Feature
Artificial intelligence (AI) is shaping the future of manufacturing. In fact, 93% of companies said they believe AI will be a pivotal technology to drive growth and innovation in the manufacturing sector, according to a recent study by Deloitte. So what are manufacturing companies doing with AI now? Here, we look at some examples of how manufacturers are using AI and machine learning (ML) to improve their production processes.
1. ASML
Chip manufacturers have been trying to put more processors in a smaller space since the first transistor, and Dutch manufacturer ASML helps them do it.
ASML worked with Google Cloud and machine learning company ML6 to store and analyze calibration data from the photolithography machines it sells to chip manufacturers, according to a case study. Between storing the data in the cloud and having AI tools to help them, the engineers could run more tests and get the results more quickly, which sped up research and development as well as the production.
However, chip manufacturers themselves have been reluctant to adopt AI in their manufacturing processes because of AI’s black-box nature coupled with intellectual property concerns, says Arnaud Hubaux, product cluster manager, ASML.
“When you are talking about high-volume manufacturing, those machines have been tuned so much that any change to those machines could lead to a potential half-life down the line,” Hubaux says.
“That could really stop production and cost customers millions. For that kind of scenario, machine learning is a thing not very widely adopted yet because of this black-box effect.”
Results
- 10x increase in performance of the application development team within six months
- 40% improvement in time to market
- Used AI for more granular chip inspections
2. Epiroc
We say “steel” like it’s a single material, but in fact there are more than 3,500 grades of the metal, each with different properties, because they’re made up of slightly different combinations of ingredients.
It’s important for Swedish construction tool manufacturer Epiroc to reproduce these metal grades accurately to make sure their customers get consistent products. To do that, it needed to be able to test and model production and then share the data with all its factories.
Epiroc used Microsoft Azure, Azure Machine Learning, Azure Data Factory and Azure Databricks to create machine learning models and predict steel density, hardness and flexibility for its drilling products, according to a case study.
“We can now more accurately assess optimal tolerance levels to prevent against structure fatigue or failure,” says Erik Nyten, enterprise architect, Epiroc.
In addition, Epiroc used automation in AI governance software from Sogeti to confirm that it meets the regulations and requirements of the countries in which it operates.
The result is that Epiroc, which had 11 analytical teams using the machine learning models, was planning to expand the project.
Results
- 30% reduction in customer rejections and product returns
- Created an “AI Factory” to share data and best practices in 60 hours
- Created ML models for a heat treatment process in six weeks
3. Georgia-Pacific
While it’s known as a paper products company, Atlanta-based Georgia-Pacific creates something else: data.“Every day, we generate about a terabyte of data,” says Steven Bakalar, VP of IT digital transformation at Georgia-Pacific, in a case study. “All of it goes through our machine learning models to tell us how we can operate better.”
Georgia-Pacific used SAS Viya running on AWS to store and analyze the data. For example, Georgia-Pacific monitored its equipment and used AI on the data to detect upcoming part failures and electrical problems before they affect the production line. As a result, the company’s facilities that used these tools experienced a 30% reduction in unplanned downtime, Bakalar says.
Georgia-Pacific also used computer vision to look for problems on the production line before workers spot them.
Results
- 5x growth in data volumes over the past year
- Monitored over 85,000 vibration sensors to predict part failures
- Ran over 15,000 ML models
See more: 10 Top AI Certifications for Manufacturing Pros
4. Siemens Gamesa Renewable Energy
When you see a wind turbine majestically turning in the air, converting wind to renewable energy, you may be surprised to learn that each giant blade is made by hand.
“Because each blade is made to order, our teams are more like artisan craftspeople building furniture than workers on an assembly line,” says Finn Mainstone, senior product manager at Siemens Gamesa Renewable Energy (since acquired by Siemens Energy).
“But as with any manual process, there is an ever-present risk of human error.”
That’s why the company turned to IBM Consulting to create a machine learning system, running on Microsoft Azure, to display a laser grid to show workers where to place each fiberglass layer, according to a case study.
Siemens Gamesa then went one step further: using an array of cameras and computer vision to detect defects. The next step is to use the AI manufacturing solution in several other Siemens Gamesa factories and potentially in all its factories worldwide.
Results
- 25% reduction in defects
- Manufacturing system expects ROI within 2.5 years
5. Toyota
Anyone who’s read “Unsafe At Any Speed” knows that just because a car is pretty doesn’t mean that it’s safe or reliable. That’s why the Toyota Research Institute (TRI) developed AI tools that help make sure car designs function in the real world.
“Generative AI tools are often used as inspiration for designers but cannot handle the complex engineering and safety considerations that go into actual car design,” said Avinash Balachandran, director of TRI’s Human Interactive Driving (HID) division.
“This technique combines Toyota’s traditional engineering strengths with the state-of-the-art capabilities of modern generative AI.”
While designers use publicly available text-to-image generative AI tools in their early designs, they typically weren’t able to include aspects such as engineering constraints. For example, designers can use the Toyota AI tools to optimize their designs based on constraints, such as drag, which affects fuel efficiency, and variables, such as ride height and cabin dimensions, which affect handling, ergonomics and safety.
Results
- Design requests to the AI system based on sketches, including properties, such as sleek, SUV-like and modern
- Reducing the number of iterations needed to reconcile design and engineering considerations
- Incorporating aerodynamic principles in the process to help design electric cars with longer ranges
See more: Forging the Future: The Fusion of Industry and Innovation
