Siberian artificial intelligence to aid welding robots
15 Sep '16
Scientists at the Tomsk Polytechnic University (TPU) in Siberia are developing an artificial intelligence algorithm expected to improve the precision of electric resistance welding. Using the new method is believed to make welded joints much stronger, the TPU website announced.
The new system could be used at companies that manufacture motor vehicles, aircraft, ships and other sizable mechanical engineering products to boost their quality.
“This welding technique utilizes high pressure and temperature to connect parts in a number of spots. For each spot, welding takes a few milliseconds. It’s a robotized process,” said Boris Pyakillya from TPU’s integrated computerized control systems chair, the chief developer in this project.
He pointed out that today industrial welding robots use a standard algorithm for all items, based on a recommended electrical current value and duration of impact on a specific metal. Individual features of each specific material, for example, alloy content, possible impurities, etc., are typically disregarded, which compromises the quality of a welded joint, making it potentially too fragile.
“We suggest that our proprietary artificial intelligence algorithm, based on a mathematical model of the welding process, be used to determine electrical current and time for each specific welding joint, taking the material’s individual features into account. The welding machine has sensors that can identify voltages in the welding circuit, as we know that the properties of surfaces to be welded can alter voltages in the circuit as welding continues,” the chief developer said.
The young researcher wants to link a welding machine up to a small computer with a special app he has developed. The app will receive voltage data and take just seconds to compute electrical current and welding duration required.
Mr. Pyakillya’s TPU team is conducting experiments to fine-tune their model. The developers hope to unveil their inaugural prototype next year.