Search

Traditional anomaly detection methods in time series data often struggle with inherent uncertainties like noise and missing values. Indeed, current approaches mostly focus on quantifying epistemic uncertainty and ignore data-dependent uncertainty. However, consideration of noise in data is important as it may have the potential to lead to more robust detection of anomalies and a better capability of distinguishing between real anomalies and anomalous patterns provoked by noise. In this paper, we propose LSTMAE-UQ (Long Short-Term Memory Autoencoder with Aleatoric and Epistemic Uncertainty Quantification), a novel approach that incorporates both aleatoric (data noise) and epistemic (model uncertainty) uncertainties for more robust anomaly detection. The model combines the strengths of LSTM networks for capturing complex time series relationships and autoencoders for unsupervised anomaly detection and quantifies uncertainties based on the Bayesian posterior approximation method Monte Carlo (MC) Dropout, enabling a deeper understanding of noise recognition. Our experimental results across different real-world datasets show that consideration of uncertainty effectively increases the robustness to noise and point outliers, making predictions more reliable for longer periodic sequential data.

We introduce a novel approach for computing gate-to-gate time automatically from audio recordings. In slalom skiing, gate-to-gate timing is a valuable metric for athletes and trainers, capturing the time elapsed between slalom gates. The availability of these measurements immediately after each run allows for prompt feedback. This study specifically concentrates on gate-to-gate timing in alpine slalom skating, serving as a foundational step towards its future application in slalom skiing. While existing methods for measuring gate-to-gate time vary in their feasibility, accuracy, and compliance with regulations, we propose a solution utilizing a convolutional neural network (CNN) to predict gate locations using the audio signals generated upon gate contact. By leveraging these predictions, we achieve fully automated computation of gate-to-gate timings. We conduct a comparative analysis between the CNN’s predictions and data obtained from an inertial measurement unit. Our findings reveal a strong predictive correlation between the two methods, with an R-squared value of 0.94 and a root mean squared error of 0.036. The majority of predictions demonstrate high accuracy, falling within a range of thousandths of a second. However, a few outliers negatively impact the overall performance. Notably, we observe no deterioration in predictive quality based on the distance between the camera and the gate. Finally, we delve into the challenges and limitations associated with our approach and provide a comprehensive discussion. To conclude, we outline potential avenues for future research and extensions of our methodology to the realm of slalom skiing.

One of the main challenges in surrogate modeling is the limited availability of data due to resource constraints associated with computationally expensive simulations. Multi-fidelity methods provide a solution by chaining models in a hierarchy with increasing fidelity, associated with lower error, but increasing cost. In this paper, we compare different multi-fidelity methods employed in constructing Gaussian process surrogates for regression. Non-linear autoregressive methods in the existing literature are primarily confined to two-fidelity models, and we extend these methods to handle more than two levels of fidelity. Additionally, we propose enhancements for an existing method incorporating delay terms by introducing a structured kernel. We demonstrate the performance of these methods across various academic and real-world scenarios. Our findings reveal that multi-fidelity methods generally have a smaller prediction error for the same computational cost as compared to the single-fidelity method, although their effectiveness varies across different scenarios.

preCICE is a free/open-source coupling library. It enables creating partitioned multi-physics simulations by gluing together separate software packages. This paper summarizes the development efforts in preCICE of the past five years. During this time span, we have turned the software from a working prototype -- sophisticated numerical coupling methods and scalability on ten thousands of compute cores -- to a sustainable and user-friendly software project with a steadily-growing community. Today, we know through forum discussions, conferences, workshops, and publications of more than 100 research groups using preCICE. We cover the fundamentals of the software alongside a performance and accuracy analysis of different data mapping methods. Afterwards, we describe ready-to-use integration with widely-used external simulation software packages, tests, and continuous integration from unit to system level, and community building measures, drawing an overview of the current preCICE ecosystem.

preCICE is a free/open-source coupling library. It enables creating partitioned multi-physics simulations by gluing together separate software packages. This paper summarizes the development efforts in preCICE of the past five years. During this time span, we have turned the software from a working prototype -- sophisticated numerical coupling methods and scalability on ten thousands of compute cores -- to a sustainable and user-friendly software project with a steadily-growing community. Today, we know through forum discussions, conferences, workshops, and publications of more than 100 research groups using preCICE. We cover the fundamentals of the software alongside a performance and accuracy analysis of different data mapping methods. Afterwards, we describe ready-to-use integration with widely-used external simulation software packages, tests, and continuous integration from unit to system level, and community building measures, drawing an overview of the current preCICE ecosystem.

Model order reduction (MOR) methods enable the generation of real-time-capable digital twins, with the potential to unlock various novel value streams in industry. While traditional projection-based methods are robust and accurate for linear problems, incorporating machine learning to deal with nonlinearity becomes a new choice for reducing complex problems. These kinds of methods are independent to the numerical solver for the full order model and keep the nonintrusiveness of the whole workflow. Such methods usually consist of two steps. The first step is the dimension reduction by a projection-based method, and the second is the model reconstruction by a neural network (NN). In this work, we apply some modifications for both steps respectively and investigate how they are impacted by testing with three different simulation models. In all cases Proper orthogonal decomposition is used for dimension reduction. For this step, the effects of generating the snapshot database with constant input parameters is compared with time-dependent input parameters. For the model reconstruction step, three types of NN architectures are compared: multilayer perceptron (MLP), explicit Euler NN (EENN), and Runge–Kutta NN (RKNN). The MLPs learn the system state directly, whereas EENNs and RKNNs learn the derivative of system state and predict the new state as a numerical integrator. In the tests, RKNNs show their advantage as the network architecture informed by higher-order numerical strategy.

Heterogeneous clusters are a widely utilized class of supercomputers assembled from different types of computing devices, for instance CPUs and GPUs, providing a huge computational potential. Programming them in a scalable way exploiting the maximal performance introduces numerous challenges such as optimizations for different computing devices, dealing with multiple levels of parallelism, the application of different programming models, work distribution, and hiding of communication with computation. We utilize the lattice Boltzmann method for fluid flow as a representative of a scientific computing application and develop a holistic implementation for large-scale CPU/GPU heterogeneous clusters. We review and combine a set of best practices and techniques ranging from optimizations for the particular computing devices to the orchestration of tens of thousands of CPU cores and thousands of GPUs. Eventually, we come up with an implementation using all the available computational resources for the lattice Boltzmann method operators. Our approach shows excellent scalability behavior making it future-proof for heterogeneous clusters of the upcoming architectures on the exaFLOPS scale. Parallel efficiencies of more than 90 % are achieved leading to 2604.72 GLUPS utilizing 24,576 CPU cores and 2048 GPUs of the CPU/GPU heterogeneous cluster Piz Daint and computing more than 6.8 × 10 9 lattice cells.

We present an Anderson acceleration-based approach to spatially couple three-dimensional Lattice Boltzmann and Navier–Stokes (LBNS) flow simulations. This allows to locally exploit the computational features of both fluid flow solver approaches to the fullest extent and yields enhanced control to match the LB and NS degrees of freedom within the LBNS overlap layer. Designed for parallel Schwarz coupling, the Anderson acceleration allows for the simultaneous execution of both Lattice Boltzmann and Navier–Stokes solver. We detail our coupling methodology, validate it, and study convergence and accuracy of the Anderson accelerated coupling, considering three steady-state scenarios: plane channel flow, flow around a sphere and channel flow across a porous structure. We find that the Anderson accelerated coupling yields a speed-up (in terms of iteration steps) of up to 40% in the considered scenarios, compared to strictly sequential Schwarz coupling.