artificial neural networks

The production of cooked ham involves a number of challenges. In production, cuts of meat are put through in a multi-stage curing and cooking process involving brine. This can lead to fluctuations in quality due to structural defects in the meat. The result: the brine is not optimally absorbed. An AI model trained on historical data intends to solve the problem.

The design of gear pumps causes gap flows, which counteract efforts to achieve exact dosing. However, due to the complex relationships between pressure, temperature, manufacturing tolerances and the material properties of the pumped medium, these gap flows cannot be reliably described in real time using physical equation systems. By using new algorithms and artificial neural networks, it was possible to solve this problem and replace a cost-intensive flow measurement method at the same time. This new approach has been tested on a pilot plant scale, and has already demonstrated that it can achieve the accuracy of a classic flow meter, while at the same time significantly decreasing the response time and increasing the application range of the dosing system.

Additive manufacturing is an increasingly important manufacturing technology with huge economical potential. However, its popularity is accompanied by high material and time losses, as defects are often detected at a very late stage. One solution for a more sustainable production is the automated detection of manufacturing defects using artificial intelligence. This article describes the digitization of the defect detection process in additive manufacturing using a system based on a neural network. In addition to the steps for automated defect detection, system performance is also discussed.

With increasing availability of data, the use of machine learning to optimize supply chains becomes attractive, as the accuracy of data analysis can be increased and simultaneously the effort can be reduced. Based on the SCOR model, exemplary approaches are described as a guidance and suitable machine learning methods are presented.

Everyone is talking about artificial intelligence and machine learning. However, knowledge about what the terms actually mean is often not yet extensively available. The article presents some basic knowledge and shows which application possibilities and added values machine learning can offer for production. Robotics, for example a bin-picking system, benefits in particular from the technologies described. Finally, the article deals with the topic of explainability of machine learning processes. For technical, legal and social reasons, decoding the “black box” is an essential task.

Dynamics and complexity of today`s production systems bring established approaches for production planning and control to their limits. Accordingly, developing new concepts and methods is a key point for research in this area. The combination of a decentralized control structure and innovative methods from the field of artificial intelligence seems promising here. Open source tools have proven their applicability to implement those methods. They are disposable and can be flexibly adapted to many problems. This contribution introduces an approach for the decentralized control of a shop floor. Here, artificial neuronal networks are used as adaptive control instruments. The simulation of these networks is performed with the open source tool Stuttgart Neural Network Simulator (SNNS) and its successor Java Neural Network Simulator (JNNS).

In an increasingly turbulent environment, convincing methods of production planning and control are needed. Many of the necessary decisions are made at shop-floor-level. They depend on the knowledge and the abilities of the workers to react on unpredictable impact and hence are not explicitly described. For a realistic, concomitant plant simulation, however, it is important, to model the control strategies as exactly as possible. This paper presents a method to identify applied control strategies by adopting artificial neural networks to data from the operating and machine data logging.