Andreas Lintermann

Forschungszentrum Jülich · Institute for Advanced Simulation (IAS)

The report reflects an agreement based on the consensus conference of the International Standardization Committee on the Objective Assessment of the Nasal Airway in Riga, 2nd Nov. 2016. The aim of the conference was to address the existing nasal airway function tests and to take into account physical, mathematical and technical correctness as a bas...

On High-Performance Computing (HPC) systems, several hyperparameter configurations can be evaluated in parallel to speed up the Hyperparameter Optimization (HPO) process. State-of-the-art HPO methods follow a bandit-based approach and build on top of successive halving, where the final performance of a combination is estimated based on a lower than...

This manuscript presents the library AI4HPC with its architecture and components. The library enables large-scale trainings of AI models on High-Performance Computing systems. It addresses challenges in handling non-uniform datasets through data manipulation routines, model complexity through specialized ML architectures, scalability through extens...

Time-marching of turbulent flow fields is computationally expensive using traditional Computational Fluid Dynamics (CFD) solvers. Machine Learning (ML) techniques can be used as an acceleration strategy to offload a few time-marching steps of a CFD solver. In this study, the Transformer (TR) architecture, which has been widely used in the Natural L...

Hyperparameter optimization (HPO) of neural networks is a computationally expensive procedure, which requires a large number of different model configurations to be trained. To reduce such costs, this work presents a distributed, hybrid workflow, that runs the training of the neural networks on multiple graphics processing units (GPUs) on a classic...

Hyperparameter Optimization of neural networks is a computationally expensive procedure, which requires a large number of different model configurations to be trained. To reduce such costs, this work presents a distributed, hybrid workflow, that runs the training of the neural networks on multiple Graphics Processing Units (GPUs) on a classical sup...

This study presents a novel approach to using a gated recurrent unit (GRU) model, a deep neural network, to predict turbulent flows in a Lagrangian framework. The emerging velocity field is predicted based on experimental data from a strained turbulent flow, which was initially a nearly homogeneous isotropic turbulent flow at the measurement area....

Hyperparameter Optimization (HPO) of Neural Networks (NNs) is a computationally expensive procedure. On accelerators, such as NVIDIA Graphics Processing Units (GPUs) equipped with Tensor Cores, it is possible to speed-up the NN training by reducing the precision of some of the NN parameters, also referred to as mixed precision training. This paper...

High-Performance Computing (HPC) enables precise analysis of large and complex Earth Observation (EO) datasets. However, the adoption of supercomputing in the EO community faces challenges from the increasing heterogeneity of HPC systems, limited expertise, and the need to leverage novel computing technologies. This paper explores the implications...

The response of a laminar lean premixed flame to excitations based on its position in the combustion chamber and the stability of a burner are numerically investigated. A finite-volume large-eddy simulation method is used to solve the compressible Navier-Stokes equations and a combined G-equation progress variable approach is used to model the flam...

Dear Colleagues, It is out pleasure to invite you to submit research contributions to the next Platform for Advanced Scientific Computing (PASC) Conference series held at the Congress Center Davos, Switzerland, 26-28 June 2023. In a few words, the PASC Conference series is an international and interdisciplinary platform for exchanging knowledge i...

Continuously increasing data volumes from multiple sources, such as simulation and experimental measurements, demand efficient algorithms for an analysis within a realistic timeframe. Deep learning models have proven to be capable of understanding and analyzing large quantities of data with high accuracy. However, training them on massive datasets...

Energy resources and production are crucial matters for economic advancement. Based on recent global challenges, an energy producer needs to generate energy under the following constraints: harvest energy with minimal environmental impact, invest into appropriate energy projects under minimization of risk and initial expenditure, produce the energy...

Deep Learning models have proven necessary in dealing with the challenges posed by the continuous growth of data volume acquired from satellites and the increasing complexity of new Remote Sensing applications. To obtain the best performance from such models, it is necessary to fine-tune their hyperparameters. Since the models might have massive am...

