geometry is determined through the geometric requirements of the assembly and the type of loading [4]. As can be seen in Fig. 1 [6], the basic joint designs are summarized as butt, corner, edge, lap and tee joints. The selection of the joint type also aligns with the requirement for the least amount of deposited weld metal to meet the strength requirements for load distribution [4]. For thick butt joints, the edges are mechanically prepared (machined, water- jet cutting etc.) to a particular geometry to provide adequate access for the weld torch and achieve even heat input flow and penetration between filler and parent metal [3,4]. These mechanically prepared geometries, which are shown in Fig. 2 [4], can be single sided or double sided for example with double V’s, and single or double U’s.
In the maritime, oil and offshore industries, the thickness of the single or double V-groove joints requires more than a single pass, usually manually or semi-autonomously deposited, resulting in low efficiency and productivity [7]. Although manual welding allows selecting different welding parameters per layer or even per pass if that is required, this depends on the experience of the welder and the limits and window of the approved Welding Procedure Specification (WPS) produced from a weld procedure qualification record.
Fig. 3 describes the terminology of multi-pass welding for a single sided V-groove open root gap assembly. The root pass refers to the initial welding pass used to join parent metals together, where a non-metallic backing strip can be used to support the root surface [3]. The hot pass is the second welding pass used to reshape the root pass, achieve sidewall fusion and fill any inconsistencies caused by improper penetration of the root pass [8]. Filler passes serve the remaining weld groove area until the cap passes are deposited to reinforce the weld groove and provide a clean finish to the top weld face. Welding passes deposited at the same height offset relative to the root face belong to the same layer. The vertical root face is used for to achieve proper fusion with the root sides.
Despite the massive utilisation of welding technology in manufacturing, the welding environment still imposes difficulties for the welders. During welding, high concentrations of fumes, gases, dust, infra-red and ultra-violet radiation are produced along with substances, such as nickel and chromium, which have an adverse effect in the human respiratory system and can lead to lung cancer and asthma [9].
Lack of space during welding can amplify the exposure to toxic fumes and increase the ambient temperature. Those unpleasant conditions can be met in enclosed spaces, as it happens in the pre-fabrication of double hull structures [10]. Nonetheless, during the maintenance of such structures, there is limited space for work. Fixed infrastructure that cannot be moved and other manufacturing processes taking place in parallel also increases the possibilities for injuries while working in a confined area.
A complex dynamic process such as welding, which is challenging to parameterize and to control [[11], [12], [13], [14]], must meet the demand for high production rates, precision, and consistent quality. However, labour turnover due to harsh environmental conditions further increase the shortage of skilled welders to fulfil the needs of manufacturing in the 21st century [15], affecting production and consistency of quality. Consequently, the life span of future assets can be reduced, and the amount of rework increased [16]. Automating welding can alleviate issues of repeatability, quality, and increased production demand. The bulk production of repeatable welds in multi-pass welding of known joint geometries can be delivered through robotic welding systems. Therefore, freeing welders to be utilized in more complex and creative tasks where a high degree of customisation is required.
When considering automated welding deployment, based on the Degrees of Freedom (DoF) that the welding systems exhibit, these can be commonly classified into either rectilinear (Fig. 4(a)-[17]) or articulated robotic solutions (Fig. 4(b)-[18]) [19]. Rectilinear robots (gantry systems) can demonstrate a constrained boxed working envelope, moving multiple axes to allow flexibility and volume coverage. Alternatively, articulated robots mimic human arms with six DoF, utilizing revolving wrists connected through joints and controlled by motors to cover the working volume with flexibility and speed.
These articulated robotic manipulators exploit interest for specific automation applications, due to their capabilities and multiple DoFs, increased pose repeatability and duty cycles. However, the realisation of fully automated robotic arc welding systems is not yet achieved, as the welding procedure, operating environment, welding joint geometry and preparation can vary significantly [20].
Considering this progress, and the conditions of common industrial welding environments, the call for flexible automated arc welding can be realized through decisional autonomy [21]. A closed-loop robotic system can alter the trajectory of the motion based on sensory feedback information (camera, laser scanner, touch-sensing). In that way, it can sense the local environment and superposes pre-programmed moves adapting to changes encountered in the environment.
The realization of a fully automated robotic welding system demands the development of a welding framework which combines sensor-driven robotic motion along with multi-pass sequence planning for the weld joint geometry. Sequence planning of multi-pass welding is imperative for automation of welding in the shipbuilding and offshore sector due to the requirement of the thickness of the joints. Moreover, this welding process is repetitive and monotonous, while manual welders can be utilized in more complex tasks.
This paper presents a new welding framework that enables automatic planning of the complete multi-pass welding sequence with different welding parameters per layer. Moreover, this approach adapts to varying single sided V-groove geometries, without human intervention, populating the number of layers and passes to minimise a desired cost function. These developments are demonstrated alongside a flexible 6-DoF sensor-driven robotic welding system. The rest of this paper is organised as follows. Section 2 introduces a multi-pass welding sequence review for V-groove joint configurations. Section 3 presents the welding framework and the developed algorithms for automated multi-pass weld planning based on the minimization of a cost function. Section 4 demonstrates the experimental setup, utilizing different butt joint geometries for a proof-of-concept validation along with inspection of the completed joints. The subsequent proof-of-concept demonstration characterization results are discussed in Section 5. Section 6 discusses future work with Section 7 concluding the paper.
