Computational Fluid Dynamics Simulation Data Research Articles

On particle dispersion statistics using unsupervised learning and Gaussian mixture models

Understanding the dispersion of particles in enclosed spaces is crucial for controlling the spread of infectious diseases. This study introduces an innovative approach that combines an unsupervised learning algorithm with a Gaussian mixture model to analyze the behavior of saliva droplets emitted from a coughing individual. The algorithm effectively clusters data, while the Gaussian mixture model captures the distribution of these clusters, revealing underlying sub-populations and variations in particle dispersion. Using computational fluid dynamics simulation data, this integrated method offers a robust, data-driven perspective on particle dynamics, unveiling intricate patterns and probabilistic distributions previously unattainable. The combined approach significantly enhances the accuracy and interpretability of predictions, providing valuable insights for public health strategies to prevent virus transmission in indoor environments. The practical implications of this study are profound, as it demonstrates the potential of advanced unsupervised learning techniques in addressing complex biomedical and engineering challenges and underscores the importance of coupling sophisticated algorithms with statistical models for comprehensive data analysis. The potential impact of these findings on public health strategies is significant, highlighting the relevance of this research to real-world applications.

Prediction of Steam Turbine Blade Erosion Using Computational Fluid Dynamics Simulation Data and Hierarchical Machine Learning

Abstract The information of the degree of blade erosion is vital for the efficient operation of steam turbines. However, it is nearly impossible to directly measure the degree of blade erosion during operation. Moreover, collecting sufficient data of eroded cases for predictive analysis is challenging. Therefore, this paper proposes a blade erosion prediction method using numerical simulation and machine learning. Pressure data of several blade erosion cases are collected from the numerical turbine simulation. The machine learning approach involves training on collected simulation data to predict the degree of erosion for the first-stage stator (1S) and the first-stage rotor blade (1R) from internal pressure data. The proposed erosion prediction model employs a two-step hierarchical approach. First, the proposed model predicts the 1S erosion degree using the k-nearest neighbor (k-NN) regression. Second, the proposed model estimates the 1R erosion degree with linear regression models. These models are tailored for each of the 1S erosion degrees, utilizing pressure data processed through fast Fourier transform (FFT). The evaluation shows that the proposed method achieves the prediction of the 1S erosion with a mean absolute error (MAE) of 0.000693 mm and the 1R erosion with an MAE of 0.458 mm. The evaluation results indicate that the proposed method can accurately capture the degree of turbine blade erosion from internal pressure data. As a result, the proposed method suggests that the erosion prediction method can be effectively used to determine the optimal timing for maintenance, repair, and overhaul (MRO).

Performance optimization of cement calciner based on CFD simulation and machine learning algorithm

In order to improve the combustion efficiency and decomposition rate of the cement calciner and reduce pollutant emission, a performance optimization method based on Computational Fluid Dynamics (CFD) numerical simulation integrated with machine learning is proposed. The Multiphase Particle-in-cell (MP-PIC) method and the chemical reaction models are employed to simulate the coal combustion and CaCO3 decomposition process, whose calculation results are combined with the industrial practical data of the cement plant, so as to establish a more comprehensive training database. On this basis, a novel Topology Particle Swarm Optimization algorithm integrating with Convolutional Neural Network and Long Short-Term Memory (RITPSO–CNN–LSTM) algorithm model is established to predict combustion efficiency, decomposition rate, and NOx emission, respectively. Results show that compared with two other relative basic algorithm models, the prediction error of the proposed algorithm model for the three targets is minimal with the average relative error of 0.045 %, 0.038 %, and 0.021 %, respectively. The addition of CFD simulation data makes the prediction model more applicable with higher stability and accuracy. Based on the prediction results, Grey Wolf Optimizer (GWO) algorithm is employed to optimize operating parameters, and finally the average optimization amount of combustion efficiency, decomposition rate, and NOx emission are 2.17 %, 2.24 %, and 6.15 ppm, respectively, which meet the optimization requirements.