Convolutional autoencoders are trained on exceptionally large actuated turbulent boundary layer simulation data (8.3 TB) on the high-performance computer JUWELS at the Jülich Supercomputing Centre. The parallelization of the training is based on a distributed data-parallelism approach. This method relies on distributing the training dataset to mult...

With the availability of large datasets and increasing high-performance computing resources, machine learning tools offer many opportunities to improve and/or augment numerical methods used in the field of computational fluid dynamics. A low-dimensional representation of a turbulent boundary layer flow field is generated by a plain and a physics-co...

Chapter “Machine-Learning-Based Control of Perturbed and Heated Channel Flows” was previously published non-open access. It has now been changed to open access under a CC BY 4.0 license and the copyright holder updated to ‘The Author(s)’.

Fluid mechanical properties of respiratory flow such as the pressure loss, temperature distribution, or the wall shear stress characterize the condition of a nasal cavity from a physical point of view. Simulations based on computational fluid dynamics methods are able to deliver in-depth details on respiration. Integrating such tools into virtual s...

Many simulation workflows require to prepare the data for the simulation manually. This is time consuming and leads to a massive bottleneck when a large number of numerical simulations is requested. This bottleneck can be overcome by an automated data processing pipeline. Such a novel pipeline is developed for a medical use case from rhinology, whe...

Physics-based analyses have the potential to consolidate and substantiate medical diagnoses in rhinology. Such methods are frequently subject to intense investigations in research. However, they are not used in clinical applications, yet. One issue preventing their direct integration is that these methods are commonly developed as isolated solution...

A reinforcement learning algorithm is coupled to a thermal lattice-Boltzmann method to control flow through a two-dimensional heated channel narrowed by a bump. The algorithm is allowed to change the disturbance factor of the bump and receives feedback in terms of the pressure loss and temperature increase between the inflow and out-flow region of...

The impact of the human nasal airway complexity on the pharyngeal airway fluid mechanics is investigated at inspiration. It is the aim to find a suitable degree of geometrical reduction that allows for an efficient segmentation of the human airways from cone-beam computed tomography images. The flow physics is simulated by a lattice Boltzmann metho...

We observe a continuously increased use of Deep Learning (DL) as a specific type of Machine Learning (ML) for data-intensive problems (i.e., ’big data’) that requires powerful computing resources with equally increasing performance. Consequently, innovative heterogeneous High-Performance Computing (HPC) systems based on multi-core CPUs and many-cor...

Benefiting and accessing high-performance computing resources can be quite difficult. Unlike domain scientists with a background in computational science, non-experts coming from, e.g., various medical fields, have almost no chance to run numerical simulations on large-scale systems. To provide easy access and a user-friendly interface to supercomp...

Benefiting and accessing high-performance computing resources can be quite difficult. Unlike domain scientists with a background in computational science, non-experts coming from, e.g., various medical fields, have almost no chance to run numerical simulations on large-scale systems. To provide easy access and a user-friendly interface to super-com...

In CFD modelling, small cells or elements are created to fill the volume to simulate the flow in. They constitute a mesh where each cell represents a discrete space that represents the flow locally. Mathematical equations that represent the flow physics are then applied to each cell of the mesh. Generating a high quality mesh is extremely important...

Using traditional computational fluid dynamics and aeroa-coustics methods, the accurate simulation of aeroacoustic sources requires high compute resources to resolve all necessary physical phenomena. In contrast, once trained, artificial neural networks such as deep encoder-decoder convolutional networks allow to predict aeroacoustics at lower cost...

A novel method, which combines both fluid‐mechanical experimental and numerical data from magnetic resonance velocimetry and Lattice‐Boltzmann (LB) simulations is presented. The LB method offers a unique and simple way of integrating the experimental data into the simulation by means of its equilibrium term. The simulation is guided by the experime...

Complex geometries pose multiple challenges to the field of computational fluid dynamics. Grid generation for intricate objects is often difficult and requires accurate and scalable geometrical methods to generate meshes for large-scale computations. Such simulations, furthermore, presume optimized scalability on high-performance computers to solve...

