Geförderte OA-Publikationen

The spatial prediction of the turbulent flow of the unsteady von Kármán vortex street behind a cylinder at Re = 1000 is studied. For this, an echo state network (ESN) with 6000 neurons was trained on the raw, low-spatial resolution data from particle image velocimetry. During prediction, the ESN is provided one half of the spatial domain of the fluid flow. The task is to infer the missing other half. Four different decompositions termed forward, backward, forward-backward, and vertical were examined to show whether there exists a favorable region of the flow for which the ESN performs best. Also, it was checked whether the flow direction has an influence on the network's performance. In order to measure the quality of the predictions, we choose the vertical velocity prediction of direction (VVPD). Furthermore, the ESN's two main hyperparameters, leaking rate (LR) and spectral radius (SR), were optimized according to the VVPD values of the corresponding network output. Moreover, each hyperparameter combination was run for 24 random reservoir realizations. Our results show that VVPD values are highest for LR ≈ 0.6, and quite independent of SR values for all four prediction approaches. Furthermore, maximum VVPD values of ≈ 0.83 were achieved for backward, forward-backward, and vertical predictions while for the forward case VVPDmax = 0.74 was achieved. We found that the predicted vertical velocity fields predominantly align with their respective ground truth. The best overall accordance was found for backward and forward-backward scenarios. In summary, we conclude that the stable quality of the reconstructed fields over a long period of time, along with the simplicity of the machine learning algorithm (ESN), which relied on coarse experimental data only, demonstrates the viability of spatial prediction as a suitable method for machine learning application in turbulence.

Two hybrid quantum-classical reservoir computing models are presented to reproduce the low-order statistical properties of a two-dimensional turbulent Rayleigh-Bénard convection flow at a Rayleigh number Ra=105 and Prandtl number Pr=10. These properties comprise the mean vertical profiles of the root mean square velocity and temperature and the turbulent convective heat flux. The latter is composed of vertical velocity and temperature and measures the global turbulent heat transfer across the convection layer; it manifests locally in coherent hot and cold thermal plumes that rise from the bottom and fall from the top boundaries. Both quantum algorithms differ by the arrangement of the circuit layers of the quantum reservoir, in particular the entanglement layers. The second of the two quantum circuit architectures, denoted H2, enables a complete execution of the reservoir update inside the quantum circuit without the usage of external memory. Their performance is compared with that of a classical reservoir computing model. Therefore, all three models have to learn the nonlinear and chaotic dynamics of the turbulent flow at hand in a lower-dimensional latent data space which is spanned by the time-dependent expansion coefficients of the 16 most energetic proper orthogonal decomposition (POD) modes. These training data are generated by a POD snapshot analysis from direct numerical simulations of the original turbulent flow. All reservoir computing models are operated in the reconstruction or open-loop mode, i.e., they receive three POD modes as an input at each step and reconstruct the 13 missing modes. We analyze different measures of the reconstruction error in dependence on the hyperparameters which are specific for the quantum cases or shared with the classical counterpart, such as the reservoir size and the leaking rate. We show that both quantum algorithms are able to reconstruct the essential statistical properties of the turbulent convection flow successfully with similar performance compared with the classical reservoir network. Most importantly, the quantum reservoirs are by a factor of four to eight smaller in comparison with the classical case.

To date, innovation management research on idea evaluation has focused on human experts and crowd evaluators. With recent advances in artificial intelligence (AI), idea evaluation and selection processes need to keep up. As a result, the potential role of AI-enabled systems in idea evaluation has become an important topic in innovation management research and practice. While AI can help overcome human capacity constraints and biases, prior research has identified also aversive behaviors of humans toward AI. However, research has also shown lay people's appreciation of AI. This study focuses on human crowdvoters’ AI adoption behavior. More precisely, we focus on gig workers, who despite often lacking expert knowledge are frequently engaged in crowdvoting. To investigate crowdvoters' AI adoption behavior, we conducted a behavioral experimental study (n = 629) with incentive-compatible rewards in a human-AI augmentation scenario. The participants had to predict the success or failure of crowd-generated ideas. In multiple rounds, participants could opt to delegate their decisions to an AI-enabled system or to make their own evaluations. Our findings contribute to the innovation management literature on open innovation, more specifically crowdvoting, by observing how human crowdvoters engage with AI. In addition to showing that the lay status of gig workers does not lead to an appreciation of AI, we identify factors that foster AI adoption in this specific innovation context. We hereby find mixed support for influencing factors previously identified in other contexts, including financial incentives, social incentives, and the provision of information about AI-enabled system's functionality. A second novel contribution of our empirical study is, however, the fading of crowdvoters’ aversive behavior over time.

The measurement of the transport of scalar quantities within flows is oftentimes laborious, difficult or even unfeasible. On the other hand, velocity measurement techniques are very advanced and give high-resolution, high-fidelity experimental data. Hence, we explore the capabilities of a deep learning model to predict the scalar quantity, in our case temperature, from measured velocity data. Our method is purely data-driven and based on the u-net architecture and, therefore, well-suited for planar experimental data. We demonstrate the applicability of the u-net on experimental temperature and velocity data, measured in large aspect ratio Rayleigh-Bénard convection at Pr = 7.1 and Ra = 2 x 10^5, 4 x 10^5, 7 x 10^5. We conduct a hyper-parameter optimization and ablation study to ensure appropriate training convergence and test different architectural variations for the u-net. We test two application scenarios that are of interest to experimentalists. One, in which the u-net is trained with data of the same experimental run and one in which the u-net is trained on data of different Ra. Our analysis shows that the u-net can predict temperature fields similar to the measurement data and preserves typical spatial structure sizes. Moreover, the analysis of the heat transfer associated with the temperature showed good agreement when the u-net is trained with data of the same experimental run. The relative difference between measured and reconstructed local heat transfer of the system characterized by the Nusselt number Nu is between 0.3 and 14.1% depending on Ra. We conclude that deep learning has the potential to supplement measurements and can partially alleviate the expense of additional measurement of the scalar quantity.

This paper explores the mechanical recycling of continuous fiber-reinforced thermoplastics (CFRTPs) waste into injection molded products, focusing on the influence of recycling parameters on fiber length and mechanical properties. CFRTPs are gaining attention for their promising attributes, including weight-specific mechanical properties, short cycle times, storability, and recyclability, making them suitable for diverse applications. However, as CFRTP production rates rise, recycling strategies become crucial for sustainability. This study investigates the processability of CFRTP waste, defines size reduction conditions, and evaluates the impact of various compounding parameters such as temperature, screw speed, and fiber volume content during extrusion. The research findings indicate that higher screw speeds lead to fiber length reduction, whereas elevated temperatures result in longer fibers. Increased fiber volume intensifies interactions, resulting in shorter lengths. Additionally, the study examines the influence of injection molding parameters such as back pressure, screw speed, and initial fiber length on the resulting fiber length and mechanical properties of injection molded specimens, emphasizing the need for precise parameter control to optimize performance in recycled CFRTPs. Key findings are that increasing the initial fiber length from 260 μm to 455 μm results in an average fiber length after injection molding of 225 μm and 341 μm, respectively. This implies that longer initial fibers are more prone to breakage. Regarding the mechanical properties, increasing back pressure from 20 bar to 60 bar results in a reduction in Young’s modulus of approximately 40 MPa. Higher screw speed also reduces modulus by approximately 70 MPa due to intensified fiber-screw interactions. However, back pressure and screw speed have neutral effects on the tensile strength and the elongation at break.