Section snippets
The current state-of-the-art in multi-pass sequence planning
Mathematically describing and approximating the shape geometry of the deposited welding beads requires developing algorithms to generate the sequence of welding parameters and as a result, the robotic motion path. In the relevant works, welding beads are represented as parallelograms and trapeziums since often the cross-section shapes of weld beads match these shapes visually. Using the same welding parameters for all the deposited weld beads, the authors in [7,22,23] simplified the welding
Proposed automated welding framework
A mathematical model, relating the cross-section area of beads with the welding parameters, pose of the torch and weaving width, was adopted from [24] and built upon to allow full-process automated welding parameter generation, optimization and robotic path planning. The flowchart in Fig. 6 describes each step of the welding framework as well as the required user input. The following sections explains in detail the implementation of the different steps along with the notation used in the
Experimental setup
A series of experiments were undertaken to prove the feasibility of the proposed welding framework for multi-pass welding, aiming to automate the generation of the robotic motion path, welding parameters allocation and optimisation per layer based on the cost function concept. A 6 DoF articulated robotic welding system was deployed, and two different types of steel of three samples were bevelled under different V-groove geometries as can be seen in Table 7. The robotic setup is shown in Fig. 13
Generated welding results and cost functions
Three manufactured samples were produced by the proposed welding framework with the cost function concept described in Section 3.4, and presented in this work with welding parameters recorded in Table 8, Table 9, Table 10. The samples were welded using the generated welding parameters and visual inspection showed no undercuts or lack of fusion between adjacent passes. This was also validated from the PAUT inspection presented in Section 4.3.1.
The welding parameters utilized to weld each sample,
Future work
Future work should focus on the full automation of the welding procedure, investigating a suitable methodology for root and hot pass welding parameter allocation based on the gap and root face of the joints. Additionally, the order in which the welding beads are deposited, which revealed distortion, should be investigated, and incorporated to avoid adverse effects on the structural integrity of the weldments. Moreover, effort should be placed on the acceptance flexibility of the framework from
Conclusion
In this paper, an algorithmic framework for automated off-line robotic multi-pass V-groove weld path planning and sequencing is developed and validated. It can generate robotic welding paths and welding parameters for varying single-sided V-groove geometries, based on the operator’s choice to minimize a cost function based on the number of welding passes, welding time and filler wire consumption.
The population of welding passes per layer is driven from the algorithm presented in Sections 3.2
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgments
This work was funded under the Maritime, Enterprise, Innovation and Research (MEIR) program and industrial sponsor Babcock International Group PLC.
References (37)
Chapter 2 - Mechanised arc welding process options for pipework fabrications
Robots in the shipbuilding industry
Robot Comput Integr Manuf
(2014)- et al.
Multivariable adaptive control of the bead profile geometry in gas metal arc welding with thermal scanning
Int J Press Vessel Pip
(2002) - et al.
Dynamic welding process monitoring based on microphone array technology
J Manuf Process
(2021) Fatigue analysis of welded joints: state of development
Mar Struct
(2003)- et al.
6.04 - Robotic welding technology
- et al.
Optimal pass planning for robotic welding of large-dimension joints with deep grooves
Procedia Cirp
(2016) - et al.
Effect of welding sequences on residual stresses
Comput Struct
(2003) - et al.
Numerical calculation of residual stress development of multi-pass gas metal arc welding
J Constr Steel Res
(2012) - et al.
A review on high-frequency pulsed arc welding
J Manuf Process
(2020)
New applications using phased array techniques
Nucl Eng Des
Springer handbook of robotics
Fabrication and welding engineering
Principles of welding: processes, physics, chemistry, and metallurgy
Joining of advanced materials
Multi-pass route planning for thick plate of low alloy high strength steel by double-sided double arc welding
AMR
Cited by (7)
A sensor enabled robotic strategy for automated Defect-Free Multi-Pass High-Integrity welding
2022, Materials and DesignCitation Excerpt :The selected V-groove geometries used for experimental validation are commonly found in marine and manufacturing configurations where the characteristics of these were captured from the optical and tactile inspection process, recorded in Table 3.6. The cost function C(80 %,5%,15 %) was selected to achieve the minimum number of passes [45]. Fig. 3.5 depict the produced welding schedule for Joints 1 and 3 for the minimum value of the assigned cost function.
Robotic welding system for adaptive process control in gas metal arc welding
2024, Welding in the WorldForming Control Method of Single-Layer Multi-Pass Welding in the Intersection for T/K/Y Joints of Jacket
2023, Steel Research InternationalGlobal and Local Cost Calculations at Welded Structures
2023, Lecture Notes in Mechanical Engineering