Single and multiphase flow leak detection in onshore/offshore pipelines and subsurface sequestration sites: An overview

Leaks may occur in existing pipelines, even when designed with quality construction and appropriate regulations. The economic impact of oil spills and natural gas dispersion from leaks can be huge. Failure to detect pipeline leaks promptly will have an adverse impact on life, the economy, the environment, and corporate reputation. Therefore, early detection of leaks, their location, and their size with high sensitivity and reliability are important for efficient hydrocarbon transportation through a pipeline, both in onshore and offshore applications. Although several studies have been conducted on leak detection using various techniques, recent literature that comprehensively investigates and summarizes the different multiphase leak detection techniques could not be found. Therefore, this paper provides a comprehensive review of the different leak detection techniques in pipelines, wellbores, and subsurface sequestration wells. This is done by studying the different multiphase flow leak detection techniques using various Computational Fluid Dynamics (CFD), Mechanistic, Machine Learning models, and digital twin techniques in the pipeline as well as in sub-surface sequestration sites. A comprehensive investigation revealed that a few studies have been conducted related to integrated multiphase flow leak experiments, computational fluid dynamics, mechanistic models, and implementing extended real-time transient monitoring using machine learning. This type of systematic investigation is deemed to be more useful for field applications. Furthermore, a new set of recommendations is provided in the last section which shows how experimental, mechanistic, and CFD simulation data can be used to drive a statistical approach based on modern deep learning and digital twin techniques. This allows for the precise understanding of the leak events such as size, location, and orientation of the leak, without sending a remotely operated underwater vehicle or aircraft to scan the whole pipeline and ocean.

Towards optimal design of patient isolation units in emergency rooms to prevent airborne virus transmission: From computational fluid dynamics to data-driven modeling

BackgroundPatient isolation units (PIUs) can be an effective method for effective infection control. Computational fluid dynamics (CFD) is commonly used for PIU design; however, optimizing this design requires extensive computational resources. Our study aims to provide data-driven models to determine the PIU settings, thereby promoting a more rapid design process. MethodUsing CFD simulations, we evaluated various PIU parameters and room conditions to assess the impact of PIU installation on ventilation and isolation. We investigated particle dispersion from coughing subjects and airflow patterns. Machine-learning models were trained using CFD simulation data to estimate the performance and identify significant parameters. ResultsPhysical isolation alone was insufficient to prevent the dispersion of smaller particles. However, a properly installed fan filter unit (FFU) generally enhanced the effectiveness of physical isolation. Ventilation and isolation performance under various conditions were predicted with a mean absolute percentage error of within 13%. The position of the FFU was found to be the most important factor affecting the PIU performance. ConclusionData-driven modeling based on CFD simulations can expedite the PIU design process by offering predictive capabilities and clarifying important performance factors. Reducing the time required to design a PIU is critical when a rapid response is required.

Capturing mesoscale structures in computational fluid dynamics simulations of gas‐solid flows

AbstractMultiphase reactors' performance depends on the mesoscale structures formed due to multiphase hydrodynamics. Examples of mesoscale structures include gas bubbles in a fluidized bed and particle clusters in a riser. Experimental investigation of these mesoscale structures is challenging and expensive. To this end, computational fluid dynamics (CFD) simulations are extensively employed; however, post‐processing CFD data to capture mesoscale structures is challenging. This article develops a density‐based spatial clustering of applications with noise (DBSCAN)‐based methodology to capture and characterize mesoscale structures from multiphase CFD simulation data. DBSCAN is an unsupervised machine‐learning algorithm, which requires the value of two hyperparameters. A simple technique to calculate these hyperparameters is provided and the performance of DBSCAN is assessed on CFD‐DEM simulations of bubbling fluidized beds and particle clustering. We demonstrate the computational complexity of DBSCAN to be , lower than the existing techniques, by testing its scalability on highly resolved grids (up to 100 million grid points).

Indoor pollution control based on surrogate model for residential buildings

Individuals typically spend most of their lives indoors, predominantly in spaces like offices and residences. Consequently, prolonged indoor exposure underscores the critical significance of maintaining optimal indoor air quality (IAQ) to safeguard one's health. The primary impediment to attaining efficient regulation of Indoor Air Quality (IAQ) is the challenge of monitoring the IAQ parameters, particularly within the immediate vicinity of an individual's breathing space. Current heating, ventilation, and air conditioning systems lack the ability to rapidly predict and optimize the quality of indoor air. The objective of this study is to acquire the distribution features of indoor pollutants and precise indoor environment data in order to efficiently forecast and enhance the IAQ. To achieve this objective, a proposed surrogate model was developed using computational fluid dynamics (CFD). Notably, the Kriging surrogate model can rapidly predict IAQ while using a limited number of CFD runs. CFD are widely used as numerical simulation methods to obtain the accurate information. Surrogate models can rapidly forecast indoor environmental conditions using CFD simulation data simultaneously. Optimization algorithms can efficiently achieve desirable indoor ambient conditions, offering highly effective and intelligent control techniques for indoor atmospheric ventilation.

