Teaching robots to weld aluminium

Leirvik, the automation expert company EGH, and Digitread have joined forces on an exciting project with the objective of automating aluminium welding using robots.


Head of Technology & Digitalisation at Leirvik

Lervik is an EPC supplier that engineers, procures and constructs (EPC) large residential areas that are used offshore and that contain hotel facilities and offices. As an EPC supplier, Leirvik usually takes charge of the entire project from detailed engineering design, procurement of all the equipment and necessary materials, and then construction in order to deliver a functioning facility or asset to their client. To achieve this, Leirvik has a design department and a shipyard that is capable of assembling the largest aluminium structures in the world.

“We also have the world’s leading expertise in welding aluminium, and we have now started work on a project called ARoW (Automatic Robot Welding) where we will work with automated and robotic welding. In other words, we want to turn the cutting-edge expertise we have today into tomorrow’s digital expertise.

When a welder is handed a drawing today, the person knows what to do and starts welding. The question is how can his knowledge be turned into standard program code that enables a robot to do the same job. We aim to find some answers to such problems in this project“, states Morten Bjelland, Head of Technolgy & Digitalisation at Leirvik.

A changing industry
There are several reasons why Leirvik wants to automate certain welding processes, one of the most important being the potential for improving efficiency. Streamlining is also necessary to ensure that Leirvik remains competitive in the future, both in existing and in new markets.

“Leirvik is actively working to find new areas where we can make use of our core competence; multidisciplinary design, project implementation and construction of large aluminium structures. Various infrastructure projects, in particular aluminium bridges and various structures for use in land-based fish farms, could prove to be other relevant business areas for us. But regardless of new focus areas, the common denominator will still be to use the expertise, technology and long-term experience we have in large aluminium constructions, EPC implementation and to find smarter ways to work“, Bjelland continues.

With Digitread on the team
Basically, Leirvik’s production processes consist of a lot of tailoring and only a few repetitive products.

“But if we break the products down into the various aluminium profiles which we use in several parts of our production, then there are repetitive processes that can be well suited for automation.

Hence, we believe that it is possible to create a profitable method of automation, but to make it happen, we must look more closely at some of the elements that must be in place to make it a success. This is where Digitread becomes a key partner for us“, he explains.

To explain in simple terms, Digitread will create a digital representation of the robot cell, place the product in the digital twin and program what the robot will do. If the simulation is correct, the process is saved as a robot code. The robot code can then be uploaded as a job package which, in turn, becomes the robot’s job description. It enables the robot to continually repeat the job to be done.

“When the project began its validation phase in January 2020, we tasked Digitread with generating welds in 3D automatically based on standardized rules. We also challenged them to make fully automatic cutting sketches for the parts to be cut out and put together. Digitread had three months to prove that they could rise to the challenge, and they more than succeeded, they surpassed themselves“, reports Bjelland.

Project meeting with Digitread.

Many challenges
In a pilot project such as ARoW, there are a great number of different challenges that must be solved so that the robot performs in the correct way. One such example is a laser sensor being used to help auto-correct the welding.

“When a welding job is simulated in the digital world, everything is perfect – but that’s not the reality in the real world. Hence, we use data from the laser sensor as a knowledge base to adjust any sources of error that may occur in real life. The laser sensor scans the object, establishes if it is positioned correctly, and if it is not, it adjusts the robot according to reality. We start with the simplest welds, and if we manage to weld the simple ones in a proper fashion then this covers the majority of welds encountered. But even if the welds themselves are simple, the process is more complex and involves a lot of trial and error. Thus, a lot of mockups are required to identify and remedy any errors and repeat the process before everything is completely correct“, he adds.

“The pilot project was completed in November 2020 and showed such good results that Leirvik has decided to continue with ARoW phase 2, where they will continue the work and develop the solution further along with with Digitread and other partners“, Bjelland concludes.

Part of the ARoW team – From left; Vebjørn Jansen, Kristian Økland, Alfred Gregersen and Morten Bjelland

Picture of the robot cell built for the pilot project.

PUBLISHED: January 07th 2021