Optimized state estimation by application of machine learning

The requirements concerning the technical availability as part of the overall equipment effectiveness increase constantly in production nowadays. Unplanned downtimes have to be prevented via efficient methods. Predictive, condition-based maintenance represents a valuable approach for fulfilling these demands, but precise models for state estimation are missing. From the manufacturers’ point of view the challenge consists in wear models with the capability of specifying the correct component’s state as well as providing reliable failure forecasts. Unfortunately, nowadays creation of wear models is based on specific stress tests or design of experiments from the manufacturer. The integration of the production phase or even data feedback and user knowledge does not take place. New potential is promised by cross-cutting technologies from ICT like cloud technologies—in general virtual platform concepts—or approaches of machine learning as enabling technologies. The objective of this paper is to adopt existing algorithms to the new application of condition monitoring in order to evaluate the applicability for automated training of robust wear models. In that context the most commonly used algorithms are described and the reader gets an impression what challenges have to be met when dealing with machine learning. A selection of about ten algorithms with 45 variants is evaluated for four different features within a packaging machine. In the outlook the embedding of the trained model in a cloud architecture is presented.     

Source identification of plant-wide faults based on k nearest neighbor time delay estimation

The identification of plant-wide faults is a very important topic as it enables plant operators to decrease rejections or to increase the product quality. This paper shows a fault propagation approach for this field of interest based on time delay estimation. Due to the fact that the estimation of time delays in multiple-input single-output systems or nonlinear systems is either impossible or very difficult with known methods, a new method based on k nearest neighbor imputation was developed and is validated in this paper theoretically. The effectiveness of the identification algorithm is demonstrated on several simulations and on an industrial hydrocracker plant.     

Digital image correlation analysis and modelling of the strain rate in metal cutting

In the last eight decades, considerable modelling and computational efforts have been made to predict the strain rate during cutting with the aim of optimizing machining processes. However, the validation of these modelling approaches on a local scale remains excessively limited due to the lack of in-situ measurements and the faulty existing quick-stop tests. This work presents the in-process analysis of the strain rate and strain in the primary shear zone using high speed Digital Image Correlation (DIC) techniques. The comparison of measured and computed results shows the suitability of the DIC techniques and the robustness of the modelling approaches.     

The implementation deficits of adaptation and mitigation: green buildings and water security in Amsterdam and Boston

Frameworks of environmental regulations are fundamental yet problematic factors in achieving climate mitigation and adaptation policy goals. Recent theoretical arguments claim the value of general legal frameworks to enable experimentation and contextual adaptation of policies. However, empirical research regarding the effects of both general and specific norms in the practice of urban intervention remains limited. In this article we empirically discern how city governments deal with the tension between control and flexibility in the implementation of urban climate change goals. We argue that policies of adaptation/mitigation face two types of implementation problems: non-adaptive implementation and non-implementation. The first stems from an excessively constraining use of rules, while the second derives from a too general and undefined regulatory framework. Analysing two empirical cases in Amsterdam, Netherlands and Boston, MA, USA, we conclude that there are three elements that affect the way actors deal with these deficits: the level of scale at which regulations are established, the degree of land ownership which provides margin of manoeuvre to public authorities, and the sense of political urgency behind mitigation and adaptation policies.     

Chemical warfare and survival strategies in bacterial range expansions

Dispersal of species is a fundamental ecological process in the evolution and maintenance of biodiversity. Limited control over ecological parameters has hindered progress in understanding of what enables species to colonize new areas, as well as the importance of interspecies interactions. Such control is necessary to construct reliable mathematical models of ecosystems. In our work, we studied dispersal in the context of bacterial range expansions and identified the major determinants of species coexistence for a bacterial model system of three Escherichia coli strains (toxin-producing, sensitive and resistant). Genetic engineering allowed us to tune strain growth rates and to design different ecological scenarios (cyclic and hierarchical). We found that coexistence of all strains depended on three strongly interdependent factors: composition of inoculum, relative strain growth rates and effective toxin range. Robust agreement between our experiments and a thoroughly calibrated computational model enabled us to extrapolate these intricate interdependencies in terms of phenomenological biodiversity laws. Our mathematical analysis also suggested that cyclic dominance between strains is not a prerequisite for coexistence in competitive range expansions. Instead, robust three-strain coexistence required a balance between growth rates and either a reduced initial ratio of the toxin-producing strain, or a sufficiently short toxin range.     

Enhancing the efficiency of alternating current driven organic light-emitting devices by optimizing the operation frequency

We report efficient and bright organic light-emitting devices operated by capacitive energy coupling. In this approach, the organic layers are enclosed between sputter-deposited hafnium dioxide layers to prevent charge carrier injection. When a sinusoidal voltage signal is applied to the electrodes, the devices emit bright green light whereas no detectable emission is generated upon application of a constant voltage. The efficiency of the process depends heavily on the frequency of the applied voltage signal. By optimizing the driving scheme, a record luminous efficacy for AC driven OLEDs of 2.7 lm/W at 500 cd/m² is achieved.     

Quantification of Entrainment Using an Optical Inline Probe

In distillation and absorption processes, entrainment above trays or packings reduces stage and column efficiency. A modified version of the optical multimode online probe enables a seminal insight and, thus, quantification of entrained droplets on various axial and radial positions. A purging mechanism inhibits precipitation of droplets on the front lenses, while telecentric optics and illumination provide distance‐independent and precise imaging of entrained droplets between two valve trays. Experiments were performed in a DN 450 cold flow (air‐water) test rig. The results were compared with noninvasive phase Doppler anemometry data.