Artificial Intelligence

3D printing has established itself as a production process and has also found its way into the fashion industry. Individualised shoes can be 3D printed, but this poses significant challenges for automated quality control, as defects are rare. Autoencoders enable to train a system with defect-free data so that detected deviations from this state can be evaluated as defects. Our research shows a ROC AUC score of 0.87, proving that this method is suitable for anomaly detection in 3D-printed shoes.

An important element for companies to deal with the demands of the world of work is the continuous and needs-specific further training of employees. The possibility of learning close to the workplace has a major role to play here.

Assistance systems are being used increasingly in the context of digital transformation. They can support employees in industrial production processes both in the learning phase and in the active work phase. In this way, competencies can be built up in a way that is close to the workplace and the process as well as demand-oriented. This paper discusses the current state of research on the possible applications of these assistance systems and illustrates them with examples. Among other things, the current challenges are also highlighted. At the end of the paper, focal points for the future development of AI in industrial learning processes and research on this are identified.

Initiated by the digitalization, the organization of production and employees are subject to change. Innovative and digitized formats should be integrated into existing training programs. Therefore, suitable and target group-specific teaching/learning formats are needed that support participative methods and digital collaboration. For this purpose, a digital teaching and learning format for the application area of automotive engineering in SMEs was designed. This prototypical teaching/learning format was created and evaluated in an iterative process through the participation of the potential. The results show that the test subjects mainly refer to the usability criteria of DIN ISO 9241-110. Recommendations for the creation of future digital teaching and learning formats for SMEs are derived from these findings.

Artificial intelligence (AI) is the next stage in the digitalization of the economy. The technology also offers great potential for small and mediusized enterprises (SMEs). However, many SMEs are still reluctant to introduce AI and are only at the beginning of digitization: only around one fifth of all SMEs in Germany have thoroughly digitized their own processes and departments. What does this mean for the use of AI in companies? What steps should businesses take now to take advantage of the opportunities AI offers? And what stumbling blocks should be avoided? This article presents practical implementation concepts for companies with different levels of digital maturity and AI deployment capabilities and shows the range of potential benefits of AI applications in different industries and with different value creation architectures in medium-sized companies.

Deep learning-based error detection methods outperform traditional methods because of the continuously increasing complexity of technical systems and inherent flexibility and scalability of Deep Learning techniques. This article introduces the KrakenBox – an autonomous Deep Learning-based error detector for industrial Cyber-Physical Systems (CPS). It exploits a lightweight, Long Short-Term Memory (LSTM) network capable of online error detection that can be deployed on an embedded platform such as NVIDIA Jetson AGX Xavier or even Google Coral Edge TPU. This article describes the architecture of the KrakenBox and demonstrates its application with two case studies.

Einer der Erfolgsfaktoren für die Realisierung von Cloud-Manufacturing-Plattformen ist die Geschwindigkeit, mit der Angebote generiert werden können. Um aus dem Kundenauftrag möglichst automatisiert die Planungsprozesse der Plattform zu unterstützen extrahieren wir Bearbeitungsprozesse aus technischen Zeichnungen mittels eines künstlichen neuronalen Netzes. Dabei werden die Probleme der Multi-Label-Bildklassifikation und eines unsymmetrischen Datensatzes behandelt. Abschließend wird die Leistungsfähigkeit des Datensatzes an den Testdaten demonstriert und ein rekursives Verfahren zur Verbesserung des Systems im laufenden Betrieb beschrieben.

To be able to use artificial intelligence, a company does not necessarily need a qualified specialist. The Fraunhofer study “Cloud-based AI Platforms - Opportunities and Limits of Services for Machine Learning as a Service” shows how small and medium-sized companies can proceed instead. This article summarizes the arguments and results of the study, citations from it are therefore not marked accordingly.

Intense global competition, shorter product life cycles and an increasing number of variants require flexible and adaptable, but at the same time economical production and logistics systems. This requires constant replanning of factories and logistics systems. Value-adding processes are being outsourced to contract logistics providers. Contract logistics planners must respond to tenders as quickly as possible and develop a proposal with an initial planning concept and a cost estimation. Despite standardization efforts in planning, the knowledge is often only implicit at the planners. This article describes the need for support by an AI-based assistance system during the planning process and how a digital platform for such an assistance system should look like.