Numerical modeling and evaluation of solid-liquid extraction with pressurized hot water extraction applied to Robinia Pseudoacacia wood

Computational Fluid Dynamics (CFD) simulation was used to study the solid-liquid extraction of Robinia Pseudoacacia wood by pressurized hot water extraction (PHWE). The proposed model couples momentum and mass transfer, including the extraction kinetics. The model is based on transport species and flow in a porous medium. The results show that the extraction of the two main flavonoids, Dihydrorobinetin (DHR) and Robinetin (Rob), increased with increasing temperature up to 120 °C, with values of 46.85±1.44 and 6.84±0.42 mg/gdB, respectively. However, at higher temperatures, these flavonoids degraded. Additionally, the exhaustive extraction cycle, ranging from 80 to 160 °C, was found to be less effective when compared to extraction at 120 °C.The kinetic parameters were included in the CFD simulation. As a result, a good agreement between the results of CFD simulation and experimental data was found with R2 > 0.97. The model predicts the temperature distribution inside the reactor during the pre-heating period, the evolution of the DHR and Rob concentrations at the reactor outlet, and the effect of degradation at high extraction temperatures.

Non-uniform state-based Markov chain model to improve the accuracy of transient contaminant transport prediction

Rapid prediction of indoor airborne contaminant distribution is crucial for controlling indoor air quality. Markov chain technology, which uses matrix multiplication, is computationally efficient in this field. The combination of computational fluid dynamics (CFD) with Markov chain technology has gained significant attention among researchers. The partitioning of the indoor zone directly affects the size and values of the transition matrix, which is the core of the Markov chain and impacts both the cost and accuracy of calculations. Although coarse matrix Markov chain models incur relatively lower computational costs, effective methods for state partitioning are lacking. This study proposes an improved method for non-uniform state partitioning based on velocity to enhance the computational accuracy of Markov chain models with coarse matrices. First, we used the airflow velocity field obtained by CFD simulation to divide the states and obtain non-uniform states. Subsequently, we obtained the transition matrix corresponding to non-uniform states and applied it to the Markov chain model for predictive calculation. We compared the predictive results of the non-uniform state-based Markov chain model with experimental and CFD simulation data, which showed that the model had a good predictive performance for transient contaminant transport and that the non-uniform state partitioning method improved predictive accuracy. Additionally, we discussed the effects of different numbers of states and time steps on computational accuracy.

Simultaneous reconstruction of 3D non-uniform temperature and velocity fields in a furnace using a bidirectional acoustic path separation tracking method

Acoustic measurement techniques applied to furnace measurements are based on the coupling of multiple fields, such as sound, temperature, and flow. To better measure high-temperature fluids within furnaces using acoustic technology, this study establishes a physical model that separates the bidirectional propagation path of sound. Additionally, a novel nonlinear acoustic ray-tracing method is proposed that allows the simultaneous measurement of temperature and flow. The combustion process of high-temperature swirling flow in a furnace is simulated, the effective path of sound propagation is tracked, and the existence of a sound shadow zone in the low-temperature jet region of the burner is discovered. After reasonably arranging sensor positions, three-dimensional nonuniform temperature and complex flow fields are simultaneously reconstructed by combining ray tracing and nonlinear acoustic tomography. The feasibility of the method is analyzed via the error analysis of the computational fluid dynamics simulation data and the reconstructed data. The proposed method provides a solution for improving the existing acoustic measurement techniques in furnaces.

