Using Machine Learning Techniques in general and Deep Learning techniques in specific needs a certain amount of data often not available in large quantities in some technical domains. The manual inspection of Machine Tool Components, as well as the manual end of line check of products, are labour intensive tasks in industrial applications that often want to be automated by companies. To automate the classification processes and to develop reliable and robust Machine Learning based classification and wear prognostics models there is a need for real-world datasets to train and test models on.

An important aspect of the overall quality of machined parts is surface roughness, which depends on cutting parameters, tool condition, and machine vibrations. Online surface roughness prediction in milling operations can reduce set up time and assist in determining economic cutting parameters. However, the adoption of existing solutions in industrial production is inhibited by lacking integration in an open and retrofittable architecture. In this contribution, a solution for surface roughness estimation by vibration monitoring is developed as part of a retrofitting kit. Wavelet packet transform is used to filter the vibration signal, then the roughness of the generated surface is estimated. The approach is tested in milling experiments.

Production systems must be able to quickly adapt to changing requirements. Especially in the field of remanufacturing, the uncertainty in the state of the incoming products is very high. Several adaptation mechanisms can be applied leading to agile and changeable production systems. Among these, adapting the degree of automation with respect to changeover times and high investment costs is one of the most challenging mechanisms. However, not only long-term changes, but also short-term adaptations can lead to enormous potentials, e.g. when night shifts can be supported by robots and thus higher labor costs and unfavorable working conditions at night can be avoided. These changes in the degree of automation on an operational level are referred to as fluid automation, which will be defined in this paper. The mechanisms of fluid automation are presented together with a case study showing its application on a disassembly station for electrical drives.

Multinational companies deal with production processes in various countries by operating global production networks. These production processes are allocated to production plants with different levels of autonomy regarding strategic and operative decisions. Typically, each plant and the whole network are managed by one or more network managers who have to deal with a decision overload in their daily business. 50% of their decisions are made in less than 9 minutes and only a small amount of decision tasks are dealt with for more than one hour. To reduce this dilemma, it was found that the distribution of decision autonomy can be enhanced. It depends on the company's strategy and complexity dimensions in global production networks. However, so far there is little evidence on how to better distribute decision autonomy in global production networks in detail. Furthermore, it is not transparent at what level of centralism a global production network should be managed without cutting the capabilities of production plants. This paper presents a methodology, which examines relevant strategy dimensions and derives guidance on how to distribute decisions in global production networks. First, the network and production strategies of global production networks are classified. Second, relevant complexity dimensions and decisions are introduced. Third, the influence of the distribution of decision autonomy on strategy dimensions is quantified by an impact model. Furthermore, the effect of complexity on the distribution of decision autonomy is quantified by an impact model. Here, the integration of empirical data was used to validate the different influences. Finally, the ideal distribution of decision autonomy for specific production plants in the global production network is derived. The methodology is applied in an industrial use case to prove its practical impact.

Autonomous systems, as found in autonomous driving and highly automated production systems, require an increased reliability in order to achieve their high economic potential. Self-aware sensors are a key component in highly reliable autonomous systems. In this paper we highlight a proof of concept (PoC) of a deep learning method that enables a LiDAR (Light detection and ranging) sensor to detect functional impairment. More specifically, a deep convolutional neural network (CNN) is developed and trained with labelled LiDAR data in the form of point clouds to classify the degree of impairment of its functionality. The results are statistically significant and can be regarded as a general classifier for objects within LiDAR data, applied to selected cases of sensor impairment. In detecting impairment and evaluating the correctness of the captured data, the sensor gains a basic form of self-awareness. The presented methods and insights pave the way for improved safety of autonomous systems by the means of more sophisticated ?self-aware? neural networks.