Multi-physics simulations are at the heart of today's engineering applications. The trend is towards more realistic and detailed simulations, which demand highly resolved spatial and temporal scales of various physical mechanisms to solve engineering problems in a reasonable amount of time. As a consequence, numerical codes need to run efficiently...

Transverse maxillary deficiency is a common pathological condition. Patients suffering from this pathology often have narrowed airways compared to healthy humans. To cure such an anatomic defective position, a new method, the Miniscrew-Assisted Rapid Maxillary Expansion (MARME), has been developed. In previous studies, the effects of this treatment...

Computational fluid dynamics methods enable to numerically predict complex flows with the help of computers. In the fields of Engineering and Physics they are already in use for decades to support design decissions and to get insight into complex physical phenomena. The simulation techniques have massively evolved over the past years and can nowada...

In late 2017, Hardkernel released the ODROID-MC1 cluster system, which is based on the ODROID-XU4 single-board computer. The cluster consists of four nodes, each equipped with a Samsung Exynos 5 Octa (5422) CPU. The system promises high computational power under low energy consumption. In this paper, the applicability of such a systems to scientifi...

The article describes how to setup a cluster system with a shared file system using the Network File System (NFS) and the Message Passing Interface (MPI) together with the cluster job scheduler SLURM on an ODROID-MC1 ARM-based mini cluster. Examples on how to run parallel computations on this system are given. An example from Computational Fluid Dy...

The immense increase of computational power in the past decades led to an evolution of numerical simulations in all kind of engineering applications. New developments in medical technologies in rhinology employ computational fluid dynamics methods to explore pathologies from a fluid-mechanics point of view. Such methods have grown mature and are ab...

Research and analysis of large amounts of data from scientific simulations, in-situ visualization, and application control are convincing scenarios for interactive supercomputing. The open-source software Jupyter (or JupyterLab) is a tool that has already been used successfully in many scientific disciplines. With its open and flexible web-based de...

Objectives: To evaluate changes in the volume and cross-sectional area of the nasal airway before and 1 year after nonsurgical miniscrew-assisted rapid maxillary expansion (MARME) in young adults. Materials and methods: Fourteen patients (mean age, 22.7 years; 10 women, four men) with a transverse discrepancy who underwent cone beam computed tom...

The Rhinodiagnost project prepares the implementation of a NOSE Service Center to deliver new extended possibilities of functional diagnostics to practicing physicians in rhinology and to advance personalized medicine. A new standardized 4-phase rhinomanometer will be used together with computational fluid dynamics simulations to obtain detailed in...

The horse-shoe vortex appearing in front of the circular cylinder placed on the boundary layer was numerically investigated. The unsteady, three-dimensional feature of the flow field was captured by a lattice-Boltzmann method, where more than 300 million grids were used. The visualized instantaneous flow field indicated a complex vortex structure b...

Respiration is an essential physiological functionality of the human organism and is responsible for supplying the body with oxygen. The nasal cavity takes care of olfaction and degustation, filters fine dust from the air as well as moisturizes and tempers the air. Therefore, it is indispensable in respiration, and a degradation of only one or a fe...

Classic hybrid methods for computational aeroacoustics use different solvers and methods to predict the flow field and the acoustic pressure field in two separate steps, which involves data exchange via disk I/O between the solvers. This limits the efficiency of the approach, as parallel I/O usually does not scale well to large numbers of cores. In...

The deposition of aerosol particles, e.g., fine dust particles, diesel aerosols, or wood dust, in the human lung is responsible for many respiratory diseases. Small respirable particles can cause inflammations of the bronchi, coughing, allergic reactions, and even lung cancer. The deposition of such aerosols in a realistic model of the upper human...

This book constitutes the thoroughly refereed post-conference proceedings of the First JARA High-Performance Computing Symposium, JARA-HPC 2016, held in Aachen, Germany, in October 2016. The 21 full papers presented were carefully reviewed and selected from 26 submissions. They cover many diverse topics, such as coupling methods and strategies in...

