A biscuit factory is using AI to ensure that individual items are perfectly baked before packing and shipping, so that customers receive a consistent high-quality product. Given the ever-increasing scale of production, the factory could no longer rely on human intelligence to keep up. But with AI, production is optimised.
The project background
Part of maintaining a perfect bake means keeping a constant temperature in the ovens 24/7, especially when it comes to batch-baking different types of biscuits. The factory had been using visual inspection, with workers checking the bake levels and adjusting oven temperatures based on their skill and experience. Staff experience is not always 100% consistent though, and each person has their own subjective judgement. In contrast, AI behaves in a non-emotional way, inspecting each biscuit objectively and controlling the oven temperatures automatically.
System requirements
In order to build an AI-driven quality control system for the production line, AI models needed to be prepared. An AI model training server was set up to capture video and collect image datasets from the production lines. Meanwhile, AI models were generated based on carefully designed neural networks, and AI training servers leveraged GPU and CPU processing power to manage all the edge AI systems. These AI models were deployed to inference systems via model management software on the training server. On-site training was necessary to improve accuracy on the actual production line.
The AI inference systems on the production line needed to be fanless to resist flour and dust damage from the environment. A small form factor was required to integrate the new edge AI system with existing legacy equipment. The edge AI system required multiple I/O interfaces to grab video/images from cameras and transmit metadata to central management.
System implementation
To collect a large dataset and train various AI models, the factory integrated Advantech’s HPC-7000 server with ASMB-815I server board and installed an RTX graphic card to make a powerful training server. The server has the capability to collect data from edge systems, perform image pre-processing before dataset labelling, and generate AI models for the production line. Once new data arrives, the process can quickly re-train models to increase accuracy.
To maintain consistency, Advantech’s MIC-720AI AI inference system is embedded with a variety of AI models that perform real-time inspection to ensure all biscuits are perfectly baked. This edge AI system is also capable of capturing video/image data from area-wide cameras and automatically inspecting bake quality and sorting biscuits into OK and NG categories.
NG category biscuits are flagged and show up on the display monitor for the attention of inspectors. Oven temperatures are adjusted as necessary. NG product images and log data are also recorded and sent to the management system for assessment.
The factory now enjoys much better efficiency and more consistent high-quality biscuit production thanks to the state-of-the-art edge AI system.
Advantech provides end-to-end AI solutions, offering a full range of edge servers and AI inference systems. Traditional industries typically reach a tipping point before they move into AI. Advantech’s product offering leverages powerful AI computing to support customer business objectives. This AI solution for the biscuit factory is a smart example of Advantech helping food manufacturers take the AI leap.
ESIS is the leading supplier of outstanding industrial electronic equipment in the Australian market since 1971. Check out our exclusive range of Fanless PCs, Industrial PCs (IPC solutions), Rugged Tablets, Nautiz Rugged Smartphones and Scanners, HMIs, Panel PCs and other electronic components. Our products are certified and made with high-quality custom materials for accurate results. Contact us for a detailed product catalogue or to share your queries.
Article courtesy of Advantech.