Propulsion–Airframe Integration for Conceptual Redesign of a Low-Boom Supersonic Transport

A low-boom supersonic transport was generated in a previous low-boom multidisciplinary optimization study that used computational fluid dynamics (CFD) off-body pressure to compute the undertrack sonic boom ground signature and calibrated low-fidelity aerodynamic analyses to compute the mission performance metrics. This low-boom aircraft, referred to as the Mach 1.7 40-PAX concept, can carry 40 passengers for a low-boom overland mission with cruise Mach 1.7 for the airport pairs over the continental United States and has the potential to achieve an undertrack sonic boom ground noise level below 70 perceived level of decibels at the start of overland cruise (SOC). The engine for the Mach 1.7 40-PAX concept was designed using the Numerical Propulsion System Simulation (NPSS) for the mission analysis and modeled as a flow-through nacelle for CFD-based sonic boom analysis. To understand how the engine plume affects the undertrack ground signature, the Mach 1.7 40-PAX concept is redesigned in this paper after replacing the flow-through nacelles with CFD engines for sonic boom analysis using aeropropulsive CFD simulation. A process is developed for approximation of the NPSS engine at SOC by a CFD engine for aeropropulsive CFD simulation. The generated CFD engine has the identical nozzle boundary conditions and approximately the same mass flow and thrust as those of the NPSS engine at SOC. Then, the outer mold line of the configuration with the CFD engines is redesigned to approximately restore the low-boom characteristics of the Mach 1.7 40-PAX concept at SOC. Finally, the CFD simulation data for the redesigned concept with the CFD engines are used to calibrate the low-fidelity aerodynamic analyses for mission analysis of this concept. The cyclic dependency of the involved disciplinary analyses is resolved using an iteration method for a consistent coupling of the mission analysis, NPSS engine analysis, and low-boom redesign using the aeropropulsive CFD simulation. This low-boom redesign study is used as an example to demonstrate how the propulsion–airframe integration could be implemented for conceptual design of supersonic transports that satisfy both the low-boom and mission performance requirements.

Computational fluid dynamics validated by micro particle image velocimetry to estimate the risk of hemolysis in arteries with atherosclerotic lesions

We aimed to verify the relationship between blood vessel shape and hemolysis risk by using a blood rheological model that reflects the physiological processes related to blood flow. Blood rheology depends on many factors including the red blood cell concentration and local shear stress, which affect the hemolysis process. We introduced a rheology model suitable for modeling hemolysis flows observed in arteries with atherosclerotic lesions in vivo based on the population balance. The model predicts the concentration of single red blood cells and the concentration and size of red blood cell agglomerates, which affect blood rheology and hemolysis. Atherosclerotic lesion geometries were obtained based on image reconstructions from tomographic projections. Computational fluid dynamics (CFD) simulation results were compared with particle image velocimetry measurements of the geometries printed with a 3D printer. Based on the CFD simulation data, a correlation function was established by combining the essential parameters of vessel shape and flow rate with the maximum shear stress, which governs the hemolysis. The chemical engineering methodology was successfully applied to the analysis of biological systems. In future, the model can be implemented in image reconstruction software using tomographic projections to quickly analyze hemolysis risk in medical practice.

Application of the DMD Approach to High-Reynolds-Number Flow over an Idealized Ground Vehicle

This paper attempts to develop a Dynamic Mode Decomposition (DMD)-based Reduced Order Model (ROMs) that can quickly but accurately predict the forces and moments experienced by a road vehicle such that they be used by an on-board controller to determine the vehicle’s trajectory. DMD can linearize a large dataset of high-dimensional measurements by decomposing them into low-dimensional coherent structures and associated time dynamics. This ROM can then also be applied to predict the future state of the fluid flow. Existing literature on DMD is limited to low Reynolds number applications. This paper presents DMD analyses of the flow around an idealized road vehicle, called the Ahmed body, at a Reynolds number of 2.7×106. The high-dimensional dataset used in this paper was collected from a computational fluid dynamics (CFD) simulation performed using the Menter’s Shear Stress Transport (SST) turbulence model within the context of Improved Delayed Detached Eddy Simulations (IDDES). The DMD algorithm, as available in the literature, was found to suffer nonphysical dampening of the medium-to-high frequency modes. Enhancements to the existing algorithm were explored, and a modified DMD approach is presented in this paper, which includes: (a) a requirement of higher sampling rate to obtain a higher resolution of data, and (b) a custom filtration process to remove spurious modes. The modified DMD algorithm thus developed was applied to the high-Reynolds-number, separation-dominated flow past the idealized ground vehicle. The effectiveness of the modified algorithm was tested by comparing future predictions of force and moment coefficients as predicted by the DMD-based ROM to the reference CFD simulation data, and they were found to offer significant improvement.

An immersive virtual reality learning environment with CFD simulations: Unveiling the Virtual Garage concept.