In this paper, a highly scalable numerical method is presented that allows to compute the aerodynamic sound from a turbulent flow field on HPC systems. A hybrid CFD-CAA method is used to compute the flow and the acoustic field, in which the two solvers are running in parallel to avoid expensive I/O operations for the acoustic source terms. Herein,...

This is a brochure of the Simulation Laboratory Highly Scalable Fluids and Solids Engineering (FSE) of the Jülich Aachen Research Alliance - High Performance Computing (JARA-HPC).

Highly resolved intrinsic geometrical shapes used in three-dimensional parallel simulations of fluid flows consume a large portion of the available memory when loaded serially on every process. This demands for a memory efficient implementation of a distributed geometry which is however a non-trivial task when complex spatial domain decomposition m...

Large-eddy simulations (LESs) of a helicopter engine jet and an axial fan are performed by using locally refined Cartesian hierarchical meshes. For the computations a high-fidelity, massively parallelized solver for compressible flow is used. To verify the numerical method, a coaxial hot round jet is computed and the results are compared to referen...

Ventricular Assist Devices (VADs) are commonly implanted to assist patients suffering from heart diseases. They provide long- and short-term support for the human heart and help patients to recover from heart attacks and from congestive heart failure. It is essential to design blood-sensitive VADs to minimize the risk of hemolysis and thrombosis. T...

We present numerical methods based on hierarchical Cartesian grids for the simulation of particle flows of different length scales. These include Eulerian-Lagrangian approaches for fully resolved moving particles with conjugate heat transfer as well as one-way coupled Lagrangian particle models for large-scale particle simulations. The domain decom...

In this work, a highly scalable numerical method is presented that allows to compute the aerodynamic sound from the flow field for large-scale problems. The acoustic perturbation equations are solved by a high-order discontinuous Galerkin method and by using the acoustic source terms obtained from an approximate solution of the Navier-Stokes equati...

Details Within this work the flow in the human nasal cavity is investigated using Computational Fluid Dynamics (CFD) methods. Different approaches exist to numerically simulate flows, e.g., Finite-Volume Methods (FVM), and Finite-Element Methods (FEM). However, when it comes to handling complex and intricate geometries like the nasal cavity, the L...

This study presents a comparison of numerical and experimental results of the steady flow field in the left main-bronchus of the upper human airways during exhalation. Stereo-particle-image velocimetry measurements were performed in multiple parallel measurement planes for a Reynolds number of Re_D = 700 based on the hydraulic diameter of the trach...

The automatic grid generation on high performance computers is a challenging task under the restriction of computational power and memory availability. The increasing demand for high grid resolutions to simulate complex flow configurations necessitates parallel grid generation on multicore machines with distributed memory. In this study, a new robu...

The flow in the human nasal cavity is of great importance to understand rhinologic pathologies like impaired respiration or heating capabilities, a diminished sense of taste and smell, and the presence of dry mucous membranes. To numerically analyze this flow problem a highly efficient and scalable Thermal Lattice-BGK (TLBGK) solver is used, which...

The human nose is a complex organ with different physiological functions. A treatment of dysfunctions of the human nose can therefore be difficult. One of the main reasons is a missing understanding of the flow phenomena which occur in the nasal cavity. The shape of the nasal cavity plays a major role in the heating and moisturizing capabilities an...

This systematic review aims first to summarize the previous areas of application of computational fluid dynamics (CFD) and then to demonstrate that CFD is also a suitable instrument for generating three-dimensional images that depict drug effects on nasal mucosa. Special emphasis is placed on the three-dimensional visualization of the antiobstructi...

The anatomy and the functionality of the human respiration system is well understood, whereas physical processes such as the details of the gas transport within the respiration cycle lack insight. The influence of the air flow on physiological functions like the sense of smell and taste is only one example which is of great interest. Furthermore, t...