The power of predictive algorithms
Deep learning methods have matured rapidly in recent years, particularly in industries with large volumes of real-world data such as insurance, retail and advanced manufacturing. Within supply chain management and manufacturing, some of the highest business impacts will be in predictive maintenance and yield optimization, followed by procurement analytics and inventory optimization.
“Predictive services are one of our initial objectives,” explains Mikael Wahlsten, Director and Product Area Manager for Photomask Generators at Mycronic. “Image classification, for example, has advanced significantly and has strong potential for improvement. In terms of image processing, we can definitely find novel ways to improve quality and enable the system to better adapt to its environment.”
Huge data gives us great potential
The quality of deep learning algorithms depends on huge data sets – in most cases requiring millions of labeled examples – in order to exceed human abilities and traditional analytical technologies. And real-world production data is something Mycronic systems have in abundance, thanks in part to the ongoing development of the Mycronic 4.0 intelligent factory.
“All of our systems today are essentially softwaredriven and increasingly integrated with other factory systems,” says Wahlsten, “The process data they create is hugely valuable as training data – which is used to train the algorithms. This holds a lot of potential when it comes to generating accurate simulations through deep learning.”
“Simulated environments,” he continues, “are particularly useful for SMT customers who need to find new adaptive methods for automated production.
The entire Mycronic 4.0 intelligent factory concept relies on factory-wide information flows – horizontal, vertical and into the cloud. This level of total automation involves so many systems, with so much complexity, which is exactly where these types of adaptive deep learning algorithms can add massive value, both within the production line and in other systems throughout the factory.”