The creators of the PreciBake artificial intelligence baking system said it worked by ‘getting to know’ products as it baked, which allowed bakers to see the oven’s content from wherever they were in the world.
The algorithm system learned what product it was baking by collecting data and adjusted to account for any discrepancies from one oven to another without requiring human intervention, Thomas Bone, managing director for PreciBake, told BakeryandSnacks.com.
The firm, which launched its systems alongside WP Bakery machines, has also developed a way in which bakery managers can track oven production through data and images of inside the machine via their smart phone or tablet.
Bone said that modern bakers were under increasing pressure to supply more varied product ranges, turning around more products in smaller batches. “They need to have solutions that provide them with quality assurance for the flexibility they need,” he said.
One size doesn’t fit all
This system could technically produce “infinite amounts” of products in different ovens, he said. Although he added that the firm tended to advise bakers to start with their top selling products and add new lines when and if needed.
He said this system was more flexible than existing solutions in its ability to adapt to heat and water inconsistencies across ovens that could emerge as machines got older. “Other systems don’t play with changes,” he said. This helped manufacturers ensure consistent, reproducible quality as they moved to different product offerings, he explained.
Bone said the self-learning software was installed overnight by the PreciBake team and began collecting and generating patterns in data gathered as the oven ran as usual. After this period, the machine could be activated by one button.
Last year, scientists looked at how bird and ant-inspired algorithms might help improve production efficiency and cut costs in bakeries. The researchers investigated two common evolutionary algorithms – Particle Swarm Optimization (PSO) which built on the movement of bird flocks and fish schools as they searched for food and Ant Colony Optimization (ACO) inspired by the way ants found the shortest routes between food source and nest.
At the time the researchers said these systems would be easy to implement and modify, allowing bakers to solve complex scheduling optimization tasks.
PreciBake's algorithms were inspired by technology already present within the car industry that enabled cars to be driven without a driver.
Bone said an app allowed bakers to see inside the ovens, which he said would be of “unbelievable importance” to manufacturers. “A good baker loves his products,” he said.
He said this and the data uploaded to a Bake IT Cloud service enabled managers to track the progress of each site from wherever they were in the world. “The more products being made, the more mistakes are made. If they can see inside, that helps them when traveling to ensure the products are being baked in the right way.”
He said the cost of such systems had come down in recent years, with the PreciBake operator ranging from approximately €2,500/€3,000 ($3,450/$4,140) for smaller in-store ovens and €5,000 ($6,900) for bigger production ovens. He said manufacturers would see this returned in turn-over as they ensured consistent quality.