Virtual reality has become a significant asset to diversify the existing toolkit supporting engineering education and training. The cognitive and behavioral advantages of virtual reality (VR) can help lecturers reduce entry barriers to concepts that students struggle with. Computational fluid dynamics (CFD) simulations are imperative tools intensively utilized in the design and analysis of chemical engineering problems. Although CFD simulation tools can be directly applied in engineering education, they bring several challenges in the implementation and operation for both students and lecturers. In this study, we develop the "Virtual Garage" as a task-centered educational VR application with CFD simulations to tackle these challenges. The Virtual Garage is composed of a holistic immersive virtual reality experience to educate students with a real-life engineering problem solved by CFD simulation data. The prototype is tested by graduate students (n = 24) assessing usability, user experience, task load and simulator sickness via standardized questionnaires together with self-reported metrics and a semi-structured interview. Results show that the Virtual Garage is well-received by participants. We identify features that can further leverage the quality of the VR experience with CFD simulations. Implications are incorporated throughout the study to provide practical guidance for developers and practitioners.

Development of an atmosphere temperature measurement system based on computational fluid dynamics and neural network algorithms

To study the impact of atmospheric temperature on architectural design, energy-saving control systems, and smart home systems, the accuracy of atmospheric temperature measurement systems must be increased by 0.1 °C or higher. However, given that existing temperature measurement systems have difficulty eliminating radiation interference, they may have a radiation error of approximately 1 °C. Therefore, we designed a novel temperature measurement system comprising three sensor probes and a radiation shield. The thin film of the platinum resistance probe was fixed at the center of a spherical copper shell. These plates can prevent all types of radiation from reaching the probe surface. The lower plate could effectively guide the airflow to the sensor probes. The radiation error can be decreased by reducing the radiation interference and increasing the air velocity. The system radiation errors under different environmental conditions were quantified using a computational fluid dynamics (CFD) approach. A neural network algorithm was used to fit the CFD simulation data to form a radiation error correction method with high accuracy and universality. Experiments conducted under different environments revealed that the average radiation error of the system was +0.09 °C. The root mean square error and mean absolute error between the experimental radiation errors and the radiation errors provided by the correction method were +0.037 °C and +0.031 °C, respectively. The correlation coefficient between the temperature of the new system and the reference temperature was 0.999. Our findings suggest that this new system may potentially reduce the radiation error to within 0.1 °C.

STUDI DISTRIBUSI UDARA PENDINGIN REEFER CONTAINER IKAN PADA KERETA API MENGGUNAKAN CFD

Fresh fish is sensitive to storage temperature. Temperatures above 2 °C accelerate the growth of bacteria and cause spoilage. The optimal storage temperature for frozen fish is -20 °C. Reefer container with a closed cooling system serves to preserve by maintaining the temperature of frozen fish. The problem of this research is the uneven distribution of cold air in the reefer containers of the train, thereby reducing the quality of frozen fish. The purpose of this research is to increase the even distribution of air, velocity, and pressure in the reefer container of the train. The solution to the problem is simulation and analysis of the condition of the reefer container train using Computational Fluid Dynamics (CFD). Reefer containers without guide plates are simulated to find out the area of the train reefer containers that are not evenly distributed. The distribution of air in the reefer container of the train is improved even by the addition of guide plates with variations in angles of 40°, 50°, and 60°. This research produces qualitative data (velocity contour and pressure contour reefer container) and quantitative (average velocity and pressure). The results of this study are CFD simulation data of uniform distribution of air, speed, and pressure on the reefer container train with a 40⁰ guide plate.

An Analysis of the Impact of Building Wind by Field Observation in Haeundae LCT Area, South Korea: Typhoon Omais in 2021

In the Haeundae area of Busan, South Korea, damage has continued to occur recently from building wind from caused by dense skyscrapers. Five wind observation stations were installed near LCT residential towers in Haeundae to analyze the effect of building winds during typhoon Omais. The impact of building wind was analyzed through relative and absolute evaluations. At an intersection located southeast of LCT (L-2), the strongest wind speed was measured during the monitoring. The maximum average wind speed for one minute was observed to be 38.93 m/s, which is about three times stronger than at an ocean observation buoy (12.7 m/s) at the same time. It is expected that 3 to 4 times stronger wind can be induced under certain conditions compared to the surrounding areas due to the building wind effect. In a Beaufort wind scale analysis, the wind speed at an ocean observatory was mostly distributed at Beaufort number 4, and the maximum was 8. At L-2, more than 50% of the wind speed exceeded Beaufort number 4, and numbers up to 12 were observed. However, since actual measurement has a limitation in analyzing the entire range, cross-validation with computational fluid dynamics simulation data is required to understand the characteristics of building winds.

