Data is the gold of Industry 4.0: it is needed to control machines and systems in highly specialized processes. Sensors monitor these processes and generate large amounts of new data. As a basis for AI and machine learning applications, data collected at the points of action is particularly interesting if it can be merged, transferred to a standardized data model and thus evaluated in its entirety. "EmulDan" starts with the architecture of data collection and provides insights for both AI-based models and improved manual control options.

"Linked Factory": networked machines, customized information

The Linked Factory data architecture is a vision developed at Fraunhofer IWU of a consistent digital representation of all products, processes and machines in companies. In addition to individual machines, logistics processes and building management systems can also be integrated. As a central element, a knowledge graph allows a wide variety of data to be linked in order to generate up-to-date information. Particularly when large amounts of data are processed and provided as information, individual data packages should support employees in their tasks and not place an additional burden on them. "Context-based provision" is the key word here.

EmulDan: Basis for machine learning

In addition to the data architecture, the AI models developed in EmulDan are an important step towards end-to-end digital twins. In production areas with a low level of automation, such machine learning applications can help planners, decision-makers and production employees to make the best possible decisions. So-called drifts, for example, where the production of a component is approaching the tolerance limit, can be detected earlier.

In EmulDan, the project partners focused on energy efficiency. In doing so, no deterioration in the classic control and measurement parameters such as production time, costs and product quality could be accepted. The partners are now presenting their results using demonstrators that clearly show the considerable energy-saving potential in hot forming, cold forming and machining production processes.

Press hardening, hot forming, cold forming

Press hardening combines the advantages of forming and heat treatment in a single step. This manufacturing process requires a lot of energy, so the question of potential savings is obvious. For data-based process control, the project partners collected all relevant production data from individual processes and process chains and created process models to predict energy requirements and component quality in various optimization scenarios. Hybrid process models based on the so-called digital twin proved to be particularly helpful. The result: potential energy savings of up to 20 percent can be achieved by combining several adapted process parameters.

EmulDan was also able to demonstrate considerable energy-saving potential for cold forming with a rotary swaging machine. The central starting point was to supplement the previously uncontrolled process parameters, which were almost exclusively based on component quality and empirical knowledge, with energy-relevant aspects. The result was a software tool for data processing that enables self-learning process correction and creates suitable models for forecasts. In the process, correlations between machine parameters, component quality and energy-related process variables are identified and transferred to process models that are based on the overall optimum. In practice, this means a better understanding of the process and the opportunity to reduce energy consumption by up to 70 percent.