AI helps the impossible become possible
When Sandvik Coromant engineers, Henrik Loikkanen and Jakob Pettersson, were tasked with creating an AI-generated, stainless steel synthesis of some of history’s most famous works of art, their metal cutting expertise was put to the ultimate test.
In partnership with Sandvik Group, Sandvik Coromant has developed a statue for the history books. Made using AI modelling and cutting-edge manufacturing solutions, the sculpture combines the dynamic poses of Michelangelo’s work, the musculature craftmanship of Auguste Rodin, the somber tones of Käthe Kollwitz, Takamura Kotaro’s Japanese influence and Augusta Savage’s inspirational defiance to unite some of history’s most famous artists from a period spanning 500 years.
Weighing 500 kilograms and standing at 150 cm tall, the Impossible Statue was officially inaugurated at Tekniska Museet, Sweden’s National Museum of Science and Technology, in April 2023.
Producing a statue in such a way has never been done before. So how did Loikkanen, Pettersson and the team craft this blend of art and science, past and future?
When AI meets art
AI has been around for some time, with intelligent machines tasked with performing activities that normally require human intelligence, such as visual perception, speech recognition, decision-making, language translation and problem-solving. Its concept dates back decades with the earliest program written in 1951 by Christopher Strachey, later director of the Programming Research Group at the University of Oxford.
But the emergence of new, conversational AI programs such as Google’s Bard and ChatGPT open up even more applications of the technology.
Today, the achievements of AI seem almost limitless — it can even create art. After establishing a 2D design that brought together the styles of the five artists, Sandvik began translating the model into a complete 3D image.
Using depth estimators to build the 3D model, human pose-estimators to refine the body, videogame algorithms to generate realistic fabric and specialised AI to re-introduce fine details that were lost in previous steps, Sandvik was ready to turn the design into reality.
Using Mastercam software, a design for a statue with over six million surfaces and complex details was converted.
There were some unique challenges involved in the statue’s design. “We needed a phenomenally precise digital simulation to help us machine the statue,” revealed Henrik Loikkanen, Technology Area Manager at Sandvik Coromant. “Digital manufacturing means we can prove out that whole complex machining process beforehand. The only time we spent on machines, therefore, is actual production time. It also meant we didn’t produce a single scrap component during the entire project.”
Machining the impossible
After finalising the statue’s design with AI and virtually simulating optimal ways to manufacturing the statue using digital twinning, it was time to commence machining. “We treated producing the statue as we’d treat machining highly-specialised, complex parts like those found in the aerospace industry,” continued Loikkanen.
“An additional challenge came from the statue’s chosen material — stainless steel from Alleima — as ISO M materials are notoriously difficult to machine. The material group is characterised by its high work hardening rates and poor chip breaking properties during machining. Careful attention, therefore, must be paid to the tools selected for machining the material.”
Sandvik Coromant chose several tools to help sculpt each intricate part of the statue. The Coromant Capto® tooling interface was used to connect the arm and head to the torso of the statue, solid round tools from the CoroMill® Plura and Dura families were responsible for finishing all the statue’s surfaces and features and CoroMill MH20, a high feed milling cutter launched in 2021, to machine the bulk of the stainless steel removed from the workpiece.
The CoroMill MH20’s ability to machine at long overhangs typically makes it beneficial to the aerospace sector. It’s designed with difficult-to-machine components in mind and, in contrast to the conventional four-edge concept, the MH20 is designed with a two-edge insert. This is especially beneficial as it means the weakest section of the insert is far away from the main cutting zone, delivering greater reliability and protection against wear. It also means that machining against a corner or wall will not impact the next edge or leading corner, ensuring an equal performance per edge.
CoroMill Plura HD is Sandvik Coromant’s first choice solution for heavy-duty applications in steel and stainless steel, offering safe and efficient solid end milling. The tool now comes with an added benefit in the form of Zertivo 2.0 coating, developed by Sandvik Coromant’s R&D team, to improve process security and productivity even further. CoroMill® Dura are designed to be a one for all tooling solution.
The end mills can be used in all processes needed to produce a component ― including roughing, finishing, semi-finishing and ramping.
A manufacturing masterpiece
Naturally, machining the impossible isn’t without its challenges and the team made several adaptations to its machining strategy. “The initial 3D model was not the type of model CAD/ CAM systems can work with,” said Loikkanen. “It had to be converted from a 3D-mesh model, which is essentially a shell of surface polygons commonly used in 3D animation studios, to a solid 3D model with density and that was a pretty challenging process.
“Once completed, the model was sectioned in 17 pieces and all interfaces between them had to be modelled with a tight fit to make the intersections invisible when the parts were put together. It took time to get everything perfect.”
In the end, a statue that seamlessly blends the works of five artists was made possible. But developing a never-seen-before statue wasn’t the only accomplishment for Sandvik Coromant. “Along the way, we implemented several techniques that can be applied to future, digital machining projects,” said Loikkanen.
“Because of the pure amount of programming work, building the statue would have been impossible without a digital twin. All the testing was done digitally to save a huge amount of time spent that would otherwise be spend on trial-and-error testing. This is certainly something we can apply in future projects to save time and reduce the number of scrap components.”
“The programming and simulation process became faster and faster for each component,” added Pettersson. “Machining took about the same time throughout, mainly because the cutting data and tooling selection never changed that much. This point brought me to one conclusion: The extensive product data and recommendations we give is accurate and is easily attained either from our web, tool guide, Tool Library or even integrated directly in CAM systems themselves. Surely, this is of great value to our customers and closes one of the more work intensive tasks in closing the digital loop around CAM-programming.”