Prediction of Mixing Uniformity of Hydrogen Injection inNatural Gas Pipeline Based on a Deep Learning Model

It is economical and efficient to use existing natural gas pipelines to transport hydrogen. The fast and accurate prediction of mixing uniformity of hydrogen injection in natural gas pipelines is important for the safety of pipeline transportation and downstream end users. In this study, the computational fluid dynamics (CFD) method was used to investigate the hydrogen injection process in a T-junction natural gas pipeline. The coefficient of variation (COV) of a hydrogen concentration on a pipeline cross section was used to quantitatively characterize the mixing uniformity of hydrogen and natural gas. To quickly and accurately predict the COV, a deep neural network (DNN) model was constructed based on CFD simulation data, and the main influencing factors of the COV including flow velocity, hydrogen blending ratio, gas temperature, flow distance, and pipeline diameter ratio were taken as input nodes of the DNN model. In the model training process, the effects of various parameters on the prediction accuracy of the DNN model were studied, and an accurate DNN architecture was constructed with an average error of 4.53% for predicting the COV. The computational efficiency of the established DNN model was also at least two orders of magnitude faster than that of the CFD simulations for predicting the COV.

Analyzing the effect of using axial impellers in large-scale bioreactors.

In high-performance industrial fermentation processes, stirring and aeration may account for significant production costs. Compared to the widely applied Rushton impellers, axial-pumping impellers are known to yield a lower power draw and at the same time improve mixing. However, their lower gas dispersion capability requires stronger agitation, compromising these benefits. Diverse advanced impeller forms have been developed to cope with this challenge. We apply alternating radial and axial impellers and demonstrate strong gas dispersion and energy-efficient mixing for the first time in a large-scale (160 m3 ) bioreactor, based on experimental and computational fluid dynamics simulation data. For equal operating conditions (stirrer speed, aeration rate), this setup yielded similar gas hold-ups and better mixing times (35%) compared to a classical Rushton-only configuration. Hence, applying a radial impeller on an upper level for improving gas dispersion maintains the benefits of axial impellers in terms of reducing energy demand (up to 50%). We conclude that this effect is significant only at large-scale, when bubbles substantially expand due to the release of the hydrostatic pressure and have time to coalesce. The work thus extends current knowledge on mixing and aeration of large-scale reactors using classical impeller types.

Geometry aware physics informed neural network surrogate for solving Navier–Stokes equation (GAPINN)

Many real world problems involve fluid flow phenomena, typically be described by the Navier–Stokes equations. The Navier–Stokes equations are partial differential equations (PDEs) with highly nonlinear properties. Currently mostly used methods solve this differential equation by discretizing geometries. In the field of fluid mechanics the finite volume method (FVM) is widely used for numerical flow simulation, so-called computational fluid dynamics (CFD). Due to high computational costs and cumbersome generation of the discretization they are not widely used in real time applications. Our presented work focuses on advancing PDE-constrained deep learning frameworks for more real-world applications with irregular geometries without parameterization. We present a Deep Neural Network framework that generate surrogates for non-geometric boundaries by data free solely physics driven training, by minimizing the residuals of the governing PDEs (i.e., conservation laws) so that no computationally expensive CFD simulation data is needed. We named this method geometry aware physics informed neural network—GAPINN. The framework involves three network types. The first network reduces the dimensions of the irregular geometries to a latent representation. In this work we used a Variational-Auto-Encoder (VAE) for this task. We proposed the concept of using this latent representation in combination with spatial coordinates as input for PINNs. Using PINNs we showed that it is possible to train a surrogate model purely driven on the reduction of the residuals of the underlying PDE for irregular non-parametric geometries. Furthermore, we showed the way of designing a boundary constraining network (BCN) to hardly enforce boundary conditions during training of the PINN. We evaluated this concept on test cases in the fields of biofluidmechanics. The experiments comprise laminar flow (Re = 500) in irregular shaped vessels. The main highlight of the presented GAPINN is the use of PINNs on irregular non-parameterized geometries. Despite that we showed the usage of this framework for Navier Stokes equations, it should be feasible to adapt this framework for other problems described by PDEs.