Artificial Intelligence (AI), Machine Learning (ML), Big Data - these terms have been accompanying the industry for quite some time now. Numerous companies would like to use these technologies. But there is a lack of use cases that represent a realistic use of artificial intelligence and describe a concrete benefit in production. This is because the field is wide in which AI can be used. plus10, a Fraunhofer spin-off, uses AI technologies in its software tools to operate complex production machines with maximum productivity. The optimization software was used in consumer goods manufacturing: at Freudenberg Home and Cleaning Solutions, high-frequency data acquisition and processing was implemented on several injection molding machines.
Software generates data-based optimization proposals
For operators of injection molding machines, it is essential to identify perfectly coordinated parameter settings as quickly as possible in order to achieve the maximum number of good parts in the best possible cycle time. However, fully automated production systems are usually very complex and difficult to operate with maximum productivity.
The AI start-up plus10 is developing self-learning software tools for data-based analysis and optimization precisely for such production plants. As a basis for the continuously operating system, thousands of machine parameters are recorded and processed every millisecond. Machine learning is used to learn in detail the machine behavior of many identical or similar machines. This is then used to automatically derive optimization proposals for each individual machine. In the background, this is based on a machine-learned purely virtual "ideal machine" that also improves itself over time. In this way, each machine develops evolutionarily in the direction of this idealized perfect machine. In this way, companies can reduce their cycle time per machine by 6 - 18 percent and thus significantly increase plant effectiveness.
AI in the production of buckets at Freudenberg
At plastics processor Freudenberg Home and Cleaning Solutions, the aim of the project was to design and implement a high-frequency Big Data infrastructure including data mapping for injection molding machines. To achieve this objective, a number of measures were implemented. As a first step, several data interfaces were implemented for each machine control. In this way, data from sensors for raw material characterization can be connected and mapped. In-mold sensors were integrated to also record the temperature and humidity of the material. Deviations in these parameter settings can cause quality losses in the end products. A quality station was set up to subject each part to individual testing. In addition, the optimization specialists developed a virtual buffer system so that each part produced could be tracked after injection molding: "The collaboration also included virtual part tracking for continuous and automated linking of all control and subsequent quality data," reports Uwe Dingert, Director Research & Development at Freudenberg Home and Cleaning Solutions.
Use of the optimization software a complete success
As a result, Freudenberg Home and Cleaning Solutions received an AI-enabled Big Data infrastructure which forms the basis for further optimization measures. A uniform data model was created that links all individual measurements to the corresponding part produced. In this way, measured quality characteristics can be assigned to successive process steps in the injection molding cycles. Uwe Dingert from Freudenberg is convinced by the joint AI project: "plus10 has built a high-frequency data infrastructure for our injection molding machines. I was very enthusiastic about the expertise and reliability."
Among other things to read at
(Image rights: Freudenberg Home and Cleaning Solutions)
.jpg)
