Bringing artificial intelligence directly to the machine
Up to now, industrial applications of artificial intelligence (AI) have been less than impressive due to latency periods and the high quantities of data involved in connecting to the cloud. The crucial factor is therefore bringing AI directly to the machine and interpreting data there – directly at the source – in real time. AI specialist Resolto is doing exactly this with its Scraitec platform, for example at a household appliance producer and a car manufacturer.
Resolto Informatik GmbH has been part of the Festo Group since 2018, and supports the leading manufacturer of automation solutions in getting its pneumatic and electric automation technology ready for Industry 4.0. Data is interpreted already in the field in close proximity to the machines. This makes it possible to save energy, shorten cycle times, and it reduces machine failures and production errors.
Festo products with AI
The Scraitec software solution from Resolto knows the healthy condition of a plant and detects every anomaly with the real-time analysis of the plant’s sensor data. Scraitec supplies early and precise prognoses, makes diagnoses and supplies recommendations for action. “The topic of analytics and artificial intelligence will have an effect on Festo’s product portfolio, since AI algorithms can be integrated both in the cloud and directly in the components from Festo,” says Tanja Maaß, Managing Director of Resolto, in describing the benefits of the collaboration.
With the Festo IoT Gateway CPX-IOT as hardware, customers can have their machine and plants monitored at field level. Field level support is provided with the ScraiField software component. This component runs constantly in close proximity to the machines in a smaller controller. A pretrained model with minimal hardware requirements that reliably interprets data streams without the need for a data connection to the central component located in the cloud (ScraiBrain) is used. The IoT gateway connects to the cloud, the Festo dashboards, as necessary. The ScraiBrain is embedded there with access to a host of preconfigured application models.
Human-in-the-loop principle
“The platform continuously learns from actual operation, integrating the knowledge of the engineers and the customer’s technical experts – we call it the ‘human-in-the-loop’ principle,” explains Maaß. The machine learning and AI product interprets information predictively in order to actively optimise parameters in plants or to send concrete instructions for action to “its” people (e.g. via smartphone).
New business models
Combining Scraitec with plants and machines to transform them into digital tools is opening up new business models for machine and plant manufacturers. New service concepts offer substantial added value through the automated, early coordination of the in-house maintenance teams.
Scraitec helps end customers to optimise the utilisation of their plants using automation. The costs for maintenance are reduced, since maintenance schedules can be adapted by predicting events and recommendations for action for known fault patterns. The platform improves all plant parameters for defined target criteria and increases the plant’s productivity.
Fault avoidance in complex production lines
As a case in point, household appliance manufacturer Miele noticed fluctuating product quality in its production process over a given period, but was unable to account for the causes. It operates complex production lines, where products are manufactured in sequential order. It is not enough here to look at the individual stations separately.
The production managers at Miele therefore wanted a system for automatically detecting anomalies in complex manufacturing flows. “Deep Learning seemed to be the right approach for this,” explains Maaß. What was needed was to develop an integrated data base that brought together different measurement systems. Additional measuring points also had to be configured for this. The Scraitec platform modelled the production lines as an integrated system, and by so doing increased throughput by 1.5%.
An automobile manufacturer is another case in point, with a pneumatic clamping system costing the car manufacturer just € 100. However, an unforeseen production stoppage can cost several hundred thousand euros. An early warning system for wear and slowing down of cycle times was therefore ideal – more precisely a learning system for the predictive maintenance of all types of clamping systems. The solution with Scraitec for real-time data analytics incorporates the Festo controller CPX-E-CEC directly. A connection to the cloud is not necessary.
