To improve cycle time planning in the automotive Body-In-White manufacturing, a better understanding of the adaptive controller's decision as well as the causes for a prolongation of resistance spot welding times is necessary. Thus, we propose an analysis using Random Forest to create a design recovery to increase the understanding of proprietary systems. The analysis suggests that a first decision for a prolongation happens during the transition of phase III to phase IV of the welding process. The first prolongation results in a moderate prolongation and a first mode, besides the reference time, in the distribution of welding times. This is consistent with the frequently proposed controllers in the literature. In later stages of the welding process, a second prolongation decision is sometimes made. If this second decision window is present at welding spots, the distribution of welding times often yields a second mode of highly prolonged times. Furthermore, the analysis suggests that the decision is based on the raw resistance curve values and not a first or second derivative of the curve. Future work is necessary to add further domain-knowledge to the analysis and construct useful decision functions based on the observed resistance curves.

Data streams represent real-world concepts that typically cannot be assumed to be static. Instead, sensors monitoring industrial resources deteriorate, network intrusion attacks exhibit new patterns, consumer behavior or public interest in news topics change. These are examples of a phenomenon commonly known as concept drift. It threatens the performance of estimation models conducting inference on associated data. Explicitly employed concept drift detection is among the most promising approaches to maintain a stable and robust performance of productively used estimators, as it identifies the times when adaptation to new concepts becomes actually necessary. Combining concept drift detectors to form ensembles can additionally increase detection performance as it aggregates their individual specializations and allows advanced processing of concept drift evidence. This survey is the first to systematically retrieve and overview research relevant in the field of concept drift detection ensembles. It identifies the research as being in its initial stages and reveals numerous gaps, which should be addressed to further exploit the substantial performance potential ensembles have over individually applied detectors. Furthermore, it contributes to the literature by discussing and describing key characteristics and principles of their design in a structured manner.

Although anomaly detection is an important task in engineering, in the practical application it is often unclear what an anomaly is and how it affects the planning and oper-ation of complex manufacturing systems. The authors propose a definition of “scientific” anomalies by an approach based on philosophy of science. The idea is to separate measurement noise and prediction errors. Based on this definition, a framework is established and applied to dynamical systems, resulting in two distinct deviation terms separating “scientific” anomalies from measurement noise. These distinct deviation terms can be used in future research to set up a metric and evaluation of the prediction model quality. The framework is demonstrated through an example by conducting an analysis to describe anomalous behavior of cold welding in holding pins.

We propose an unsupervised-learning-based method for anomaly detection and root cause analysis for an industrial press machine. A skip-connected autoencoder with 55% performance improvement measured by reconstruction root mean square error to vanilla variant in average is used to train the collected multivariant time series data in different schemes. We then conduct a stacked evaluation method for both machine- level anomalies with the root cause localization and anomaly on specific cylinder tracks. Both real-world and synthetic anomalies embedded in real data are used for evaluation. The result shows that the multi-models training scheme and the relatively short window length can gain better performance, i.e., fewer anomaly false alarms and misses.

Many studies have explored how individual differences can affect users' susceptibility to cybersickness in a VR application. However, the lack of strategy to integrate the influence of each factor on cybersickness makes it difficult to utilize the results of existing research. Based on the fuzzy logic theory that can represent the effect of different factors as a single value containing integrated information, we developed two approaches including the knowledge-based Mamdani-type fuzzy inference system and the data-driven Adaptive neuro-fuzzy inference system (ANFIS) to involve three individual differences (Age, Gaming experience and Ethnicity). We correlated the corresponding outputs with the simulator sickness questionnaire (SSQ) scores in a simple navigation scenario. The correlation coefficients obtained through a 4- fold cross validation were found statistically significant with both fuzzy logic approaches, indicating their effectiveness to influence the occurrence and the level of cybersickness. Our work provides insights to establish customized experiences for VR navigation by involving individual differences.

TechViz has been a supporter of Formula Student at KIT for several years reflecting the companys long-term commitment to enhance engineering and education by providing students with powerful VR system software to connect curriculum to real-world applications. Incorporating immersive visualisation and interaction environment into Formula Student vehicle design is proven to deliver race day success, by helping to detect faults and optimise product life cycle. The TechViz LESC system helps to improve the car design despites the short limit of time, thanks to the direct visualisation in the VR system of the CAD mockup and the ease of usage for non-VR experts.